Theory of Elasticity

Aldo Maceri

Theory of Elasticity

123
Prof. Dr.-Ing. Aldo Maceri
Universitá Roma Tre
Departimento di Ingegneria Meccanica e
Industriale
Via della Vasca Navale, 79
00146 Roma
Italy
aldo.maceri@fastwebnet.it

ISBN 978-3-642-11391-8 e-ISBN 978-3-642-11392-5

DOI 10.1007/978-3-642-11392-5
Springer Heidelberg Dordrecht London New York

Library of Congress Control Number: 2010923688

© Springer-Verlag Berlin Heidelberg 2010

This work is subject to copyright. All rights are reserved, whether the whole or part of the material is
concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting,
reproduction on microfilm or in any other way, and storage in data banks. Duplication of this publication
or parts thereof is permitted only under the provisions of the German Copyright Law of September 9,
1965, in its current version, and permission for use must always be obtained from Springer. Violations
are liable to prosecution under the German Copyright Law.
The use of general descriptive names, registered names, trademarks, etc. in this publication does not
imply, even in the absence of a specific statement, that such names are exempt from the relevant protective
laws and regulations and therefore free for general use.

Cover design: eStudio Calamar S.L.

Printed on acid-free paper

Springer is part of Springer Science+Business Media (www.springer.com)

Preface

The Theory of elasticity studies the behavior of those bodies that recover their ini-
tial state when the causes which produce deformations are removed. Its results
constitute the foundations of the Theory of structures and then are of maximum
importance for engineers.
The Theory of elasticity moves freely within an unified mathematical frame-
work that provides the analytical tools for calculating stresses and deformations in
a strained elastic body. All the elastic problems can be exactly analyzed employing
the classical Mathematical analysis, with the exception of the unilateral problems
for which the employment of the Functional analysis is mandatory.
The Theory of elasticity was founded by the famous mathematician Cauchy in the
eighteenth-century. During its historical development this scientific sector proposed
to the mathematicians various problems that have contributed or entirely generated
the development of complex mathematical theories, as the Variational calculus and
the Finite element method.
The matter analyzed in this book is

– three-dimensional problems (Chap. 1), and particularly the problem of Saint

Venant (Chap. 1),
– two-dimensional problems, as panels, plates, shells (Chap. 3),
– one-dimensional problems, as ropes, beams, arches (Chap. 4),
– thermal stress problems (Chap. 5),
– stability problems (Chap. 6),
– anisotropic problems, that constitute the basic tool for the analysis of structures in
composite material (Chap. 7),
– nonlinear elastic problems, as finite elasticity and unilateral problems (Chap. 8).

In this book I have constantly kept in mind the practical application of the theo-
retical results. So I have always tried to give to engineers, in a simple form, a clear
indication of the necessary fundamental knowledge of the Theory of elasticity. In
the past some techniques of calculation were developed for particular elastic prob-
lems that cannot be organized in mathematical theories but are extremely simple to
apply. Such technical theories have always furnished results experimentally verified

v
vi Preface

with good approximation and then among them I have presented those that are still
useful tools of verification in the Structural design.
Throughout the analysis of the elastic problems my constant focus has been to
achieve the maximum clarity and because of this I have sacrificed various bright
discussions. I have developed the treatment of the subjects in classical way, but to
the light of the modern Mathematical theory of the elasticity and with more accented
relief to the connections with the Thermodynamics. Just for this, to give a clear
justification of the fundamental equation of the Thermoelasticity I have applied a
technique of analysis proper of the Fluid dynamics. However in the discussion of
the unilateral problems, where the Functional analysis is compulsory, I have related
in details the mathematical aspects of the theoretical analysis.

Roma, Italy Aldo Maceri

October 2009
Contents

1 The Three-Dimensional Problem . . . . . . . . . . . . . . . . . . . . 1

1.1 Analysis of Strain . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1 Components of Displacement . . . . . . . . . . . . . . . . 1
1.1.2 Infinitesimal Deformation . . . . . . . . . . . . . . . . . . 2
1.1.3 Elongation and Shearing Strain . . . . . . . . . . . . . . . 4
1.1.4 Small Deformations . . . . . . . . . . . . . . . . . . . . . 5
1.1.5 Components of Strain . . . . . . . . . . . . . . . . . . . . 9
1.1.6 Principal Direction of Strain . . . . . . . . . . . . . . . . 14
1.1.7 Invariants of Strain . . . . . . . . . . . . . . . . . . . . . 21
1.1.8 Plane State of Strain . . . . . . . . . . . . . . . . . . . . . 23
1.1.9 Equations of Compatibility . . . . . . . . . . . . . . . . . 24
1.1.10 Measurement of Strain . . . . . . . . . . . . . . . . . . . 25
1.2 Analysis of Stress . . . . . . . . . . . . . . . . . . . . . . . . . . 27
1.2.1 Stress Vector . . . . . . . . . . . . . . . . . . . . . . . . . 27
1.2.2 Normal Stress – Shearing Stress . . . . . . . . . . . . . . 29
1.2.3 Components of Stress . . . . . . . . . . . . . . . . . . . . 30
1.2.4 Symmetry of τ – Differential Equations
of Equilibrium – Cauchy’s Boundary Conditions . . . . . . 31
1.2.5 Symmetry of Stress Vector . . . . . . . . . . . . . . . . . 38
1.2.6 Relations Between Normal or Shearing Stress
and Components of Stress . . . . . . . . . . . . . . . . . . 39
1.2.7 Principal Direction of Stress . . . . . . . . . . . . . . . . 40
1.2.8 Invariants of Stress . . . . . . . . . . . . . . . . . . . . . 42
1.2.9 Mohr’s Circle . . . . . . . . . . . . . . . . . . . . . . . . 43
1.2.10 Mohr’s Principal Circles . . . . . . . . . . . . . . . . . . 57
1.2.11 Determination of the Maximum Normal Stress
or Shearing Stress by the Mohr’s Principal Circles . . . . . 61
1.2.12 Plane State of Stress . . . . . . . . . . . . . . . . . . . . . 63
1.2.13 Uniaxial State of Stress . . . . . . . . . . . . . . . . . . . 65
1.2.14 Measurement of Stress . . . . . . . . . . . . . . . . . . . 66
1.3 Principle of Virtual Works . . . . . . . . . . . . . . . . . . . . . 66
1.3.1 Principle of Virtual Works . . . . . . . . . . . . . . . . . 66

vii
viii Contents

1.4 Relations Between Stress and Strain . . . . . . . . . . . . . . . . 70

1.4.1 Tensile Breaking Test . . . . . . . . . . . . . . . . . . . . 70
1.4.2 Homogeneous and Isotropic Materials – Navier’s Relations 75
1.4.3 Bounds for the Elastic Modulus . . . . . . . . . . . . . . . 78
1.5 The Elastic Equilibrium Problem . . . . . . . . . . . . . . . . . . 80
1.5.1 Classical Formulations . . . . . . . . . . . . . . . . . . . 80
1.5.2 Variational Formulations . . . . . . . . . . . . . . . . . . 90
1.6 Strain Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
1.6.1 Elements of Thermodynamics . . . . . . . . . . . . . . . 95
1.6.2 Thermodynamics of the Problem of the Elastic Equilibrium 101
1.6.3 Strain Work . . . . . . . . . . . . . . . . . . . . . . . . . 103
1.6.4 The Elastic Potential . . . . . . . . . . . . . . . . . . . . 107
1.6.5 Work Theorems . . . . . . . . . . . . . . . . . . . . . . . 110
1.7 Strength Criterions . . . . . . . . . . . . . . . . . . . . . . . . . 115
1.7.1 Structural Safety . . . . . . . . . . . . . . . . . . . . . . . 115
1.7.2 The Maximum Shearing Stress Criterion . . . . . . . . . . 116
1.7.3 The Octahedral Shearing Stress Criterion . . . . . . . . . . 117
1.7.4 The Energetic Criterion . . . . . . . . . . . . . . . . . . . 118
1.7.5 The Intrinsic Curve Criterion . . . . . . . . . . . . . . . . 124
2 The Problem of Saint Venant . . . . . . . . . . . . . . . . . . . . . . 127
2.1 Geometry of Areas . . . . . . . . . . . . . . . . . . . . . . . . . 127
2.1.1 Centroid . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
2.1.2 Inertia Centroidal Ellipse . . . . . . . . . . . . . . . . . . 131
2.1.3 Antipolarity . . . . . . . . . . . . . . . . . . . . . . . . . 141
2.1.4 Inertia Centroidal Kernel . . . . . . . . . . . . . . . . . . 159
2.2 The Problem of Saint Venant . . . . . . . . . . . . . . . . . . . . 164
2.2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 164
2.2.2 State of Stress . . . . . . . . . . . . . . . . . . . . . . . . 169
2.3 Bending . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
2.3.1 Right Bending of Axis x . . . . . . . . . . . . . . . . . . 173
2.3.2 Right Bending of Axis y . . . . . . . . . . . . . . . . . . 187
2.3.3 Deviated Bending . . . . . . . . . . . . . . . . . . . . . . 190
2.4 Axial Load . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
2.4.1 Centroidal Axial Load . . . . . . . . . . . . . . . . . . . . 195
2.4.2 Non Centroidal Axial Load . . . . . . . . . . . . . . . . . 201
2.4.3 Material Non Resistant to Traction . . . . . . . . . . . . . 215
2.5 Torsion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
2.5.1 The Exact Solution . . . . . . . . . . . . . . . . . . . . . 219
2.5.2 The Circular Cross Section . . . . . . . . . . . . . . . . . 226
2.5.3 The Stress Concentration . . . . . . . . . . . . . . . . . . 232
2.5.4 Closed Thin Walled Cross Section . . . . . . . . . . . . . 240
2.5.5 Open Thin Walled Cross Section . . . . . . . . . . . . . . 250
2.5.6 Non Uniform Torsion . . . . . . . . . . . . . . . . . . . . 264
2.6 Shear . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 278
2.6.1 The Exact Solution . . . . . . . . . . . . . . . . . . . . . 278
Contents ix

2.6.2 The Approximate Solution . . . . . . . . . . . . . . . . . 294

2.6.3 The Circular Cross Section . . . . . . . . . . . . . . . . . 304
2.6.4 Open Thin Walled Cross Section . . . . . . . . . . . . . . 308
2.6.5 Closed Thin Walled Cross Section . . . . . . . . . . . . . 351
3 The Two-Dimensional Problems . . . . . . . . . . . . . . . . . . . . 359
3.1 Panels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 359
3.1.1 The Problem of the Panel . . . . . . . . . . . . . . . . . . 359
3.1.2 Rectangular Panels . . . . . . . . . . . . . . . . . . . . . 364
3.1.3 Circular Panels . . . . . . . . . . . . . . . . . . . . . . . 370
3.1.4 Effect of a Hole . . . . . . . . . . . . . . . . . . . . . . . 387
3.2 Plates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 392
3.2.1 Small Deflections of Thin Plates . . . . . . . . . . . . . . 392
3.2.2 Thin Plates on Elastic Foundation . . . . . . . . . . . . . 402
3.3 Shells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403
3.3.1 Membranes . . . . . . . . . . . . . . . . . . . . . . . . . 403
3.3.2 Thin Shells . . . . . . . . . . . . . . . . . . . . . . . . . 411
3.4 Plane Strain Problems . . . . . . . . . . . . . . . . . . . . . . . . 422
3.4.1 Plane Strain Problems . . . . . . . . . . . . . . . . . . . . 422
4 The One-Dimensional Problems . . . . . . . . . . . . . . . . . . . . 427
4.1 Ropes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 427
4.1.1 The Funicular . . . . . . . . . . . . . . . . . . . . . . . . 427
4.1.2 The Ropes . . . . . . . . . . . . . . . . . . . . . . . . . . 430
4.2 Beams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 433
4.2.1 The Deflected Beam . . . . . . . . . . . . . . . . . . . . . 433
4.2.2 The Analogy of Mohr . . . . . . . . . . . . . . . . . . . . 449
4.2.3 Principle of Virtual Works . . . . . . . . . . . . . . . . . 462
4.2.4 Strain Energy . . . . . . . . . . . . . . . . . . . . . . . . 463
4.2.5 Deflected Beams on Elastic Foundation . . . . . . . . . . 481
4.3 Arches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 482
4.3.1 Arches with Small Curvature . . . . . . . . . . . . . . . . 482
4.3.2 Arches with Great Curvature . . . . . . . . . . . . . . . . 488
5 Thermoelasticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 493
5.1 Mechanics of Continuous Media . . . . . . . . . . . . . . . . . . 493
5.1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 493
5.1.2 Classical Thermodynamics . . . . . . . . . . . . . . . . . 494
5.1.3 The Equations of Balance . . . . . . . . . . . . . . . . . . 501
5.1.4 Thermodynamics of the Irreversible Processes . . . . . . . 503
5.2 Fluid Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . 506
5.2.1 The Mathematical Model . . . . . . . . . . . . . . . . . . 506
5.2.2 The Characteristic Numbers . . . . . . . . . . . . . . . . 510
5.2.3 Non Dissipative Flows . . . . . . . . . . . . . . . . . . . 524
5.2.4 Dissipative Flows . . . . . . . . . . . . . . . . . . . . . . 537
5.3 Mechanics of Solids . . . . . . . . . . . . . . . . . . . . . . . . . 552
5.3.1 The Dynamic Thermoelastic Problem . . . . . . . . . . . 552
x Contents

5.3.2 The Thermoelastic Dissipation . . . . . . . . . . . . . . . 562

5.3.3 The Uncoupled Thermoelastic Problem . . . . . . . . . . 563
5.3.4 Thermoelasticity . . . . . . . . . . . . . . . . . . . . . . 566
5.3.5 The One-Dimensional Problem . . . . . . . . . . . . . . . 573
6 Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 593
6.1 Stability of the Elastic Equilibrium . . . . . . . . . . . . . . . . . 593
6.1.1 The Buckling . . . . . . . . . . . . . . . . . . . . . . . . 593
6.1.2 The Ultimate Strength . . . . . . . . . . . . . . . . . . . . 595
6.2 Energy Method . . . . . . . . . . . . . . . . . . . . . . . . . . . 596
6.2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 596
6.2.2 Some Elementary Application . . . . . . . . . . . . . . . 598
6.3 Static Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . 607
6.3.1 The Beam Axially Loaded . . . . . . . . . . . . . . . . . 607
6.3.2 Approximate Analysis . . . . . . . . . . . . . . . . . . . 607
6.3.3 Exact Analysis . . . . . . . . . . . . . . . . . . . . . . . 618
6.3.4 Effect of the Imperfections . . . . . . . . . . . . . . . . . 622
6.3.5 Limit Slenderness . . . . . . . . . . . . . . . . . . . . . . 624
6.3.6 Other Ways of Buckling . . . . . . . . . . . . . . . . . . . 627
6.4 Second Type Instability . . . . . . . . . . . . . . . . . . . . . . . 629
6.4.1 The Snapping . . . . . . . . . . . . . . . . . . . . . . . . 629
7 Anisotropy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 635
7.1 The Three-Dimensional Anisotropic Problem . . . . . . . . . . . 635
7.1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 635
7.1.2 Constituent Links . . . . . . . . . . . . . . . . . . . . . . 636
7.1.3 The Anisotropic Elastic Body . . . . . . . . . . . . . . . . 639
7.1.4 Energetic Aspects . . . . . . . . . . . . . . . . . . . . . . 641
7.1.5 The Anisotropic Saint Venant’s Problem . . . . . . . . . . 642
7.2 The Macroscopic Anisotropy . . . . . . . . . . . . . . . . . . . . 649
7.2.1 Composite Materials . . . . . . . . . . . . . . . . . . . . 649
7.2.2 Structural Anisotropy . . . . . . . . . . . . . . . . . . . . 656
8 Nonlinear Elasticity . . . . . . . . . . . . . . . . . . . . . . . . . . . 663
8.1 Nonlinear Problems . . . . . . . . . . . . . . . . . . . . . . . . . 663
8.1.1 The Nonlinearity Causes . . . . . . . . . . . . . . . . . . 663
8.2 Finite Deformations . . . . . . . . . . . . . . . . . . . . . . . . . 664
8.2.1 Three-Dimensional Problem . . . . . . . . . . . . . . . . 664
8.2.2 Large Deflections of Thin Plates . . . . . . . . . . . . . . 669
8.3 Unilateral Problems . . . . . . . . . . . . . . . . . . . . . . . . . 673
8.3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 673
8.3.2 Contact Problems . . . . . . . . . . . . . . . . . . . . . . 674
8.3.3 Unilateral Constraints . . . . . . . . . . . . . . . . . . . . 684
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 705
Author Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 707
Subject Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 709
Chapter 1
The Three-Dimensional Problem

1.1 Analysis of Strain

1.1.1 Components of Displacement

Let the solid body C, elastic or not, occupies with continuity some region V of the
three-dimensional space. Let us employ a Cartesian orthogonal reference frame O,
x, y, z and denote with x, y, z the unit vectors of the coordinate axes x, y, z (Fig. 1.1.1).
When C deforms it, its typical point P = (x, y, z) moves in the position P =
(x , y , z ) and in total C will occupy, after deformation, a new region of space V
(Fig. 1.1.2).
We denote with s the vectorial function that to every point P = (x, y, z) of V
associates the vector s(P) having for ends P and P and directed in the verse that
goes from P to P (Fig. 1.1.2). P is the point occupied by P after deformation.
The vector s(P) is called displacement of the point P. Evidently, if the function s is
known, the deformation of C is perfectly understood.
We call components of displacement and we denote with the symbols u, v, w the
components of s according to the coordinate axes x, y, z. Therefore,

s = ux + vy + wz in V,

that is

∀(x, y, z) ∈ V s(x, y, z) = u(x, y, z)x + v(x, y, z)y + w(x, y, z) z.

Furthermore

x = x + u(x, y, z)
∀(x, y, z) ∈ V y = y + v(x, y, z)
z = z + w(x, y, z).

Obviously, if the functions u, v, w are known, the deformation of C is also

known.

A. Maceri, Theory of Elasticity, DOI 10.1007/978-3-642-11392-5_1, 1


C Springer-Verlag Berlin Heidelberg 2010
2 1 The Three-Dimensional Problem

Fig. 1.1.1 y

y
V
O
x
x
z
z

V V′
P = (x,y,z)

y
s(P)
x P′
z

Fig. 1.1.2

We suppose that the frontier ∂V of V is a regular surface S of 3 , which we

decompose in two parts Su and Sp . In every point of Su there are constraints that
impose to u, v, w and/or their first derivatives to assume assigned values. We notice
that Su can be the void set. The external superficial load is applied on Sp .
In this book, we will consider only compatible deformations in the following
sense. Fractures of material are excluded, i.e. we exclude that a point of V can be
moved, to happened deformation, in two distinct points P and P . The overlaps of
material are excluded, i.e. we exclude that two distinct points P and Q of V can be
moved, to happened deformation, in a same point P . The respect of such conditions
is evidently translated, as for the analytical aspects, in hypothesis of regularity1.1.1
of the components of the displacement (i.e. u, v, w and their partial derivatives are
supposed univocal and continuous).

1.1.2 Infinitesimal Deformation

Let us consider a point P = (x, y, z) of V and denote with I a neighborhood of P (i.e.
an open set of 3 containing P). The deformation of I admits a simple mathematical
model and as more precise as smaller it is I.
To prove this we denote with P̄ = (x̄, ȳ, z̄) any point of I. By hypothesis u, v, w are
continuous in V together with their first order partial derivatives. Then, for a known

1.1.1 This section of the Mathematical theory of elasticity (formulation of the problem of the elas-
tic equilibrium in classical terms) involves delicate matters of the modern Mathematical analysis
whose study is still incomplete.
1.1 Analysis of Strain 3

theorem of the classical Mathematical analysis, u, v, w are differentiable in P. As a

consequence, we have, approximately

∂u
u(x̄, ȳ, z̄) = u(x, y, z) + (x, y, z)(x̄ − x)
∂x
∂u ∂u
+ (x, y, z) (ȳ − y) + (x, y, z) (z̄ − z)
∂y ∂z
∂v
v(x̄, ȳ, z̄) = v(x, y, z) + (x, y, z) (x̄ − x)
∂x
(1.1.1)
∂v ∂v
+ (x, y, z) (ȳ − y) + (x, y, z) (z̄ − z)
∂y ∂z
∂w
w(x̄, ȳ, z̄) =w(x, y, z) + (x, y, z) (x̄ − x)
∂x
∂w ∂w
+ (x, y, z) (ȳ − y) + (x, y, z) (z̄ − z).
∂y ∂z

Furthermore, the smaller the distance of P̄ from P, the best is the approximation
of these relations.
By employing the procedures of the Analytical geometry, these relations allow
to prove that a line passing through P turns it (to happened deformation) into a line
and that a plane passing through P turns it (to happened deformation) into a plane.
Precisely, let r be a line passing through P and let P be the position of P after
the deformation (Fig. 1.1.3). We can prove that all the points of V ranged before
the deformation on r are found after the deformation on a curve c passing through
P . Furthermore, we can prove that with good approximation a small segment of
r containing P turns after the deformation to a small segment of r containing P ,
where r is the tangent to c at P (Fig. 1.1.3). So we have

[1.1.1] In a neighborhood of P the lines are turned into lines and the planes they are
transformed into planes. 

Remark 1.1.1 In mathematical terms, we have linearized u, v, w in a neighborhood

of P. In fact, we have assumed for u, v, w linear expressions in x, y, z.

r r’

P’
P
s(P)

Fig. 1.1.3 c
4 1 The Three-Dimensional Problem

1.1.3 Elongation and Shearing Strain

Let P be a point of C, r be a line (of the three-dimensional space) passing
through P, Q be any point of r (Fig. 1.1.4) different from P and belonging to a
neighborhood of P. As usual, we denote with P the point occupied by P to hap-
pened deformation and with Q the point occupied by Q to happened deformation
(Fig. 1.1.4).
To happened deformation, the points of r are lined up on a line r passing through

P (Fig. 1.1.4). This way, the segment (which can be small as we like) of r of ends
P and Q is turned into the segment of r of extreme P and Q . We choose the nota-
tions1.1.2 l = dist(P, Q), l = dist(P , Q ) and we call elongation according the
direction r (at the point P) the dimensionless real number1.1.3

l − l
εr = . (1.1.2)
l
Let P be a point of C, r and t be two lines (of the space) passing through P, P
be the point occupied by P to happened deformation, r and t be the transform of r
and t (Fig. 1.1.5). Let we denote with ϑ the angle between r and t, and with ϑ  the
angle between r and t . We call shearing strain between the directions r and t (at
the point P) the dimensionless real number (Fig. 1.1.5)

γrt = ϑ − ϑ  . (1.1.3)

Q s(Q)
Q′

P′
P s(P)

Fig. 1.1.4
r r′

r t
r′ t′

s(P) Q

Fig. 1.1.5

1.1.2 We denote with dist(A,B) the distance between any two points A, B.
1.1.3 The definition of elongation according the direction r (at the point P) in use in the
dist(P , Q )−dist(P, Q)
Mathematical theory of the elasticity is εr = limQ→F dist(P, Q) .
1.1 Analysis of Strain 5

Remark 1.1.2 The shearing strain is measured in radians. 

Remark 1.1.3 Obviously γrt = γtr . 

1.1.4 Small Deformations

We will say that we are in the field of the small deformations when

∂u ∂u ∂u ∂v ∂v ∂v ∂w ∂w ∂w
u, v, w, , , , , , , , ,
∂x ∂y ∂z ∂x ∂y ∂z ∂x ∂y ∂z
have (in every point of V) modulus near to zero. Obviously, if the deformations are
small, the products of the first order partial derivatives are negligible in comparison
to the first order partial derivatives.1.1.4
In this book we will always suppose to be in the field of the small deformations,
except in Chaps. 6 and 8.
Let us consider now the deformation of a neighborhood I of any point P =
(x, y, z) of V. Let P̄ = (x̄, ȳ, z̄) be an any point of I and

∂u
u1 (x̄, ȳ, z̄) = (x, y, z) (x̄ − x)
∂x  
1 ∂u ∂v
+ (x, y, z) + (x, y, z) (ȳ − y)
2 ∂y ∂x
 
1 ∂u ∂w
+ (x, y, z) + (x, y, z) (z̄ − z)
2 ∂z ∂x
 
1 ∂v ∂u
v1 (x̄, ȳ, z̄) = (x, y, z) + (x, y, z) (x̄ − x)
2 ∂x ∂y
∂v
+ (x, y, z)(ȳ − y) (1.1.4)
∂y
 
1 ∂v ∂w
+ (x, y, z) + (x, y , z) (z̄ − z)
2 ∂z ∂y
 
1 ∂w ∂u
w1 (x̄, ȳ, z̄) = (x, y, z) + (x, y, z) (x̄ − x)
2 ∂x ∂z
 
1 ∂w ∂v
+ (x, y, z) + (x, y, z) (ȳ − y)
2 ∂y ∂z
∂w
+ (x, y, z)(z̄ − z)
∂z

1.1.4 In fact, if a,b are real numbers such that 0 < a < 1 and 0 < b < 1, then we have 0 < ab < a,
0 < ab < b.
6 1 The Three-Dimensional Problem

 
1 ∂u ∂v
u2 (x̄, ȳ, z̄) = (x, y, z) − (x, y, z) (ȳ − y)
2 ∂y ∂x
 
1 ∂u ∂w
+ (x, y, z) − (x, y, z) (z̄ − z)
2 ∂z ∂x
 
1 ∂v ∂u
v2 (x̄, ȳ, z̄) = (x, y, z) − (x, y, z) (x̄ − x)
2 ∂x ∂y
  (1.1.5)
1 ∂v ∂w
+ (x, y, z) − (x, y, z) (z̄ − z)
2 ∂z ∂y
 
1 ∂w ∂u
w2 (x̄, ȳ, z̄) = (x, y, z) − (x, y, z) (x̄ − x)
2 ∂x ∂z
 
1 ∂w ∂v
+ (x, y, z) − (x, y, z) (ȳ − y)
2 ∂y ∂z
u3 (x̄, ȳ, z̄) = u(x, y, z)
v3 (x̄, ȳ, z̄) = v(x, y, z) (1.1.6)
w3 (x̄, ȳ, z̄) = w(x, y, z) .

It is obvious that u3 , v3 , w3 are the components of a rigid translation of I.

Besides, from the hypothesis of small deformations, u2 , v2 , w2 are the compo-
nents of a rigid rotation1.1.5 of I around a line passing through P.
The components u1 , v1 , w1 take the name of displacement’s components related
to the pure deformation.
From (1.1.1), (1.1.4), (1.1.5), and (1.1.6) we have

u(x̄, ȳ, z̄) = u1 (x̄, ȳ, z̄) + u2 (x̄, ȳ, z̄) + u3 (x̄, ȳ, z̄)
v(x̄, ȳ, z̄) = v1 (x̄, ȳ, z̄) + v2 (x̄, ȳ, z̄) + v3 (x̄, ȳ, z̄) (1.1.7)
w(x̄, ȳ, z̄) = w1 (x̄, ȳ, z̄) + w2 (x̄, ȳ, z̄) + w3 (x̄, ȳ, z̄) .

In terms of displacement vectors, if we consider the vectors

 
s1 (P̄) = u1 (P̄), v1 (P̄), w1 (P̄) ,
 
s2 (P̄) = u2 (P̄), v2 (P̄), w2 (P̄) ,
 
s3 (P̄) = u3 (P̄), v3 (P̄), w3 (P̄) ,
from (1.1.7) we have (Fig. 1.1.6)

s(P̄) = s1 (P̄) + s2 (P̄) + s3 (P̄)

1.1.5 Precisely, if α, β, γ are real numbers with modulus very smaller than 1, then u = γ γ + β z,
v = −γ x + α z, −βx − αy are, with good approximation, the components of the displacement of
an any point (x, y, z) of a rigid body because of three anticlockwise rotations around axes passing
through (0, 0, 0). The first is of intensity (infinitesimal) α and around the axis parallel to x, the
second is of intensity (infinitesimal) β and around the axis parallel to y, the third is of intensity
(infinitesimal) γ and around the axis parallel to z.
1.1 Analysis of Strain 7

Fig. 1.1.6

so that
[1.1.2] In the field of the small deformations, the deformation of the neighborhood
of a point is the sum of a pure deformation, of a rigid rotation and of a rigid
translation. 
Taking into consideration the precedent, it follows that (Fig. 1.1.6):

– the pure deformation s1 transforms the points of I belonging to a line r passing

through P in points of a line r1 passing through P,
– the rigid rotation s2 transforms the points of I belonging to r1 in points of a line r2
passing through P,
– the rigid translation s3 transforms the points of I belonging to r2 in points of a line
r3 passing through P and parallel to r2 ,
– the line r3 coincides exactly with the line r on which, because (1.1.1), the points
of I belonging to r turn to happened deformation,
– the vector s2 is practically orthogonal to r2 .

Remark 1.1.4 As an example of the [1.1.2], we can consider the deflection of the
cantilever of Fig. 1.1.7.
Clearly the true deformation of the neighborhood I of P can be obtained with
a pure deformation of I, with a rigid rotation of I and with a rigid translation of I
(Fig. 1.1.8). 

F
P

Fig. 1.1.7 P’
8 1 The Three-Dimensional Problem

Fig. 1.1.8

Remark 1.1.5 Evidently, the variations of volume and/or of form of the neighbor-
hood of a point are due only to the pure deformation. 
Remark 1.1.6 The angle α 1 between r and r1 (Fig. 1.1.9) is very small. To prove
this, we consider a point P̄ = (x̄, ȳ, z̄) of a neighborhood of P = (x, y, z), different
from P, and the line r passing through P and P̄. We denote with rx , ry , rz the direction
cosines of r. Obviously if l = dist(P, P̄) it results x̄−x = l rx , ȳ−y = l ry , z̄−z = l rz
(Fig. 1.1.9). We denote with r1x , r1y , r1z the direction cosines of r1 and with r [resp.
r1 ] the unit vector of r [resp. r1 ]. Taking into account that u, v, w are near to zero,
we have

x̄ − x [x̄ + u1 (P̄)] − x
cos α1 = r × r1 = rx r1x + ry r1y + rz r1z =
dist(P, P̄) dist(P, P̄1 )
ȳ − y [ȳ + v1 (P̄)] − y z̄ − z [z̄ + w1 (P̄)] − z
+ +
dist(P, P̄) dist(P, P̄1 ) dist(P, P̄) dist(P, P̄1 )
   
x̄ − x u1 (P̄) x̄ − x ȳ − y v1 (P̄) ȳ − y
= + + +
dist(P, P̄) dist(P, P̄1 ) dist(P, P̄1 ) dist(P, P̄) dist(P, P̄1 ) dist(P, P̄1 )
 
z̄ − z w1 (P̄) z̄ − z ∼ (x̄ − x)2
+ + =
dist(P, P̄) dist(P, P̄1 ) dist(P, P̄1 ) dist(P, P̄) dist(P, P̄1 )
(ȳ − y) 2 (z̄ − z) 2
+ +
dist(P, P̄) dist(P, P̄1 ) dist(P, P̄) dist(P, P̄1 )
 1
= [(x̄ − x)2 + (ȳ − y)2 + (z̄ − z)2 ] [(x̄ − x)2 + (ȳ − y)2 + (z̄ − z)2 ] 2
1

·[(x̄ + u1 (P̄) − x)2 + (ȳ + v1 (P̄) − y)2 + (z̄ + w1 (P̄) − z)2 ] 2

∼ (x̄ − x)2 + (ȳ − y)2 + (z̄ − z)2

= 1 1
= 1.
[(x̄ − x)2 + (ȳ − y)2 + (z̄ − z)2 ] 2 [(x̄ − x)2 + (ȳ − y)2 + (z̄ − z)2 ] 2

As a consequence, α 1 is near to zero. 

P
Fig. 1.1.9
1.1 Analysis of Strain 9

Fig. 1.1.10

Remark 1.1.7 Let (Fig. 1.1.10)

• A = (xA , yA , zA ) and B = (xB , yB , zB ) be points af a neighborhood of P = (x, y, z),

such that dist(A, P) = dist(B, P) = l,
• a [resp. b] be the line passing through A [resp. B] and P,
• ax , ay , az [resp. bx , by , bz ] be the direction cosines of a [resp. b],
• a [resp. b] the unit vector of a [resp. b].

Resulting l ax = xA − x, l ay = yA − y, l az = zA − z, l bx = xB − x, l by =
yB − y, l bz = zB − z, we have

s1 (A) × b = u1 (A) bx + v1 (A) by + w1 (A) bz

   
∂u 1 ∂u ∂v 1 ∂u ∂w
= (P) l ax bx + (P) + (P) l ay bx + (P) + (P) l az bx
∂x 2 ∂y ∂x 2 ∂z ∂x
   
1 ∂v ∂u ∂v 1 ∂v ∂w
+ (P) + (P) l ax by + (P) l ay by + (P) + (P) l az by
2 ∂x ∂y ∂y 2 ∂z ∂y
   
1 ∂w ∂u 1 ∂w ∂v ∂w
+ (P) + (P) l ax bz + (P) + (P) l ay bz + (P)l az bz
2 ∂x ∂z 2 ∂y ∂z ∂z
= u1 (B) ax + v1 (B) ay + w1 (B) az = s1 (B) × a.

So we have proved that the (orthogonal) projection on b of the displacement

of A in the pure deformation is equal to the (orthogonal) projection on a of the
displacement of B in the pure deformation. 

1.1.5 Components of Strain

Let (Fig. 1.1.11)

– P̄ = (x̄, ȳ, z̄) be a point of a neighborhood of P = (x, y, z), different from P,

– r be the line passing through P and P̄, having rx , ry , rz as direction cosines,
10 1 The Three-Dimensional Problem

Fig. 1.1.11

– r1 be the line transformed of r in the pure deformation s1 ,

– α 1 the angle between r and r1 and α 2 the angle between r1 and r2 .

As we have seen in the Remark 1.1.5, α 1 is very small. As a consequence, if

Q̄ is the orthogonal projection of P̄1 on r (Fig. 1.1.11), it results1.1.6 dist(P̄1 , P) =
dist(Q̄, P).
Because the rigid rotation is infinitesimal, α 2 is very small. As a consequence, it
results (Fig. 1.1.11) dist(P̄2 , P) = dist(P̄1 , P). It follows of it that (Fig. 1.1.11)

dist(P̄ , P ) = dist(Q̄, P)

from which, putting l = dist(P, P̄) and keeping in mind of the (1.1.4)

Fig. 1.1.12

1.1.6 If α is very small, we have with good approximation α = sinα = tgα, cosα = 1 (Fig. 1.1.12).
1.1 Analysis of Strain 11

dist(P̄ , P ) − dist(P, P̄) dist(Q̄, P) − dist(P, P̄)

εr = =
dist(P, P̄) l
dist(P̄, Q̄) s1 (P̄) × r u1 (P̄) rx + v1 (P̄) ry + w1 (P̄) rz
= = =
l  l  l
 
∂u 1 ∂u ∂v 1 ∂u ∂w
= (P) rx + (P) + (P) ry + (P) + (P) rz rx
∂x 2 ∂y ∂x 2 ∂z ∂x
   
1 ∂v ∂u ∂v 1 ∂v ∂w
+ (P) + (P) rx + (P) ry + (P) + (P) rz ry
2 ∂x ∂y ∂y 2 ∂z ∂y
   
1 ∂w ∂u 1 ∂w ∂v ∂w
+ (P) + (P) rx + (P) + (P) ry + (P)rz rz
2 ∂x ∂z 2 ∂y ∂z ∂z
       
∂u ∂v ∂w ∂u ∂v
= (P) rx2 + (P) ry2 + (P) rz2 + (P) + (P) rx ry
∂x ∂y ∂z ∂y ∂x
   
∂u ∂w ∂v ∂w
+ (P) + (P) rx rz + (P) + (P) ry rz .
∂z ∂x ∂z ∂y
(1.1.8)

From (1.1.8) it follows, being the direction cosines of the axis x equal to (1, 0,
0), those of the axis y equal to (0, 1, 0) and those of the axis z equal to (0, 0, 1), that
in the point P

∂u
εx =
∂x
∂v
εy = (1.1.9)
∂y
∂w
εz = .
∂z

Let now (Fig. 1.1.13)

– A = (xA , yA , zA ) and B = (xB , yB , zB ) be two any points af a neighborhood of

P = (x, y, z), different from P and such that dist(A, P) = dist(B, P) = l,
– a [resp. b] be the line passing through A [resp. B] and P,
– ax , ay , az [resp. bx , by , bz ] be the direction cosines of a [resp. b],
– ϑ be the angle between a and b,
– a1 [resp. b1 ] be the line transformed of a [resp. b] in the pure deformation s1 ,
– a1x , a1y , a1z [resp. b1x , b1y , b1z ] be the direction cosines of a1 [resp. b1 ],
– α 1 [resp. β 1 ] be the angle between a [resp. b] and a1 [resp. b1 ],
– a2 [resp. b2 ] be the line transformed of a1 [resp. b1 ] in the rigid rotation s2 ,
– α be the angle between a1 and a2 (and then also between b1 and b2 ),
– a3 = a [resp. b3 = b ] be the line transformed of a2 [resp. b2 ] in the rigid
translation s3 ,
– P be the transformed of P,
– ϑ  be the angle between a and b .
12 1 The Three-Dimensional Problem

Fig. 1.1.13

Obviously the angle between a1 and b1 it is equal to ϑ  , so ϑ = ϑ  + α1 + β1

(Fig. 1.1.13). Furthermore, from the hypothesis of small deformations, it follows (as
already highlighted) that α 1 and β 1 are very small. From this it follows with good
approximation1.1.7

cosϑ  ∼
= cosϑ + ( − sinϑ)(ϑ  − ϑ) = cosϑ + γab sinϑ. (1.1.10)

If we denote with A1 [resp. B1 ] the point transformed of the point A [resp. B] in

the pure deformation s1 , we clearly have

dist(A1 , P) − dist(A, P) dist(B1 , P) − dist(B, P)

εa = , εb =
dist(A, P) dist(B, P)
and then

dist(A1 , P) = l (1 + εa ), dist(B1 , P) = l (1 + εb ). (1.1.11)

On the other hand, since the angle between a1 and b1 is equal to ϑ  , we have

cosϑ  = a1x b1x + a1y b1y + a1z b1z

1.1.7 If xbelongs to a neighborhood of x and f is a real function of a real variable, differentiable in

x, it results with good approximation
 
df
f (x ) ∼
= f (x) + (x) (x − x),
dx
1.1 Analysis of Strain 13

from which, taking into account the (1.1.11) and (1.1.4) and neglecting the
infinitesimal ones of superior order, we have

(1 + εa )(1 + εb )cosϑ 

dist(A1 , P) dist(B1 , P) xA + u1 (A) − x xB + u1 (B) − x yA + v1 (A) − y
= · +
l2 dist(A1 , P) dist(B1 , P) dist(A1 , P)

yB + v1 (B) − y zA + w1 (A) − z zB + w1 (B) − z
· + ·
dist(B1 , P) dist(A1 , P) dist(B1 , P)

ax + u1 (A) bx + u1 (B) ay + v1 (A) by + v1 (B)

= · + ·
l l l l
az + w1 (A) bz + w1 (B)
+ · = ax bx + ay by + az bz
l l
ax u1 (B) ay v1 (B) az w1 (B) bx u1 (A) by v1 (A) bz w1 (A)
+ + + + + +
l l l l l l
 
∂u 1 ∂u ∂v 1 ∂u ∂w
= cosϑ + ax (P)bx + (P) + (P) by + (P) + (P) bz
∂x 2 ∂y ∂x 2 ∂z ∂x
 
1 ∂v ∂u ∂v 1 ∂v ∂w
+ ay (P) + (P) bx + (P) by + (P) + (P) bz
2 ∂x ∂y ∂y 2 ∂z ∂y
 
1 ∂w ∂u 1 ∂w ∂v ∂w
+ az (P) + (P) bx + (P) + (P) by + (P) bz
2 ∂x ∂z 2 ∂y ∂z ∂z
 
∂u 1 ∂u ∂v 1 ∂u ∂w
+ bx (P)ax + (P) + (P) ay + (P) + (P) az
∂x 2 ∂y ∂x 2 ∂z ∂x
 
1 ∂v ∂u ∂v 1 ∂v ∂w
+ by (P) + (P) ax + (P) ay + (P) + (P) az
2 ∂x ∂y ∂y 2 ∂z ∂y
 
1 ∂w ∂u 1 ∂w ∂v ∂w
+ bz (P) + (P) ax + (P) + (P) ay + (P) az
2 ∂x ∂z 2 ∂y ∂z ∂z

∂u ∂v ∂w
= cosϑ + 2 (P) ax bx + 2 (P) ay by + 2 (P) az bz
∂x ∂y ∂z
   
∂u ∂v ∂u ∂w
+ (P) + (P) (ax by + ay bx ) + (P) + (P) (ax bz + az bx )
∂y ∂x ∂z ∂x
 
∂v ∂w
+ (P) + (P) (ay bz + az by ).
∂z ∂y
14 1 The Three-Dimensional Problem

From here, keeping in mind of the (1.1.10) and the fact that approximately ϑ  =
ϑ, we get

γab sinϑ + (εa + εb )cosϑ

= cosϑ  − cosϑ + (εa + εb )cosϑ ∼ = cosϑ  − cosϑ + (εa + εb )cosϑ 
∼ ∂u ∂v
= (1 + εa )(1 + εb )cosϑ  − cosϑ = 2 (P) ax bx + 2 (P) ay by
∂x ∂y
 
∂w ∂u ∂v
+ 2 (P) az bz + (P) + (P) (ax by + ay bx )
∂z ∂y ∂x
   
∂u ∂w ∂v ∂w
+ (P) + (P) (ax bz + az bx ) + (P) + (P) (ay bz + az by ).
∂z ∂x ∂z ∂y
(1.1.12)

The (1.1.12) for the coordinate axes x, y, z furnishes, being (1, 0, 0), (0, 0, 1), and
(0, 1, 0) their direction cosines
∂u ∂v
γxy = +
∂y ∂x
∂u ∂w
γxz = + (1.1.13)
∂z ∂x
∂v ∂w
γyz = + .
∂z ∂y
From (1.1.8), (1.1.9), (1.1.12), and (1.1.13) it follows

εa = εx a2x + εy a2y + εz a2z + γxy ax ay + γxz ax az + γyz ay az (1.1.14)

γab sinϑ + (εa + εb )cosϑ = 2 εx ax bx + 2 εy ay by + 2 εz az bz

(1.1.15)
+γxy (ax by + ay bx ) + γxz (ax bz + az bx ) + γyz (ay bz + az by ).

The (1.1.14) and (1.1.15) are called quadratic expressions of ε and γ in func-
tion of the direction cosines. Knowing the value at P of the elongations and of the
shearing strains according to the coordinate axes, the (1.1.14) and (1.1.15) allow
to determine the elongations and the shearing strains at P according to two any
directions a, b.
It is for this motive that εx , εy , εz , γ xy , γ xz , γ yz are called components of strain
and the (1.1.9) and (1.1.13) are called relations between strain and displacements.
Of course, in the field of the small deformations, in which we are, ε and γ they
always are negligible with reference to 1 and the products of ε and γ are always
negligible with reference to ε and γ .

1.1.6 Principal Direction of Strain

Let P be a point of the body C, r be a line passing through P (Fig. 1.1.14). We

have seen that, when the body get deformed, in a neighborhood of P r turns into a
1.1 Analysis of Strain 15

Fig. 1.1.14

line r passing through P (being P the point in which P turns) and that in the pure
deformation s1 r turns into a line r1 passing through P. When r1 = r happens, we
will say that r is a principal direction of strain (Fig. 1.1.14).
Furthermore we will call principal elongation the elongation according to a
principal direction of strain.
We notice that if a, b are two principal directions of strain, it results γab = 0. In
fact it is obvious that in such case it results ϑ  = ϑ (Fig. 1.1.15).
In the analysis of strain, it is important to know if in a point of the body some
principal directions of strain exist and, in the affirmative case, how many there are.
To this aim, let us consider the secular (i.e. algebraic of third order) equation in
the unknown one ε

εx − ε 2 γxy
1
2 γxz
1

2 γyx
1
εy − ε 2 γyz
1 = 0. (1.1.16)

2 γzx 2 γzy εz − ε
1 1

If we put

J1 = εx + εy + εz
1 2 1 1
J2 = εx εy + εx εz + εy εz − γxy − γxz2 − γyz2
4 4 4
1 1 1 1
J3 = εx εy εz + γxy γxz γyz − εx γyz − εy γxz2 − εz γxy
2 2
4 4 4 4

Fig. 1.1.15
16 1 The Three-Dimensional Problem

we have

εx − ε 2 γxy
1
2 γxz
1

2 γyx
1
εy − ε 2 γyz
1 = −ε3 + J1 ε2 − J2 ε + J3 .

2 γzx 2 γzy εz − ε
1 1

For the algebraic equation of third order the possible cases are the three following

1. Three real numbers εa , εb , εc exist such that εa = εb , εa = εc , εb = εc and

−ε3 + J1 ε2 − J2 ε + J3 = −(ε − εa )(ε − εb )(ε − εc ).

2. Two real numbers εa , εb exist such that εa = εb and

−ε3 + J1 ε2 − J2 ε + J3 = −(ε − εa )(ε − εb )2 .

3. A real number εa exists such that

−ε3 + J1 ε2 − J2 ε + J3 = −(ε − εa )3 .

In the case 1 we will say that the secular equation admits three simple roots. In
the case 2 we will say that the secular equation admits a simple root and a double
root. In the case 3 we will say that the secular equation admits a triple root.
Let us study the case 1. We consider the solution εa of Eq. (1.1.16). It is possible
to prove that the problem of find the three real numbers ax , ay , az such that

1 1
εx ax + γxy ay + γxz az = εa αx
2 2
1 1
γyx ax + εy ay + γyz az = εa ay
2 2 (1.1.17)
1 1
γzx ax + γzy ay + εz az = εa az
2 2
a2x + a2y + a2z =1

admits one and an only one solution.

Well we assert that the line individualized by the direction cosines ax , ay, az is a
principal direction of strain and εa is the relative principal elongation.
Let in fact (Fig. 1.1.16):

– (ax , ay, az ) be the solution of the problem (1.1.17),

– a be the line passing through P = (x, y, z) and having direction cosines ax , ay , az ,
– P̄ = (x̄, ȳ, z̄) be a point of a belonging to a neighborhood of P and different from P,
– l = dist(P, P̄).
1.1 Analysis of Strain 17

Fig. 1.1.16

Evidently, l > 0 and

x̄ − x ȳ − y z̄ − z
ax = , ay = , az = .
l l l
In the pure deformation s1 , it happens that a turns in a1 and P̄ turns in P̄1 =
(x̄1 , ȳ1 , z̄1 ) (Fig. 1.1.16). Since for the (1.1.4), (1.1.9), and (1.1.13) it results

1 1
u1 (x̄, ȳ, z̄) = εx (x̄ − x) + γxy (ȳ − y) + γxz (z̄ − z)
2 2
1 1
v1 (x̄, ȳ, z̄) = γyx (x̄ − x) + εy (ȳ − y) + γyz (z̄ − z)
2 2
1 1
w1 (x̄, ȳ, z̄) = γzx (x̄ − x) + γzy (ȳ − y) + εz (z̄ − z),
2 2
we have

1 1
u1 (x̄, ȳ, z̄) = l εx ax + γxy ay + γxz az
2 2
1 1
v1 (x̄, ȳ, z̄) = l γyx ax + εy ay + γyz az
2 2
1 1
w1 (x̄, ȳ, z̄) = l γzx ax + γzy ay + εz az .
2 2

From these relationships, keeping in mind that (ax , ay , az ) is the solution of the
problem (1.1.17), we draw

u1 (P̄) = l εa ax
v1 (P̄) = l εa ay (1.1.18)
w1 (P̄) = l εa az
18 1 The Three-Dimensional Problem

which implicates that s1 (P) is parallel to a. Since this reasoning is true for every
point P̄ of a, it results a1 = a. Consequently, a is a principal direction of strain.
Let us verify now that the elongation at P according the direction a is exactly
equal to εa . Since the deformations are small, we have

x̄ − x = x̄ + u1 (P̄) + u2 (P̄) + u3 (P̄) − x

= x̄1 + u2 (P̄) + u(P̄) − x = x̄1 − x + u2 (P̄) ∼

= x̄1 − x

ȳ − y = ȳ + v1 (P̄) + v2 (P̄) + v3 (P̄) − y

= ȳ1 + v2 (P̄) + v(P) − y = ȳ1 − y + v2 (P̄) ∼

= ȳ1 − y

z̄ − z = z̄ + w1 (P̄) + w2 (P̄) + w3 (P̄) − z

= z̄1 + w2 (P̄) + w(P) − z = z̄1 − z + w2 (P̄) ∼
= z̄1 − z.

This way dist(P̄ , P ) = dist(P̄1 , P). As a consequence, taking into account the
(1.1.18), we get (Fig. 1.1.16)

l − l dist(P̄, P ) − l dist(P̄1 , P) − l
= =
l l l
 1
(x̄ + u1 (P̄) − x)2 + (ȳ + v1 (P̄) − y)2 + (z̄ + w1 (P̄) − z)2 2 − l
=
 l 1
2
= (1 + εa ) ax + (1 + εa ) ay + (1 + εa )2 a2z − 1 = εa .
2 2 2 2

We can show that in any point they always exist at least three principal directions
of strain twos twos orthogonal, which we call principal triplet of strain. Precisely

[1.1.3] In the case 1 they are three and only three principal directions of strain
a, b, c. These are twos twos orthogonal and the relative principal elongations are
εa , εb , εc .
In the case 2 they are ∞1 principal directions of strain. Precisely, it exists a prin-
cipal direction of strain a and every line of the plane α normal to a is principal
direction of strain. Moreover, every line distinguished from a and not belonging to
α is not a principal direction of strain. The principal elongation according to a is
equal to εa and, for every line t belonging to α, the principal elongation according
to t is equal to εb .
In the case 3 every line r is principal direction of strain and the principal
elongation according to r is equal to εa .
In general, if ξ is a principal direction of strain, if ξx , ξy , ξz are its direction
cosines and if εξ is the principal elongation according to ξ , it results
1.1 Analysis of Strain 19

1 1
εx ξx + γxy ξy + γxz ξz = εξ ξx
2 2
1 1
γyx ξx + εy ξy + γyz ξz = εξ ξy
2 2
1 1
γzx ξx + γzy ξy + εz ξz = εξ ξz
2 2
ξx2 + ξy2 + ξz2 = 1. 

Remark 1.1.8 In every point of the body they are at least three principal directions
of strain twos twos orthogonal. To find the principal elongations we need to resolve
the secular equation. To find the principal directions of strain we need, for each
principal elongation, to resolve the problem (1.1.17). 
Remark 1.1.9 If a, b, c are lines passing through P, twos twos orthogonal and such
that γab = γac = γbc = 0, then P, a, b, c is an orthogonal set of principal axes of
strain.
In fact, if we assume P, a, b, c as frame of reference, the secular equation is
written
εa − ε 0 0
0 εb − ε 0 = 0;
0 0 εc − ε

so the three principal elongations are εa , εb , εc . To draw the principal direction of

strain individualized by the principal elongation εa , it is necessary to determine the
solution of the problem (1.1.17), which is written

εa ax = εa ax
εb ay = εa ay
εc az = εa az
a2x + a2y + a2z = 1.

Since (1, 0, 0) is solution of such problem, which admits an unique solution, we

obtain that a is a principal direction of strain. Likewise, we prove that b and c are
principal directions of strain. 
Remark 1.1.10 We can prove that the maximum elongation (at the point P) is equal
at the biggest of the three principal elongations.
In fact, if P, a, b, c is a principal reference of deformation, r is an any line passing
through P and ra , rb , rc are the direction cosines of r, from (1.1.14) we have

εr = εa ra2 + εb rb2 + εc rc2 . (1.1.19)

Therefore, the problem is to find the maximum one that the function εr assumes
in the set of the ordered triplet (ra , rb , rc ) such that ra2 + rb2 + rc2 = 1. Since it must
be ra2 + rb2 = 1 − rc2 ≤ 1, we must find the maximum one that the function
20 1 The Three-Dimensional Problem
 
εr = εa ra2 + εb rb2 + εc 1 − ra2 − rb2 = (εa − εc ) ra2 + (εb − εc ) rb2 + εc (1.1.20)

assumes in the circle of center (0, 0) and radius 1. In practice it deals with determin-
ing the maximum one among the relative extreme and the maximum one assumed
on the frontier (that is on the circumference of center (0, 0) and radius 1). It is known
that in a point of relative extreme all the first order partial derivatives must be zero.
So, since from (1.1.20) it results

∂εr ∂εr
= 2 (εa − εc ) ra , = 2 (εb − εc ) rb
∂ra ∂rb
it is undeniable that the function εr assumes a relative extreme to the more in (0, 0).
Then one of the values to be compared is

εr (0, 0) = εc . (1.1.21)

We appraise now the maximum one assumed on the circumference of center

(0, 0) and radius 1. Clearly the scalar parametric equations are ra = cos t, rn = sin t
and the interval of variation of the parameter t is [0, 2π]. So the problem consists of
determining the maximum one in the interval [0, 2π] of the function of one variable

εr (ra (t) , rb (t)) = (εa − εc ) cos2 t + (εb − εc ) sin2 t + εc . (1.1.22)

Naturally it is the maximum one among its relative extremes and the values that
it assumes in 0 and in 2π. The (1.1.22) assumes in 0 and in 2π the value

εr (1, 0) = εa . (1.1.23)

About the relative extremes, in the points in which they are assumed the first
order derivative one is zero. Then, being

d
εr (ra (t) , rb (t)) = − (εa − εc ) 2 cos t sin t
dt (1.1.24)
+ 2 (εb − εc ) sin t cos t = (εb − εa ) sin 2t

we distinguish two cases. If εa = εb , it results εr (ra (t) , rb (t)) = εa and εr con-

stantly assumes on the circumference of center (0, 0) and radius 1 the value εa . If
εa = εb , because (1.1.24) the possible points of relative extreme are the solutions
of the equation sin 2t = 0, i.e. 2 t1 = π , 2 t2 = π , 2 t3 = 2 π . Since in t1 = 0 itis
assumed the value (1.1.23), it is enough to see what values are assumed in t2 = π 2
and in t3 = π . We have
 π π
εr cos , sin = εb , εr (cos π , sin π ) = εa . (1.1.25)
2 2
1.1 Analysis of Strain 21

This way, on the circumference of center (0, 0) and radius 1 εr either assumes
a maximum value εa or, since (1.1.23) and (1.1.25), assumes a maximum value
max {εa , εb }. As a consequence, from the (1.1.21) it follows

max εr (ra , rb , rc ) = max {εa , εb , εc } . 

ra2 +rb2 +rc2 =1

1.1.7 Invariants of Strain

There are physical quantity that don’t vary when we pass from a Cartesian orthog-
onal frame of reference O, x1 , x2 , x3 to another Cartesian orthogonal frame of
reference O, x1 , x2 , x3 .
The quantity individualized by a real number in Tensor analysis is called scalar
(or tensor of order zero). It is an example the distance between two points.
The quantity individualized by three ordered real numbers in Tensor analysis is
called vector (or tensor of order one). So to a vector we can associate a direction. It
is an example the speed of a mobile point.
The quantity individualized by six orderd real numbers in Tensor analysis is
called tensor (or tensor of order two). To a tensor we can associate two directions
(or a plane and a direction).
The change of frame of reference is accomplished by an orthogonal transforma-
tion
3
xi = aij xj + ci ∀i = 1, 2, 3 (1.1.26)
j=1

  3
where aij is an orthogonal matrix (i.e. for any i, j = 1, 2, 3 it is k=1 aki akj = δij
where δij is the Kronecker delta) and c1 , c2 , c3 are the coordinates of the old origin
in the new frame of reference.
The tensor of order one is an ordered triplet (v1 , v2 , v3 ) of real numbers. It
changes with the law


3
vi = aij vj ∀i = 1, 2, 3
j=1

under the transformation (1.1.26).

The tensors of order two what we meet inMechanics
 of continua generally are
real and symmetrical matrixes of order three wij . They change with the law


3 
3
wij = aik ajm wkm ∀i, j = 1, 2, 3
k=1 m=1

under the transformation (1.1.26).

22 1 The Three-Dimensional Problem

Remark 1.1.11 A tensorial equation has the same look in every Cartesian orthogonal
frame of reference. 

We call strain tensor the second order tensor

⎡ ⎤
εx 2 γxy 2 γxz
1 1
⎢1 ⎥
⎢ 2 γyx εy ⎥
2 γyz ⎦ .
1
⎣
2 γzx 2 γzy εz
1 1

During such analysis we verify that J1 , J2 , J3 are invariant when we turn from
a Cartesian orthogonal frame of reference O, x, y, z to another Cartesian orthogonal
frame of reference O , x , y , z . Let’s call them respectively linear invariant of strain,
quadratic invariant of strain, cubic invariant of strain. Insofar, with reference to a
principal triplet of strain O, a, b, c, we have
εx + εy + εz = εa + εb + εc
1 2 1 1
εx εy + εx εz + εy εz − γxy − γxz2 − γyz2 = εa εb + εa εc + εb εc
4 4 4
1 1 1 1
εx εy εz + γxy γxz γyz − εx γyz2 − εy γxz2 − εz γxy
2
= εa εb εc .
4 4 4 4
Let us consider now a principal triplet of strain P, a, b, c and an infinitesimal cube
having edge l and faces parallel to the coordinate planes (Fig. 1.1.17). As we have
seen, after the deformation the cube is turned in a parallelepiped whose edges have
lengths la , lb , lc and are twos twos orthogonal.
Of course
la − l lb − l lc − l
εa = , εb = , εc = .
l l l
Then, if we denote with Vf the volume of the deformed cube and with
Vi the volume of the cube before of the deformation, it results with good
approximation

Vf − Vi la lb lc − l 3 l (1 + εa ) l (1 + εb ) l (1 + εc ) − l3
= =
Vi l3 l3
= εa + εb + εc = J1 .

P b c′

a
P′

a′ b′
Fig. 1.1.17
1.1 Analysis of Strain 23

Then, the body after the deformation has volume V equal to


V =V+ J1 dV.
V

We also call J1 cubic elongation coefficient.

1.1.8 Plane State of Strain

Let us consider any point P of the body. We will say that the state of strain is plane
at P when a line a passing through P exists such that for every point  P̄
 of a near to P
(i.e. belonging to a neighborhood of P) the displacement vector s1 P̄ (related to the
pure deformation) has zero intensity. The plane α passing through P and normal to
a is said, then, plane of the deformation. We will immediately see that every  point Q̄
of the neighborhood of P has in the pure deformation a displacement s1 Q̄ parallel
to α.
In fact, let us suppose that the state of strain is plane at P. We denote with α the
plane of the deformation and with a the line passing through P and orthogonal to α.
Since the state of strain
 is plane, every point P̄ of a has (in the pure deformation) a
displacement s1 P̄ = 0. Then the pure deformation transforms every point of a in
itself. So a is a principal direction of strain and εa = 0.
Vice versa let us suppose that the line a passing through P is a principal direction
of strain and εa = 0. We assume a principal reference of  deformation
 a, b, c and
consider any point P = (a, b, c) of the body. If Q̄ = ā, b̄, c̄ is any point of a
neighborhood of P, it results
 
u1 Q̄ = εa (ā − a) = 0
   
v1 Q̄ = εb b̄ − b
 
w1 Q̄ = εc (c̄ − c) .
 
As a consequence, s1 Q̄ is parallel to the plane (b, c). All the points that are on
a normal to the plane (b, c), since the second and third coordinates are equal, have
an equal displacement s1 .
So we have proved that

[1.1.4] If the state of strain is plane at P, every point near to P suffers in the pure
deformation a displacement parallel to the plane of the deformation α. Furthermore,
for any line r orthogonal to the plane of the deformation, it results

A, B ∈ r ⇒ s1 (A) = s1 (B) . 

[1.1.5] If the state of strain is plane at P then a principal elongation is zero, and
vice versa. 
24 1 The Three-Dimensional Problem

Let us suppose now that the state of strain is plane in P. Then, because of [1.1.5],
a principal elongation is zero. As a consequence, the (1.1.16) must be satisfied by
ε = 0. Insofar
εx 2 γxy
1
2 γxz
1

2 γyx
1
εy 2 γyz
1 = 0. (1.1.27)
2 γzx 2 γzy εz
1 1

Vice versa, let us suppose that (1.1.27) is true. As a consequence, the (1.1.16) is
satisfied by ε = 0. Then a principal elongation is zero. Thus, because of [1.1.5], the
state of strain is plane in P.
So we have proved that

[1.1.6] If the state of strain is plane then (1.1.27) is satisfied, and vice versa. 

1.1.9 Equations of Compatibility

The regularity of the components of the displacement u, v, w guarantees in the

respects of the inside compatibility of the deformation. In fact, the univocality and
the continuity of u, v, w and their derivatives exclude the lacerations and the overlaps
of material. Vice versa, in general the regularity of the components of the deforma-
tion εx , εy , εz , γxy , γxz , γyz does not guarantee in the respects of the inside congruence
of the deformation. In fact, if we assume as components of the deformation a generic
sextuple of functions continuous with theirs partial derivatives (in V), a situation
similar to Fig. 1.1.18 can happen.
In analytical terms, if in the equations

∂u
= εx
∂x
∂v
= εy
∂y
∂w
= εz
∂z
(1.1.28)
∂u ∂v
+ = γxy
∂y ∂x
∂u ∂w
+ = γxz
∂z ∂x
∂v ∂w
+ = γyz
∂z ∂y

the functions εx , εy , εz , γxy , γxz , γyz are continuous with theirs partial derivatives
(in V), the problem to determine three functions u, v, w able to satisfy in V the
Eq. (1.1.28) could not admit a solution.
1.1 Analysis of Strain 25

Fig. 1.1.18

Let us consider the followings six compatibility equations

∂ 2 εx ∂ 2 εy ∂ 2 γxy
+ =
∂y2 ∂x2 ∂x∂y
∂ 2 εx ∂ 2 εz ∂ 2 γxz
+ =
∂z2 ∂x2 ∂x∂z
∂ 2 εy ∂ 2 εz ∂ 2 γyz
+ =
∂z2 ∂y2 ∂y∂z
(1.1.29)
∂ 2 εx ∂ ∂γxy ∂γyz ∂γzx
2 = − +
∂y∂z ∂x ∂z ∂x ∂y
∂ 2 εy ∂ ∂γyz ∂γzx ∂γxy
2 = − +
∂z∂x ∂y ∂x ∂y ∂z
∂ 2 εz ∂ ∂γzx ∂γxy ∂γyz
2 = − + .
∂x∂y ∂z ∂y ∂z ∂x

We have1.1.8

[1.1.7] Equations (1.1.28) are a necessary and sufficient condition for the inside
compatibility of the deformation (i.e. for the exclusion of the lacerations and the
overlaps of material). 

1.1.10 Measurement of Strain

The deformation of a body can experimentally be appraised in a Laboratory mate-
rial tests. The displacement of a point of the surface of a body is measured

1.1.8 The necessity is obvious. The sufficiency involves delicate matters of the modern
Mathematical analysis whose study is still incomplete.
26 1 The Three-Dimensional Problem

through transducers.1.1.9 They take the name of deflectometer (if can appreciate
displacements of the order of centimeters) or of comparators (if can appreciate
displacements of the order of millimeters).
The electric transducers nowadays have supplanted those mechanic, although
they were very sensitive.1.1.10
They have the great advantage to be able to be link to the computer (which auto-
matically elaborates the experimental data) or more simply to a tracer (for instance
for the automatic tracing of the curve load-displacement).
The transducer that allows to measure the elongation in a point of the surface
of a body is called extensometer. Naturally the value noticed by the extensometer
furnishes not the value of the elongation εa in the point P but the average value of
the elongation εa in a small area in which the point P belongs.
Among the electric extensometers, the most important are those to resistance
(straingauges). They are easy to use and, since numerous types exist, they can be
used in various applications. The Fig. 1.1.19 shows a straingauge to thread. It is
constituted by a metallic thread integral to a support that is collated, with a proper
sticker, on the body. To happened deformation, because of the elongation according
to the direction of the lines of the thread, the length of these changes and with it the
electric resistance. A bridge of Wheatstone1.1.11 (Fig. 1.1.20) measures its variation.
After having calibrated the system, the measured variation furnishes the value of
the elongation (averaged on the area engaged by the straingauge) according to the
direction (of the lines) of the thread.
When we want to determine the state of strain in a point of a body we can
use the rosette straingauge. It is constituted of more straingauges prepared on an
unique support. The grates are tilted from opportune angles so that it results simple

Fig. 1.1.19

1.1.9 We call transducer a device sensitive to the greatness to measure and that, giving an elec-
tric signal or another indication (as for instance the position of the needle in a dial), allows to
measure her.
1.1.10 A mechanical comparator was quietly able to appreciate the thousandth one of millimeter.
1.1.11 Sir Charles Wheatstone, Gloucester 1802 – Paris 1875.
1.2 Analysis of Stress 27

Fig. 1.1.20

Fig. 1.1.21

the computation of the principal deformations. In Fig. 1.1.21 we show a rosette

straingauge with three grates.
If we have to measure the state of strain in a internal point of the body we can
employ embedded straingauges.

1.2 Analysis of Stress

1.2.1 Stress Vector

We consider a solid deformable body C and we denote with V its volume, i.e. the
region of the three-dimensional space occupied by C (Fig. 1.2.1). We suppose that
the surface (or boundary) S of V (Fig. 1.2.1) is regular, so in every point S admits

z S

z
x V
O
x
y

Fig. 1.2.1 y
28 1 The Three-Dimensional Problem

one and only one tangent plane. We decompose S in two parts Su and Sp . We denote
with V the region of space occupied by the configuration deformed of C and with
S the surface of V . Let us refer to an orthogonal set of Cartesian axes O, x, y, z and
denote with x, y, z their unit vectors (Fig. 1.2.1).
We suppose that on the body some forces act. We distinguish such forces in
loads (or external forces) and in stresses (or internal forces). Furthermore we
distinguish the loads in volumetric loads (or loads of mass) and in superficial
loads (or loads of contact). We distinguish the superficial loads (that can be
applied only by contact of other bodies) in distributed loads and in concentrated
loads.
The volumetric load (for instance the force of gravity) is a vectorial function
defined in V. It associates to every point P = (x, y, z) of V the vector

X (x, y, z) x + Y (x, y, z) y + Z (x, y, z) z.

The functions X, Y, Z are said components of the volumetric load. They are
measured in Kg/cm3 .
The distributed superficial load is a vectorial function defined in Sp . It associates
to every point P = (x, y, z) of Sp the vector

px (x, y, z) x + py (x, y, z) y + pz (x, y, z) z.

The functions px , py , pz are said components of the distributed superficial load.

They are measured in Kg/cm2 .
The concentrated superficial load is a vector

Fx x + Fy y + Fz z

applied in a point Q of Sp . The numbers Fx , Fy , Fz are said components of the

concentrated superficial load. They are measured in Kg.
As for the stress, let us consider any point Q = (x, y, z) of V and a regular surface
Σ passing through Q and such to divide V in two parts A and B (Fig. 1.2.2). We
model the action of the material of the part B on the material of the part A with a
superficial load distributed on Σ, that is with a vectorial function defined in Σ. So,

A Q
A
B

Fig. 1.2.2
1.2 Analysis of Stress 29

denoting with α n the plane tangent to Σ in Q and with n the normal one in Q to α n ,
we model the action at Q of the material of the part B on the material of the part A
with the vector (Fig. 1.2.2)

tn (x, y, z) = tnx (x, y, z) x + tny (x, y, z) y + tnz (x, y, z) z.

We call the vector tn (x, y, z) stress vector (or tension vector) at the point Q acting
on the plane α n (Q). The functions tnx , tny , tnz are called components of the stress
vector on the axes x, y, z. They are measured in Kg/cm2 .
As imposed by the principle of action and reaction of the Physics, the action of
the material of the part A on the material of the part B it has in the point Q the same
line of action, equal intensity and opposite verse of the stress vector tn (Fig. 1.2.3).
It is necessary to highlight that in a solid body, unlike what happens in the
liquids,1.2.1 the stress vector tn as a rule is not orthogonal to the plane α n .
It is equally important to observe that in the point Q the stress vector tn usually
varies a lot when the plane α n varies (Fig. 1.2.4).

Fig. 1.2.3

n
B

Fig. 1.2.4

1.2.2 Normal Stress – Shearing Stress

We consider any point Q of V and we divide V in two parts with an any plane α n
passing through Q. We denote with n the line passing through Q and orthogonal to

the liquids the model is such that the stress vector tn acting on α n has as line of action the
1.2.1 In
orthogonal n to α n , is of compression and of constant intensity for every n (hydrostatic pressure).
30 1 The Three-Dimensional Problem

Fig. 1.2.5

α n . We remove one of the two parts and we apply in Q the stress vector tn acting on
α n that the removed part was applying. The stress vector tn has Q as first end point
and B as second end point (Fig. 1.2.5).
We consider the plane orthogonal to n and passing through B. Such plane inter-
sects n in the point B . We call normal stress acting on α n and denote with the
symbol σ n the vector having Q as first end point and B as second end point
(Fig. 1.2.5).
We consider the line parallel to n and passing through B. Such line intersects α n
in the point B . We call shearing stress acting on α n and denote with the symbol τ n
the vector having Q as first end point and B as second end point (Fig. 1.2.5).
Clearly σ n and τ n are measured in kg/cm2 and it results

tn = σ n + τ n .

We observe that σ n , which is orthogonal to the ideal plane of cut α n , is the force
with which (in Q) the two parts of the body try to be detached (or to penetrate it).
Analogously, the τ n , which is contained in the plane α n , is the force with which
(in Q) the two parts of the body try to shear it.

1.2.3 Components of Stress

Let Q be any point of V. As seen, ty is the stress vector at Q acting on the plane
passing through Q and parallel to the coordinate plane x, z (whose orthogonal line it
is y). As seen, we can decompose ty in a σ y (parallel to y) and in a τ y . Since τ y lies
in the plane x, z, we can decompose τ y in a component τ yx (according to x) and in a
component τ yz (according to z). So the stresses acting on the coordinate planes are

σx , τxy , τxz , σy , τyx , τyz , σz , τzx , τzy . (1.2.1)

The (1.2.1) play a primary role in the stress analysis. We call them components
of stress.
1.2 Analysis of Stress 31

Let us adopt the following

[Cauchy’s1.2.2 CONVENTION ] We divide V in two parts with a plane passing
through Q and parallel to a coordinate plane. We remove one of the two parts in
which V has been separated. We trace in the remained one the line n passing through
Q, orthogonal to the plane of cut and directed so that goes out of the material. If n is
oriented according to a coordinate axis, the σ and τ positive are oriented according
to the coordinate axes; otherwise opposite. 

1.2.4 Symmetry of τ – Differential Equations of Equilibrium –

Cauchy’s Boundary Conditions
As a consequence of the hypothesis of small deformations, the deformed configu-
ration V  of the body C is very near to the initial one V. So we can approximate V 
with V and S  with S. For this motive we can apply to the initial configuration of the
body the principle of dissection (of the Physics). We have
[1.2.1] Let the body C be in equilibrium (i.e. the external1.2.3 loads constitute a
system of forces equivalent to zero1.2.4 ). We remove from the body C any part C0 .
We denote with V0 the region of space occupied by C0 and suppose that the sur-
face (or boundary) S0 of V0 be regular. We apply on C0 the volumetric loads that
insisted on before we isolated C0 from the remainder of C. If Sp and S0 have com-
mon points, we apply on Sp ∩ S0 the (distributed and/or concentrated) superficial
loads which insisted on before we isolated C0 from the remainder of C. Furthermore
we apply on S0 − Sp ∩ S0 the stress vectors that the removed part of C charged on
the isolated one C0 . Clearly such stress vectors are now a superficial distributed
load for C0 . Then the body C0 is in equilibrium (i.e. the forces that we have applied
on C0 and the possible constraint reactions acting on Su ∩ S0 are a system equiva-
lent to zero). Furthermore the fields of displacements of the bodies C and C0 differ
in V0 at most from a rigid translation and/or from a rigid rotation. Then, since the
(1.1.28) and the footnote 1.1.5, the bodies C and C0 have in V0 just the same state of
strain. 
Remark 1.2.1 The assumption V  =V and S =S allows to calculate the value of the
stresses that arise in the body C, and in general in every structure, on the initial con-
figuration with very good approximation. We will adopt such assumption anywhere
in this book, except that in the problem of the stability of the equilibrium. 

1.2.2 Augustin Louis Cauchy, Paris 1789 – Sceaux 1857. It is one of the greatest mathematicians of
all time.
1.2.3 That is volumetric loads, superficial loads acting on S and constraint reactions acting on S .
p u
1.2.4 In the three-dimensional space we call equivalent to zero (or balanced) a system of forces such
that its resultant has zero intensity and its moment with respect to any axis of the space has zero
intensity.
32 1 The Three-Dimensional Problem

From [1.2.1] we extract the following property of shearing stress symmetry

[1.2.2] If the body C is in equilibrium, at each point of V it results

τxy = τyx , τxz = τzx , τyz = τzy . (1.2.2)

Proof We choose in arbitrary way a point P = (x, y, z) of V and three real numbers dx,
dy, dz, positive and very near to zero. We cut C according planes parallel to the coor-
dinate planes and passing through P = (x, y, z) and P = (x + dx, y + dy, z + dz). This
way we remove from C the infinitesimal parallelepiped C0 recognized in Fig. 1.2.6.
We apply on it, as required by [1.2.1], the volumetric loads X, Y, Z (positive if equiv-
erse to the coordinate axes) and the stress vectors that the removed material charged
on the isolated one C0 .
Since C0 is infinitesimal we assume that (Fig. 1.2.6)
– σ and τ are constant on every face of C0 ,
– σ and τ are equal to the values assumed in P in the faces passing through P,
– σ and τ are equal to the values assumed in A = (x + dx, y, z) in the face passing
through A and P ,
– σ and τ are equal to the values assumed in B = (x, y + dy, z) in the face passing
through B and P ,
– σ and τ are equal to the values assumed in C = (x, y, z + dz) in the face passing
through C and P .

Furthermore, since C0 is infinitesimal, we can approximate the value assumed in

A [resp. B] [resp. C] with the sum of the value assumed in P and of the value that
the differential one related to P assumes in A [resp. B] [resp. C], i.e.

Fig. 1.2.6
1.2 Analysis of Stress 33

∂σx ∂σx
σx (A) = σx (x + dx, y, z) = σx (x, y, z) + (x, y, z) dx = σx +
∂x ∂x
∂τxy
τxy (A) = τxy (x + dx, y, z) = τxy + dx
∂x
∂τxz
τxz (A) = τxz (x + dx, y, z) = τxz + dx
∂x
∂τyx
τyx (B) = τyx (x, y + dy, z) = τyx + dy
∂y
∂σy
σy (B) = σy (x, y + dy, z) = σy + dy
∂y
∂τyz
τyz (B) = τyz (x, y + dy, z) = τyz + dy
∂y
∂τzx
τzx (C) = τzx (x, y, z + dz) = τzx + dz
∂z
∂τzy
τzy (C) = τzy (x, y, z + dz) = τzy + dz
∂z
∂σz
σz (C) = σz (x, y, z + dz) = σz + dz.
∂z
In conclusion, using the convention of Cauchy, the situation is that of
Fig. 1.2.6.
By reason of the principle of dissection, C0 must be in equilibrium to the rotation
around the axis xP parallel to x and passing through P, that is the forces applied on
C0 must have equal to zero the moment referred to xP . So it must result

   
∂τyz   dz ∂σy dz
τyz + dy dx dz dy + σy dx dz − σy + dy dx dz
∂y 2 ∂y 2
   
dy ∂σz dy ∂τzy
− (σz dx dy) + σz + dz dx dy − τzy + dz dx dy dz
2 ∂z 2 ∂z
 
dy ∂τxz dy   dz
− (τxz dy dz) + τxz + dx dx dy + τxy dy dz
2 ∂x 2 2
 
∂τxy dz dz dy
− τxy + dx dy dz − (Ydx dy dz) + (Z dx dy dz) = 0.
∂x 2 2 2

It follows, neglecting the infinitesimal ones of order greater than 1, the third one
of the (1.2.2).
From the equilibriums to the rotation of C0 around the parallel ones to y and z
passing through P, the others two of (1.2.2) in the same manner follow. 

Remark 1.2.2 As a consequence of the [1.2.2], the nine components of stress (1.2.1)
reduce it to six different functions. 
34 1 The Three-Dimensional Problem

Remark 1.2.3 As a consequence of the [1.2.2], if r, s are any two lines of 3 passing
through P, it results τrs = τsr (that is the orthogonal component on s of the τr acting
on the plane of normal r is equal to the orthogonal component on r of the τs acting
on the plane of normal s). This is easily proven by observing that in the [1.2.2] the
choice of the frame of reference is arbitrary. 

From [1.2.1] we also can extract the following differential equations of equilib-
rium
[1.2.3] If the body C is in equilibrium, at each point of V it results

∂σx ∂τxy ∂τxz

+ + +X =0
∂x ∂y ∂z
∂τyx ∂σy ∂τyz
+ + +Y =0 (1.2.3)
∂x ∂y ∂z
∂τzx ∂τzy ∂σz
+ + + Z = 0.
∂x ∂y ∂z
Proof Proceeding as in the proof of the [1.2.2], we consider the equilibrium of the
infinitesimal parallelepiped C0 recognized in Fig. 1.2.6. Since C0 must also be in
equilibrium to the translation according x, the sum of the forces acting on C0 and
parallel to x must be equal to zero. That is

∂σx ∂τyx
σx + dx dy dz − σx dy dz + τyx + dy dx dz
∂x ∂y
∂τzx
− τyx dx dz + τzx + dz dx dy − τzx dx dy + X dx dy dz = 0.
∂z

This relation, taking into account the [1.2.2], shows that the first of (1.2.3) is true.
The second and the third of (1.2.3) analogously follow from the equilibrium of C0
to the translation according to y and to the translation according to z. 
From [1.2.1] we also can extract the following Cauchy’s relations
[1.2.4] Let the body C be in equilibrium, P be any point of V, αn be any plane passing
through P. We denote with A one of two parts in which αn cut C. We isolate A and
denote with n the line passing through P, orthogonal to αn and directed going out
from the material. We denote with nx , ny , nz the direction cosines of n and consider
the stress vector tn at P acting on αn . Then it results

tnx = σx nx + τxy ny + τxz nz

tny = τyx nx + σy ny + τyz nz (1.2.4)
tnz = τzx nx + τzy ny + σz nz .
1.2 Analysis of Stress 35

Fig. 1.2.7

Proof If n is parallel to a coordinate axis, the (1.2.4) are obviously true. If n is not
parallel to none of the coordinate axes, the possible cases are the two followings

(a) n is parallel to a coordinate plane

(b) n is not parallel to any of the coordinate planes.

In the case a, without loss of generality, we can suppose that n be orthogonal to

the axis z (Fig. 1.2.7). We denote with dx and dz two real numbers, positive and
small as we like. We consider the points P1 = (x + dx, y, z), P3 = (x, y, z + dz).
The plane passing through P1 and parallel to α n intersects the line passing through
P and parallel to axis y at P2 = (x, y + dy, z). Clearly dy is a real number
very near to zero (Fig. 1.2.7). We consider the points P4 = (x, y + dy, z + dz),
P5 = (x + dx, y, z + dz) and denote with C0 the infinitesimal part (of the body C)
PP1 P2 P3 P4 P5 . We direct the normal n to the plane P1 P2 P4 P5 so that it goes out of
C0 (Fig. 1.2.7).
The (1.2.4) are immediately obtained by observing that1.2.5

– nz = 0,
– meas2 (P, P1 , P3 , P5 ) = ny meas2 (P1 , P2 , P4 , P5 ) ,
– meas2 (P, P2 , P3 , P4 ) = nx meas2 (P1 , P2 , P4 , P5 )

and neglecting the infinitesimal ones of superior order.

In the case b we denote with dx a real number, positive and small as we like. We
consider the point P1 = (x + dx, y, z) and the plane β n passing through P1 and par-
allel to α n (Fig. 1.2.8). β n intersects the lines passing through P and parallel to axis
y and z at P2 = (x, y + dy, z), P3 = (x, y, z + dz). Clearly dy, dz are real numbers

Fig. 1.2.8

1.2.5 We denote with measn () the n-dimensional measure of .

36 1 The Three-Dimensional Problem

very near to zero (Fig. 1.2.8). We denote with C0 the infinitesimal element isolated
cutting the body C with the planes PP1 P3 , PP1 P2 , PP2 P3 , P1 P2 P3 (Fig. 1.2.8). We
direct the normal n to the plane P1 P2 P3 so that it goes out of C0 (Fig. 1.2.8).
We denote with nx , ny , nz the direction cosines of n and, as in precedence, with
tnx , tny , tnz , the components on the axes x, y, z of the stress vector acting on the plane
α n . Obviously the tnx , tny , tnz positive are oriented according to the coordinate axes.
Since C0 is infinitesimal we can assume that

– on the faces PP1 P3 , PP1 P2 , PP2 P3 , σ and τ have values constant and equal to the
values assumed in P,
– on the face P1 P2 P3 , tnx , tny , tnz have values constant and equal to the values
assumed in P.

Insofar, in conclusion, using the convention of Cauchy, the situation is that of

Fig. 1.2.9.
We put now A = meas2 (P1 P2 P3 ), Ax = meas2 (PP2 P3 ), Ay = meas2 (PP1 P3 ),
Az = meas (PP1 P2 ), so that the following known relationships subsist

Ax = A nx , Ay = A ny , Az = A nz .

By reason of the principle of dissection, C0 must also be in equilibrium to the

translation according x, so that the sum of the forces acting on C0 and parallel to x
must be equal to zero. As a consequence

− τyx Ay − σx Ax − τzx Az + X meas3 (PP1 P2 P3 ) + tnx A = 0

from which, neglecting the infinitesimal ones of order greater then 1, it follows

tnx A = σx A nx + τyx A ny + τzx A nz .

So the first of (1.2.4) is true. From the equilibriums to the translation of C0

according to y and z, the others two of (1.2.4) follow. 

Fig. 1.2.9
1.2 Analysis of Stress 37

Remark 1.2.4 From (1.2.3), in every point P of V to calculate the value of the stress
vector tn (and therefore of σ n and τ n ) related to an every plane α n is enough to know
the values undertaken in P by σx , τxy , τxz , σy , τyx , τyz , σz , τzx , τzy . 

From [1.2.1] we also can extract the following Cauchy’s boundary conditions

[1.2.5] Let the body C be in equilibrium, Q be any point of Sp , αn be the plane

tangent to Sp at Q, px , py , pz be the values in Q of the components of the superficial
distributed load. We denote with n the line passing through Q, orthogonal to αn and
directed going out from the material. We denote with nx , ny , nz the direction cosines
of n. Then it results

px = σx nx + τxy ny + τxz nz
py = τyx nx + σy ny + τyz nz (1.2.5)
pz = τzx nx + τzy ny + σz nz .

Proof If n is parallel to a coordinate axis, the (1.2.5) are obviously true. If n is not
parallel to none of the coordinate axes, the possible cases are the two following

(a) n is parallel to a coordinate plane

(b) n is not parallel to any of the coordinate planes.

In the case a, without loss of generality, we can suppose that n is orthogonal to

the axis z (Fig. 1.2.7). Then, reasoning as in the proof of the [1.2.4], we draw the
(1.2.5).
In the case b we denote with P = (x, y, z) a point of the line passing through Q
and parallel to n, inside V and such that dist(P, Q) is as small as we like. We denote
(Fig. 1.2.9)

– with P1 the intersection between α n and the line passing through P and parallel to
axis x,
– with P2 the intersection between α n and the line passing through P and parallel to
axis y,
– with P3 the intersection between α n and the line passing through P and parallel to
axis z,

and we put dx = dist (P, P1 ), dy = dist (P, P2 ), dz = dist (P, P3 ), so that P1 =

(x + dx, y, z), P2 = (x, y + dy, z), P3 = (x, y, z + dz), and dx, dy, dz are real numbers
very near to zero.
We denote with C0 the infinitesimal element of V isolated cutting the body C with
the planes PP1 P3 , PP1 P2 , PP2 P3 . We approximate C0 with the infinitesimal element
C0 of 3 individualized from the planes PP1 P3 , PP1 P2 , PP2 P3 , αn (Fig. 1.2.9).
Since C0 is infinitesimal we can assume that
38 1 The Three-Dimensional Problem

– on the faces PP1 P3 , PP1 P2 , PP2 P3 , σ and τ have values constant and equal to the
values assumed in Q,
– on the face P1 P2 P3 we apply a superficial distributed load px , py , pz which have
values constant and equal to the values assumed in Q. Obviously, the px , py , pz
positive are oriented according to the coordinate axes,
– on the infinitesimal element C0 we apply a volumetric load X, Y, Z which have
values constant and equal to the values assumed in Q. Obviously, the X, Y, Z
positive are oriented according to the coordinate axes.

Insofar, in conclusion, using the convention of Cauchy, the situation is that of

Fig. 1.2.9. To this point the reasoning done in the proof of [1.2.4] furnishes the
(1.2.5). 

1.2.5 Symmetry of Stress Vector

The symmetry of the τ can be generalized. In fact it results

[1.2.6] Let P be any point of V and n, s any two lines of 3 passing through P. Then
the component on s of the stress vector acting on the plane of normal n is equal to
the component on n of the stress vector acting on the plane of normal s.
Proof We denote with nx , ny , nz [resp. sx , sy , sz ] the direction cosines of n [resp. s]
and with n [resp. s] the unit vector of n [resp. s] (Fig. 1.2.10). We denote with tns
the orthogonal component on s of the stress vector tn (Fig. 1.2.11) and with tsn the
orthogonal component on n of the stress vector ts (Fig. 1.2.12).

Fig. 1.2.10

Fig. 1.2.11
1.2 Analysis of Stress 39

Fig. 1.2.12

Obviously

tns = 1 · tns = s × tn = tnx sx + tny sy + tnz sz

(1.2.6)
tsn = 1 · tsn = n × ts = tsx nx + tsy ny + tsz nz .

From here and from the (1.2.4) we get

   
tns = σx nx + τxy ny + τxz nz sx + τyx nx + σy ny + τyz nz sy
   
+ τzx nx + τzy ny + σz nz sz = σx sx + τxy sy + τxz sz nx
   
+ τyx sx + σy sy + τyz sz ny + τzx sx + τzy sy + σz sz nz = tsn . 

1.2.6 Relations Between Normal or Shearing Stress

and Components of Stress

Let P be any point of V and n, s any two lines of 3 passing through P. We denote
with nx , ny , nz [resp. sx , sy , sz ] the direction cosines of n [resp. s]. Employing the
(1.2.4), from the (1.2.6) we have

tns = σx nx sx + σy ny sy + σz nz sz
    (1.2.7)
+ τxy nx sy + ny sx + τxz (nx sz + nz sx ) + τyz ny sz + nz sy .

If s = n, evidently tns = tnn = σn and the (1.2.7) furnishes the important expression

σn = σx n2x + σy n2y + σz n2z + 2τxy nx ny + 2τxz nx nz + 2τyz ny nz . (1.2.8)

Let us consider now the case in which s belongs to the plane α n of normal n
(Fig. 1.2.13). We denote with Q the second end point of the stress vector tn and
consider the plane β passing through Q and orthogonal to s. The plane β intersects
s at point Q1 and the vector PQ1 is clearly just the orthogonal projection tns of tn on
s (Fig. 1.2.13). On the other hand we denote with n1 the line passing through Q and
orthogonal to α n (so that n1 lies on β). The line n1 intersects α n at point Q2 and the
vector PQ2 is clearly just the shearing stress τ n (Fig. 1.2.13).
40 1 The Three-Dimensional Problem

Fig. 1.2.13

In the plane α n we now trace the line r passing through Q2 and orthogonal to
s. Obviously the line r intersects s only at point Q1 , so that the vector PQ1 is also
the orthogonal projection τ ns of τ n on s (Fig. 1.2.13). As a consequence, since the
(1.2.7), we get the important expression

τns = σx nx sx + σy ny sy + σz nz sz
    (1.2.9)
+ τxy nx sy + ny sx + τxz (nx sz + nz sx ) + τyz ny sz + nz sy .

Remark 1.2.5 As underlined in the Sect. 1.2.6, τns is the orthogonal projection on s
of the (total) τ n acting on the plane of normal n. 

1.2.7 Principal Direction of Stress

Let P be a point of the body, ξ be a line passing through P, tξ the stress vector acting
on the plane of normal ξ . We will say that ξ is a principal direction of stress if tξ lies
on ξ . If ξ is a principal direction of stress we will say that σ ξ is a principal stress.

Remark 1.2.6 It is obvious that if ξ is a principal direction of stress it results

tξ = σξ , τξ = 0

and that for any line s belonging to the plane having normal ξ it results τξ s = 0. 

In the analysis of stress it is important to know if in a point of the body there are
some principal directions of stress and, in the affirmative case, how many there are.
To this aim let us consider the secular (i.e. algebraic of third order) equation in
the unknown one σ

σx − σ τxy τxz
τyx σy − σ τyz = 0. (1.2.10)
τzx τzy σz − σ
1.2 Analysis of Stress 41

Since

σx − σ τxy τxz  
τyx σy − σ τyz = (σx − σ ) σy − σ (σz − σ ) + 2τxy τxz τyz
τzx τzy σz − σ
 
− (σx − σ ) τyz
2 − σ − σ τ 2 − (σ − σ ) τ 2
y xz z xy
  2  
= σ + σx + σy + σz σ − σx σy + σx σz + σy σz − τxy − τxz − τyz σ
3 2 2 2
 
+ σx σy σz + 2τxy τxz τyz − σx τyz
2 − σ τ2 − σ τ2
y xz z xy

if we put
I1 = σx + σy + σz
I2 = σx σy + σx σz + σy σz − τxy
2
− τxz
2
− τyz
2

I3 = σx σy σz + 2τxy τxz τyz − σx τyz

2
− σy τxz
2
− σz τxy
2
,
the secular equation becomes

− σ 3 + I1 σ 2 − I2 σ + I3 = 0.

For the algebraic of third order equation it is known that the possible cases are
the three following

(1) Three real numbers σξ , ση , σζ exist such that σξ = ση , σξ = σζ , ση = σζ and

   
−σ 3 + I1 σ 2 − I2 σ + I3 = − σ − σξ σ − ση σ − σζ .

(2) Two real numbers σ ξ , σ η exist such that σξ = ση and

  2
−σ 3 + I1 σ 2 − I2 σ + I3 = − σ − σξ σ − ση .

(3) A real number σ ξ exists such that

 3
−σ 3 + I1 σ 2 − I2 σ + I3 = − σ − σξ .

In the case 1 we will say that the secular equation admits three simple roots. In
the case 2 we will say that the secular equation admits a simple root and a double
root. In the case 3 we will say that the secular equation admits a triple root.
Let us study the case 1. We consider the solution σ ξ of Eq. (1.1.10). It is possible
to prove that the problem of find three real numbers ξx , ξy , ξz such that

σx ξx + τxy ξy + τxz ξz = σξ ξx
τyx ξx + σy ξy + τyz ξz = σξ ξy
(1.2.11)
τzx ξx + τzy ξy + σz ξz = σξ ξz
ξx2 + ξx2 + ξx2 = 1

admits one and only one solution.

42 1 The Three-Dimensional Problem

Well we assert that the line individualized by the direction cosines ξx , ξy , ξz is a

principal direction of stress and σ ξ is the relative principal stress. In fact, because
of (1.2.4) and (1.2.11), we easily get

tξ x = σξ ξx
tξ y = σξ ξy
tξ z = σξ ξz .
 
As a consequence, tξ2 = σξ2 ξx2 + ξy2 + ξz2 = σξ2 and from this tξ = σξ , τξ =
 1/2
tξ2 − σξ2 = 0.
We can show that in any point they always exist at least three principal directions
of stress twos twos orthogonal, which we call principal triplet of stress. Precisely
[1.2.7] In the case 1 there are three and only three principal directions of stress ξ, η,
ζ. These are twos twos orthogonal and the relative principal stresses are σξ , ση , σξ
(so that they are twos twos distinct). We determine ξ, η, ζ by solving the problem
(1.2.11) and the two analogous related to ση , σξ .
In the case 2 there are ∞1 principal directions of stress. Precisely it exists a
direction principal of stress ξ and, if αξ is the plane normal to ξ, every line of αξ is
a principal direction of stress. Every line distinguished from ξ and not belonging to
αξ is not principal direction of stress. The principal stress acting on the plane αξ is
equal to σξ and, for every line t belonging to αξ , the relative principal stress (acting
on the plane of normal t) is equal to ση .
In the case 3 every line r is principal direction of stress and the relative principal
stress (acting on the plane of normal r) is equal to σξ . In such case we said that the
state of stress is hydrostatics. 
Remark 1.2.7 In every point of the body there are at least three principal directions
of stress twos twos orthogonal. To find the principal stresses we need to resolve the
secular equation. To find the principal directions of stress we need, for each principal
stress, to resolve a problem analogous to (1.2.11) one. 

1.2.8 Invariants of Stress

In analogy to the study performed in the Sect. 1.1.7, we call stress tensor the second
order tensor

σx τxy τxz
τyx σy τyz .
τzx τzy σz

We easily verify that the coefficients I1 , I2 , I3 are invariant when we turn from a
Cartesian orthogonal frame of reference O, x, y, z to another Cartesian orthogonal
frame of reference O , x , y , z . Let’s call them respectively linear invariant of stress,
1.2 Analysis of Stress 43

quadratic invariant of stress, cubic invariant of stress. Insofar, with reference to a

principal deformation reference 0, ξ , η, ζ , we have

σx + σy + σz = σξ + ση + σζ
σx σy + σx σz + σy σz − τxy − τxz
2 2
− τyz
2
= σξ ση + σξ σζ + ση σζ
σx σy σz + 2τxy τxz τyz − σx τyz
2
− σy τxz
2
− σz τxy
2
= σξ ση σζ .

1.2.9 Mohr’s Circle

We turn now our attention to the procedure of the Stress analysis called Mohr’s1.2.6
circle. It introduces the important advantages of the rapidity, simplicity and
synthesis proper of the graphic methods.
Let P be any point of the body, 0, x, y, z every orthogonal set of Cartesian axes,
r the line parallel to z and oriented as z (Fig. 1.2.14). We consider the set Λ of the
planes having support r (i.e. containing r), any plane α n of Λ and denote with n the
line normal to α n and directed going out from the material.
We denote with m the line one orthogonal to n and r and directed so that the
frames of reference O, x, y, z and P, n, m, r can be overlapped with a translation and
a rotation around the axis r (Fig. 1.2.15).

Fig. 1.2.14

n
Fig. 1.2.15 x

1.2.6 Christian Otto Mohr, Wesselburen 1835 – Dresden 1918.

44 1 The Three-Dimensional Problem

We consider the stress vector tn acting on α n and his components σ n and τ n .

We decompose τ n in the component τnm on m and in the component τ nr on r
(Fig. 1.2.14).
Finally, we assume that σn , τnm , τnr are positive if σn , τnm , τnr have the same
direction of the axes n, m, r respectively (Mohr’s circle convention).
We associate to the plane α n the ordered couple of real numbers (σn , τnm ). We
translate in P the frame of reference 0, x, y, z and denote with φ the angle of which
we must turn (in the verse counterclockwise) the axis x to overlap it to the axis n
(Fig. 1.2.15). It is obvious that when φ varies among 0 and 2π we obtain all and
only the planes of Λ. We will now prove that the set of the points (σn , τnm ) obtained
varying α n in Λ in all the possible ways is the circumference of a circle, that we will
call circle of Mohr.
First of all we remember that from the known formulas

sin2ϕ = 2 sinϕ cosϕ

cos2ϕ = cos2 ϕ − sin2 ϕ = 1 − 2 sin2 ϕ = 2 cos2 ϕ − 1

it follows

sin2ϕ = 2 sinϕ cosϕ

1 − cos2ϕ
sin2 ϕ = (1.2.12)
2
1 + cos2ϕ
cos2 ϕ = .
2

We observe (Fig. 1.2.16) that the direction cosines of n and m (in the frame
of reference 0, x, y, z) are, respectively, (cos ϕ, sin ϕ, 0) and (− sin ϕ, cos ϕ, 0).
Insofar, for the (1.2.8) and (1.2.9)

σn = σx cos2 ϕ + σy sin2 ϕ + 2 τxy sinϕ cosϕ

 
τnm = −σx sinϕ cosϕ + σy sinϕ cosϕ + τxy cos2 ϕ − sin2 ϕ .

n
Fig. 1.2.16
1.2 Analysis of Stress 45

From this it follows, taking into account the (1.2.12), that

1 + cos2ϕ 1 − cos2ϕ
σn = σx + σy + τxy sin2ϕ
2 2
σx + σy σx − σy
= + cos2ϕ + τxy sin2ϕ
2 2
σx σy
τnm = − sin2ϕ + sin2ϕ + τxy cos2ϕ
2 2
σx − σy
=− sin2ϕ + τxy cos2ϕ
2
from which

2 2
σx + σy σx − σy
σn − = cos2ϕ + τxy sin2ϕ
2 2
2
σx − σy
τnm
2
= − sin2ϕ + τxy cos2ϕ
2
from which

2 2
σx + σy σx − σy
σn − + τnm
2
= + τxy
2
.
2 2

Insofar, when ϕ varies among 0 and 2π , the point (σn , τnm ) describes in the plane
of Fig. 1.2.17 the circumference of center

σx + σy
,0
2

Fig. 1.2.17
46 1 The Three-Dimensional Problem

and radius
 1/2
2
σx − σy
+ τxy
2
.
2

As for the practical employment of the Mohr’s circle, the following proposition
subsists
   
[1.2.8] Let consider in the plane σn , τnm the points R = σx , τxy and P = σy , τxy . Let
denote with C the intersection among the segment PR and the axis σn (Fig. 1.2.18).
Let consider the Mohr’s circle of center C and radius dist(C, R). Let consider the
plane αn of Λ having oriented normal n (Fig. 1.2.18). We trace the line n passing
through R and parallel to n. We denote with Q the point intersection among n and
the circumference. We consider the point Q of the circumference symmetrical of Q
with reference to the axis σn . Then the coordinates of Q furnish in value and sign
(in the Mohr’s circle convention) the σn and the τnm acting on αn (Fig. 1.2.18). If
we fix a point (σn , τnm ) of the Mohr’s circumference, performing to the inverse such
construction, we determine the plane αn of Λ on which σn and τnm have the preset
values. 

Remark 1.2.8 It should be underlined that every Mohr’s circle has the center on the
axis σ n . 

Fig. 1.2.18

Fig. 1.2.19
1.2 Analysis of Stress 47

Problem 1.2.1 In the point A of the body the stress components (Fig. 1.2.19) are:

σx = 1, 000 kg cm2

σy = − 1, 000 kg cm2
σz = 0

τxy = − 1, 000 kg cm2
τxz = 0

τyz = 500 kg cm2 .

You shall trace the Mohr’s circle for the set of planes containing axis z and you
shall calculate σ n and τnm acting on the plane individualized from ϕ = π/4.
Solution To trace the Mohr’s circle for the set of planes containing axis z, σx , σy , σxy
are necessary. We notice that the state of stress in A is assigned with reference to
the convention of Cauchy, so for the infinitesimal cube of material of center A the
situation is that of Fig. 1.2.20. In the Fig. 1.2.20, z is orthogonal to the plane of
the drawing and only σ n and τnm are indicated. This premised, in the plane σ n , τnm
we bring the points R = (1, 000, −1, 000) and P = (−1, 000, 1, 000) (Fig. 1.2.21).
Consequently the center C of the Mohr’s circle is the origin of the frame of reference
and the radius one is CR. Having traced the Mohr’s circumference, we employ the
[1.2.8] to find again the results of Fig. 1.2.20.
As for the inferior face of the infinitesimal cube (Fig. 1.2.22), which is one of the
planes containing axis z, we draw the normal n going out from the material and we
orientate m so that x, y can be overlapped to n, m. Then σ n , τnm positive are (for the
Mohr’s circle convention) oriented as the axes n and m of Fig. 1.2.22. Then we trace

1000

1000
1000
1000
1000 1000
1000
1000
Fig. 1.2.20

Fig. 1.2.21
48 1 The Three-Dimensional Problem

Fig. 1.2.22

m
n

Fig. 1.2.23

Fig. 1.2.24

1000
1000

on the Mohr’s circle the parallel one to n (= x) passing through R (Fig. 1.2.23). This
way we find on the circumference the point Q (Fig. 1.2.23). The symmetrical of Q
with reference to axis of σ n is obviously Q = R = (1, 000, −1, 000), from which
we get σn = 1, 000, τnm = −1, 000. Insofar the situation is that of Fig. 1.2.24. We
have discovered again the situation of Fig. 1.2.20.
As for the right face of the infinitesimal cube (Fig. 1.2.25), which is one of the
planes containing axis z, we draw the normal n going out from the material and we
orientate m so that x, y can be overlapped to n, m. Then σ n , τnm positive are (for the
Mohr’s circle convention) oriented as the axes n and m of Fig. 1.2.25. Then we trace
on the Mohr’s circle the parallel one to n (= y) passing through R (Fig. 1.2.26). This

Fig. 1.2.25
1.2 Analysis of Stress 49

Fig. 1.2.26

Fig. 1.2.27 1000

1000

way we find on the circumference the point Q (Fig. 1.2.26). The symmetrical of Q
with reference to axis of σn is obviously Q = P = (−1, 000, 1, 000), from which
we get σn = −1, 000, τnm = 1, 000. Insofar the situation is that of Fig. 1.2.27. We
have discovered again the situation of Fig. 1.2.20.
As for the left face of the infinitesimal cube (Fig. 1.2.31), which is one of the
planes containing axis z, we draw the normal n going out from the material and we
orientate m so that x, y can be overlapped to n, m. Then σ n , τnm positive are (for the
Mohr’s circle convention) oriented as the axes n and m of Fig. 1.2.31. Then we trace
on the Mohr’s circle the parallel one to n (= y) passing through R (Fig. 1.2.32). This
way we find on the circumference the point Q (Fig. 1.2.32). The symmetrical of Q
with reference to axis of σ n is obviously Q = P = (−1, 000, 1, 000), from which
we get σn = −1, 000, τnm = 1, 000. Insofar the situation is that of Fig. 1.2.33. We
have discovered again the situation of Fig. 1.2.20.
As for the superior face of the infinitesimal cube (Fig. 1.2.28), which is one of the
planes containing axis z, we draw the normal n going out from the material and we
orientate m so that x, y can be overlapped to n, m. Then σ n , τnm positive are (for the
Mohr’s circle convention) oriented as the axes n and m of Fig. 1.2.28. Then we trace
on the Mohr’s circle the parallel one to n (= x) passing through R (Fig. 1.2.29). This
way we find on the circumference the point Q (Fig. 1.2.29). The symmetrical of Q
with reference to axis of σ n is obviously Q = R = (1, 000, −1, 000), from which
we get σn = 1, 000, τnm = −1, 000. Insofar the situation is that of Fig. 1.2.30. We
have discovered again the situation of Fig. 1.2.20.

Fig. 1.2.28
50 1 The Three-Dimensional Problem

Fig. 1.2.29

Fig. 1.2.30 1000

1000

Fig. 1.2.31

n
m

Fig. 1.2.32

Fig. 1.2.33 1000

1000

Now we calculate σ n and τnm acting on the plane individualized from ϕ = π/4
(Fig. 1.2.34). To this aim we trace on the Mohr’s circle the parallel one to n pass-
ing through R (Fig. 1.2.34). This way we find on the circumference the point Q
(Fig. 1.2.35). The symmetrical of Q with reference to axis of σ n is obviously
Q = (−1, 000, −1, 000), from which we get σn = −1, 000, τnm = 1, 000. Insofar
the situation is that of Fig. 1.2.35. 
1.2 Analysis of Stress 51

Fig. 1.2.34

Fig. 1.2.35

Problem 1.2.2 In the point A of the body the stress components (Fig. 1.2.19) are

σx = 0, σy = 1, 000 Kg/cm2 , σz = −1, 000 Kg/cm2 , τxy = 0, τxy = 0, τyz = 0.

You shall trace the Mohr’s circle for the set of planes containing axis x and you
shall calculate σ n and τnm acting on the planes individualized from ϕ1 = π/4, ϕ2 =
π/4 + π/2, ϕ3 = π/4 + π , ϕ4 = π/4 + 3π/2.
Solution To trace the Mohr’s circle for the set of planes containing axis z they are
necessary σy , σz , τyz . We notice that the state of stress in A is assigned with reference
to the convention of Cauchy, so for the infinitesimal cube of material of center A
and side l the situation is that of Fig. 1.2.36. In the Fig. 1.2.36 x is orthogonal to
the plane of the drawing and we have indicate only the σ n and τnm . This premised,
in the plane σ n , τnm we bring the points R = (1, 000, 0) and P = (−1, 000, 0)
(Fig. 1.2.37). Consequently the Mohr’s circle is that of Fig. 1.2.37.

Fig. 1.2.36
52 1 The Three-Dimensional Problem

Fig. 1.2.37

Fig. 1.2.38

Fig. 1.2.39

Let us calculate σ n and τnm acting on the plane individualized from ϕ3 = π/4
(Fig. 1.2.38). To this aim we trace on the Mohr’s circle the parallel one to n pass-
ing through R (Fig. 1.2.39). This way we find on the circumference the point Q
(Fig. 1.2.39). The symmetrical of Q with reference to axis of σ n is obviously
Q = (0, −1, 000), from which we get σn = 0, τnm = −1, 000. Insofar the situation
is that of Fig. 1.2.40.
Let us calculate σ n and τnm acting on the plane individualized from ϕ4 = π/4 +
π/2 (Fig. 1.2.41). To this aim we trace on the Mohr’s circle the parallel one to n
passing through R (Fig. 1.2.42). This way we find on the circumference the point
Q (Fig. 1.2.42). The symmetrical of Q with reference to axis of σ n is obviously
Q = (0, 1, 000), from which we get σn = 0, τnm = 1, 000. Insofar the situation is
that of Fig. 1.2.43.

Fig. 1.2.40
1.2 Analysis of Stress 53

Fig. 1.2.41

Fig. 1.2.42

Fig. 1.2.43

Let us calculate σ n and τnm acting on the plane individualized from ϕ3 = π/4 +
π (Fig. 1.2.44). To this aim we trace on the Mohr’s circle the parallel one to n
passing through R (Fig. 1.2.45). This way we find on the circumference the point
Q (Fig. 1.2.45). The symmetrical of Q with reference to axis of σ n is obviously
Q = (0, −1, 000), from which we get σn = 0, τnm = −1, 000. Insofar the situation
is that of Fig. 1.2.46.
Let us calculate σ n and τnm acting on the plane individualized from ϕ3 = π/4 +
3π /2 (Fig. 1.2.47). To this aim we trace on the Mohr’s circle the parallel one to n
passing through R (Fig. 1.2.48). This way we find on the circumference the point
Q (Fig. 1.2.48). The symmetrical of Q with reference to axis of σ n is obviously
Q = (0, −1, 000), from which we get σn = 0, τnm = −1, 000. Insofar the situation
is that of Fig. 1.2.49.

Fig. 1.2.44
54 1 The Three-Dimensional Problem

Fig. 1.2.45

Fig. 1.2.46

Fig. 1.2.47

Fig. 1.2.48

Fig. 1.2.49

We complete the study of the state of stress in the point A of the body with the
following considerations.
We ideally cut the infiniresimal cube of material of Fig. 1.2.36 according to the
planes (of the set containing axis x) individualized from ϕ1 , ϕ2 , ϕ3 , ϕ4 , as indicated
in Fig. 1.2.50. We isolate the inside part of material. Evidently, the actions suffering
(from the removed material) are those of Fig. 1.2.51. Overall the situation is that of
Fig. 1.2.52, except the τ with index x.
1.2 Analysis of Stress 55

Fig. 1.2.50

Fig. 1.2.51

Fig. 1.2.52

We notice that the equilibrium to the translation according to z of the infinitesimal

element of material delineated in Fig. 1.2.52 is satisfied, since we obviously have
!
l2  π l 2
l 2 1/2
l2 l2
1, 000 − 1, 000 cos l + = 1, 000 − 1, 000 = 0. 
2 4 2 2 2 2

Problem 1.2.3 In the point A of the body the stress components (Fig. 1.2.19) are

σx = 0, σy = 0, σz = 0, τxy = 0, τxz = 1, 000 kg/cm2 , τyz = 0.

You shall trace the Mohr’s circle for the set of planes containing axis y and you
shall calculate σ n and τ nm acting on the planes individualized from ϕ1 = π/4,
ϕ2 = π/4 + π/2, ϕ3 = π/4 + π , ϕ4 = π/4 + 3π/2.
56 1 The Three-Dimensional Problem

Fig. 1.2.53
1000

1000 1000

1000

Solution To trace the Mohr’s circle for the set of planes containing axis y σ z , σ x, τ zx
are necessary. We notice that the state of stress in A is assigned with reference to the
convention of Cauchy, so for the infinitesimal cube of material of center A and side
l the situation is that of Fig. 1.2.53. In the Fig. 1.2.53 y is orthogonal to the plane of
the drawing and we have indicate only the σ n and τ nm . This premised, in the plane
σ n , τ nm we bring the points R = (0, 1, 000) and P = (0, −1, 000). Consequently
the Mohr’s circle is that of Fig. 1.2.54.
Let us calculate σ n and τ nm acting on the plane individualized from ϕ1 = π/4
(Fig. 1.2.55). To this aim we trace on the Mohr’s circle the parallel one to n pass-
ing through R (Fig. 1.2.56). This way we find on the circumference the point Q
(Fig. 1.2.56). The symmetrical of Q with reference to axis of σ n is obviously
Q = (1, 000, 0), from which we get σn = 1, 000, τnm = 0. Insofar the situation
is that of Fig. 1.2.57.
Analogously proceeding, we calculate the normal stress and the shearing stress
acting on the planes individualized from ϕ2 = π/4 + π/2, ϕ3 = π/4 + π , ϕ4 =
π/4 + 3π/2. So we find the situation of Fig. 1.2.58. 

Fig. 1.2.54

Fig. 1.2.55
1.2 Analysis of Stress 57

Fig. 1.2.56

Fig. 1.2.57

Fig. 1.2.58

1.2.10 Mohr’s Principal Circles

Let us suppose to know, in the frame of reference of Fig. 1.2.59, the state of stress at
the point P of the body. We consider the Mohr’s circle for the set of planes of support
z (Fig. 1.2.59). We denote with σ 1 [resp. σ 2 ] the maximum [resp. minimum] normal
stress furnished by the Mohr’s circle (points A and B of Fig. 1.2.59).

Fig. 1.2.59
58 1 The Three-Dimensional Problem

In this matter it is important to know if σ 1 [resp. σ 2 ] is a principal stress.

Regarding σ 1 , it is obviously enough to see if the normal one to the plane of support
z corresponding to the point A is a principal direction of stress. Hence it is enough
to see if on the plane of support z corresponding to the point A the τ is zero.
To this aim, because of [1.2.8], it is enough to individualize on the Mohr’s circle
the point A symmetrical of A with reference to the axis σ n and successively the
line passing through R and A (Fig. 1.2.60). Such line individualizes the plane ᾱ (of
the set of the planes of support z) correspondent to A = (σ1 , 0) (Fig. 1.2.61). Then
on ᾱ it results τnm = 0. However the τ nz (on which the Mohr’s circle doesn’t give
indications) could be different from zero. Therefore in general the maximum [resp.
minimum] normal stress furnished by the Mohr’s circle is not a principal stress.
On this matter the following property subsists
[1.2.9] If the support of the set of planes is a principal direction of stress, then the
maximum and minimum normal stresses furnished by the Mohr’s circle are principal
stresses.
Proof With reference to the Fig. 1.2.59, let us suppose that z is a principal direction
of stress. To demonstrate the [1.2.9], it is enough to prove that on every plane of the
set of planes of support z it results τnz = 0. In fact then on the plane ᾱ of Fig. 1.2.61
the τ n will have intensity zero, and the same will happen on the plane related to the
minimum normal stress σ2 .
Let therefore α be a plane of the set of support z and n be the normal one
to α (Fig. 1.2.62). We suppose, reasoning by absurd, that τ n admits orthogonal
component on z different from zero. Well, for the symmetry of the τ , it results

Fig. 1.2.60

Fig. 1.2.61
1.2 Analysis of Stress 59

Fig. 1.2.62

τzn = τnz = 0. Then on the plane of normal z (i.e. on the plane x, y) the τ z has
orthogonal component on n different from zero. But this is absurd since, being for
hypothesis z a principal direction of stress, it is τz = 0 and consequently τnz = 0.
Insofar τzn = 0, however n is chosen. 

We will call principal circle of Mohr a circle of Mohr for which the support of
the set of planes is a principal direction of stress.
Remark 1.2.9 Recalling the Remark 1.2.7, we can assert that in every point of the
body there are three principal circles of Mohr. 

Problem 1.2.4 In the point A of the body the stress components (Fig. 1.2.19) are

σx = 0, σy = 1, 000 kg/cm2 , σz = 0, τxy = 0, τxz = 0, τyz = 0.

You shall determine the principal stresses and the principal directions of stress.
Solution Since τxy = 0 and τxz = 0, on the plane having x as normal it is τx = 0. So x
is a principal direction of stress. As a consequence, σ1 = σx = 0 is a principal stress.
Likewise, since τxy = 0 and τyz = 0, y is another principal direction of stress. Then
the second principal stress is σ2 = σy = 1, 000 kg/cm2 . Likewise, since τxz = 0
and τyz = 0, z is another principal direction of stress. Then the third principal stress
is σ3 = σz = 0. So the triplet x, y, z is a principal triplet of stress.
In Fig. 1.2.63 we can see the principal circle of Mohr related to the sets of planes
of support x, y, z. Of them, that related to the set of planes of support x is the circle
of radius 500 kg/cm2 . That related to the set of planes of support y reduces him to a
point (and precisely the origin of the frame of reference). That related to the set of
planes of support z coincides with that related to the set of planes of support x. 

Fig. 1.2.63
60 1 The Three-Dimensional Problem

Problem 1.2.5 In the point A of the body the stress components (Fig. 1.2.19) are

σx = 0, σy = 1, 000 kg/cm2 , σz = −1, 000 kg/cm2 , τxy = 0, τxz = 0, τyz = 0.

You shall determine the principal stresses and the principal directions of stress.

Solution As in the preceding problem, the τ on the coordinate planes are zero. It
follows of it that the coordinate axes are principal directions of stress and that
the principal stresses are σ1 = σx = 0, σ2 = σy = 1, 000 kg/cm2 , σ3 = σz =
−1, 000 kg/cm2 .
The three principal circles of Mohr, related to the set of planes of support x, y,
z, are given in Fig. 1.2.64. Of them, that related to the set of planes of support x is
the circle σ2 σ3 ; that related to the set of planes of support y is the circle σ1 σ3 ; that
related to the set of planes of support z is the circle σ 1 σ 2 . 
Problem 1.2.6 In the point A of the body the stress components (Fig. 1.2.19) are

σx = 0, σy = 0, σz = 0, τxy = 0, τxz = 1, 000 kg/cm2 , τyz = 0.

You shall determine the principal stresses and the principal directions of stress.

Solution By absurd, we suppose that x is a principal direction of stress. Then τx = 0

and, as a consequence, τxz = 0, that is absurd. So x is not a principal direction of
stress.
By absurd, we suppose that z is a principal direction of stress. Then τz = 0
and, as a consequence, τzx = 0, that is absurd. So z is not a principal direction of
stress.
Since τxy = 0 and τyz = 0, y is a principal direction of stress. Then a principal
stress is σn = σy = 0.
We get the other two principal stresses with the circle of Mohr for the set
of planes of support y. We obtain (Fig. 1.2.54) σξ = 1, 000 kg/cm2 , σζ =
−1, 000 kg/cm2 .
As the principal direction of stress ξ [resp. ζ ], it is enough consider in Fig. 1.2.65
the line passing through R and A = A = (1, 000, 0) [resp. B = B = ( − 1, 000, 0)].

Fig. 1.2.64
1.2 Analysis of Stress 61

Fig. 1.2.65

Fig. 1.2.66

The triplet ξ , η, ζ is a frame of reference orthogonal, Cartesian and principal of

stress.
The three principal circles of Mohr are given in Fig. 1.2.66. Obviously the circle
related to the set of planes of support ξ is ση σζ , that related to the set of planes of
support η is σξ σζ , that related to the set of planes of support ζ is σξ ση . 

1.2.11 Determination of the Maximum Normal Stress or Shearing

Stress by the Mohr’s Principal Circles

We have

[1.2.10] Let A be any point of the body, α n be any plane passing through A, n be the
oriented normal to αn , σn [resp. τ n ] be the normal [resp. shearing] stress acting on
α n . We draw the threeprincipal circles of Mohr in a same plane σ n , τ n . In this plane
the point P = ση , τη is external to the two smaller principal circles of Mohr and
inside to the maximum one, so that P always belongs to the zone of plane etched in
Fig. 1.2.67.

Proof We assume a frame of reference ξ , η, ζ principal of stress and draw the

three principal circles of Mohr in a same plane σ n , τ n . We suppose σξ < ση < σζ
(Fig. 1.2.67) and denote with nξ , nη , nζ the direction cosines of n.
Reasoning by absurd, let us suppose that P be external to the circle σ ξ , σ ζ
(Fig. 1.2.68). Then we have, with the notations of Fig. 1.2.69
62 1 The Three-Dimensional Problem

Fig. 1.2.67

Fig. 1.2.68

Fig. 1.2.69

   
dist2 P, Cξ ζ = dist2 (P, K) + dist2 K, Cξ ζ

from which

2 2
σξ + σζ σζ − σξ
σn − + τn2 = dist2 (P, K) +
2 2
from which, taking into account the (1.2.4)
1.2 Analysis of Stress 63

2 2
σξ + σζ σζ − σξ
dist2 (P, K) = σn − + τn2 −
2 2
2 2
σξ + σζ σξ + σζ σζ − σξ
= σn2 + − 2 σn + τn2 −
2 2 2
σξ2 σζ2
σξ σζ   σζ2 σξ2 σξ σζ
= + tn2 + + − σξ + σζ σn − − +
4 4 2 4 4 2
   2  2
= tnξ + tnη + tnζ + σξ σζ − σξ + σζ σn = σξ nξ + ση nη
2 2 2

 2   
+ σζ nζ + σξ σζ − σξ + σζ σξ n2ξ + ση n2η + σζ n2ζ
= ση2 n2η + σξ σζ − σξ ση n2η − σξ σζ n2ζ − σζ σξ n2ξ − σζ ση n2η
 
= ση2 n2n + σξ σζ − σξ ση n2η − σζ ση n2η − σξ σζ 1 − n2η
= ση2 n2η + σξ σζ − σξ ση n2η − σζ ση n2η − σξ σζ + σξ σζ n2η
    
= n2η ση2 − σξ ση − σζ ση + σξ σζ = n2η ση − σξ ση − σζ .

This is absurd because, having supposed P external to the circle σ ξ , σ ζ , it is

dist(P, K) > 0 while, being σξ < ση < σζ
  
n2η ση − σξ ση − σζ ≤ 0.

Insofar P cannot be external to the circle σξ , σζ .

With analogously reasoning, we prove that P = (σn , τn ) is external to the two
smaller principal circles of Mohr. 
Remark 1.2.10 From the [1.2.10], the three principal circles of Mohr immediately
furnish the level of the severity of the state of stress in the point P. As we will see
afterwards, such level is measured from various indicators, among which the most
important are τ max and σ max . We denote with τ max [resp. σ max ] the maximum of
the numerical set described by τ n [resp. σ n ] when the plane α n (passing through P)
varies in all the possible ways. Evidently, as underlined in Fig. 1.2.69, because of
the [1.2.10], τ max is equal to the radius of the maximum principal circle of Mohr
and σ max is equal to the greatest of the three principal stresses. 

1.2.12 Plane State of Stress

Let P be any point of the body. We said that the state of stress is plane in P if it
exists a plane α passing through P and such that, for any plane α n passing through
P, the stress vector tn (acting in P on α n ) lies in α n . If the state of stress is plane in
P, we call α plane of stress.
64 1 The Three-Dimensional Problem

We have
[1.2.1] If the state of stress is plane in P, then an unloaded plane (that is whose
stress vector has zero intensity) passing through P exists. Besides if an unloaded
plane passing through P exists, then the state of stress is plane in P.
Proof Let the state of stress be plane in P. We denote with α the plane of stress and
with a the normal one to α. Since the state of stress is plane, the stress vector ta
(acting on the plane α) lies in α. Then σa = 0. So to prove the thesis it is enough to
show that τa = 0. Proceeding by absurd, we suppose τa = 0 and we denote with b
the line of action of τ a and with β the plane of normal b. For the symmetry of the
τ , it results (Fig. 1.2.70) τba = τab = τa = 0. This is absurd. In fact from a side
tba = τba , from the other (because of the hypothesis) tb must be parallel to α and
as a consequence tba = 0. This involves τba = 0. This way we have verified that
σa = 0 and τa = 0. Then we have ta = 0 and this is really what we intend when we
say that the plane α is unloaded.
We suppose vice versa that a plane α exists unloaded and passing through P.
We denote with a its normal. By hypothesis ta = 0 and then τa = 0. Insofar a is a
principal direction of stress. Clearly, being ta = 0, it results σa = 0. We assume now
a frame of reference a, b, c principal of stress. Let α n be any plane passing through
P and n be the normal one to α n . We denote with na , nb , nc the direction cosines of
n and with tn the stress vector acting on α n . The equations of Cauchy furnish

tna = σa na = 0
tnb = σb nb
tnc = σc nc .

This proves that the stress vector tn lies in the plane b, c (which, being of normal
a, is really the plane α). Then all the stress vectors are contained in α. Then the state
of stress is plane and α is the plane of stress. 

We also have
[1.2.2] If in A the state of stress is plane, it results

Fig. 1.2.70
1.2 Analysis of Stress 65

σx τxy τxz
τyx σy τyz = 0
τzx τzy σz
and vice versa.

Proof Let us suppose that in A the state of stress is plane. Then, because of [1.2.11],
an unloaded plane exists. From this we get, reasoning as in the proof of [1.2.11],
that a principal stress is zero. So, recalling that the principal stresses are solution of
the secular equation, we have

σx − 0 τxy τxz
τyx σy − 0 τyz = 0.
τzx τzy σz − 0

Vice versa let us suppose that

σx τxy τxz
τyx σy τyz = 0.
τzx τzy σz
Then the secular equation

σx − σ τxy τxz
τyx σy − σ τyz =0
τzx τzy σz − σ

admits the solution σ = 0. So a principal stress is zero. Clearly, the plane on which
such principal stress act is unloaded. As a consequence, because of [1.2.11], the
thesis follows. 

1.2.13 Uniaxial State of Stress

The state of stress in the point A of the body is said uniaxial if a line r exists such
that, for any plane α n passing through A, the stress vector acting on α n lies on r. If
it exists, r is said axis of stress.
Evidently

[1.2.3] If in A the state of stress is plane and it admits two planes of stress α and
β, with α = β, then the state of stress is uniaxial. The axis of stress is the line α ∩ β.
Furthermore if in A the state of stress is uniaxial with axis r, then the state of stress
is plane and every plane containing r is plane of stress.
66 1 The Three-Dimensional Problem

1.2.14 Measurement of Stress

The acting loads and the stresses of a body can experimentally be measured in
a Laboratory material tests. For the measure of the acting loads, instruments
called cells of load are used, that employ electric extensometers to resistance.
The force to be measured is applied on the summit of the cell of load. This
is constituted by a spherical (generally of great radius) surface so that the con-
centric locating and the axiality of the load are insured. The deformation of
the cell of load is in relief from the installed straingauges and translated, in
conclusion, in an electric signal proportional to the intensity of the applied
load.
For the experimental determination of the state of stress in a point, a technique
consists of the experimental determination of the state of strain in that point. Of here
and from the stress–strain relationships (what we will discuss in the succession) we
can calculate the values of the stresses in the point.
Another technique, often very useful, is the photoelasticity. It is necessary
however to realize a model of the body (to analyze) in an opportune mate-
rial (for instance bakelite) and to observe it in polarized light. The application
of the load then makes to appear clear and obscure fringes on the model.
The state of stress in the model is visualized as such. Clear indications are
obtained on the parts of the body that are stressed the most. The photoelastic-
ity also allows to measure the value of the stresses in the inside points of the
body.

1.3 Principle of Virtual Works

1.3.1 Principle of Virtual Works

We consider a solid body C that occupies with continuity a region V of the three-
dimensional space. As usual, we denote with S the surface of the body (that is
the boundary of V). We now derive a relationship in which stresses and strains of
the body appear. Such relationship goes under the name of principle of the vir-
tual works. It is, and this will be immediately clarified, an abstract relationship (i.e.
a mathematical relationship deprived of physical meaning). However, within the
Theory of elasticity, it is one of the most powerful tools of theoretical investigation
and calculation. It is important to underline that we do not make hypotheses neither
on the nature neither on the mechanical characteristics of the material of which the
body C is made.
Let’s consider the following problem A (or force system). On the body we apply
volumetric loads X, Y, Z and distributed superficial loads px , py , pz , which by
hypothesis constitute a balanced system of forces (Fig. 1.3.1). We denote with σ x ,
1.3 Principle of Virtual Works 67

Fig. 1.3.1

O
x

σ y , σ z , τ xy , τ xz , τ yz a sextuple of functions satisfying (in V) the differential equations

of equilibrium

∂σx ∂τxy ∂τxz

+ + +X =0
∂x ∂y ∂z
∂τyx ∂σy ∂τyz
+ + +Y =0 (1.3.1)
∂x ∂y ∂z
∂τzx ∂τzy ∂σz
+ + +Z =0
∂x ∂y ∂z

and (on S) the Cauchy’s boundary conditions

px = σx nx + τxy ny + τxz nz
py = τyx nx + σy ny + τyz nz (1.3.2)
pz = τzx nx + τzy ny + σz nz .

In the (1.3.2) nx , ny , nz denote the direction cosines of the normal to Sp , oriented

going out from the material (Fig. 1.3.2).
Then in the problem A every infinitesimal element of the body, inside
or at the boundary, is in equilibrium. Obviously this does not mean that
such stresses are indeed those born in the body on which we have applied
the load. In fact every infinitesimal element of material, being submitted to

Fig. 1.3.2
68 1 The Three-Dimensional Problem

σx , σy , σz , τxy , τxz , τyz and to the external load X, Y, Z, px , py , pz , deforms

it, but the εx , εy , εz , γxy , γxz , γyz born in such deformation can be not able
to satisfy the compatibility equations (1.1.28). Therefore, the stresses that
we employ in the problem A are balanced but (in the general case) not
compatible.
Let’s consider the problem B (or displacement system) constituted by the body
C when it deforms itself as imposed by an assigned field of displacements ũ, ṽ, w̃
(where such functions are regular). We do not specify the cause that excites the
deformation of the body. It can be from a load that is applied on the body but it
also can be from a thermal load, due to the presence of a source of heat. Or still
from a rigid rotation, where we have displacements but not strain. We make the
hypothesis, in the problem B, of small deformations, so that in the problem B the
ε̃, γ̃ are connected to the ũ, ṽ, w̃ by

∂ ũ ∂ ṽ ∂ w̃
ε̃x = , ε̃y = ε̃z = ,
∂x ∂y ∂z
(1.3.3)
∂ ũ ∂ ṽ ∂ ũ ∂ w̃ ∂ ṽ ∂ w̃
γ̃xy = + , γ̃xz = + , γ̃yz = + .
∂y ∂x ∂z ∂x ∂z ∂y

We call external virtual work and denote with Lve the (abstract) work1.3.1 that the
loads of problem A (that is px , py , pz , X, Y, Z) make when we impress to their points
of application the displacements that such points have in the problem B.
We call internal virtual work and denote with Lvi the real number

 
σx ε̃x + σy ε̃y + σz ε̃z + τxy γ̃xy + τxz γ̃xz + τyz γ̃yz dV.
V

It subsists the following property, that we call virtual works principle.

[1.3.1] In the previous hypotheses the external virtual work is equal to the internal
virtual work.

Proof It results, keeping in mind of the (1.3.2)

  
Lve = (px dσ )(ũ) + (py dσ )(ṽ) + (pz dσ )(w̃)
S
 
S

S

+ (X dV)(ũ) + (Y dV)(ṽ) + (Z dV)(w̃)

 V V
 V
 
= px ũ + py ṽ + pz w̃ dσ + (X ũ + Y ṽ + Z w̃) dV
S V

1.3.1 In Physics we call work the scalar product of a force and of the displacement of its point of
application.
1.3 Principle of Virtual Works 69

 
= σx nx + τxy ny + τxz nz ũ + (τyx nx + σy ny + τyz nz ) ṽ
S

  
+ τzx nx + τzy ny + σz nz w̃ dσ + (X ũ + Y ṽ + Z w̃) dV
 V
   
= σx ũ + τxy ṽ + τxz w̃ nx + τyx ũ + σy ṽ + τyz w̃ ny
S

  
+ τzx ũ + τzy ṽ + σz w̃ nz dσ + (X ũ + Y ṽ + Z w̃) dV.
V

From this we have, since the formulas of Gauss1.3.2

 
∂   ∂  
Lve = σx ũ + τxy ṽ + τxz w̃ + τyx ũ + σy ṽ + τyz w̃
V ∂x ∂y
 
∂  
+ τzx ũ + τzy ṽ + σz w̃ dV + (X ũ + Y ṽ + Z w̃) dV
∂z V

∂ ũ ∂σx ∂ ṽ ∂τxy ∂ w̃ ∂τxz
= σx + ũ + τxy + ṽ + τxz + w̃
V ∂x ∂x ∂x ∂x ∂x ∂x
∂ ũ ∂τyx ∂ ṽ ∂σy ∂ w̃ ∂τyz ∂ ũ
+ τyx + ũ + σy + ṽ + τyz + w̃ + τzx
∂y ∂y ∂y ∂y ∂y ∂y ∂z
∂τzx ∂ ṽ ∂τzy ∂ w̃ ∂σz
+ ũ + τzy + ṽ + σz + w̃ + X ũ + Y ṽ + Z w̃ dV
∂z ∂z ∂z ∂z ∂z
 
∂σx ∂τxy ∂τxz ∂τyx ∂σy ∂τyz
= ũ + + + X + ṽ + + +Y
V ∂x ∂y ∂z ∂x ∂y ∂z
∂τzx ∂τzy ∂σz ∂ ũ ∂ ṽ ∂ w̃
+ w̃ + + + Z + σx + σy + σz
∂x ∂y ∂z ∂x ∂y ∂z

∂ ũ ∂ ṽ ∂ ũ ∂ w̃ ∂ ṽ ∂ w̃
+ τxy + + τxz + + τyz + dV,
∂y ∂x ∂z ∂x ∂z ∂y

from which, because of (1.3.1) and (1.3.3)


 
Lve = σx ε̃x + σy ε̃y + σz ε̃z + τxy γ̃xy + τxz γ̃xz + τyz γ̃yz dV.
V

1.3.2 Karl Friedrich Gauss, Brunswick 1777 – Gottingen 1855. He is one of the greatest mathe-
maticians that has ever had the humanity. The Gauss’s formulas transform the surface integral in
volume integral.
 Precisely,
 if f it is a function defined in V , if f and the boundary S of V are regular
and if n = nx , ny , nz is the normal one to S, it results (Fig. 1.3.2) that the surface integral on S of
f nx [resp. f ny ] [resp. f nz ] is equal to the volumetric integral on V of the x [resp. y] [resp. z] partial
derivative of f.
70 1 The Three-Dimensional Problem

The thesis follows. 

Remark 1.3.1 If the body C is rigid, from the definitions of ε and γ it immediately
follows that and ε = γ = 0 anywhere. Then Lvi = 0 and from this it follows
Lve = 0. In the case of a deformable body, if ũ, ṽ, w̃ are the components of a rigid
translation or a rigid rotation, it results Lvi = 0 and from this follows Lve = 0. This
is an obvious consequence of the (1.1.5), (1.1.6), and (1.3.3). 

1.4 Relations Between Stress and Strain

1.4.1 Tensile Breaking Test

We consider a steel test specimen having in a portion of lenght L = 100 mm a
cross section constant and circular with  = 10 mm (Fig. 1.4.1). We submit such
specimen to a tensile breaking test, meaning an increasing effort of traction until
to breaking. When the test machine (that generally is of hydraulic
 type) applies the
tensile load P, the specimen suffers a deformation L − L /L, where L denotes
the distance (corresponding to the actual value of P) between the points H and K
(Fig. 1.4.1). This way we get the classical diagram load–deformation of the tensile
fracture test.
It is opportune to specify that in such type of test machine, in proximity of the
breaking, there are phenomena of hydraulic nature that alter notably the diagram
load–deformation.
Therefore, to get the real diagram load–deformation and then a best understand-
ing of the phenomenon, we employ a test machine with direct1.4.1 application of the
load (Fig. 1.4.2). So we obtain a diagram load–deformation that presents the course
of Fig. 1.4.3.
Until to the value Pp the deformation suffered by the specimen proportionally
varies to the applied load. We call the load Pp proportionality load.
When P is greater then Pp , the behavior of the material is non linear and the
elongations grow with more rapid course of the load.
Until Pe value the specimen returns to the primitive length if we carry back to
zero the applied load, so permanent deformations are not verified. We call the load
Pe limit elastic load.
When P is greater then Pe , permanent deformations happen, which we call plastic
deformations. With increasing the load a little over Pe , we arrive to a value Ps (that

Fig. 1.4.1

1.4.1 This is an ancient type of test machine. The load was increased manually adding small weights
(one to the time) on the dish of the test machine.
1.4 Relations Between Stress and Strain 71

Fig. 1.4.2

we call yelding load) in correspondence of which the deformations rapidly increase.

Practically it results Ps = Pe .
Increasing P over Ps , we finally arrive to a value Pr (what we call breaking load)
that is the maximum load supportable from the material.
When P attains the maximum value Pr (point E of Fig. l.4.3), a characteristic
phenomenon happens which we call necking. When P attains the value Pr , then the
deformations do not happen more uniformly on the whole length of the specimen
but they are pronounced in a brief tract that quickly grows thinner up to the breaking
(Fig. 1.4.4).
We denote with Ω the section that the specimen has before of the tensile test and
with z the axis of the specimen. We will prove (in the problem of the axial barycen-
tric load of the Saint Venant’s solid) that σ z is constant in Ω, so that in the linear
elastic field it results P = σz . For these reason we agree to call proportionality
stress the quantity

Pp
σp = ;


Fig. 1.4.3

Fig. 1.4.4
72 1 The Three-Dimensional Problem

yelding stress [resp. breaking stress] the quantity

 
Ps Pr
σ0 = resp. σr = ;
 

moreover we call breaking elongation the pure number

L − L
L
where L is the distance between H and K (Fig. 1.4.1) after the breaking. We measure
it by touching the two slugs of the specimen.
In Sch. 1.4.1 the typical values (at the temperature of 20◦ C) of the σ r of some
material ones are given.
In conclusion, the curve load-deformation of Fig. 1.4.4 has an elastic part (curve
OAB) and a plastic part (curve BEC).
If in any point W of the elastic curve OAB we decrease the load up to zero and
then we report it to the value related to W, the pen of the autoplotter again draws
(in descent and in ascent) the elastic curve OAB with good approximation, returning
before in O and after in W.
If, going away from the point W, we increase the load, the pen draws the curve
OABEC of Fig. 1.4.3.
If in any point D of the plastic curve BEC (Fig. 1.4.3) we decrease the load up
to zero and then we report it to the value related to D, the following phenomenon
happens.
When we decrease the load up to zero, the pen draws the curve DF, assimilable
with good approximation to a line parallel to the line passing through O and A (of the
linear elasticity). Insofar when the specimen is unloaded, it remains permanently (or
plastically) deformed. Precisely when we decrease the load up to zero, the distance
between the points H and K of Fig. 1.4.1 does not return to the initial value L but
sensitively remains greater of it. If now we increase the load up to the value related
to the point D, the pen again draws the line DF with good approximation, returning
in D. If, going away from the point D, we increase the load, the pen draws the curve
DEC of Fig. 1.4.3.
The elastic deformations are studied in the Theory of elasticity. The plastic
deformations are studied in the Theory of plasticity.
In this book, we deal only of the tract OAB of Fig. 1.4.1, limiting us to the
Theory of elasticity. For the linear tract OA (i.e. in the field of the linear elastic-
ity), Hooke1.4.2 proposed, from the results of the previous tensile breaking test, the
famous law
σ
ε= .
E

1.4.2 Robert Hooke, Wight 1635 – London 1703.

 1.4

Sch 1.4.1 RELATIONS BETWEEN STRESS AND STRAIN

σ r (tract.) (kg/cm2 ) σ r (comp.) (kg/cm2 ) E (kg/cm2 ) ν G (kg/cm2 )

Weldable iron 3,300 ~ 4,000 ≥ 1,800 2,000,000 0.3 770,000

Omogeneous iron 3,400 ~ 5,000 ≥ 1,900 2,150,000 0.3 830,000
Smelting steel 5,000 ~ 20,000 ≥ 2,800 2,000,000 0.18 850,000
Mold steel for springs ≥ 10,000 2,200,000 0.29 850,000
Hardened steel for springs ≥ 17,000 2,200,000 0.29 850,000
Cast iron 2,400 8,500 1,050,000 0.31 400,000
Hardened nickel-chromium
steel (1.5% nic.–0.2% chr.) 6,000 ~ 8,000
Hardened nickel-chromium
Relations Between Stress and Strain

steel (4.5% nic.–1.1% chr.) 12,000 ~ 14,000

Chromium dead soft steel 16,000 ~ 18,000
Annealed copper 2,500
Cold finished copper 3,900 1,200,000 0.35 440,000
Brass 4,800 ~ 5,200 1,050,000
Superforged bronze 4,400 ~ 5,600 1,200,000
Ordinary bronze 1,500 800,000
Hardened aluminum 5,050 724,000 0.2 290,000
Melted aluminium 930 ~ 1,500 675,000 0.2 270,000
Laminated aluminium 1,500 726,000 0.2 290,000
Zinc 2,350 873,000 ~ 1,200,000
Tin 350 400,000
Melted lead 125 50,000
Glass 250 700,000 0.2 290,000
Paper 40 ~ 180
Leather 250 ~ 400 1,250 ~ 2,250
Hemp for cords 1,200 ~ 1,350 6,000 ~ 15,000
Pine (load parallel to fibers) 790 280 90,000 ~ 108,000
73
 74

Sch 1.4.1 (Continued)

σ r (tract.) (kg/cm2 ) σ r (comp.) (kg/cm2 ) E (kg/cm2 ) ν G (kg/cm2 )

Fir (load parallel to fibers) 750 245 92,000 ~ 110,000

Oak (load parallel to fibers) 965 345 100,000 ~ 108,000
Beech (load parallel to fibers) 1,340 320 128,000 ~ 180,000
Ordinary granite 300 ~ 800
Hard granite 800 ~ 1,200
Very hard granite 1,200 ~ 2,000
Basalt 1,000 ~ 2,000
Slate 500 ~ 1,500
Marble 500 ~ 1,800
Soft tuff 60 ~ 300
Hard tuff 300 ~ 1,500
Brick (weak cooking) 100 ~ 200
Brick (middle cooking) 200 ~ 300
Brick (strong cooking) 300 ~ 900
Brick masonry 140
Cement 1-sand 3 after 28 days 150 ~ 350
Concrete as rich of cement 150 ~ 600 250,000 0.12
1

Cement mortar 40 ~ 60
Lime mortar 15 ~ 30
Beryllium 3,150,000 0.05 1,500,000
Boron 4,200,000 0.2 1,750,000
Graphite 3,850,000 0.2 1,600,000
Tungsten 4,000,000 0.28 1,550,000
Silica 700,000 0.2 290,000
Molybdenum 3,360,000 0.34 1,250,000
Epoxy resin 350,000 0.34 130,000
Nickel 2,100,000 0.3 810,000
Titanium 1,200,000 0.31 460,000
The Three-Dimensional Problem
1.4 Relations Between Stress and Strain 75

Remark 1.4.1 The materials for which the tensile breaking test furnishes a diagram
of the type of that of Fig. 1.4.3 (which is obtained with the steel) are said ductil. All
the metals are ductil. Other materials (and the glass is one of these) exist however
for which in the tensile breaking test plastic deformations are not verified. Such
materials are said brittle. Their diagram load-deformation is practically constituted
only by the linearly elastic tract OA.
Obviously it is not advisable to employ brittle materials in the constructions. The
structure in fact can break suddenly, without manifesting the proximity of the crisis
preliminarily and gradually with great displacements.
It is also important to notice that for the steel (and this generally happens for
all the metals) the test of compression practically furnishes a diagram polarly
symmetrical of that of Fig. 1.4.3.
Other materials of practical interest in the constructions have however trac-
tion behavior deeply different from the compression behavior. We will study them
afterwards. 
Remark 1.4.2 We give in Fig. 1.4.5 the diagram load-deformation obtained by a stan-
dard test machine for an usual steel bar to concrete reinforcing. The values resulting
by the test are: as for as the yelding stress σ0 = 5,500 kg/cm2 , as for as the breaking
stress σr = 6,500 kg/cm2 , as for as the breaking elongation L/L = 0.25. 

1.4.2 Homogeneous and Isotropic Materials – Navier’s Relations

The tensile breaking test and numerous other experiences performed in the
Laboratories material tests have suggested that a linear link between stress and
strain (that is of the type of the Hooke’s law) could describe, at least in the field
of small deformations, the behavior of a structure. This conjecture was correct.
Furthermore it conducted to a good mathematical model of the problem of the elastic
equilibrium. All the values furnished from this model have always received bright
experimental confirmations.
So we assume that in any point (x, y, z) of the volume V occupied by the body it
results

εx (x, y, z) = c11 (x, y, z) σx (x, y, z) + c12 (x, y, z) σy (x, y, z)

+ c13 (x, y, z) σz (x, y, z) + c14 (x, y, z) τxy (x, y, z)
+ c15 (x, y, z) τxz (x, y, z) + c16 (x, y, z) τyz (x, y, z)
εy (x, y, z) = c21 (x, y, z) σx (x, y, z) + c22 (x, y, z) σy (x, y, z)
+ c23 (x, y, z) σz (x, y, z) + σ24 (x, y, z) τxy (x, y, z)
+ c25 (x, y, z) τxz (x, y, z) + c26 (x, y, z) τyz (x, y, z)
εz (x, y, z) = c31 (x, y, z) σx (x, y, z) + c32 (x, y, z) σy (x, y, z)
+ c33 (x, y, z) σz (x, y, z) + c34 (x, y, z) τxy (x, y, z)
+ c35 (x, y, z) τxz (x, y, z) + c36 (x, y, z) τyz (x, y, z)
76 1 The Three-Dimensional Problem

γxy (x, y, z) = c41 (x, y, z) σx (x, y, z) + c42 (x, y, z) σy (x, y, z)

+ c43 (x, y, z) σz (x, y, z) + c44 (x, y, z) τxy (x, y, z)
+ c45 (x, y, z) τxz (x, y, z) + c46 (x, y, z) τyz (x, y, z)
γxz (x, y, z) = c51 (x, y, z) σx (x, y, z) + c52 (x, y, z) σy (x, y, z)
+ c53 (x, y, z) σz (x, y, z) + c54 (x, y, z) τxy (x, y, z)
+ c55 (x, y, z) τxz (x, y, z) + c56 (x, y, z) τyz (x, y, z)
γyz (x, y, z) = c61 (x, y, z) σx (x, y, z) + c62 (x, y, z) σy (x, y, z)
+ c63 (x, y, z) σz (x, y, z) + c64 (x, y, z) τxy (x, y, z)
+ c65 (x, y, z) τxz (x, y, z) + c66 (x, y, z) τyz (x, y, z).

We call these relationships constituent link and call linearly elastic a material
satisfying them.
If all the cij (x, y, z) are constant in regard of x, y, z we said that the material is
homogeneous.
If the material is homogeneous we call elastic constants of the material the 36
real numbers cij . We will prove later on, with energetic considerations, that the dis-
tinct elastic constants are at best 21. Such generic material is also said anisotropic
in (x, y, z).
If, in any point (x, y, z) of V, every value cij (x, y, z) does not depend on the choice
of the Cartesian frame of reference O, x, y, z, we said that the material is isotropic
in the point (x, y, z).
If the material is homogeneous and isotropic we can show, with energetic
considerations, that the distinct elastic constants are at best 2.
Remark 1.4.3 Let us consider any point P = (x, y, z) of the body, cut out an infinites-
imal cube containing P and load its faces with a system of σ and τ . Obviously, if the
material is isotropic in P, wathever orientation the infinitesimal cube has, it deforms
it always equally. In fact the previous constituent link always furnishes the same val-
ues of ε and γ independently from the choice of the Cartesian frame of reference.
Instead, if the material is anisotropic in P the deformation of the infinitesimal cube
generally depends on the orientation of the cut. 
For a homogeneous, isotropic and linearly elastic material, Navier1.4.3 proposed
for the constituent link the following form (Navier’s direct relationships)
  
εx = E1 σx − ν σy + σz
 
εy = E1 σy − ν (σx + σz )
  
εz = E1 σz − ν σx + σy
(1.4.1)
γxy = G1 τxy
γxz = G1 τxz
γyz = G1 τyz

1.4.3 Louis Marie Henri Navier, Dijon 1785 – Paris 1836.

1.4 Relations Between Stress and Strain 77

where
E
G= . (1.4.2)
2(1 + ν)

In the (1.4.1) E takes the name of modulus of Young1.4.4 (or longitudinal elastic
modulus), ν takes the name of modulus of Poisson1.4.5 (or transversal contraction
modulus), G takes the name of tangent modulus. Their typical values for some
common construction material are given in Sch. 1.4.1.
From the first three of the (1.4.1), adding member to member and denoting with
T the linear invariant of stress and with  the linear invariant of strain, we have

1 1 − 2ν
= (T − ν2T) = T . (1.4.3)
E E
From the first of the (1.4.1) and from the (1.4.3) then we get
 
σx = Eεx + ν σy + σz + σx − σx = Eεx + νT − νσx

from which
E ν ν E
σx = εx + T = 2Gεx + ·  = 2Gεx + λ
1+ν 1+ν 1 + ν 1 − 2ν
where
Eν
λ= . (1.4.4)
(1 + ν)(1 − 2ν)

We call λ constant of Lamè1.4.6 .

This way from the (1.4.1) we obtain the following inverse Navier relationships
 
σx = 2Gεx + λ εx + εy + εz 
σy = 2Gεy + λ  εx + εy + εz
σz = 2Gεz + λ εx + εy + εz
(1.4.5)
τxy = Gγxy
τxz = Gγxz
τyz = Gγyz .

We have
[1.4.1] If the material is homogeneous, isotropic and linearly elastic, then in any
point of the body a principal triplet of strain is a principal triplet of stress and vice
versa.

1.4.4 Thomas Young, Milverton 1773 – London 1829.

1.4.5 Siméon Denis Poisson, Pithiviers 1781 – Paris 1840.
1.4.6 Gabriel Lamé, Tours 1795 – Paris 1870.
78 1 The Three-Dimensional Problem

Proof Let the orthogonal set of Cartesian axes O, x, y, z be a principal triplet of

strain. Then γxy = γxz = γyz = 0. As a consequence, because of (1.4.5), τxy =
τxz = τyz = 0, so that O, x, y, z is also a principal triplet of stress.
Let now O, x, y, z be a principal triplet of stress. Then τxy = τxz = τyz = 0. As
a consequence, because of (1.4.1), γxy = γxz = γyz = 0, so that (as seen in Remark
1.1.9) O, x, y, z is also a principal triplet of strain. 

1.4.3 Bounds for the Elastic Modulus

We consider an infinitesimal cube of material with the faces parallel to the coordi-
nate planes and submitted to only σ x of traction and therefore positive (Fig. 1.4.6).
Evidently, being τxy = τxz = τyz = 0, the triplet x, y, z is principal of stress and
therefore, since [1.4.1], also principal of strain. Consequently the cube deforms itself
in a parallelepiped with faces parallel to the coordinate planes. Moreover from the
(1.4.1) we have
σx ν ν
εx = , εy = − σx , εz = − σx .
E E E
For all the known materials, in the situation of Fig. 1.4.6 the
 edge parallel to x
axis extends (εx > 0) and those parallel to y and z axes shorten εy < 0, εz < 0 .
As a consequence

E > 0, ν > 0.

Since for all the known materials a hydrostatic pressure tends to diminish the
bulk, it is clear from (1.4.3) that

1 − 2ν > 0

so that1.4.7 ν < 1/2.

The quantity
E
μ= ,
3(1 − 2ν)

ΔL /L
Fig. 1.4.5

1.4.7 The result 0 < ν < 1/2 is notably important in the Mathematical theory of plates.
1.4 Relations Between Stress and Strain 79

Fig. 1.4.6

that since (1.4.3) represents the ratio of the compressive hydrostatic pressure to the
specific bulk diminution, is called bulk modulus.
Remark 1.4.4 The result 0 < ν < 1
2 implies that for all the known materials we
have, coarsely

E
G∼
= . 
2

Remark 1.4.5 The variation of volume of the infinitesimal cube of Fig. 1.4.6 is
  σx
V = εx + εy + εz V = (1 − 2ν)V > 0.
E

Then the infinitesimal parallelepiped of Fig. 1.4.6 has variation of volume. But,
because its edges are parallel to principal directions of strain, it doesn’t have
variation of form. 

Remark 1.4.6 Let us consider an infinitesimal cube of material with the faces parallel
to the coordinate planes and submitted to the alone τ yz , that we suppose positive
(Fig. 1.4.7). Being τxy = τxz = 0, the axis x is a principal direction of stress. The
Mohr’s circle for the set of planes of support x furnishes then the two principal
stress ση = τyz , σζ = −τyz (Fig. 1.4.8). The third principal stress is obviously
σξ = σx = 0. The Mohr’s circle of Fig. 1.4.8 furnishes the principal directions of
stress η, ζ . The third principal direction of stress is ξ = x. Then, since [1.4.1], ξ , η,
ζ is principal triplet of strain. As a consequence

εξ = εx = 0
  
εη = E1 ση − ν σξ + σζ = 1+ν
E τyz > 0
  
εζ = E1 σζ − ν σξ + ση = − 1+ν
E τyz < 0.

Fig. 1.4.7
80 1 The Three-Dimensional Problem

Fig. 1.4.8

Fig. 1.4.9

Then the cube deforms it as shown in Fig. 1.4.9. Moreover, its variation of
volume is
  1+ν  
V = εξ + εη + εζ V = τyz − τyz V = 0.
E
In conclusion the infinitesimal cube of Fig. 1.4.7 has variation of form but doesn’t
have variation of volume. 

1.5 The Elastic Equilibrium Problem

1.5.1 Classical Formulations

We consider a solid deformable body C and we denote with V his volume, i.e. the
portion of the three-dimensional space occupied by C (Fig. 1.5.1). We suppose that
the surface (or boundary) S of V is regular, so in every point of S the tangent plane
exists. Let us refer to an orthogonal set of Cartesian axes O, x, y, z and decompose
S in two parts Sp and Su (Fig. 1.5.1).
1.5 The Elastic Equilibrium Problem 81

Fig. 1.5.1

We suppose that

– the material is homogeneous, isotropic and linearly elastic, so that the (1.4.1) are
true;
– we are in the field of small deformations, so that the (1.1.28) and [1.2.1] are true;
– the body is constrained in the points of Su ;
– in any instant t, the body is submitted to a volumetric load X, Y, Z and, in the
points of Sp , to a distributed superficial load px , py , pz . We suppose that the load
X, Y, Z, px , py , pz and the constraint’s reactions1.5.1 constitutes a system of forces
equivalent to zero.

We consider the elastic equilibrium problem, which is the problem to find

deformed configurations in which the constrained body supports in equilibrium
the applied load. Clearly the possible deformed configurations in equilibrium must
satisfy

– the inside compatibility, that is the (1.1.29),

– the outside compatibility, which are the conditions that the constraints impose to
the points of Su ,
– the inside equilibrium, that is the (1.2.3),
– the boundary equilibrium, that is the (1.2.5) in the points of Sp .

So mathematically the elastic equilibrium problem is a boundary problem, i.e.

the problem to find functions satisfying in V a system of differential equations and
satisfying in the boundary S of V some assigned conditions.
Cauchy undertook as unknown quantities of the elastic equilibrium problem the
functions u, v, w. First of all, we observe that the inside compatibility (1.1.29) is
banally satisfied. As the system of differential equations, from the (1.2.3), taking
into account the (1.4.1), we have

∂ 2Gν   ∂   ∂
2Gεx + εx + εy + εz + Gγxy + (Gγxz ) + X = 0
∂x 1 − 2ν ∂y ∂z

1.5.1 We call constraint’s reactions the superficial forces, applied in the points of Su , that the
constraints act on the body C.
82 1 The Three-Dimensional Problem

∂   ∂ 2Gν   ∂  
Gγyx + 2Gεy + εx + εy + εz + Gγyz + Y = 0
∂x ∂y 1 − 2ν ∂z
∂ ∂   ∂ 2Gν  
(Gγzx ) + Gγzy + 2Gεz + εx + εy + εz + Z = 0
∂x ∂y ∂z 1 − 2ν

from which, taking into account the (1.1.28)

∂ 2u ∂ 2u ∂ 2u 1 ∂ ∂u ∂v ∂w X
+ 2 + 2 + + + + =0
∂x2 ∂y ∂z 1 − 2ν ∂x ∂x ∂y ∂z G
∂ 2v ∂ 2v ∂ 2v 1 ∂ ∂u ∂v ∂w Y
+ 2+ 2 + + + + =0 (1.5.1)
∂x2 ∂y ∂z 1 − 2ν ∂y ∂x ∂y ∂z G
∂ 2w ∂ 2w ∂ 2w 1 ∂ ∂u ∂v ∂w Z
+ 2 + 2 + + + + = 0.
∂x 2 ∂y ∂z 1 − 2ν ∂z ∂x ∂y ∂z G

As the boundary conditions, they are assigned on Su directly in terms of u, v, w

and are assigned on Sp with the (1.2.5), which we explicit in terms of u, v, w by
means the (1.4.1) and (1.1.28).
So the Cauchy’s formulation of the elastic equilibrium problem is the following
boundary problem.

[1.5.1] In the hypothesis of regularity of S and of X, Y, Z, px , py , pz , find the functions

u, v, w that are real, definite and regular in V and such that

– the differential equations (1.5.1) are satisfied in V

– the boundary conditions are satisfied in S = Su ∪ Sp . 

Beltrami undertook as unknown quantities of the elastic equilibrium problem the

functions σx , σy , σz , τxy , τxz , τyz . As the system of differential equations, from the
first of (1.1.29) and from the (1.4.1) we get
   
∂ 2 (1 + ν) σx − ν σx + σy + σz ∂ 2 (1 + ν)σy − ν σx + σy + σz
+ 2
∂y2 E ∂x E
∂ 2(1 + ν)τxy
2
=
∂x∂y E
from which
 
∂ 2 σx ∂ 2 σy
(1 + ν) +
∂y2 ∂x2
 2  (1.5.2)
∂   ∂2   ∂ 2 τxy
−ν σ x + σ y + σ z + σx + σy + σz = 2(1 + ν) .
∂x 2 ∂y2 ∂x∂y
1.5 The Elastic Equilibrium Problem 83

Adding member up member the x partial derivative of the first of (1.2.3), the y
partial derivative of the second of (1.2.3) and the z partial derivative of the third of
(1.2.3) with changed signs, we have

∂ 2 σx ∂ 2 σy ∂ 2 σz ∂X ∂Y ∂Z ∂ 2 τxy
+ − + + − + 2 =0
∂x2 ∂y2 ∂z2 ∂x ∂y ∂z ∂x∂y

and of here, keeping in mind of the (1.5.2)

 
∂ 2 σx ∂ 2 σy ∂ 2 σx ∂ 2 σy ∂ 2 σz
(1 + ν) + + + −
∂y2 ∂x2 ∂x2 ∂y2 ∂z2
 2 
∂   ∂2   (1.5.3)
−ν σx + σy + σz + 2 σx + σy + σz
∂x2 ∂y
∂Z ∂X ∂Y
= (1 + ν) − − .
∂z ∂x ∂y

Analogously, from the second of (1.1.29) and from the (1.4.1) we get

   
∂ 2 (1 + ν)σx − ν σx + σy + σz ∂ 2 (1 + ν)σz − ν σx + σy + σz
+ 2
∂z2 E ∂x E
∂ 2(1 + ν)τxz
2
=
∂x∂z E

from which

∂ 2 σx ∂ 2 σz
(1 + ν) +
∂z2 ∂x2
 2  (1.5.4)
∂   ∂2   ∂ 2 τxz
−ν σ x + σ y + σz + σx + σy + σz = 2(1 + ν) .
∂z2 ∂x2 ∂x∂z

Adding member up member the x partial derivative of the first of (1.2.3), the
y partial derivative of the second of (1.2.3) with changed signs and the z partial
derivative of the third of (1.2.3), we have

∂ 2 σx ∂ 2 σy ∂ 2 σz ∂X ∂Y ∂Z ∂ 2 τxz
− + + − + + 2 =0
∂x2 ∂y2 ∂z2 ∂x ∂y ∂z ∂x∂z
84 1 The Three-Dimensional Problem

and of here, keeping in mind of the (1.5.4)

 
∂ 2 σx ∂ 2 σz ∂ 2 σx ∂ 2 σy ∂ 2 σz
(1 + ν) + + − +
∂z2 ∂x2 ∂x2 ∂y2 ∂z2
 2 
∂   ∂2   (1.5.5)
−ν σx + σ y + σz + σ x + σy + σ z
∂z2 ∂x2
∂Y ∂Z ∂X
= (1 + ν) − − .
∂y ∂z ∂x

Analogously, from the third of (1.1.29) and from the (1.4.1) we get
   
∂ 2 (1 + ν)σy − ν σx + σy + σz ∂ 2 (1 + ν)σz − ν σx + σy + σz
+ 2
∂z2 E ∂y E
∂ 2 2(1 + ν)τyz
=
∂y∂z E

from which
 
∂ 2 σy ∂ 2 σz
(1 + ν) +
∂z2 ∂y2
 2  (1.5.6)
∂   ∂2   ∂ 2 τyz
−ν σx + σ y + σ z + σx + σy + σz = 2(1 + ν) .
∂z2 ∂y2 ∂y∂z

Adding member up member the x partial derivative of the first of (1.2.3) with
changed signs, the y partial derivative of the second of (1.2.3) and the z partial
derivative of the third of (1.2.3), we have

∂ 2 σx ∂ 2 σy ∂ 2 σz ∂X ∂Y ∂Z ∂ 2 τyz
− + + − + + + 2 =0
∂x2 ∂y2 ∂z2 ∂x ∂y ∂z ∂y∂z

and of here, keeping in mind of the (1.5.4)

 
∂ 2 σy ∂ 2 σz ∂ 2 σx ∂ 2 σy ∂ 2 σz
(1 + ν) + + + +
∂z2 ∂y2 ∂x2 ∂y2 ∂z2
 2 
∂   ∂2   (1.5.7)
−ν σx + σ y + σz + σ x + σy + σ z
∂z2 ∂y2
∂X ∂Y ∂Z
= (1 + ν) − − .
∂x ∂y ∂z
1.5 The Elastic Equilibrium Problem 85

Adding member up member the (1.5.3), (1.5.5), and (1.5.7) we have

 
∂2   ∂2   ∂2  
(1 − ν) σx + σy + σz + 2 σx + σy + σz + 2 σx + σy + σz
∂x2 ∂y ∂z
∂X ∂Y ∂Z
= −(1 + ν) + + .
∂x ∂y ∂z
(1.5.8)

From (1.5.3) it follows

∂2   ∂2   ∂2  
σx + σy + σz + 2 σx + σy + σz + 2 σx + σy + σz
∂x 2 ∂y ∂z
∂ σz
2 ∂ σz
2 ∂ σz
2 ∂ 
2 
− (1 + ν) + + − 2 σx + σy + σz
∂x 2 ∂y 2 ∂z 2 ∂z
∂ 2  ∂ 
2  ∂2  
= (1 + ν) σ x + σ y + σ z + σx + σy + σz + σx + σy + σz
∂x2 ∂y2 ∂z2
∂ 2 σz ∂ 2 σz ∂ 2 σz ∂2  
− 2 − − − σx + σy + σz
∂x ∂y 2 ∂z 2 ∂z 2

∂ 2   ∂2  
−ν σx + σy + σz + 2 σx + σy + σz
∂x 2 ∂y
∂ 2  ∂ 
2 
+ 2 σx + σy + σz − 2 σx + σy + σz
∂z ∂z
!
∂ 2 σx ∂ 2 σy ∂ 2 σx ∂ 2 σy ∂ 2 σz
= (1 + ν) + + + −
∂y2 ∂x2 ∂x2 ∂y2 ∂z2
 2 
∂   ∂2  
−ν σ x + σ y + σ z + σx + σy + σ z
∂x2 ∂y2
∂Z ∂X ∂Y
= (1 + ν) − −
∂z ∂x ∂y

and of here, keeping in mind of the (1.5.8)

∂ 2 σz ∂ 2 σz ∂ 2 σz 1 ∂2  
+ + + σx + σy + σz
∂x 2 ∂y 2 ∂z 2 1 + ν ∂z 2
(1.5.9)
ν ∂X ∂Y ∂Z ∂Z
=− + + −2 .
1 − ν ∂x ∂y ∂z ∂z
86 1 The Three-Dimensional Problem

Analogously, from (1.5.5) it follows

∂2   ∂2   ∂2  
σx + σy + σz + 2 σx + σy + σz + 2 σx + σy + σz
∂x 2 ∂y ∂z
 
∂ σy
2 ∂ σy
2 ∂ σy
2 ∂ 
2 
− (1 + ν) + + − 2 σx + σy + σz
∂x 2 ∂y 2 ∂z 2 ∂y
∂2   ∂2   ∂2  
= (1 + ν) σx + σ y + σz + σx + σy + σz + σx + σy + σz
∂x 2 ∂y2 ∂z 2
"
∂ σy
2 ∂ σy
2 ∂ σy
2 ∂ 
2 
− 2 − − − 2 σx + σy + σz
∂x ∂y2 ∂z2 ∂y
∂2   ∂2  
−ν σx + σy + σz + σx + σy + σz
∂x2 ∂y2
∂2   ∂2  
+ 2 σx + σy + σz − 2 σx + σy + σz
∂z ∂y
!
∂ 2 σx ∂ 2 σz ∂ 2 σx ∂ 2 σy ∂ 2 σz
= (1 + ν) + + − +
∂z2 ∂x2 ∂x2 ∂y2 ∂z2
 2 
∂   ∂2  
−ν σx + σy + σz + 2 σx + σy + σz
∂z2 ∂x
∂Y ∂Z ∂X
= (1 + ν) − −
∂y ∂z ∂x

and of here, keeping in mind of the (1.5.8)

∂ 2 σy ∂ 2 σy ∂ 2 σy 1 ∂2  
+ + + σx + σy + σz
∂x 2 ∂y 2 ∂z 2 1 + ν ∂y2
(1.5.10)
ν ∂X ∂Y ∂Z ∂Y
=− + + −2 .
1 − ν ∂x ∂y ∂z ∂y

Analogously, from (1.5.7) it follows

∂2   ∂2   ∂2  
σx + σy + σz + σx + σy + σz + σx + σy + σz
∂x2 ∂y2 ∂z2
∂ 2 σx ∂ 2 σx ∂ 2 σx ∂2  
− (1 + ν) + + − σx + σy + σz
∂x2 ∂y2 ∂z2 ∂x2
1.5 The Elastic Equilibrium Problem 87

∂2   ∂2   ∂2  
= (1 + ν) σx + σ y + σ z + σx + σ y + σz + σx + σy + σz
∂x 2 ∂y 2 ∂z 2

∂ σx
2 ∂ σx
2 ∂ σx
2 ∂ 2  
− 2 − − − 2 σx + σy + σz
∂x ∂y2 ∂z2 ∂x
∂2   ∂2  
−ν σ x + σy + σz + σx + σy + σz
∂x2 ∂y2
∂2   ∂2  
+ 2 σx + σy + σz − 2 σx + σy + σz
∂z ∂x
!
∂ σy
2 ∂ σz
2 ∂ 2 σx ∂ 2 σy ∂ 2 σz
= (1 + ν) + + + +
∂z2 ∂y2 ∂x2 ∂y2 ∂z2
 2 
∂   ∂2  
−ν σx + σy + σz + 2 σx + σy + σz
∂z2 ∂y
∂X ∂Y ∂Z
= (1 + ν) − −
∂x ∂y ∂z

and of here, keeping in mind of the (1.5.8)

∂ 2 σx ∂ 2 σx ∂ 2 σx 1 ∂2  
+ + + σx + σy + σz
∂x 2 ∂y 2 ∂z 2 1 + ν ∂x 2
(1.5.11)
ν ∂X ∂Y ∂Z ∂X
=− + + −2 .
1 − ν ∂x ∂y ∂z ∂x

Moreover, from the fourth of (1.1.29) and from the (1.4.1) we get

 
∂ 2 (1 + ν)σx − ν σx + σy + σz
2
∂y∂z E
(1.5.12)
∂ 2(1 + ν)τxy
2 ∂ 2 2(1 + ν)τyz ∂ 2 2(1 + ν)τxz
= − 2 + .
∂x∂z E ∂x E ∂x∂y E

The z partial derivative of the second of (1.2.3) and the y partial derivative of the
third of (1.2.3) furnish the relationships

∂ 2 τxy ∂ 2 σy ∂ 2 τyz ∂Y
+ + + =0
∂x∂z ∂y∂z ∂z 2 ∂z
∂ 2 τxz ∂ 2 τzy ∂ 2 σz ∂Z
+ + + =0
∂x∂y ∂y 2 ∂y∂z ∂y
88 1 The Three-Dimensional Problem

and these, together with (1.5.12), furnish

∂ 2 σx ν ∂2   ∂ 2 τxy ∂ 2 τxz ∂ 2 τyz

− σx + σy + σz = + −
∂y∂z 1 + ν ∂y∂z ∂x∂z ∂x∂y ∂x2
∂ 2 σy ∂ 2 τyz ∂Y ∂ 2 τzy ∂ 2 σz ∂Z ∂ 2 τyz
=− − − − − − −
∂y∂z ∂z 2 ∂z ∂y 2 ∂y∂z ∂y ∂x2
from which

∂ 2 τyz ∂ 2 τyz ∂ 2 τyz 1 ∂2   ∂Y ∂Z

+ + + σx + σy + σz = − − . (1.5.13)
∂x 2 ∂y 2 ∂z 2 1 + ν ∂y∂z ∂z ∂y
In perfect analogy, from the fifth of (1.1.29) we get

∂ 2 τxz ∂ 2 τxz ∂ 2 τxz 1 ∂2   ∂X ∂Z

+ + + σx + σy + σz = − − (1.5.14)
∂x2 ∂y2 ∂z2 1 + ν ∂x∂z ∂z ∂x
and from the sixth of (1.1.29) we get

∂ 2 τxy ∂ 2 τxy ∂ 2 τxy 1 ∂2   ∂X ∂Y

+ + + σx + σy + σz = − − . (1.5.15)
∂x 2 ∂y 2 ∂z 2 1 + ν ∂x∂y ∂y ∂x
About the boundary conditions, they are assigned on Su in terms of u, v, w and
are assigned on Sp directly in terms of σx , σy , σz , τxy , τxz , τyz . Obviously, to satisfy
the boundary conditions in Su , we must use at first the (1.4.5) and afterwards the
(1.1.28).
So the Beltrami’s1.5.2 formulation of the elastic equilibrium problem is the
following boundary problem
[1.5.2] In the hypothesis of regulatity of S and of X, Y, Z, px , py , pz , find the functions
σx , σy , σz , τxy , τxz , τyz that are real, definite and regular in V and such that

– the differential Equations (1.5.9), (1.5.10), (1.5.11), (1.5.13), (1.5.14), and

(1.5.15) are satisfied in V
– the boundary conditions are satisfied in S = Su ∪ Sp . 

We can prove the following equivalence theorem

[1.5.3] If the triplet u, v, w is a solution of the problem [1.5.1] then the func-
tions σx , σy , σz , τxy , τxz , τyz gotten by u, v, w through the (1.4.1) and (1.1.28) are a
solution of the problem [1.5.2]. Furthermore if the sextuple σx , σy , σz , τxy , τxz , τyz
is a solution of the problem [1.5.2], then the functions u, v, w gotten by
σx , σy , σz , τxy , τxz , τyz through the (1.4.5) and (1.1.28) are a solution of the problem
[1.5.1]. 

1.5.2 Eugenio Beltrami, Cremona 1835 – Roma 1900.

1.5 The Elastic Equilibrium Problem 89

We also can prove the following existence theorem

[1.5.4] The problem [1.5.1] (and then, since the [1.5.3], also the problem [1.5.2])
admits at least a solution. 
Moreover we have the following uniqueness theorem, obtained by Kirchhoff1.5.3
[1.5.5] The problem [1.5.1] (and then, since the [1.5.3], also the problem [1.5.2])
admits only one solution. 
Another general result of very high importance in this matter is the so-called
principle of superposition. It is a simple consequence of the mathematical linearity
of the elastic equilibrium problem. Precisely
[1.5.6] We consider any two real numbers c, c and the following three elastic equi-
librium problems. The first problem is the body C loaded by X̄, Ȳ, Z̄, p̄x , p̄y , p̄z . The
second problem is the body C loaded by X, Y, Z, px , py , pz . The third problem is the
body C loaded by X, Y, Z, px , py , pz such that

X = c̄ X̄ + c X, Y = c̄ Ȳ + c Y, Z = c̄ Z̄ + c Z
px = c̄ p̄x + c px , py = c̄ p̄y + c py , pz = c̄ p̄z + c pz .

Then if ū, v̄, w̄ [resp. u, v, w] [resp. u, v, w] is the unique solution of the first [resp.
second] [resp. third] elastic equilibrium problem, it results

u = c̄ ū + c̄¯ ū,
¯ v = c̄ v̄ + c̄¯ v̄,
¯ w = c̄ w̄ + c̄¯ w̄.
¯  (1.5.16)

Remark 1.5.1 From (1.5.16), taking into account the (1.1.28) and (1.4.5), it follows

σx = c̄ σ̄x + c σ x , εx = c̄ ε̄x + c ε x
σy = c̄ σ̄y + c σ y , εy = c̄ εy + c ε y
σz = c̄ σ̄z + c σ z , εz = c̄ ε̄z + c ε z
τxy = c̄ τ̄xy + c τ xy , γxy = c̄ γ̄xy + c γ xy
τxz = c̄ τ̄xz + c τ xz , γxz = c̄ γ̄xz + c γ xz
τyz = c̄ τ̄yz + c τ yz , γyz = c̄ γ̄yz + c γ yz . 

Remark 1.5.2 We have built the mathematical model of the elastic equilibrium prob-
lem without making use of the law according to which at the instant t the load has
assumed the values X, Y, Z, px , py , pz . Therefore the solution of the problem of the
elastic equilibrium is independent from the path of load. 

1.5.3 Gustave Robert Kirchhoff, Kònigsberg 1824 – Berlin 1887.

90 1 The Three-Dimensional Problem

1.5.2 Variational Formulations

The equations established in Sect. 1.5.1 hold in every point of the domain V, so that
they have local character. Nevertheless the elastic equilibrium problem also admits
formulations that have a global character. It deals with variational principles for
energy, very convenient for the mathematical analysis and the approximate solution.
Such formulations are also called weak formulations.
Let us suppose that the open set V has a Lipschitz-continuous boundary S. We
consider the Sobolev1.5.4 space H2 (V) and denote with W the subset of (H2 (V))3
constituted by the triplet (u, v, w) satisfying the boundary conditions in Su .
We consider the following virtual works problem
 
[1.5.7] Find (u, v, w) ∈ W such that ∀ û, v̂, ŵ ∈ W
 
   
px û + py v̂ + pz ŵ dσ + X û + Y v̂ + Z ŵ dV
S V
  
∂u ∂u ∂v ∂w ∂ û
= 2G +λ + +
V ∂x ∂x ∂y ∂z ∂x
   
∂v ∂u ∂v ∂w ∂ v̂ ∂w ∂u ∂v ∂w ∂ ŵ
+ 2G + λ + + + 2G +λ + +
∂y ∂x ∂y ∂z ∂y ∂z ∂x ∂y ∂z ∂z

∂u ∂v ∂ û ∂ v̂ ∂u ∂w ∂ û ∂ ŵ
+G + + + + +
∂y ∂x ∂y ∂x ∂z ∂x ∂z ∂x

∂v ∂w ∂ v̂ ∂ ŵ
+ + + dV. 
∂z ∂y ∂z ∂y

We also consider the following minimum potential energy problem

 
1.5.8 Find (u, v, w) ∈ W such that ∀ û, v̂, ŵ ∈ W
  
1 ∂u ∂u ∂v ∂w ∂u
2G +λ + +
2 V ∂x ∂x ∂y ∂z ∂x
   
∂v ∂u ∂v ∂w ∂v ∂w ∂u ∂v ∂w ∂w
+ 2G + λ + + + 2G +λ + +
∂y ∂x ∂y ∂z ∂y ∂z ∂x ∂y ∂z ∂z
!"
2 2 2
∂u ∂v ∂u ∂w ∂v ∂w
+G + + + + + dV
∂y ∂x ∂z ∂x ∂z ∂y
 
 
− px u + py v + pz w dσ − (X u + Y v + Z w) dV
S V

1.5.4 Sergei
Sobolev , St. Petersburg 1908 – St. Petersburg 1989. This mathematician gave a very
big contribution to the Functional analysis.
1.5 The Elastic Equilibrium Problem 91
  
1 ∂ û ∂ û ∂ v̂ ∂ ŵ ∂ û
− 2G +λ + +
2 ∂x ∂x ∂y ∂z ∂x
 V   
∂ v̂ ∂ û ∂ v̂ ∂ ŵ ∂ v̂ ∂ ŵ ∂ û ∂ v̂ ∂ ŵ ∂ ŵ
+ 2G + λ + + + 2G +λ + +
∂y ∂x ∂y ∂z ∂y ∂z ∂x ∂y ∂z ∂z
!"
∂ û ∂ v̂ 2 ∂ û ∂ ŵ 2 ∂ v̂ ∂ ŵ 2
+G + + + + + dV
∂y ∂x ∂z ∂x ∂z ∂y
 
   
− px û + py v̂ + pz ŵ dσ − X û + Y v̂ + Z ŵ dV.
S V

Remark 1.5.3 We call potential energy functional the functional Jp that to every
(u, v, w) ∈ W associates the real number
    
1 ∂u ∂u ∂v ∂w ∂u ∂v ∂u ∂v ∂w ∂v
2G +λ + + + 2G + λ + +
2 V ∂x ∂x ∂y ∂z ∂x ∂y ∂x ∂y ∂z ∂y
 
∂w ∂u ∂v ∂w ∂w
+ 2G +λ + +
∂z ∂x ∂y ∂z ∂z
!"
∂u ∂v 2 ∂u ∂w 2 ∂v ∂w 2
+G + + + + + dV
∂y ∂x ∂z ∂x ∂z ∂y
 
 
− px u + py v + pz v + pz w dσ − (Xu + Yv + Zw) dV.
S V

So the problem [1.5.8] can be written

   
(u, v, w) ∈ W: Jp (u, v, w) ≤ Jp û, v̂, ŵ ∀ û, v̂, ŵ ∈ W. 
 2 6
 We denote with U the subset of H (V) constituted by the functions
σx , σy , σz , τxy , τxz , τyz satisfying in V the (1.2.3) and in Sp the (1.2.4). We denote
with a, b, c the values assigned to u, v, w in the points of Su by the constraints of the
body C. We finally consider the following minimum complementary energy problem
 
[1.5.9] Find σx , σy , σz , τxy , τxz , τyz ∈ U such that

  1  σx   
∀ σ̂x , σ̂y , σ̂z , τ̂xy , τ̂xz , τ̂yz ∈ U σx − ν σy + σz
V 2 E
"
σy   σz    τxy 2
τ2 τyz
2
+ σy − ν (σx + σz ) + σz − ν σx + σy + + xz + dV
E E G G G

   
− σx nx + τxy ny + τxz nz a + τyx nx + σy ny + τyz nz b
su
92 1 The Three-Dimensional Problem

   1 σ̂x   
+ τzx nx + τzy ny + σz nz c dS ≤ σ̂x − ν σ̂y + σ̂z
V 2 E
"
σ̂y    σ̂z    τ̂xy 2
τ̂ 2 τ̂yz
2
+ σ̂y − ν σ̂x + σ̂z + σ̂z − ν σ̂x + σ̂y + + xz + dV
E E G G G

   
− σ̂x nx + τ̂xy ny + τ̂xz nz a + τ̂yz nz + σ̂y ny + τ̂yz nz b
s
u  
+ τ̂zx nz + τ̂zy nz + σ̂z nz c dS. 

Remark 1.5.4 We call complementary energy functional the functional Jc that to

every (u, v, w) ∈ U associates the real number

1  σx    σy   σz   
σx − ν σy + σz + σy − ν (σx + σz ) + σz − ν σx + σy
V 2 E E E
" 
τxy
2
τ 2 τyz
2
 
+ + xz + dV − σx nx + τxy ny + τxz nz a
G G G Su
    
+ τyx nx + σy ny + τyz nz b + τzx nx + τzy ny + σz nz c dS. 

The following equivalence theorems subsist

[1.5.10] Every solution of the problem [1.5.7] is solution of the problem [1.5.8] and
vice versa.
 
Proof Let (u, v, w) be a solution of the problem {1.5.7} and û, v̂, ŵ be any element
of W. We put
     
∂ û − u ∂ v̂ − v ∂ ŵ − w
1 = + + ,
∂x ∂y ∂z
∂u ∂v ∂w
2 = + + ,
∂x ∂y ∂z
 #   !     !  
∂ û − u ∂ û − u ∂ v̂ − v ∂ v̂ − v
A= 2G + λ1 + 2G + λ1
V ∂x ∂x ∂y ∂y
  !   ⎡      2
∂ ŵ − w ∂ ŵ − w ∂ û − u ∂ v̂ − v
+ 2G + λ1 +G ⎣ +
∂z ∂z ∂y ∂x
    
 2      2 ⎫
 ⎤
⎬
∂ û − u ∂ ŵ − w ∂ v̂ − v ∂ ŵ − w
+ + + + ⎦ dV,
∂z ∂x ∂z ∂y ⎭
1.5 The Elastic Equilibrium Problem 93

 #   !   !
∂ û − u ∂u ∂ v̂ − v ∂v
B= 2G + λ1 + 2G + λ1
V ∂x ∂x ∂y ∂y
  !     
∂ ŵ − w ∂w ∂ û − u ∂ v̂ − v ∂u ∂v
+ 2G + λ1 +G + +
∂z ∂z ∂y ∂x ∂y ∂x
    
∂ û − u ∂ ŵ − w ∂u ∂w
+ + +
∂z ∂x ∂z ∂x
     !"
∂ v̂ − v ∂ ŵ − w ∂v ∂w
+ + + dV,
∂z ∂y ∂z ∂y
 #       
∂u ∂ û − u ∂v ∂ v̂ − v
C= 2G + λ1 + 2G + λ1
V ∂x ∂x ∂y ∂y
        
∂w ∂ ŵ − w ∂u ∂v ∂ û − u ∂ v̂ − v
+ 2G + λ1 +G + +
∂z ∂z ∂y ∂x ∂y ∂x
    
∂u ∂w ∂ û − u ∂ ŵ − w
+ + +
∂z ∂x ∂z ∂x
     !"
∂v ∂w ∂ v̂ − v ∂ ŵ − w
+ + + dV,
∂z ∂y ∂z ∂y

      
D= px û − u + py v̂ − v + pz ŵ − w dσ
S

      
+ X û − u + Y v̂ − v + Z ŵ − w dV.
V

Employing the potential energy functional Jp , we have

 
Jp û, v̂, ŵ − Jp (u, v, w)
      
= Jp û − u + u, v̂ − v + v, ŵ − w + w − Jp (u, v, w) (1.5.17)
A B C
= + + − D.
2 2 2
Let us observe now that we can easily verify that B = C and that, since (u, v, w)
is solution of the problem [1.5.7], it results C − D = 0. From this and from the
(1.5.17) it follows

  A
Jp û, v̂, ŵ − Jp (u, v, w) =
2
and as a consequence (u, v, w) is solution of the problem [1.5.8].  
As the vice versa, let (u, v, w) be a solution of the problem [1.5.8] and û, v̂, ŵ
be any element of W. We prove, with the tools of the Functional analysis, that
94 1 The Three-Dimensional Problem

the potential energy  functional Jp is Gateaux differentiable in W and it results

∀ (u, v, w) , û, v̂, ŵ ∈ W
  
Jp = (u, v, w) , û, v̂, ŵ
      
∂u ∂ û ∂v ∂ v̂ ∂w ∂ ŵ
= 2G + λ2 + 2G + λ2 + 2G + λ2
V ∂x ∂x ∂y ∂y ∂z ∂z

∂u ∂v ∂ û ∂ v̂ ∂u ∂w ∂ û ∂ ŵ
+G + + + + +
∂y ∂x ∂y ∂x ∂z ∂x ∂z ∂x
 
∂v ∂w ∂ v̂ ∂ ŵ  
+ + + dV − px û + py v̂ + pz ŵ dσ
∂z ∂y ∂z ∂y S

 
− X û + Y v̂ + Z ŵ dV.
V

Then, since the potential energy functional Jp is convex, we get

  
Jp = (u, v, w) , û, v̂, ŵ = 0

and with it the thesis. 

[1.5.11] If the triplet u, v, w is solution of the problem [1.5.7] then the functions
σx , σy , σz , τxy , τxz , τyz gotten by u, v, w through the (1.4.1) and (1.1.28) are solution of
the problem [1.5.9]. Furthermore if the sextuple σx , σy , σz , τxy , τxz , τyz is solution of
the problem [1.5.9], then the functions u, v, w gotten by σx , σy , σz , τxy , τxz , τyz through
the (1.4.5) and (1.1.28) are solution of the problem [1.5.7].
Proof The thesis is obvious. 
[1.5.12] In stronger hypotheses of regularity, every solution of the problem [1.5.1]
is solution of the problem [1.5.7] and vice versa.
Proof Let (u, v, w) solution of problem [1.5.1]. Then, proceeding as in the proof of
the [1.3.1], we verify that (u, v, w) is solution of the problem [1.5.7].

 Vice versa,
 let (u, v, w) solution of problem [1.5.7]. Integrating by parts, we have
∀ û, v̂, ŵ ∈ W

  
∂ ∂u ∂u ∂v ∂w
0= 2G +λ + + û
V ∂x ∂x ∂x ∂y ∂z
   
∂v ∂u ∂v ∂w ∂w ∂u ∂v ∂w
+ 2G + λ + + v̂ + 2G +λ + + ŵ
∂y ∂x ∂y ∂z ∂z ∂x ∂y ∂z
     
∂ ∂u ∂v ∂ ∂u ∂v ∂ ∂u ∂w
+ G + û + G + v̂ + G + û
∂y ∂y ∂x ∂x ∂y ∂x ∂z ∂z ∂x
     
∂ ∂u ∂w ∂ ∂v ∂w ∂ ∂v ∂w
+ G + ŵ + G + v̂ + G + ŵ dV
∂x ∂z ∂x ∂z ∂z ∂y ∂y ∂z ∂y
1.6 Strain Energy 95
   
  ∂ ∂u ∂u ∂v ∂w
− px û + py v̂ + pz ŵ dσ − 2G +λ + + û
S V ∂x ∂x ∂x ∂y ∂z
   
∂ ∂v ∂u ∂v ∂w ∂ ∂w ∂u ∂v ∂w
+ 2G + λ + + v̂ + 2G +λ + + ŵ
∂y ∂y ∂x ∂y ∂z ∂z ∂z ∂x ∂y ∂z
∂ ∂u ∂v ∂ ∂u ∂v ∂ ∂u ∂w
+ Gû + + Gv̂ + + Gû +
∂y ∂y ∂x ∂x ∂y ∂x ∂z ∂z ∂x
∂ ∂u ∂w ∂ ∂v ∂w
+ Gŵ + + Gv̂ +
∂x ∂z ∂x ∂z ∂z ∂y

∂ ∂v ∂w  
+ Gŵ + dV − X û + Y v̂ + Z ŵ dV
∂y ∂z ∂y V
 
∂u ∂u ∂v ∂w ∂v ∂u
= 2G +λ + + nx + G + ny
S ∂x ∂x ∂y ∂z ∂x ∂y
 
∂u ∂w ∂v ∂u
+G + nz − px û + G + nx
∂z ∂x ∂x ∂y

∂v ∂u ∂v ∂w ∂v ∂w
+ 2G + λ + + ny + G + nz − py v̂
∂y ∂x ∂y ∂z ∂z ∂y

∂w ∂u ∂w ∂v
+ G + nx + G + ny
∂x ∂z ∂y ∂z

∂w ∂u ∂v ∂w
+ 2G +λ + + nz − pz ŵ dσ
∂z ∂x ∂y ∂z
 # 
2 2 2
 !
∂ u ∂ u ∂ u 1 ∂ ∂u ∂v ∂w
− G + 2 + 2 + + + + X û
V ∂x2 ∂y ∂z 1 − 2v ∂x ∂x ∂y ∂z
  !
∂ 2v ∂ 2v ∂ 2v 1 ∂ ∂u ∂v ∂w
+ G + 2 + 2 + + + + Y v̂
∂x2 ∂y ∂z 1 − 2v ∂y ∂x ∂y ∂z
  ! "
∂ 2w ∂ 2w ∂ 2w 1 ∂ ∂u ∂v ∂w
+ G + 2 + 2 + + + + Z ŵ dV. 
∂x2 ∂y ∂z 1 − 2v ∂z ∂x ∂y ∂z

Finally, we can prove1.5.5 the following existence and uniqueness theorem

[1.5.13] The problem [1.5.7] (and then, since the [1.5.10] and [1.5.11], also the
problems [1.5.8] and [1.5.9]) admits one and only one solution. 

1.6 Strain Energy

1.6.1 Elements of Thermodynamics

It is common experience that an iron thread heats itself when alternatively and
repeatedly folded up. Therefore to study the effect arising when we apply forces
on a solid body the Thermodynamics appears necessary.

1.5.5 By using the Functional analysis.

96 1 The Three-Dimensional Problem

In Physics we meet various forms of energy, among which the kinetic energy, the
internal energy, the electromagnetic energy, the chemical energy, the nuclear energy.
Limiting us to the model of the continuous one, we denote with t the time, with V(t)
the region of the three-dimensional space occupied at the instant t by the body (solid,
liquid or gas) in evolution, with Σ(t) the boundary of V(t), that is the surface1.6.1 that
separates the system (i.e. the body) in evolution from the environment (i.e. the rest
of the universe). In the space fields of forces (electromagnetic and/or gravitational)
can be present.
A system is said in thermodynamic equilibrium if

– the principle of dissection is valid for every part of the universe, that is of the
system and of the environment,
– chemical reactions in action are not present,
– all the points of the universe have the same temperature.1.6.2

It is evident that under conditions of thermodynamic equilibrium there is no

tendency from the system or from the environment to change state. This can be
individualized assigning a small number of parameters, that we call variables of
state.
We call transformation any evolution that brings the system from a state of ther-
modynamic equilibrium to another state of thermodynamic equilibrium.1.6.3 If the
initial state coincides with that final, we call the evolution closed transformation or
cycle.
Remark 1.6.1 If a system is in thermodynamic equilibrium, exchanging the roles
between system and environment, it is obvious that also the environment is in ther-
modynamic equilibrium. Then in any transformation in the initial and final states
the environment also is in thermodynamic equilibrium. 

Let us suppose that the system pass through a transformation that conducts it
from an initial state of thermodynamic equilibrium (at the time ti ) to a final state
of thermodynamic equilibrium (at the time tf ). During the transformation the envi-
ronment transmits a heat1.6.4 Q to the body, that is to the system. Besides, during

1.6.1 The surface of separation is function of the time t. We call impermeable [resp. adiabatic] [resp.
unergodic] a surface of separation that blocks the flow (that is the exchange) of mass [resp. heat]
[resp. work] between system and environment.
1.6.2 The temperature is preliminarily defined by experimental way introducing a reproducible
scale, that we call thermometer.
1.6.3 If it does not depart from a state of thermodynamic equilibrium or if it does not reach a state
of thermodynamic equilibrium, the evolution is called process.
1.6.4 It does not make sense to speak of heat contained in a body. It can only be talked of heat
(or thermal energy) exchanged among two bodies, that is between environment and system. This
exchange can happen for conduction or for radiation. The heat can be transmitted for conduction
from the environment to the system, or vice versa, when and only when there is a difference of
temperature between the environment and the system. In Statistic thermodynamics it is seen that
the heat transmitted for conduction is an exchange of mechanical energy at microscopic state.
1.6 Strain Energy 97

the transformation the environment and/or the present gravitational fields and/or the
present electromagnetic fields performs on the system a work1.6.5 L.
It is possible to pass from the state of thermodynamic equilibrium at the time ti to
that at the time tf with various transformations. It is experimentally proved that if we
change the transformation then they vary both Q and L. All the known experimental
results assure however that L + Q is always constant. So we have the following
important law1.6.6 of the Physics known as First principle of Thermodynamics.

[1.6.1] A state function1.6.7 exists, that we call energy of the system and we denote
with the symbol E, such that, whatever a transformation of the system is performed
by the state of thermodynamic equilibrium at the time ti to that at the time tf , it
results1.6.8

L + Q = E.  (1.6.1)

We see in experimental Physics that E is the sum of all the forms of energy
possessed by the system. If the system is in motion, we denote with the symbol
Ek (t) the kinetic energy1.6.9 of the system at the instant t

1
Ek (t) = ρ v2 dV
2 V(t)

and with Ek (t) the variation of the kinetic energy during the transformation, that
is Ek (t) = Ek (t) − Ek (ti ).
In absence of other macroscopic forms of energy, we call internal energy of the
system, and we denote with the symbol Ei , the quantity1.6.10

Ei = E − Ek . (1.6.2)

1.6.5 It does not make sense to speak of work contained in a body. It can only be talked of work exe-
cuted by a body on another (as exchange of mechanical energy between environment and system)
and of work executed by a field of forces on a body.
1.6.6 The laws of the Physics are conjectures confirmed by all the known experimental data.
1.6.7 That is a function that only depends by the variables of state.
1.6.8 We denote with E the variation of the energy of the system during the transformation, that
 
is E = E tf − E (ti ) . If  is permeable we must add to first member of (1.6.1) the exchange
of chemical energy associate to the exchange of mass. It is obvious that the [1.6.1] furnishes the
energy of the system for less than a constant. So to E(ti ) we can assign any value.
1.6.9 We denote with ρ [resp.v] the density [resp. velocity].
1.6.10 The internal energy is a microscopic energy of the system. Precisely, in Statistic thermody-
namics we see that internal energy is constituted by all the forms of energy existing at microscopic
level. They are the kinetic energy of translation, of rotation and of vibration of the elementary par-
ticles and the energy of position of the elementary particles, which are reciprocally attracted with
force that depends on the mutual distance.
98 1 The Three-Dimensional Problem

Insofar, since E is a state function, also Ei is a state function.1.6.11 Furthermore,

putting Ei (t) = Ei (t) − Ei (ti ), it results

E = Ek + Ei . (1.6.3)

Remark 1.6.2 The first principle of the Thermodynamics is also said principle of
the energy conservation. In fact the mechanical and thermal energy that the system
receives from the environment remains stored, without losses, in the system still in
form of energy. Precisely the total energy E (ti ) = Ek (ti ) + Ei (ti ) that the system
has at the instant ti , up to the value E(tf ) related to the state of thermodynamic
equilibrium at the instant tf increases. Equivalently it is also used to say that during
a transformation the total energy of a system cannot be produced (meaning created)
nor destroyed. 

Problem 1.6.1 You shall study the motion of a little sphere of rubber, of center C
and diameter d. The sphere is only subject to his weight and it falls on a rigid plane
surface S.

Solution We denote with z(t) [resp. v(t)] the height [resp. speed] of C at the instant
t. We suppose that the motion begins at the instant t1 , so that v (t1 ) = 0. Moreover
we suppose that d is negligible in comparison to z(t1 ). Since v (t1 ) = 0, it results
Ek (t1 ) = 0. At the instant t2 in which the sphere touches the plane S its speed,
and therefore Ek , is maximum. Assuming Ei (t1 ) = 0, since at the instant t2 the
sphere is still undeformed it results Ei (t2 ) = 0. Successively, at the instant t3 the
sphere, after the deformation, arrests his own motion. Consequently Ek (t3 ) = 0.
Still subsequently, the sphere begins an ascending motion and practically repur-
chases the initial form. So at the instant t4 in which it detaches him from the plane S,
Ei (t4 ) = 0. Obviously, also at the instant t5 in which z (t5 ) = z (t1 ), it is Ei (t5 ) = 0.
The system in evolution is constituted by the sphere. We can admit that the sphere
and the environment are in every instant in state of thermodynamic equilibrium, so
the evolutions of the system from the instant ti to the instant ti+1 are,∀i ∈ {1, . . . , 4},
transformations which we denote with Gi . We admit besides that during such trans-
formations they are negligible the thermal exchanges between the sphere and the
environment, so that Q = 0.
We denote with m the mass of the sphere and with g the gravitational acceleration.
Clearly the work accomplished on the sphere from the gravitational field in the
transformations G1 and G4 and from the rigid plane in the transformations G2 and
G3 is

L1 = m g z (t1 ) , L2 = L3 = 0, L4 = −m g z (t1 ) .

1.6.11 We notice that the internal energy cannot be separate in a mechanical part and in a thermal
part.
1.6 Strain Energy 99

Then, because of the first principle of the Thermodynamics, it results

Li = E (ti+1 ) − E (ti ) ∀i ∈ {1, . . . , 4} ,

so that the total energy E = Ek + Ei of the sphere is such that

Ek (t1 ) = Ei (t1 ) = E (t1 ) = 0, Ek (t2 ) = E (t2 ) = E (t1 ) + m g z (t1 ) = m g z (t1 ) ,

Ei (t3 ) = E (t3 ) = E (t2 ) + L2 = Ek (t2 ) = m g z (t1 ) ,
Ek (t4 ) = E (t4 ) = E (t3 ) + L3 = E (t3 ) = m g z (t1 ) ,
Ek (t5 ) = E (t5 ) = E (t4 ) − L4 = m g z (t1 ) = 0 = E (t1 ) .

It is understood that during the motion from the instant t1 to the instant t2 the
work accomplished from the gravitational field increases the kinetic energy of the
sphere until it reaches the maximum value Ek (t2 ). The kinetic energy transforms
itself completely in internal energy at the instant t3 in which the center C stops him.
Subsequently Ei (t3 ) integrally turns into kinetic energy at the instant t4 in which
the sphere detaches him from the rigid plane in the ascending motion. Now this one
gradually decreases up to reach, at the instant t5 , the value Ek (t5 ) = Ek (t1 ) = 0.
In this last transformation the work accomplished from the gravitational field on the
sphere is clearly negative. 

In numerous problems of Engineering the system is called thermal machine and

the environment is considered as composed of three parts, that are said warm source,
cold source and utilizer. All the experimental attempts performed till now to inte-
grally turn with a cycle an heat transferred from the warm source to the thermal
machine into a work transferred from the thermal machine to the utilizer have
failed.1.6.12 This way the following other important law of the Physics, known as
Second principle of Thermodynamics was obtained
[1.6.2] It is impossible to build a thermal machine that converts completely with
a cycle the heat absorbed by a warm source in a work surrendered to the uti-
lizer. The thermal machine must necessarily surrender a flow of heat Q1 to the cold
source.1.6.13 
A transformation is said reversible if, after it, it is possible to bring the system
and the environment under the initial conditions.
Remark 1.6.3 Every real transformation is irreversible. However some real trans-
formations are able, with good approximation, to be schematized as reversble
transformation. 

It subsists the following property, obtained by Clausius1.6.14

1.6.12 Toa such transformation the first principle of the Thermodynamics does not oppose it.
1.6.13 Q constitutes a loss of energy.
1
1.6.14 Rudolf Julius Emmanuel Clausius, Kòslìn 1822 – Bonn 1888.
100 1 The Three-Dimensional Problem

[1.6.3] If a system in thermodynamic equilibrium is submitted to a reversible cycle

r, it results

dQ
= 0.  (1.6.4)
r T

The (1.6.4) assures that, if the system passes by a reversible transformation from
a state of thermodynamic equilibrium at the instant ti to a state of thermodynamic
equilibrium at the instant tf , then the integral one

dQ
ti ,tf T

is independent from the particular reversible transformation. Consequently, a state

function exists, which we call entropy1.6.15 of the system and which we denote with
the symbol S, such that

  dQ
S tf − S (ti ) = .
ti ,tf T

After that, the second principle of Thermodynamics admits the following equiv-
alent formulation
[1.6.4] In an irreversibile transformation it results
 
S tf > S (ti ) ; (1.6.5)

in a reversibile transformation it results

 
S tf = S (ti ) .  (1.6.6)

Remark 1.6.4 From the (1.6.5) and (1.6.6) it follows that the entropy can be pro-
duced but not destroyed. Besides it is a proper greatness to measure the degree of
irreversibility of a transformation. It is important to observe that if the entropy of
the system stays unchanged, then the transformation is reversible. 

Problem 1.6.2 At the instant ti a gas contained in a tank and the environment
are in thermodynamic equilibrium. Their temperature is respectively Tg and Ta
(Fig. 1.6.1). The surface Σ of separation between system and environment, that
is the wall of the tank, is impermeable and unergodic; moreover Tg < Ta . You shall
study the evolution of the gas.

Fig. 1.6.1

1.6.15 Unlike the internal energy, it is not possible to attribute to the entropy a physical meaning.
1.6 Strain Energy 101

Fig. 1.6.2

Solution Since Tg < Ta it begins a transformation in which the environment

transmits to the gas a quantity Q of heat. The gas, in consequence, heats and its
temperature and pressure increase. This way, since Ek = 0 and L = 0, for the first
principle of the Thermodynamics it results Ei = Q. By heating the gas we have
increased its pressure, that is its state of stress. As a consequence the internal enery
is increased. 

Problem 1.6.3 A gas is contained in a metallic cylinder closed by a mobile piston

(Fig. 1.6.2). At the instant ti the gas and the environment are in thermodynamic equi-
librium. The surface Σ of separation between system and environment is the wall
of the cylinder and of the piston. The wall of the cylinder is impermeable, adiabatic
and unergodic; the wall of the piston is mobile, impermeable and adiabatic. You
shall study the transformation that undergoes the gas when the piston is pressed.
Solution When the piston is pressed the environment accomplishes on the system a
work L and the gas is compressed. Since Ek = 0 and Q = 0, for the first principle
of the Thermodynamics Ei = L. Insofar, applying a force on the piston (Fig. 1.6.2),
we have accomplished a work onto the gas. The state of stress of the body, that is
the pressure of the gas, is increased. Simultaneously, its internal energy is increased
of a quantity exactly equal to the work accomplished for compressing the gas. 

1.6.2 Thermodynamics of the Problem of the Elastic Equilibrium

Now, turning the attention to the problem of the elastic equilibrium, we observe
that the system in evolution is constituted by the three-dimensional solid body C.
We can suppose that at the initial instant ti the system and the environment are in
state of thermodynamic equilibrium. In fact the deformation of the body, that is its
thermodynamic transformation, begins in the instant ti in which we begin to apply
the load.
When we admit as true the Navier’s relationships (1.4.1), we declare that the
elongations and the shearing strains depend only from the stress. Clearly this is true
if during the transformation the temperature of all the points of the body is always
equal. As a consequence1.6.16 in the elastic equilibrium problem

Q = 0.

1.6.16 In the Thermoelasticity and in the Plasticity transformations with Q # 0 are studied.
102 1 The Three-Dimensional Problem

Still evidently, the hypothesis that the material remains in the elastic field
involves that the transformation (from the state of thermodynamic equilibrium at
the initial instant ti to the state of thermodynamic equilibrium at the final instant tf )
is reversibile.1.6.17
In the elastic equilibrium problem we suppose that

– the body is unloaded up to the initial instant ti ,

– at the initial instant ti we begin to apply on the body a volumetric and a
superficial load whose  intensity varies up to reach at the final  instant
 tf the

values
 X x,
 y, z, tf , Y x, y, z, tf , Z x, y, z, tf , px x, y, z, t f , py x, y, z, tf ,
pz x, y, z, tf ,
– at the initial instant ti the load has intensity zero,
– after tf the value of the load applied on the body indefinitely stays equal to the
value assumed in tf .

To build the mathematical three-dimensional model of the elastic equilibrium

problem we did not take into account the forces of inertia1.6.18

∂ 2u ∂ 2v ∂ 2w
ρ , ρ , ρ
∂t2 ∂t2 ∂t2
and with this we have supposed that in every instant the accelerations

∂ 2u ∂ 2v ∂ 2w
, ,
∂t2 ∂t2 ∂t2
are zero anywhere. To point out such hypothesis we will say that the load is statically
applied. Furthermore, since the constraints, in such hypothesis in every instant also
the velocities

∂u ∂v ∂w
, ,
∂t ∂t ∂t

are zero anywhere.

Insofar in the elastic equilibrium problem the kinetic energy, that is a macro-
scopic energy, is zero in every instant. So
 
∀t ∈ ti , tf Ek (t) = 0.

Then, since the (1.6.2), the total energy E of the body coincides in every instant
with its internal energy Ei , which is a microscopic energy. So

1.6.17 In the Plasticity and in the Thermoelasticity the transformations are irreversible.
1.6.18 The forces of inertia are of primary importance in the Dynamics of structures.
1.6 Strain Energy 103
 
∀t ∈ ti , tf E (t) = Ei (t) . (1.6.7)

Afterwards we call strain work and we denote with the symbol L the work that
the environment performs on the system, which is the work that the load performs
on the body.
Finally we assume

Ei (ti ) = E (ti ) = 0.

With such premises, the first principle of the Thermodynamics assures that in the
elastic equilibrium problem at the final instant tf the strain work L is exactly equal
to the internal energy of the body. Precisely
   
L = Ei tf = E tf . (1.6.8)

1.6.3 Strain Work

We call load path the law according to which the loads applied from the surrounding
environment on the body vary. It is obvious that every load path individualizes a
transformation of the system.
From the (1.6.8) it follows that the strain work performed for conducting the
body from the state of thermodynamic equilibrium at the initial instant ti to that at
the final instant tf is independent from the load path.
We can therefore choose that at the initial instant ti the volumetric and superficial
loads are anywhere zero and assume, to appraise L, as load path

t − ti  
X (x, y, z, t) = X x, y, z, tf ,
tf − ti
t − ti  
Y (x, y, z, t) = Y x, y, z, tf , (1.6.9)
tf − ti
t − ti  
Z (x, y, z, t) = Z x, y, z, tf ,
tf − ti

t − ti  
px (x, y, z, t) = px x, y, z, tf ,
tf − ti
t − ti  
py (x, y, z, t) = py x, y, z, tf , (1.6.10)
tf − ti
t − ti  
pz (x, y, z, t) = pz x, y, z, tf .
tf − ti
 
Since the transformation is reversible, at every instant t ∈ ti , tf the  body
 is
in thermodynamic equilibrium. As a consequence at every instant t ∈ ti , tf the
components of the displacement u (x, y, z, t) , v (x, y, z, t) , w (x, y, z, t) are the unique
104 1 The Three-Dimensional Problem

solution of the problem of the elastic equilibrium related to the condition of

load X (x, y, z, t) , Y (x, y, z, t) , Z (x, y, z, t) , px (x, y, z, t) , py (x, y, z, t) pz (x, y, z, t) .
Then, taking into account the (1.6.9) and (1.6.10), from the principle of superpo-
sition [1.5.4] we have

t − ti  
u (x, y, z, t) = u x, y, z, tf ,
tf − ti
t − ti  
v (x, y, z, t) = v x, y, z, tf , (1.6.11)
tf − ti
t − ti  
w (x, y, z, t) = w x, y, z, tf .
tf − ti

In the classical Mechanics the work of a force is defined as scalar product

between the force and the displacement of its point of application, in the hypoth-
esis that it changes neither the intensity nor the direction nor the verse of the force
while its point of application moves it. Instead in the elastic equilibrium problem
the load is constituted by external forces whose intensity varies while their points of
application move them. Such intensity varies from an initial value at the time ti to a
final value at the time tf .
Clearly in such cases we must define the work L as the integral from ti to tf of
the elementary works dL performed during the interval [t, t + dt], where t ∈ ti , tf
and dt is positive and small as we like. This way operating, we assume that during
the interval [t, t + dt] the force is constant and in consequence we can appreciate the
infinitesimal work dL with the previous definition of classical Mechanics. In fact, if
dt is very small, the error that we commit supposing the force constant during the
interval [t, t + dt] is negligible.
With such procedure Clapeyron1.6.19 got the following result

[1.6.5] In the elastic equilibrium problem the strain work L is exactly equal to half
of the scalar product between the external forces acting at the final instant tf and the
displacements that their points of application have reached at the final instant tf .

Proof We must appraise

 tf
L= dL.
ti

Remembering that for the hypothesis of small deformations we can perform the
calculations reporting us to the undeformed configuration S = S (ti ) , V = V (ti ) ,
using the (1.6.9), (1.6.10), and (1.6.11) we have

1.6.19 Benoit Paul Emile Clapeyron, Paris 1799 – Paris 1864.

1.6 Strain Energy 105
 
tf '  
px (x, y, z, t) dσ u (x, y, z, t + dt) − u (x, y, z, t)
ti S(t)
  
+ py (x, y, z, t) dσ v (x, y, z, t + dt) − v (x, y, z, t)
  (
+ pz (x, y, z, t) dσ w (x, y, z, t + dt) − w (x, y, z, t)
 tf 
'  
+ X (x, y, z, t) dV u (x, y, z, t + dt) − u (x, y, z, t)
t V(t)
i  
+ Y (x, y, z, t) dV v (x, y, z, t + dt) − v (x, y, z, t)
  (
+ Z (x, y, z, t) dV w (x, y, z, t + dt) − w (x, y, z, t)
 tf   
'  ∂u
= px (x, y, z, t) dσ dt (x, y, z, t)
ti S(t) ∂t
   
  ∂v   ∂w
+ py (x, y, z, t) dσ dt (x, y, z, t) + pz (x, y, z, t) dσ dt (x, y, z, t)
∂t ∂t
 tf   
'  ∂u
+ X (x, y, z, t) dV dt (x, y, z, t)
ti V(t) ∂t
   
  ∂v   ∂w
+ Y (x, y, z, t) dV dt (x, y, z, t) + Z (x, y, z, t) dV dt (x, y, z, t)
∂t ∂t
 tf  
∂u ∂v
= px (x, y, z, t) (x, y, z, t) + py (x, y, z, t) (x, y, z, t)
ti S ∂t ∂t
  tf 
∂w  ∂u
+ pz (x, y, z, t) (x, y, z, t) dσ dt + X (x, y, z, t) (x, y, z, t)
∂t ti V ∂t

∂v ∂w
+ Y (x, y, z, t) (x, y, z, t) + Z (x, y, z, t) (x, y, z, t) dV dt
∂t ∂t
 tf  
t − ti   1  
= px x, y, z, tf u x, y, z, tf
ti S tf − ti tf − ti
t − ti   1  
+ py x, y, z, tf v x, y, z, tf
tf − ti tf − ti

t − ti   1  
+ pz x, y, z, tf w x, y, z, tf dσ dt
tf − ti tf − ti
 tf  
t − ti   1  
+ X x, y, z, tf u x, y, z, tf
ti V tf − ti tf − ti
t − ti   1  
+ Y x, y, z, tf v x, y, z, tf
tf − ti tf − ti

t − ti   1  
+ Z x, y, z, tf w x, y, z, tf dV dt
tf − ti tf − ti
 tf 
t − ti     
=  2 px x, y, z, tf u x, y, z, tf
ti t f − ti S
        (
+ py x, y, z, tf v x, y, z, tf + pz x, y, z, tf w x, y, z, tf dσ dt
106 1 The Three-Dimensional Problem
 
tf t − ti     
+  2 X x, y, z, tf u x, y, z, tf
ti t f − ti V
        (
+ Y x, y, z, tf v x, y, z, tf + Z x, y, z, tf w x, y, z, tf dV dt

        
= px x, y, z, tf u x, y, z, tf + py x, y, z, tf v x, y, z, tf
S

        
+ pz x, y, z, tf w x, y, z, tf dσ + X x, y, z, tf u x, y, z, tf
V

        ( tf t − ti
+ Y x, y, z, tf v x, y, z, tf + Z x, y, z, tf w x, y, z, tf dV  2 dt
ti tf − ti

1         
= px x, y, z, tf u x, y, z, tf + py x, y, z, tf v x, y, z, tf
2 S

    1     
+ pz x, y, z, tf w x, y, z, tf dσ + X x, y, z, tf u x, y, z, tf
2 V
       
+ Y x, y, z, tf v x, y, z, tf + Z x, y, z, tf w x, y, z, tf dV.

Insofar it results
 tf 
1         
L= dL = px x, y, z, tf u x, y, z, tf + py x, y, z, tf v x, y, z, tf
ti 2 S

    1     
+pz x, y, z, tf w x, y, z, tf dσ + X x, y, z, tf u x, y, z, tf
2 V
       
+Y x, y, z, tf v x, y, z, tf + Z x, y, z, tf w x, y, z, tf dV. 
(1.6.12)

About the strain work we also have

[1.6.6] For the strain work (and as consequence for the energy) the principle of
superposition [1.5.4] is not valid.
Proof We consider a solid body homogeneous, isotropic and linearly elastic. We
submit such body, in the field of small deformations, to three different conditions
of load. This way we individualize three problems of the elastic equilibrium. In the
first [resp. second] problem, of which we denote with (u, v, w) [resp. (ũ, ṽ, w̃)] the
unique solution at the instant tf , the load is statically applied and its value varies
from zero (at the instant ti ) to px [resp. p̃x ] (at the instant tf ). In the third problem the
load is statically applied and its value varies from zero (at the instant ti ) to px + p̃x (at
the instant tf ). Then, since the principle of superposition [1.5.4], the unique solution
of the third problem at the instant tf is (u + ũ, v + ṽ, w + w̃). Because of [1.6.5], the
strain work L1 [resp. L2 ] [resp. L3 ] of the first [resp. second] [resp. third] problem is
 
1 1
L1 = px u dσ , L2 = p̃x dσ ,
2 S 2 S
1.6 Strain Energy 107
 
1 px ũ + p̃x u
L3 = (px + p̃x ) (u + ũ) dσ = L1 + L2 + dσ .
2 S S 2

Insofar L3 = L1 + L2 and this proves the thesis. 

1.6.4 The Elastic Potential

We consider a solid body homogeneous, isotropic and linearly elastic. We submit

such body, in the field of small deformations, to a load statically applied and we
make to increase its value from zero (at the instant ti ) to X, Y, Z, px , py , pz (at the
instant tf ). For this elastic equilibrium problem we denote with u, v, w the unique
solution at the instant tf .
We obtain now an operational expression of the strain work L in function of the
stresses and strains arised in the body at the final instant tf . Because of the (1.6.8),
such relationship will  also furnish the value of the total energy E tf and of the
internal energy Ei tf .
We denote with σx , σy , σz , τxy , τxz , τyz , εx , εy , εz , γxy , γxz , γyz the components of
stress and strain at the final instant tf . By using the formulas of Gauss and the (1.2.5),
(1.2.3), and (1.1.28), from the (1.6.8) and [1.6.5] it follows

  1  
Ei tf = L = px u + py v + pz w dσ
2 S
 
1 1  
+ (Xu + Yv + Zw) dV = σx nx + τxy ny + τxz nz u
2 2 S
 V    
+ τyx nx + σy ny + τyz nz v + τzx nx + τzy ny + σz nz w dσ

1
+ (Xu + Yv + Zw) dV
2
 V
1 ∂   ∂  
= σx u + τxy v + τxz w + τyx u + σy v + τyz w
2 V ∂x ∂y
∂  
+ τzx u + τzy v + σz w + (Xu + Yv + Zw) dV
∂z
 
1 ∂σx ∂τxy ∂τxz ∂τyx ∂τy ∂τyz
= + + +X u+ + + +Y v
2 V ∂x ∂y ∂z ∂x ∂y ∂z
∂τzx ∂τzy ∂σz ∂u ∂v ∂w
+ + + + Z w + σx + σy + σz
∂x ∂y ∂z ∂x ∂y ∂z

∂u ∂v ∂u ∂w ∂v ∂w
+τxy + + τxz + + τyz + dV
∂y ∂x ∂z ∂x ∂z ∂y

1 
= σx εx + σy εy + σz εz + τxy γxy + τxz γxz + τyz γyz dV.
V 2
(1.6.13)
108 1 The Three-Dimensional Problem

In the Theory of elasticity we call elastic potential and we denote with the symbol
ϕ the function

1 
ϕ= σx εx + σy εy + σz εz + τxy γxy + τxz γxz + τyz γyz . (1.6.14)
2
Such way, from the (1.6.13) and (1.6.14) it follows

 
Ei tf = ϕ dV. (1.6.15)
V

Remark 1.6.5 From the (1.6.15) it obviously follows that the elastic potential ϕ
has the physical meaning of density of internal energy. In other words, the internal
energy stored in an infinitesimal element dV it is equal to dEi = ϕ dV. 

We have

[1.6.7] The components of stress [resp. strain] are the partial derivatives of the
elastic potential with respect to the components of strain [resp. stress].

Proof From the (1.6.14) and (1.4.1) it follows

1  σx    σy  
ϕ= σx − ν σy + σz + σy − ν (σz + σx )
2 E E
σz    τxy τxz τyz 
+ σz − ν σx + σy + τxy + τxz + τyz
E G G G
1  2   1  2 
= σx + σy + σz − 2ν σx σy + σx σz + σy σz +
2 2
τxy + τxz
2
+ τyz
2
2E 2G
(1.6.16)

from which

∂ϕ 2σx − 2νσy − 2νσz

= = εx
∂σx 2E
∂ϕ 2σy − 2νσx − 2νσz
= = εy
∂σy 2E
∂ϕ 2σz − 2vσx − 2vσy
= = εz
∂σz 2E
(1.6.17)
∂ϕ 2τxy
= = γxy
∂τxy 2G
∂ϕ 2τxz
= = γxz
∂τxz 2G
∂ϕ 2τyz
= = γyz .
∂τyz 2G
1.6 Strain Energy 109

Likewise, from the (1.6.14) and (1.4.5) it follows

1 '     
ϕ= 2Gεx + λ εx + εy + εz εx + 2Gεy + λ εx + εy + εz εy
2   (
+ 2Gεz + λ εx + εy + εz εz + Gγxy γxy +Gγxz γxz + Gγyz γyz
 
λ  2 G 2 
= G εx + εy + εz +
2 2 2
εx + εy + εz + γxy + γxz2 + γyz2
2G 2 (1.6.18)
 
λ 2λ  
=G 1+ εx + εy + εz +
2 2 2
εx εy + εx εz + εy εz
2G 2G + λ
G 2 
+ γxy + γxz2 + γyz2
2
from which
 
∂ϕ λ 2λ    
=G 1+ 2εx + εy + εz = 2Gεx + λ εx + εy + εz = σx
∂εx 2G 2G + λ
 
∂ϕ λ 2λ  
=G 1+ 2εy + (εx + εz ) = 2Gεy + λ εx + εy + εz = σy
∂εy 2G 2G + λ
 
∂ϕ λ 2λ    
=G 1+ 2εz + εx + εy = 2 Gεz + λ εx + εy + εz = σz
∂εz 2G 2G + λ

∂ϕ 2Gγxy
= = τxy
∂γxy 2
∂ϕ 2Gγxz
= = τxz (1.6.19)
∂γxz 2
∂ϕ 2Gγyz
= τyz . 
∂γyz 2

[1.6.8] The values of the elastic potential do not depend on the choice of the frame
of reference.
Proof We denote with J1 [resp. J2 ] the linear [resp. quadratic] invariant of strain, so
that

J1 = εx + εy + εz
1 2 1 1 (1.6.20)
J2 = εx εy + εx εz + εy εz − γxy − γxz2 − γyz2 .
4 4 4
From (1.6.18) and (1.6.20) it follows

λ  2
ϕ =G 1+ εx + εy + εz − 2εx εy − 2εx εz − 2εy εz
2G

2λ   G 2 
+ εx εy + εx εz + εy εz + γxy + γxz2 + γyz2
2G + λ 2
110 1 The Three-Dimensional Problem

λ  2 λ 2λ
=G 1+ εx + εy + εz + G 1 + −2 +
2G 2G 2G + λ
  G 2 
· εx εy + εx εz + εy εz + γxy + γxz2 + γyz2
2
λ   G 2 
=G 1+ J12 − 2G εx εy + εx εz + εy εz + γxy + γxz2 + γyz2
2G 2
λ
=G 1+ J 2 − 2GJ2
2G 1

with which the thesis is achieved. 

[1.6.9] The elastic potential is a real quadratic form, function of the real variables
εx , εy , εz , γxy , γxz , γyz and positive defined.
Proof From the (1.6.14) and (1.4.5) it follows

1 '     
ϕ= 2Gεx + λ εx + εy + εz εx + 2Gεy + λ εx + εy + εz εy
2   (
+ 2Gεz + λ εx + εy + εz εz + Gγxy γxy +Gγxz γxz + Gγyz γyz
 
λ  2 G 2 
= G εx + εy + εz +
2 2 2
εx + εy + εz + γxy + γxz2 + γyz2
2G 2

with which the thesis is achieved. 

Remark 1.6.6 From the (1.4.1) and (1.6.17) we get

∂εx ∂ 2ϕ ∂εy
c12 = = = = c21
∂σy ∂σx ∂σy ∂σx
and, proceeding in similar way

∀i, j ∈ {1, . . . , 6} cij = cji .

This way we have proved that of the 36 elastic constants c11 , . . . , c66 of every
homogeneous, isotropic and linearly elastic material, to the more 21 are distinct. 

1.6.5 Work Theorems

We have already examined in the Sect. 1.5 the two energetic functionals called
potential energy and complementary energy. We have seen that the unique solu-
tion (u, v, w) of the elastic equilibrium problem is a point of absolute minimum  of
such functionals in the space W of the kinematically admissible triplet û, v̂, ŵ .
For the elastic equilibrium problem they hold other energetic theorems that we
now examine.
1.6 Strain Energy 111

We consider a solid body homogeneous, isotropic and linearly elastic, in the

field of small deformations. We consider the elastic equilibrium problem A [resp.
Ã], related to the final instant tf , in which such body is submitted to a load
X, Y, Z, px , py , pz [resp. X̃, Ỹ, Z̃, p̃x , p̃y , p̃z ]. We denote with (u, v, w) [resp. (ũ, ṽ, w̃)]
the unique solution of this problem. We denote with σx , σy , σz , τxy , τxz , τyz [resp.
σ̃x , σ̃y , σ̃z , τ̃xy , τ̃xz , τ̃yz ] the related state of stress and with εx , εy , εz , γxy , γxz , γyz [resp.
ε̃x , ε̃y , ε̃z , γ̃xy , γ̃xz , γ̃yz ] the related state of strain. We call mutual work1.6.20 among
the systems A and Ã, and we denote with the symbol LAÃ , the work done by the
external forces of the problem A when we engrave to their points of application the
displacements that such points have in the problem Ã.
As the mutual work, it subsists the following result, obtained by Betti1.6.21
[1.6.0] The mutual works are equal.
Proof By definition it result

   
LAÃ = (px dσ ) (ũ) + py dσ (ṽ) + (pz dσ ) (w̃)

S
(1.6.21)
 
+ (XdV) (ũ) + (YdV) (ṽ) + (ZdV) (w̃)
V

   
LÃA = (p̃x dσ ) (u) + p̃y dσ (v) + (p̃z dσ ) (w)
S
 (1.6.22)
      
+ X̃dV (u) + ỸdV (v) + Z̃dV (w) .
V

We apply the virtual work principle to problems A and Ã. Assuming the problem
à as system of displacement and the problem A as system of forces, we have
 
 
px ũ + py ṽ + pz w̃ dσ + (X ũ + Y ṽ + Z w̃) dV
S
 V
(1.6.23)
 
= σx ε̃x + σy ε̃y + σz ε̃z + τxy γ̃xy + τxz γ̃xz + τyz γ̃yz dV;
V

assuming the problem à as system of forces and the problem A as system of

displacement, we have
 
   
p̃x u + p̃y v + p̃z w dσ + X̃u + Ỹv + Z̃w dV
S
 V
(1.6.24)
 
= σ̃x εx + σ̃y εy + σ̃z εz + τ̃xy γxy + τ̃xz γxz + τ̃yz γyz dV.
V

1.6.20 The mutual work is also called dragging work.

1.6.21 Enrico Betti, Pistoia 1823 – Soiana 1892.
112 1 The Three-Dimensional Problem

From the (1.6.21) and (1.6.23) it follows


 
LAÃ = σx ε̃x + σy ε̃y + σz ε̃z + τxy γ̃xy + τxz γ̃xz + τyz γ̃yz dV; (1.6.25)
V

and from the (1.6.22) and (1.6.24) it follows


 
LÃA = σ̃x εx + σ̃y εy + σ̃z εz + τ̃xy γxy + τ̃xz γxz + τ̃yz γyz dV. (1.6.26)
V

From the (1.4.5) it follows

σx ε̃x + σy ε̃y + σz ε̃z + τxy γ̃xy + τxz γ̃xz + τyz γ̃yz

 
= (2Gεx + λ) ε̃x + 2Gεy + λ ε̃y + (2Gεz + λ) ε̃z
+ Gγxy γ̃xy + Gγxz γ̃xz + Gγyz γ̃yz
= 2Gεx ε̃x + 2Gεy ε̃y + 2Gεz ε̃z + λ ˜ + Gγxy γ̃xy + Gγxz γ̃xz + Gγyz γ̃yz
     
= 2Gε̃x + λ ˜ εx + 2Gε̃y + λ ˜ εy + 2Gε̃z + λ ˜ εz

+ Gγ̃xy γxy + Gγ̃xz γxz + Gγ̃yz γyz

= σ̃x εx + σ̃y εy + σ̃z εz + τ̃xy γxy + τ̃xz γxz + τ̃yz γyz

so that from here and from the (1.6.25) and (1.6.26) the thesis follows. 

Moreover we have the following result, obtained by Castigliano,1.6.22 that allows

to calculate a displacement of every point of S as partial derivative of the strain
work.

[1.6.1] The partial derivative of the strain work with respect to an external force
F concentrated in a point of S is equal to the component s̃, according the force
direction, of the displacement s of such point

∂L
= s̃.
∂F

Proof We consider the general case of a body rigidly constrained and submitted, at
the final instant tf , to n concentrated superficial forces F1 , . . . , Fn (Fig. 1.6.3).
 We
denote with L the strain work performed by such forces in the interval ti , tf . We
denote with P1 , . . . , Pn the points of S in which the forces F1 , . . . , Fn are applied

1.6.22 Alberto Carlo Castigliano, Asti 1847 – Milano 1884.

1.6 Strain Energy 113

Fig. 1.6.3 Fi
Pi
si
P1′ Pi′
s1
F1 sn
Pn′ Pn
P1
Fn

and with s1 , . . . , sn their displacements at the final instant tf (Fig. 1.6.3). We denote
with s̃1 , . . . , s̃n the components of s1 , . . . , sn according the directions of F1 , . . . , Fn
respectively. Because of the [1.6.5], we have

1 1
L= F1 s̃1 + . . . + Fn s̃n
2 2
so that

1 1
L= F1 s̃1 (F1 , . . . , Fn ) + . . . + Fn s̃n (F1 , . . . , Fn ) . (1.6.27)
2 2

Now we fix i ∈ {1, . . . , n} and increase only the Fi of a dFi small as we like.
From the (1.6.27) we get that the strain work L∗ performed by the external forces
F1 , . . . , Fi−1 , Fi + dFi , Fi+1 , . . . , Fn is given by

L∗ = L (F1 , . . . , Fi−1 , Fi + dFi , Fi+1 , . . . , Fn )

that approximately is equal to

∂L ∂L
L∗ = L (F1 , . . . , Fi−1 , Fi , Fi+1 , . . . , Fn ) + dFi = L + dFi . (1.6.28)
∂Fi ∂Fi

Since L∗ can be calculated with any load path, we choose to apply the load in
two phases. In the first phase we apply on the body the F1 , . . . , Fn . In the second
phase we maintain constant the F1 , . . . , Fn and make grow only the dFi from 0 to
the final value dFi (Fig. 1.6.4). The strain work L∗ performed is therefore equal
to the strain work L performed in the first phase increased of the strain work per-
formed in the second phase. This last result is therefore equal to dL. We denote with
ds1 , . . . , dsn the displacements of the points P1 , . . . , Pn in the problem of Fig. 1.6.4.
114 1 The Three-Dimensional Problem

Fig. 1.6.4

In the second phase the F1 , . . . , Fn have constant intensity so that they perform
a work


n
Fj ds̃j
j=1

and the dFi perform a work that, since the [1.6.5], is equal to

1
dFi ds̃i .
2

Insofar


n
1  n
dL = Fj ds̃j + dFi ds̃i ∼
= Fj ds̃j . (1.6.29)
2
j=1 j=1

Now we denote with A the problem of Fig. 1.6.3 and with B the problem of
Fig. 1.6.4. Because of the [1.6.10] it results


n
Fj ds̃j = LAB = LBA = dFi s̃i . (1.6.30)
j=1

From the (1.6.29) and (1.6.30) we get

dL = dFi s̃i

from which, taking into account the (1.6.28)

∂L
= s̃i ∀i ∈ {1, . . . , n} .
∂Fi
The thesis follows. 
1.7 Strength Criterions 115

1.7 Strength Criterions

1.7.1 Structural Safety

The considerations that we do in this book have for principal purpose to allow us
to establish if a body submitted to a load is in structural safety. More precisely we
consider a body submitted to a load such that

– the body geometry and the material with which it is done are known,
– the material is homogeneous and isotropic,
– we consider the value of the load at final instant tf , that we call design load,
– we suppose that the intensity of the load stays constant for t > tf ,
– the design load is such that the material stays in the linearly elastic field and the
body stays in the field of small displacements.

We have seen in the Sect. 1.5 that in this situation the body can absorb the load
in one and only one way, finding a deformed configuration always equal from the
instant tf onwards. We will say that the body is in structural safety if it can assume
and preserve such configuration deformed without danger of collapse.
Obviously the expression structural safety has a meaning that varies dependently
from the employment of the body. In fact, the verification of the safety of a body
can consist of verifying that there is no danger of breakup or that the displacements
of certain points of the body are not only small but they do not overcome any preset
values.
Let’s us deal with the problem to appraise the structural safety as ability of resis-
tance of a body in the respects of the collapse. To the actual state of the knowledge,
the Theoretical physics does not offer any indications that are of some utilities
practice to this intention. Therefore to appraise the structural safety of a body it
is necessary to make a macroscopic analysis of the problem. Such analysis can be
done with the Method of the admissible stress.1.7.1

1.7.1 Up to 1980 approximately, to appraise the structural safety in the respects of the collapse the
Method of the admissible stress has only been employed. Since then, the Method of the ultimate
limit states has been used. The second of such methods allows a more precise calculation. With it
the threshold not to be overcome is the collapse. However the first of such methods allows a much
more prudential calculation. With it the threshold not to be overcome is the yelding (i.e. practically
the point B of Fig. 1.4.4). As a consequence, if because of exceptional events the threshold of crisis
is reached

– if the calculation has been made with the second of such methods the body collapses,
– if the calculation has been made with the first of such methods the body still has the ability
of resistance of the whole plastic field, so that it absorbs in safety the exceptional load
suffering plastic deformations.

Therefore, if the structure is desired to be operational for very long time, the Method of the
admissible stress still today constitutes a profit tool for the structural design.
116 1 The Three-Dimensional Problem

The Method of the admissible stress declares the body in structural safety if every
point of the body is in structural safety. We say that a point of the body is in structural
safety if in such point a strength criterion is satisfied.
There are several strength criterions, all based on coarse conjectures. However
such conjectures were always confirmed in the Laboratories material tests and in
the constructive practice.
In the structural design for caution the calculation is always performed amplify-
ing the design load with a safety factor s, which is a positive real number greater
than 1, whose value essentially depends on the importance of the structure to realize.
The safety factor is employed in both the methods exposed in footnote 1.7.1.
In every strength criterion it is called admissible stress the quantity

σ0
σa = (1.7.1)
s

where σ 0 denotes the yelding stress furnished by the tensile breaking test, already
discussed in the Sect. 1.4. When we calculate the stresses of the body, the reliability
of the gotten values does not only depend on the accuracy of the method adopted
to resolve the elastic equilibrium problem but also on the accuracy with which the
load applied on the structure and the mechanical characteristics of the material are
esteemed. Particularly the value of σ 0 is chosen with probabilistic considerations
after having effected in the Laboratory material tests the tensile breaking test on
several specimens all equal ones and all taken from the same lot.

1.7.2 The Maximum Shearing Stress Criterion

The maximum shearing stress criterion proposed by Tresca1.7.2 is a proper strength

criterion for metallic materials. We consider every point P of the body. We consider
the τ n acting on the plane α n passing through P. We call maximum shearing stress
and denote with the symbol τmax the greatest of the τ n which are obtained by varying
α n in all the possible ways. We have seen in the Sect.1.2.11 that τmax is equal to the
radius of the maximum principal circle of Mohr. Insofar, denoting with σξ , ση, σζ
the three principal stresses in P, it results

1 ' ( ' (
τmax = max σξ , ση , σζ − min σξ , ση , σζ .
2
Tresca assumed as parameter of crisis the τmax amplified by the safety factor s
and as threshold of crisis τ0 the value of τmax furnished by the tensile breaking test
at the yelding. Clearly in the tensile breaking test at the yelding in the point P the

1.7.2 Henry Edouard Tresca, Dunkerque 1814 – Paris 1885.

1.7 Strength Criterions 117

Fig. 1.7.1

principal stresses are 0, 0, σ0 , so that the three principal circle of Mohr are those of
Fig. 1.7.1. As a consequence

σ0
τ0 = .
2

In conclusion Tresca declares the material in structural safety in the point P if

s τmax ≤ τ0
that is if
 ' ( ' (
s max σξ , ση , σζ − min σξ , ση , σζ ≤ σ0 .

1.7.3 The Octahedral Shearing Stress Criterion

The octahedral shearing stress criterion is another proper strength criterion for
metallic materials. We consider every point P of the body. We choose as frame
of reference the principal triplet of stress P, ξ , η, ζ (Fig. 1.7.2). We consider the line
noct passing through P whose direction cosines are
1
noctξ = noctη = noctζ = √ .
3

Fig. 1.7.2
118 1 The Three-Dimensional Problem

We call octahedral shearing stress and denote with τoct the τ acting on the plane
αoct having normal noct and passing through P. We denote with toct the stress vector
acting on αoct , so that
2
toct = σoct
2
+ τoct
2
. (1.7.2)

Since the frame of reference is principal of stress, taking into account the (1.2.4)
and (1.2.8), from the (1.7.2) we have
 2
τoct
2
= toct
2
− σoct
2
= toctξ
2
+ toctη
2
+ toctζ
2
− σoct
2
= σξ noctξ
 2  2  2
+ ση noctη + σζ noctζ − σξ n2octξ + ση n2octη + σζ n2octζ
σξ2 + ση2 + σζ2 σξ + ση + σζ 2
= −
3 3
2 2 
= σξ + ση2 + σζ2 − σξ ση − σξ σζ − ση σζ
9
so that
√  1
2 2 2
τoct = σξ + ση2 + σζ2 − σξ ση − σξ σζ − ση σζ . (1.7.3)
3
In this criterion we assume as parameter of crisis the τoct amplified by the safety
factor s and as threshold of crisis τ0 the value of τoct furnished by the tensile break-
ing test at the yelding. Clearly in the tensile breaking test at the yelding in the point
P the principal stresses are 0, 0, σ0 , so that from the (1.7.3) we have
√
2
τ0 = σ0 .
3
In conclusion in this criterion we declare the material in structural safety in the
point P if

s τoct ≤ τ0

that is if
 
s σξ2 + ση2 + σζ2 − σξ ση − σξ σζ − ση σζ ≤ σ0 . (1.7.4)

1.7.4 The Energetic Criterion

This criterion, proposed by von Mises,1.7.3 develops energetic considerations with

which declares the material in structural safety in the point P if a condition is sat-
isfied that results exactly equal to the (1.7.4). We consider any point P of the body.
We choose as frame of reference the principal triplet of stress P, ξ , η, ζ . We consider

1.7.3 Richard von Mises, Leopoli 1883 – Boston 1953.

1.7 Strength Criterions 119

the invariant of stress T = σξ + ση + σζ and the elastic potential ϕ. From (1.6.16)

we have
1  2  
ϕ= σξ + ση2 + σζ2 − 2ν σξ ση + σξ σζ + ση + σζ
2E #
1 T T 2 T T 2 T T 2
= σξ − + + ση − + + σζ − +
2E 3 3 3 3 3 3

T T T T T T T T
− 2ν σξ − + ση − + + σξ − + σζ − +
3 3 3 3 3 3 3 3

T T T T
+ ση − + σζ − +
3 3 3 3
#
2 2
1 T T T T
= σξ − +2 σξ −
2E 3 3 3 3
T 2 T 2 T T T 2 T 2 T T
+ ση − + +2 ση − + σζ − + +2 σζ −
3 3 3 3 3 3 3 3
T T T 3 T T T T
− 2ν σξ − ση − + + σξ − + ση −
3 3 3 3 3 3 3
T T T 2 T T T T
+ σξ − σζ − + + σξ − + σζ −
3 3 3 3 3 3 3
!"
T T T 2 T T T T
+ ση − σζ − + + ση − + σζ −
3 3 3 3 3 3 3
#
1 T 2 T 2 T 2 T 2 T 2 T 2
= σξ − + + ση − + + σζ − +
2E 3 3 3 3 3 3

T T T 2 T T T 2
− 2ν σξ − ση − + + σξ − σζ − + +
3 3 3 3 3 3
!"
T T T 2
+ ση − σζ − +
3 3 3
#
1 T 2 T 2 T 2
= σξ − + ση − + σζ −
2E 3 3 3
 
T T T T T T
− 2ν σξ − ση − + σξ − σζ − + ση − σζ −
3 3 3 3 3 3
1 − 2ν T 2
+
2E 3
1  2 T 2 T T 2 T
= σξ + − 2σξ + ση2 + − 2ση
2E 3 3 3 3
T 2 T T 2 T  T 2
+ σζ2 + − 2σζ − 2ν σξ ση + − σξ + ση + σξ σζ +
3 3 3 3 3
!"
T  T 2 T  1 − 2ν T 2
− σξ + σζ + ση σζ + − ση + σζ +
3 3 3 2E 3
120 1 The Three-Dimensional Problem

#  "
1 T2 T2 1 − 2ν T 2
= σξ2 + ση2 + σζ2 − − 2ν
σξ ση + σξ σζ + ση σζ − +
2E 3 3 2E 3
1+ν  
(1.7.5)
= σξ2 + ση2 + σζ2 − σξ ση − σξ σζ − ση σζ
3E
1 − 2ν 2
+ T .
6E
With the position

1+ν  2 
ϕf = σξ + ση2 + σζ2 − σξ ση − σξ σζ − ση σζ (1.7.6)
3E
from (1.7.5) and (1.7.6) we get

1 − 2ν 2
ϕ = ϕf + T . (1.7.7)
6E
von Mises assumed as parameter of crisis the aliquot part ϕf of the density of energy
stored in a neighborhood of P, amplified by the safety factor s. Moreover he assumed
as threshold of crisis ϕf 0 the value of ϕf furnished by the tensile breaking test at the
yelding. Clearly in the tensile breaking test at the yelding in the point P the principal
stresses are 0, 0, σ0 , so that from the (1.7.6) we have

1+ν 2
ϕf 0 = σ .
3E 0
In conclusion in this criterion we declare the material in structural safety in the
point P if

1 + ν  2  2  2        
sσξ + sση + sσζ − sσξ sση − sσξ sσζ
3E
    1 + ν 2
− sση sσζ ≤ σ
3E 0
that is if
 
s2 σξ2 + ση2 + σζ2 − σξ ση − σξ σζ − ση σζ ≤ σ02

that is if
 1
2
s σξ2 + ση2 + σζ2 − σξ ση − σξ σζ − ση σζ ≤ σ0 . (1.7.8)

Remark 1.7.1 We notice that the elastic potential ϕ cannot be assumed as parameter
of crisis. In fact, reasoning by absurd, let us assume ϕ as parameter of crisis and

σ02
ϕ0 =
2E
1.7 Strength Criterions 121

as threshold of crisis. We suppose that in P the state of stress is hydrostatic, that is

σx = c, σy = c, σz = c, τxy = 0, τxz = 0, τyz = 0,

where c is a negative real number. Evidently x, y, z are three principal directions of

stress, so that the (1.7.6) furnishes
ϕf = 0

from which, taking into account the (1.7.7), it follows

1 − 2ν
ϕ= (3c)2 .
6E

To the growth of c in absolute value, therefore, it is ϕf = 0 and ϕ grows. When

ϕ reaches the value ϕ0 the material collapse in P. This is absurd. In fact it has been
experimentally shown in the Laboratory material tests that this event cannot verify.
Precisely pieces of material have been disposed in steel boxes containing a liquid.
By reaching elevated values of the pressure of the liquid, the pieces of material have
never reached the breakup. 

Remark 1.7.2 The parameter of crisis ϕf of von Mises is also called disfiguration
aliquot of the elastic potential and ϕ − ϕf is sometimes denominated volume aliquot
of the elastic potential and denoted with the symbol ϕv .
Evidently it is possible to think that the real state of stress σξ , ση , σζ is sum of a
hydrostatics aliquot σ ξ , σ η , σ ζ and of a disfiguration aliquot σ ξ , σ η , σ ζ

T T
σξ = + σξ − = σ ξ + σ ξ
3 3
T T
ση = + ση − = σ η + σ η
3 3
T T
σζ = + σζ − = σ ζ + σ ζ .
3 3

It is easy to verify that to hydrostatics aliquot a variation of volume is associated

but not of form and to disfiguration aliquot a variation of form is associated but not
of volume. Well we call hydrostatics aliquot of the elastic potential and denote with
ϕv the density of energy relative to the hydrostatics aliquot σ ξ , σ η , σ ζ of the stress.
Clearly
   2  2  
2
ϕv = 1
2E σ ξ + σ η + σ ζ − 2v σ ξ σ η + σ ξ σ ζ + σ η σ ζ
= 1−2ν 2
6E T .
122 1 The Three-Dimensional Problem

Insofar, since from the (1.7.7) it follows ϕ = ϕf + ϕv , the parameter of crisis

ϕf of von Mises can be interpreted as density of energy relative to the disfiguration
aliquot σ ξ , σ η , σ ζ of the stress. 
Remark 1.7.3 Let us consider the state of stress

σx = 0
σy = 0
σz = 0
τxy = c ∈  − {0}
τxz = 0
τyz = 0

whose principal stresses are obviously 0,0,c. We notice that every previous strength
criterion furnishes as safety condition
σ0
c ≤ σa = . 
s
Problem 1.7.1 In the point P of a structure the state of stress is

σx = 0, σy = 0, σz = 1,000 kg/cm2
τxy = 0, τxz = 200 kg/cm2 , τyz = 0.
The structure, realized with a steel having σ0 = 4,500 kg/cm2 , has σa =
2,500kg/cm2 . You shall verify the structural safety in P.
Solution First of all, being
 1/2
τy = τyx
2
+ τyz
2
= 0,

y is a principal direction of stress. It follows of it that ση = σy = 0 is a principal

stress and that the others two principal stresses are furnished by the circle of Mohr
for the set of planes having support y. We have
 1
σz − σz2 + 4τxz
2 2
σξ =
 2 1
σz + σz2 + 4τxz
2 2
σζ = .
2
Then, since σζ and σξ are the greatest and smallest principal stress, the maximum
shearing stress criterion furnishes as safety condition
⎛  1  1 ⎞
1 ⎝ σz + σz2 + 4τxz
2 2 σz − σz2 + 4τxz
2 2
⎠ ≤ σa
−
2 2 2 2
1.7 Strength Criterions 123

Fig. 1.7.3

that is
 1
2 2
s σz2 + 4τxz ≤ σa .

Insofar in P the structure is in structural safety, since

 1
2
1,0002 + 4 · 2002 ≤ 2,500. 

Problem 1.7.2 In the point P of a structure the state of stress is

σx = 0, σy = 0, σz = 1,000 kg/cm2
τxy = 0, τxz = 200 kg/cm , 2
τyz = 300 kg/cm2 .

The structure, realized with a steel having σ0 = 4,500 kg/cm2 , has σa =

2,500 kg/cm2 . You shall verify the structural safety in P.
Solution It is convenient to change the frame of reference. In the plane x, y we denote
with x the line having the direction of τ z and with y the line normal to x and to z
(Fig. 1.7.3). Therefore
 1
2
τzy = τz = 3002 + 2002 , τzx = 0

We put z = z and assume the frame of reference P, x , y , z . Since the circle

of Mohr for the set of planes having support z degenerates in a point that is the
reference origin (Fig. 1.7.4), it results

σx = 0, σy = 0, τx y = 0.

Fig. 1.7.4
124 1 The Three-Dimensional Problem

Obviously

σz = 0, τz y = τzy , τz x = τzx .

This way we known the state of stress in the point P in the new frame of reference.
We can proceed now as in the Problem 1.7.1. 

1.7.5 The Intrinsic Curve Criterion

This criterion, proposed by Mohr, is suitable for non metallic construction materials
and for the ground. It is available for every type of material a curve said intrinsic
curve determined effecting three-dimensional tests in the Laboratory material tests.
For a lot of specimens carried to crisis we draw the maximum principal circles of
Mohr. We call intrinsic curve their envelope. Such curve is qualitatively always of
the type of Fig. 1.7.5. In fact the crisis is not verified for hydrostatics compres-
sion. Let P be a point of the body and σξ , ση , σζ be the three principal stress in
P. This criterion declare that the material is in structural safety in P if, amplifying
the load s times, the maximum principal circle of Mohr is inside to the intrinsic

Fig. 1.7.5

Fig. 1.7.6
1.7 Strength Criterions 125

curve (Fig. 1.7.6). The crisis happens when the maximum principal circle of Mohr
is tangent to the intrinsic curve.
Remark 1.7.4 Other strength criterions have also been studied. They, however,
turned out to be less efficient. Among these they must be underlined the maximum
normal stress criterion and the maximum elongation criterion. As usual the thresh-
old of crisis of these criterions is the value of the parameter of crisis furnished by
the tensile breaking test at the yelding. 
Chapter 2
The Problem of Saint Venant

2.1 Geometry of Areas

2.1.1 Centroid

We consider a set of areas contained in a plane α. For a distributed area we know

from the Measure theory that, if it is measurable, its area is measured from the
non-negative real number

A = dA.
A

When, in the system of areas, distributed areas that have negligible dimensions
of encumbrance in comparison to the mutual distances are present (Fig. 2.1.1), it is
convenient to approximate such distributed areas with concentrated areas.
The area concentrated in a point is an abstract greatness, in the sense that it
cannot geometrically be represented. Using the concentrated areas the calculations
are a lot simplified.2.1.1
In this book we consider systems of areas constituted from a distributed area A
and/or n areas A1 , . . ., An , respectively concentrated in the points P1 , . . ., Pn . The
total area Atot
 
n
Atot = dA + Ai
A i=1

is always supposed greater than zero.

We consider a system of distributed and/or concentrated areas. Let a be an ori-
ented line of α, Q be a point of a, b be a line normal to a and passing through Q, P

2.1.1 An important case in which the concentrated areas are employed is the reinforced concrete.
In it the procedure of calculation substitutes the cross section of the beam, constituted from a
concrete area and from an iron area, with a system constituted from a distributed area (that is the
part of concrete submitted to compression) and several concentrated areas (which are the iron areas
amplified with a convenient factor and concentrated in their barycenters).

A. Maceri, Theory of Elasticity, DOI 10.1007/978-3-642-11392-5_2, 127


C Springer-Verlag Berlin Heidelberg 2010
128 2 The Problem of Saint Venant

Fig. 2.1.1

Fig. 2.1.2

be a point of α. We still denote with b the orthogonal abscissa on the axis b of P and
with bi the orthogonal abscissa on the axis b of a point Pi of α (Fig. 2.1.2). We call
static moment with respect to the line a of the system of areas and we denote with
the symbol Sa the real number
 
n
Sa = bdA + Ai bi .
A i=1

It is obvious that

– Sa is measured in cm3 ,
– Sa doesn’t depend on the orientation of a,
– Sa doesn’t depend on the particular choice of Q,
– changing the orientation of b, Sa changes the sign but not the absolute value.

Besides, if we decompose the system assigned in k subsystems, it is easy to verify

that2.1.2

Sa = (Sa )1 + · · · + (Sa )k

where (Sa )i denotes the static moment with respect to line a of the subsystem i of
the system of areas.

1 , . .-., Bk such
2.1.2 In the Theory of integration we prove that, decomposing an area B in k subareas B
- -
that ∪i=1
k B and meas (B ∩ B ) = 0 ∀i, j{1, · · · , k}, it results
i 2 i j B fdB = B1 fdB + · · · + Bk fdB.
2.1 Geometry of Areas 129

We consider a system of distributed and/or concentrated areas and any orthogonal

cartesian frame of reference. We denote with Pi = (xi , yi ) the point where the area
Ai is concentrated and consider the real numbers

Sy Sx
xG = , yG = . (2.1.1)
Atot Atot
We call centroid or barycenter of the system of areas and we denote with the
symbol G the point

G = (xG, yG ).

We can prove that the position of G with respect to the areas of the system doesn’t
depend on the particular choice of the cartesian orthogonal reference system 0, x, y.
We have
[2.1.1] The static moment with respect to an axis of a system of areas is equal to the
static moment with respect to the same axis of the system constituted by an unique
area, of intensity Atot and concentrated in the centroid of the assigned system of
areas.
Proof It is enough to observe that, since the (2.1.1)

Sx = Atot yG · 

[2.1.2] The static moment of a system of areas with respect to a centroidal axis is
zero. If the static moment of a system of areas with respect to an axis x is zero, then
the centroid of the system of areas belongs to x.
Proof Obvious. 
Problem 2.1.1 You shall determine the centroid of the system of areas of Fig. 2.1.3,
where in the point P1 an area A1 = 200 cm2 is concentrated.
Solution We introduce a frame of reference 0, x, y, chosen so that the calculations
are simple (Fig. 2.1.4). We have, decomposing the area A in a subarea B1 and in a
subarea B2

20 20 cm

10

20

10
Fig. 2.1.3
130 2 The Problem of Saint Venant

Fig. 2.1.4

Atot = B1 + B2 + A1 = 200 + 400 + 200 = 800 cm2

Sx = (Sx )1 + (Sx )2 + A1 y1 = 7, 000 + 2, 000 + 8, 000 = 17, 000 cm3
Sy = (Sy )1 + (Sy )2 + A1 x1 = 6, 000 + 8, 000 + 0 = 14, 000 cm3 .

So we have

G = (17.5, 21.25). 

We consider a system of areas. We call axis of symmetry with respect to the

direction y for the system a line x such that (Fig. 2.1.5)

– for every parallel to y having non void intersection with a distributed area
it happens that, whatever point P of the intersection is chosen, another
point Q of the intersection exists such that the segments OP and OQ
have equal length, where O is the point common to x and to the parallel
one to y,
– for every point Pi in which an area Ai is concentrated, on the parallel one to y
passing through Pi a point Pj exists in which an area Aj = Ai is concentrated
and such that the segments WPi and WPj have equal length, where W is the point
common to x and to the parallel one to y.

Fig. 2.1.5
2.1 Geometry of Areas 131

Fig. 2.1.6

We can prove that

[2.1.3] If x is axis of symmetry of a system of areas, then the centroid of the system
of areas belongs to x. 
Problem 2.1.2 You shall determine the centroid of a circle, that of a rectangle, that
of a triangle, that of a trapeze.
Solution About the circle, let x and y two orthogonal diameters. Since x [resp. y]
is axis of symmetry with respect to y [resp. x] for the circle, since the [2.1.3] the
centroid of the circle must belong to x [resp. y]. As a consequence, the centroid of a
circle coincides with its center.
About the rectangle, reasoning as for the circle, we consider its medians, i.e. the
lines passing through the middle points of two opposite sides. Since each median is
axis of symmetry with respect to the other for the rectangle, because of the [2.1.3]
the centroid of a rectangle is the point of intersection of the two medians.
About the triangle, we consider its medians, i.e. the lines passing through a vertex
and the middle point of the opposite side. Since each median is axis of symmetry
with respect to the opposite side for the triangle, because of the [2.1.3] the centroid
of a triangle is the point of intersection of the medians.
About the trapeze, we consider its median, i.e. the line passing through the middle
points of the parallel bases. Since the median is axis of symmetry with respect to
the bases for the trapeze (Fig. 2.1.6), because of the [2.1.3], the centroid of a trapeze
belongs to its median. 

2.1.2 Inertia Centroidal Ellipse

We consider a system of distributed and/or concentrated areas, a line a and a point

Q of a. We consider the line b normal to a and passing through Q. We orientate a
and b and still denote with b [resp. bi ] the coordinate according to the axis b of the
generic point P [resp. Pi ].
We call moment of inertia of the system of areas with respect to the line a, and
we denote with the symbol Ia , the real number
 
n
Ia = b2 dA + Ai b2i . (2.1.2)
A i=1
132 2 The Problem of Saint Venant

We notice that
– since the Theory of integration, Ia is always non negative,
– Ia is measured in cm4 ,
– Ia doesn’t depend on the orientation of b,
– Ia doesn’t depend on the orientation of a,
– Ia doesn’t depend on the choice of Q.

Besides it is easy to verify that if we decompose the system assigned in k sub-

systems, in the hypothesis specified in the note 2.1.2 since the Theory of integration

Ia = (Ia )1 + · · · + (Ia )k (2.1.3)

where (Ia )i denotes the moment of inertia with respect to the line a of the subsystem
i of the system of areas.
We consider now two lines x, y oriented and not parallel. We denote with O the
point of intersection and with α the angle formed by their positive rays (Fig. 2.1.7).
We still denote with x [resp. xi ] and y [resp. yi ] the coordinates of the generic point
P [resp. Pi ] in the cartesian frame of reference 0, x, y (Fig. 2.1.7). We finally set

h = y sina, k = x sina, hi = yi sina, ki = xi sina. (2.1.4)

We call centrifugal moment of the system of areas with respect to the lines x, y
and we denote with the symbol Ixy the real number
 
n
Ixy = h k dA + Ai hi ki .
A i=1

Evidently, in the particular case of axes x, y orthogonal, it results

 
n
Ixy = x y dA + Ai xi yi .
A i=1

We notice that
– Ixy is measured in cm4 ,
– if we change the orientation of x or of y, the sign of Ixy changes but not the absolute
value.

Fig. 2.1.7
2.1 Geometry of Areas 133

Fig. 2.1.8

Besides it is easy to verify that if we decompose the system assigned in k sub-

systems, in the hypothesis specified in the note 2.1.2 since the Theory of integration

Ixy = (Ixy )1 + · · · + (Ixy )k (2.1.5)

where (Ixy ) denotes the centrifugal moment with respect to the lines x, y of the
subsystem i of the system of areas.
Problem 2.1.3 You shall calculate the moment of inertia of a rectangle with respect
to the centroidal line parallel to a side and the centrifugal moment of the rectangle
with respect to the centroidal lines parallel to the sides.
Solution We consider the rectangle of Fig. 2.1.8. We denote with x, y the centroidal
lines parallel to the sides of the rectangle. We easily obtain

bh3
Ix = , Ixy = 0. 
12

Problem 2.1.4 You shall calculate the moment of inertia of a circle with respect to
a diameter and the centrifugal moment of the circle with respect to two orthogonal
diameters.
Solution We consider the circle of Fig. 2.1.9. We denote with x [resp. y] the
horizontal [resp. vertical] diameter and with R the radius. We easily obtain

π R4
Ix = , Ixy = 0. 
4
Problem 2.1.5 You shall calculate the moment of inertia of the triangle of Fig. 2.1.10
with respect to the axes x (passing through B and G) and y (parallel to AC).
Solution We easily obtain

b3 h 2 bh3
Ix = sin α, Iy = . 
48 36

Fig. 2.1.9
134 2 The Problem of Saint Venant

Fig. 2.1.10

We consider a system of distributed and/or concentrated areas and a line a. We

denote with G the centroid of the system and with a the parallel one to a passing
through G. We have the following transport theorem (Fig. 2.1.11)

[2.1.4] The moment of inertia of a system of areas with respect to a line a is

equal to the sum of the moment of inertia of the system with respect to the cen-
troidal parallel line aG and of the moment of inertia with respect to the line a
of a system constituted by an unique area, having intensity Atot and concentrated
in G:

Ia = IaG + Atot dist2 (G, a).

Proof Let O be a point of a, b be the line normal to a and passing through O, O be

the point common to a and b, a be the centroidal line parallel to a. We orientate b.
We orientate a and a in the same way. We still denote with the symbol b [resp. b ]
the second coordinate of the point P in the frame of reference 0, a, b [resp. O, a , b].
We denote with the symbol d the second coordinate of the point O in the frame of
reference 0, a, b (Fig. 2.1.12).
Since

b = b + d

Fig. 2.1.11
2.1 Geometry of Areas 135

Fig. 2.1.12

we have

- 
n
Ia = 
A (b + d) dA +
2 Ai (bi + d)2
i=1
-  2
n  2

= A (b ) dA + i=1 Ai (bi )
   
- 
n - 
n
+ A dA + Ai d2 + 2d A b dA + Ai bi
i=1 i=1

= Ia + Atot d2 + 2dSa .

From this, since the [2.1.2], the thesis follows. 

The transport theorem subsists for the centrifugal moments too

[2.1.5] The centrifugal moment of a system with respect to the axes x, y is equal
to the sum of the centrifugal moment of the system with respect to the centroidal
parallel and equiverse axes xG , yG and of the centrifugal moment with respect to
the axes x, y of the system constituted by an unique area having intensity Atot and
concentrated in G

Ixy = IxG yG + Atot dist(G, x) dist(G, y).

Proof Let x,y be two arbitrary orientated lines not parallel, O be their common
point, x [resp. y ] be the centroidal line parallel and equiverse to x [resp. y]. We
denote with x , y [resp. xi , yi ] the coordinates of the point P [resp. Pi ] in the cartesian
frame of reference G, x , y and with xG , yG those of G in the reference O, x, y. Finally
we put

h = y sinα, k = x sinα, hi = yi sinα, ki = xi sinα,

hG = dist (G, x) = yG sina, kG = dist(G, y) = xG sina.
136 2 The Problem of Saint Venant

Fig. 2.1.13

Obviously (Fig. 2.1.13)

h = h + hG , k = k + kG

so that
 
n
 
Ixy = (h + hG )(k + kG )dA + Ai (hi + hG )(ki + kG )
A i=1

 !  !

n 
n
 
= h k dA + Ai hi ki + dA + Ai hG kG
A i=1 A i=1

 !  !

n 
n

+ hG k dA + Ai ki + kG 
h dA + Ai hi
A i=1 A i=1

= Ix y + Atot hG kG + hG Sy + kG Sx .

We observe that, since the [2.1.2], Sx and Sy are zero. Insofar

Ixy = Ix y +Atot hG kG . 

Fig. 2.1.14
2.1 Geometry of Areas 137

Problem 2.1.6 We consider the system of areas of Fig. 2.1.14. We denote with G the
centroid of the system and with y the centroidal line normal to x and oriented as in
Fig. 2.1.14. You shall calculate the value of l such that Ix = Iy .
Solution First of all, we decompose the system in the four areas of Fig. 2.1.15 and
we denote with A1 , . . ., A4 and G1 , . . ., G4 the partial areas and the their centroids.
We denote with xGi , yGi the parallel ones to x, y passing through Gi and equiverse
to x, y. Since x is axis of symmetry for the system with respect to y, G belongs to
x. We introduce the frame of reference of Fig. 2.1.16 and denote with l1 the first
coordinate of G in such reference. We have

Sz = Atot l1 = (10 · 30 + 10 · 60 + 10 · 30 + 3.14 · 102 )l1 = (Sz )1 + (Sz )2 + (Sz )3

+(Sz )4 = 10 · 30( − 25) + 10 · 60( − 5) + 10 · 30( − 25) + 3.14 · 102 (10 + l).

Therefore

l1 = 0.207 l − 9.815.

Now keeping in mind of (2.1.3) and (2.1.5) and of the transport theorems we get
(Fig. 2.1.15)
   10 · 603
Ix = (Ix )1 + (Ix )2 + (Ix )3 + (Ix )4 = IxG1 + A1 · 252 +
1 12
   3.14 · 104 30 · 103 10 · 603
+ IxG3 + A3 · 252 + = + 30 · 10 · 252 +
3 4 12 12
30 · 103 3.14 · 104
+ + 30 · 10 · 252 + = 56.78 · 104 cm4
12 4

Fig. 2.1.15

Fig. 2.1.16
138 2 The Problem of Saint Venant
  
Iy = (Iy )1 + (Iy )2 + (Iy )3 + (Iy )4 = IyG1 + A1 (l1 + 25)2
1
     
+ IyG2 + A2 (l1 + 5)2 + IyG3 + A3 (l1 + 25)2
2 3
  
+ IyG4 + A4 (10 + l − l1 )2
4

10 · 303 60 · 103
= +10 · 60 · (0.207 l +15.185)2 + +60 · 10 · (0.207l −4.815)2
6 12
3.14 · 104
+ + 3.14 · 102 · (0.793 l + 19.815)2
4
= 249.106 l2 + 12.444 · 16 l + 33.34 · 104 .

So we must find a value of l such that

249.106 l2 + 12.444 · 16 l + 33.34 · 104 = 56.78 · 104 . (2.1.6)

The (2.1.6) admits the solutions 14.58 and –64.53 cm. The negative solution is
to discard because l denote a distance and then a non negative number. 

After this we can introduce the concept of central inertia ellipse. We consider
a system of distributed and/or concentrated areas satisfying at least one of the two
following conditions
– a distributed area of the system exists having positive measure,
– there are at least three points not ranged in a line in which non zero areas of the
system are concentrated.

Obviously such system has

Atot > 0

and, for every line a2.1.3

Ia > 0.

We call inertia principal axes two axes ξ , η such that

– ξ , η are orthogonal,
– ξ , η are centroidal,
– Iξ η = 0.

About the existence of the inertia principal axes, let x, y be two axes orthogonal
and centroidal and let x , y be the axes gotten rotating x, y of an angle α (Fig. 2.1.17).

-
2.1.3 From the Integration theory it follows that, since meas2 (B) > 0, it results By
2 dA > 0.
2.1 Geometry of Areas 139

Fig. 2.1.17

We easily2.1.4 prove that

Ix + Iy Ix − Iy
Ix  = + cos2α − Ixy sin2α
2 2
Ix + Iy Ix − Iy
Iy  = − cos2α + Ixy sin2α
2 2
Ix − Iy
Ix  y  = sin2α + Ixy cos2α.
2

Therefore we have to see if some values of α in [–π /2, π /2] exist such that

Ix − Iy
sin2α + Ixy cos2α = 0. (2.1.7)
2
The possible cases are the following two
1 it is true that Ix = Iy and Ixy = 0
2 it is false that Ix = Iy and Ixy = 0.

In the case 1 evidently every α belonging to [–π /2, π /2] is solution of the (2.1.7).
Consequently every couple (ξ , η) of centroidal and orthogonal axes is a couple of
inertia principal axes. Iξ = Iη is also obvious.
In the case 2 evidently the possible cases are the following three
a Ix = Iy and Ixy = 0
b Ix = Iy and Ixy = 0
c Ix = Iy and Ixy = 0.

Fig. 2.1.18

2.1.4 We have from the Fig. 2.1.18 that x = y sin α + x cos α and from the Fig. 2.1.19 that y cos α =
y + x sin α.
140 2 The Problem of Saint Venant

Fig. 2.1.19

In the case a, evidently, α = π /4 is unique solution of the (2.1.7). Iξ = Iη is also

obvious.
In the case b, evidently, α = 0 is unique solution of the (2.1.7). Iξ = Iη is also
obvious.
In the case c, obviously
1 2Ixy
α= arctg −
2 Ix − Iy
is the unique solution of the (2.1.7). We can prove that Iξ # Iη .
So in the case 2 one and only one couple ξ , η of inertia principal axes exists and
it results Iξ = Iη .
Let a be an inertia principal axis. We call principal inertia radius of the system
of areas with respect to the inertia principal axis a, and we denote with the symbol
ρ a , the positive real number
1
Ia 2
ρa = .
Atot

Let (ξ , η) be a couple of inertia principal axes of the system of areas. We call

inertia centroidal ellipse of the system of areas the ellipse of equation (Fig. 2.1.20)

ξ2 η2
+ = 1.
ρη2 ρξ2

Remark 2.1.1 We observe that in the case 1 the axes ξ , η can be chosen however
provided that centroidal and orthogonal. Besides, since ρ ξ = ρ η , the ellipse is a
circle. In the case 2, instead, only one possibility of choice exists for the axes ξ , η.
Besides, since ρ ξ = ρ η , the ellipse is not a circle. 

Fig. 2.1.20
2.1 Geometry of Areas 141

Remark 2.1.2 Sometimes it is convenient when the inertia centroidal ellipse is a cir-
cle. This happens in a compressed member of a truss to assure the stability safety.
If the member is done by two channel irons, we can calculate the installation dis-
tance with the procedure employed in Problem 2.1.6. So the inertia centroidal ellipse
becomes a circle. 

2.1.3 Antipolarity

Let us consider a system of areas and denote with G his centroid. We call inertia
centroidal frame of reference of the system a cartesian orthogonal frame of reference
having for origin G and for axes a couple ξ , η of inertia principal axes.
Let P = (ξp , ηp ) be any point distinct from G (Fig. 2.1.21). We call antipolar of
P with respect to the inertia centroidal ellipse and we denote with the symbol p the
line of equation

ξ ξP ηηP
+ 2 = −1, (2.1.8)
ρη2 ρξ

where ρ ξ , ρ η denote the principal inertia radius with respect to the inertia principal
axes ξ , η. Obviously p is not centroidal. Equally obvious is that P doesn’t belong
to p.
We call antipolarity the law of correspondence that to every point P = G asso-
ciates his antipolar (2.1.8).
The antipolarity has many important properties. First of all, it is an one to one
function. In fact, if r is a non centroidal line, we can prove that it exists one and
only one point R distinct from G and having r as antipolar. We call R antipole of the
line r.
Moreover let a be a non centroidal line, b be a line non parallel to a, O be their
common point, A be the antipole of a (Fig. 2.1.22). We orientate a and b, assume

Fig. 2.1.21

Fig. 2.1.22
142 2 The Problem of Saint Venant

Fig. 2.1.23

the cartesian frame of reference 0, a, b and denote with aA , bA the coordinates of A

in such frame of reference.
Well, we can prove that (Fig. 2.1.23)

Ia = Sa dist (A, a).

Iab = Sa dist (A, b)

where the orthogonal distance dist(A, a)[resp. dist(A, b)] is considered positive if A
belongs to the positive halfplane individualized from a [resp. b], negative otherwise.
Clearly Ia is always positive, but Iab can assume any sign.
As a consequence the following first fundamental property of the antipolarity
holds

[2.1.6] The inertia moment of a system of areas with respect to a non centroidal line
a is equal to the static moment with respect to the line a of a system constituted by an
unique area, having intensity Sa and concentrated in the antipole A of the line a. The
centrifugal moment of a system of areas with respect to two any non parallel lines a
and b, with a non centroidal, is equal to the static moment with respect to the line b
of a system constituted by an unique area, having intensity Sa and concentrated in
the antipole A of the line a. 

Let r and s be two non centroidal lines, R [resp. S] be the antipole of r [resp. s].
Well, we can prove that if R belongs to s, then S belongs to r (Fig. 2.1.24). As a
consequence it holds the following second fundamental property of the antipolarity

Fig. 2.1.24
2.1 Geometry of Areas 143

Fig. 2.1.25

[2.1.7] Let A be a point distinct from G, a the antipolar of A. Then the antipole of
the lines non centroidal and passing through A belong to a. Moreover any point of
a is antipole of a line non centroidal and passing through A (Fig. 2.1.25). 

We call diameter of the inertia centroidal ellipse any centroidal line. Let a
be a diameter of the inertia centroidal ellipse. We can prove the following third
fundamental property of the antipolarity

[2.1.8] Every line parallel to a and distinct from a has antipole on a diameter b
distinct from a. Every point of b is antipole of a line parallel to a and distinct from
a. Moreover every line parallel to b and distinct from b has antipole on the diameter
a (Fig. 2.1.26). 

As a consequence of the [2.1.8], the diameters of the inertia centroidal ellipse go

two by two. We call every couple couple of conjugate diameters.
The antipolarity also has other remarkable properties. We can in fact prove that

[2.1.9] If x is an axis of symmetry with respect to the direction y for the system
of areas, then the diameters parallel to x and y constitute a couple of conjugate
diameters of the inertia centroidal ellipse. 

[2.1.10] If the inertia centroidal ellipse is not degenerated in a circle, the unique
couple of conjugate and orthogonal diameters is the couple of the inertia principal
axes. If the inertia centroidal ellipse is degenerated in a circle, every couple of
centroidal and orthogonal axes is a couple of conjugate diameters and the diameters
of every couple of conjugate diameters are orthogonal. 

Fig. 2.1.26
144 2 The Problem of Saint Venant

Fig. 2.1.27

[2.1.11] Let t be a tangent to the inertia centroidal ellipse, T be the point of tan-
gency, T  be the point symmetrical of T with respect to G. Then T  is the antipole of t
(Fig. 2.1.27).2.1.5 
Remark 2.1.3 From the [2.1.11] a simple way follows to find the conjugate diameter.
In fact, given a diameter a of the ellipse, the conjugated diameter b is immediately
built drawing a line parallel to a and tangent to the ellipse. Then the diameter b is
obtained connecting G with the point of tangency (Fig. 2.1.28). 

We call conjugate diameters also the segments of the conjugate diameters inside
to the ellipse (Fig. 2.1.29).
Let (a, b) be a couple of conjugate diameters (Fig. 2.1.30). We call inertia radius
with respect to the diameter a, and we denote with the symbol ρ a , the semidiameter
conjugate to the diameter a (Fig. 2.1.30).
To determine the length of ρ a , taking into account the [2.1.4], [2.1.6], [2.1.11]
we get (Fig. 2.1.30)

Ia1 = Ia + Atot ρa2 sin2 ϕ = (Atot ρa sinϕ)2ρa sinϕ

Fig. 2.1.28

Fig. 2.1.29

2.1.5 In Analytical geometry we prove that the tangent to ellipse in T  is parallel to t.

2.1 Geometry of Areas 145

Fig. 2.1.30

from which

Ia = Atot (ρa sinϕ)2

from which
1
1 Ia 2
ρa = .
sinϕ Atot

Remark 2.1.4 In the structural design, it is not necessary to trace the inertia cen-
troidal ellipse to perform the calculations. It is sufficient to employ any couple of
conjugate diameters (Fig. 2.1.29). 

Problem 2.1.7 You shall find a couple of conjugate diameters of the inertia
centroidal ellipse for the system of areas of Fig. 2.1.31.
Solution First of all we find the centroid of the system. Assuming the frame of
reference of Fig. 2.1.32 we have

Sx 400
yG = = = 10 cm
Atot 40

Fig. 2.1.31

Fig. 2.1.32
146 2 The Problem of Saint Venant

Fig. 2.1.33

Sy 600
xG = = = 15 cm.
Atot 40

We now consider the non centroidal line a1 of Fig. 2.1.33 and determine its
antipole A1 . Clearly the line b passing through A1 and G is the diameter conju-
gated to the direction of a1 . Then, denoting with a the centroidal line parallel to a1 ,
the couple a, b is a couple of conjugate diameters of the inertia centroidal ellipse of
the system of areas of Fig. 2.1.31.
To determine A1 , we assume the frame of reference of Fig. 2.1.34. We have

Ia1 = Sa1 hA1 , Ia1 y1 = Sa1 kA1 ,

Sa1 = Atot · 10 = 400 cm3 .

Beside

Ia1 y 1 8, 000
ka1 = = = 20 cm
Sa1 400
Ia1 8, 000
ha1 = = = 20 cm.
Sa1 400
The conjugate diameters a, b are therefore those of Fig. 2.1.36.
We now detach the semidiameters. Denoting with α the angle between a and b,
we have (Fig. 2.1.35)
Ia = Atot (ρa sinα)2

Fig. 2.1.34
2.1 Geometry of Areas 147

Fig. 2.1.35

and, by the Pythagoras’s2.1.6 theorem

sinα(102 + 52 )1/2 = 10

and, since the transport theorem (Fig. 2.1.34)

Ia1 = 8, 000 cm2 = Ia + Atot · 102 = Ia + 4, 000.

As a consequence
1 1
1 Ia 2 (125) 2 1 1
ρa = = (100) 2 = (125) 2 cm.
sinα Atot 10
Insofar we can detach the semidiameters on b (Fig. 2.1.36).
To detach the semidiameters on a we have, in perfect analogy

Ib = Atot (ρb sinα)2

and (Fig. 2.1.37)

d1 = 20 sinα, d2 = 20 sina, d3 = 10 sinα, d4 = 10 sinα.

Fig. 2.1.36

2.1.6 Pythagoras, Samos 571 b.C. – Croton 496 b.C.

148 2 The Problem of Saint Venant

Fig. 2.1.37

As a consequence
1 1
1 Ib 2 Ib 2
ρb = =
sin α Atot sin2 αAtot

 1 1
10 d12 + 10 d22 + 10 d32 + 10 d42 2
10, 000 2 1
= = = (250) 2 cm.
40sin2 α 40
With this the problem is solved (Fig. 2.1.38). 

Problem 2.1.8 You shall find the principal diameters of the inertia centroidal ellipse
of a rectangle.
Solution We consider a rectangle. First of all, we observe that the centroidal lines
x, y parallel to the sides (Fig. 2.1.39) are the one axis of symmetry with respect to
the direction of the other. Consequently x, y is a couple of conjugate diameters of
the inertia centroidal ellipse of the rectangle. Besides such diameters, being orthog-
onal, are a couple of principal diameters. Therefore, we only have to detach the

Fig. 2.1.38

Fig. 2.1.39
2.1 Geometry of Areas 149

semidiameters. We easily have (Fig. 2.1.39)

1 1
Ix 2 b h3 /12 2
h
ρx = = =√
bh bh 12
1 1
Iy 2 h b3 /12 2
b
ρy = = =√ .
bh bh 12

So the couple of principal diameters is that of Fig. 2.1.40. 

Problem 2.1.9 You shall find the principal diameters of the inertia centroidal ellipse
of a circle.
Solution Reasoning as in the Problem 2.1.8 we see that every couple of orthogo-
nal and centroidal axes individualizes a couple of principal diameters of the inertia
centroidal ellipse. Therefore this one is a circle. To detach the semidiameters we
observe that (Fig. 2.1.41)
π R4
Ix = π R2 ρx2 = .
4
Consequently
R
ρx = .
2
Insofar the inertia centroidal ellipse of a circle is a circle of radius R/2. A couple
of semidiameters is shown in Fig. 2.1.42. 

Fig. 2.1.40

Fig. 2.1.41

Fig. 2.1.42
150 2 The Problem of Saint Venant

Problem 2.1.10 You shall find a couple of conjugate diameters of the inertia
centroidal ellipse of a triangle.

Solution We consider the triangle of Fig. 2.1.43 and its centroid G. We denote with
r the line passing through B and G; with s the parallel to AC (Fig. 2.1.43). Since r
is axis of simmetry with respect to r, r, s is a couple of conjugate diameters of the
inertia centroidal ellipse of the triangle. Denoting with b [resp. h] the base [resp.
height] of the triangle (Fig. 2.1.44), it result (Problem 2.1.5)

b3 h 2 bh3
Ir = sin α, Is = .
48 36

Consequently (Fig. 2.1.45)

1
2 Is 2 h
ρs = =√
b h sin2 α 18sinα

1
2 Ir 2 b
ρr = =√ . 
b h sin2 α 24
Others important properties of the antipolarity are the following

Fig. 2.1.43

Fig. 2.1.44
2.1 Geometry of Areas 151

Fig. 2.1.45

[2.1.12] Any non centroidal line x and its antipole X are aside opposite with respect
to G (Fig. 2.1.48).

Proof Let
– x be a not centroidal line (Fig. 2.1.46),
– x0 be its centroidal parallel line (Fig. 2.1.46),
– X be the antipole of x with respect to the inertia centroidal ellipse of the system of
areas,
– y0 be the line passing through G and X (Fig. 2.1.47),
– be the point common to x and y0 ,
– α be the angle between x0 and y0 ,
– lX be the length of the segment whose end points are O and X (Fig. 2.1.48),
– lG be the length of the segment whose end points are O and G.

Fig. 2.1.46

Fig. 2.1.47
152 2 The Problem of Saint Venant

Fig. 2.1.48

Obviously the orthogonal distance (with sign) of X from x is equal to

dist(X, x) = lX sinα

and the orthogonal distance (with sign) of G from x is equal to

dist(G, x) = lG sinα.

Denoting with ρx0 the inertia radius with respect to axis x0 of the system of areas
(Fig. 2.1.49), we have

Atot lG sinα lX sinα = Sx lX sinα = Ix

(2.1.9)
= Ix0 + Atot (lG sinα)2 = Atot (ρx0 sinα)2 + Atot (lG sinα)2

from which

lG lX = ρx20 + lG
2

from which

lG (lX − lG ) = ρx20 (2.1.10)

from which

lG (lX − lG ) > 0. (2.1.11)

Fig. 2.1.49
2.1 Geometry of Areas 153

Since y0 is orientated as in Fig. 2.1.47, it results lG > 0. Consequently, taking

into account the (2.1.11), we have

l X − lG > 0

from which

lX > lG > 0.

The thesis follows. 

[2.1.13] If the antipolar x goes near to G, then the antipole X goes away from G. If
the antipolar x goes away from G, then the antipole X goes near to G (Fig. 2.1.50).
Proof Since the (2.1.10) it results

lG (lX − lG ) = const.

The thesis follows. 

[2.1.14] Let a be a line, b the diameter conjugate to direction of a. It results Iab = 0.
Proof In fact if a is not centroidal, since the [2.1.6] we have Iab = Sa · 0. If a is
centroidal, denoting with a a non centroidal line parallel to a, we have Ia b = 0.
Then, because of the transport theorem, it results Iab = Ia b = 0. 
[2.1.15] Let a be a line, b be a diameter of the inertia centroidal ellipse. If Iab = 0,
then b is the diameter conjugated to the direction of a.
Proof In fact let a be a non centroidal line parallel to a, A be the antipole of a .
Then, because of the transport theorem and of the [2.1.6], it results Ia b = Iab + 0 =
0 = Sa x. But Sa = 0 since a is non centroidal. Then the distance x of A from b is
zero. The thesis follows. 
Problem 2.1.11 Given a couple of conjugate diameters of the inertia centroidal
ellipse and a non centroidal line x, you should determine the antipole X of x.

Fig. 2.1.50
154 2 The Problem of Saint Venant

Fig. 2.1.51

Solution We proceed graphically. Preliminarily we remember that, because of a

theorem of Euclid,2.1.7 for the straight triangle of Fig. 2.1.51 it results ab = h2 .
Now we consider the couple r, s of conjugate diameters of the inertia centroidal
ellipse and the non centroidal line x of Fig. 2.1.52. We consider the intersection S1
[resp. R1 ] between r [resp. s] and x. Since the [2.1.8], the antipole X of x is the point
intersection of the antipolars r1 of R1 and s1 of S1 . Since S1 belongs to diameter
r, for the [2.1.8] the antipolar s1 is parallel to the conjugate diameter r. To find
its position, we use the (2.1.10) by the graphic way of Fig. 2.1.53. So we turn the
semidiameter ρ s on the perpendicular one to r passing through G (Fig. 2.1.53). We
connect the point P1 obtained with S1 . Then we draw the line passing through P1
and orthogonal to the segment P1 S1 (Fig. 2.1.53). We denote with Q1 the point
intersection of such line and r. Clearly the line passing through Q1 and parallel
to s is the antipolar s1 of S1 (Fig. 2.1.53). Analogously proceeding, we draw the
antipolar r1 of R1 . Finally, the point intersection of the antipolars s1 and r1 furnishes
the antipole X (Fig. 2.1.54). 

Fig. 2.1.52

Fig. 2.1.53

2.1.7 Euclid, Alexandria 300 b.C.

2.1 Geometry of Areas 155

Fig. 2.1.54

Remark 2.1.5 Performing in inverse order the construction of Fig. 2.1.54, it

is possible to graphically determine the antipolar of a point distinct by the
centroid. 

Problem 2.1.12 You should determine a couple of conjugate diameters of the inertia
centroidal ellipse for the system of areas of Fig. 2.1.55. In the point P1 an area A1
is concentrated.
Solution We proceed still graphically. To simplify the calculations we approximate
the distributed area with rectangles and/or triangles (Fig. 2.1.56).
In the way already seen we determine the centroid G2 of the rectangle 2, that
G3 of the rectangle 3, that G4 of the triangle 4, that G of the whole system, a
couple of conjugate diameters of the inertia centroidal ellipse of the subarea 2
[resp. 3], [resp. 4] (Fig. 2.1.57). After that, we consider an horizontal line x that
doesn’t pass through P1 , neither through G2 , neither through G3 , neither through
G4 (Fig. 2.1.58).

Fig. 2.1.55

Fig. 2.1.56

Fig. 2.1.57
156 2 The Problem of Saint Venant

Fig. 2.1.58

Evidently, denoting with X the antipole of x, with x0 the centroidal line parallel
to x and with y0 the line passing through X and G, (x0 , y0 ) is a couple of con-
jugate diameters of the inertia centroidal ellipse of the assigned system of areas
(Fig. 2.1.59).
To determine X, we orientate x and arbitrarily choose a point O of x. After that
we conduct from O the perpendicular one to x, that we denote with the symbol t and
that we orientate (Fig. 2.1.60). Then, in the graphical way already seen in Problem
2.1.9, we determine the antipole X2 [resp. X3 ] [resp. X4 ] of x with respect to the
inertia centroidal ellipse of the subarea 2 [resp. 3] [resp. 4] (Fig. 2.1.61).

Fig. 2.1.59

Fig. 2.1.60

Fig. 2.1.61
2.1 Geometry of Areas 157

We now set, in the frame of reference 0, x, t of Fig. 2.1.60

     
G = (xG , tG ) , X = (xX , tX ) P1 = xp1 , tp2 , G2 = xG2 , tG2 , G3 = xG3 , tG3 ,
       
G4 = xG4 , tG4 , X2 = xX2 , tX2 , X3 = xX3 , tX3 , X4 = xX4 , tX4 ,

and denote with A2 [resp. A3 ] [resp. A4 ] the area of the subarea 2 [resp. 3] [resp. 4].
We have
Ix (Ix )1 + (Ix )2 + (Ix )3 + (Ix )4
tX = =
Sx Atot tG
A1 tP2 1 + (Sx )2 tX2 + (Sx )3 tX3 + (Sx )4 tX4
=
Atot tG
A1 tP1 + A2 tG2 tX2 + A3 tG3 tX3 + A4 tG4 tX4
2
=
Atot tG

where (Sx )i [resp. (Ix )i ] denote the static [resp. inertia] moment with respect to axis
x of the area Ai . Besides Sx denotes the static moment with respect to axis x of the
whole system of areas and Atot = A1 + A2 + A3 + A4 .
Analogously

Ixt (Ixt )1 + (Ixt )2 + (Ixt )3 + (Ixt )4

xX = =
Sx Atot tG
A1 tP1 xP1 + (Sx )2 xX2 + (Sx )3 xX3 + (Sx )4 xX4
=
Atot tG
A1 tP1 xP1 + A2 tG2 xX2 + A3 tG3 xX3 + A4 tG4 xX4
=
Atot tG

where (Ixt )i denotes the centrifugal moment with respect to axes x, t of the area Ai .
Having so determinated the coordinates of X, we know the couple (x0 , y0 ) of
conjugate diameters of the inertia centroidal ellipse of the assigned system of areas
(Fig. 2.1.59).
We have now to determine the semidiameters ρxo , ρyo (Fig. 2.1.62). About ρxo
from the (2.1.10) we have (Fig. 2.1.63)
1
ρxo = [lG (lX − lG )] 2 .

Fig. 2.1.62
158 2 The Problem of Saint Venant

Fig. 2.1.63

Fig. 2.1.64

To calculate ρy0 we consider a line y parallel to y0 not passing through G2 neither

through G3 neither through G4 (Fig. 2.1.64). Then, in the graphical way already seen
in Problem 2.1.9, we determine the antipole Y2 [resp. Y3 ] [resp. Y4 ] of y with respect
to the inertia centroidal ellipse of the subarea 2 [resp. 3] [resp. 4] (Fig. 2.1.65). Then
we have (Fig. 2.1.66)

Iy = A1 d12 + A2 d2 h2 + A3 d3 h3 + A4 d4 h4 .

Finally (Fig. 2.1.67)

1  1
Iy 0 2 Iy − Atot dG2 2
ρyo = = . 
Atot sin2 α Atot sin2 α

Fig. 2.1.65
2.1 Geometry of Areas 159

Fig. 2.1.66

Fig. 2.1.67

2.1.4 Inertia Centroidal Kernel

Let C be a closed curve, A be the region of the plane contained by C (Fig. 2.1.68).
We say that C is convex if for any two points P and Q of C the segment PQ is
contained in A. The closed curve of Fig. 2.1.69 is not convex.
We consider a system of distributed and/or concentrated areas (Fig. 2.1.70).
We denote with C the convex closed curve containing the system of areas and

Fig. 2.1.68

Fig. 2.1.69
160 2 The Problem of Saint Venant

Fig. 2.1.70

Fig. 2.1.71

having smallest length. We denote with A the region of the plane contained by C
(Fig. 2.1.71). We denote with W the set of the non centroidal lines whose intersec-
tion with A is a point or void. We call inertia centroidal kernel the set of the antipoles
of the lines belonging to W.
We can prove that
– the boundary of the kernel is the set of the antipoles of the tangents to A,
– the boundary of the kernel is a convex closed curve,
– the centroid is inside the kernel.

Problem 2.1.13 You should determine the inertia centroidal kernel of the rectangle
of Fig. 2.1.72.
Solution Evidently the convex closed curve containing the system of areas and
having the smallest length is the contour of the rectangle. We have therefore to
determine the set of the antipoles of the tangents to the contour of the rectangle.
Evidently the tangents to the contour of the rectangle are the lines a1 , a2 , a3 , a4
of Fig. 2.1.75 as well as the lines of the type b1 , b2 , b3 , b4 of Fig. 2.1.73. Of here
and from the [2.1.7] it follows that the contour of the kernel is the contour of the
rhombus of Fig. 2.1.74, where A1 denotes the antipole of a1 . To determine A1 we
consider the centroidal line a parallel to a1 and denote with x the distance of A1

Fig. 2.1.72
2.1 Geometry of Areas 161

Fig. 2.1.73

Fig. 2.1.74

Fig. 2.1.75

from a (Fig. 2.1.75). Clearly A1 belongs to the line normal to a1 and passing through
G and

h Ia b h3
x = ρa2 = =
2 bh 12 b h
so that x = h/6. Insofar the inertia centroidal kernel of the rectangle is the rhombus
of Fig. 2.1.76. 

Problem 2.1.14 You should determine the inertia centroidal kernel of the circle of
Fig. 2.1.77.

Fig. 2.1.76

Fig. 2.1.77
162 2 The Problem of Saint Venant

Fig. 2.1.78

Solution Evidently the convex closed curve containing the system of areas and
having the smallest length is the circumference of the circle. We have therefore
to determine the set of the antipoles of the tangents to the circumference. Since
the antipole of a tangent a to the circumference is on the centroidal line normal
to the tangent (Fig. 2.1.78), it follows that the inertia centroidal kernel is a circle.
To determine its radius r, we consider the centroidal line a0 parallel to a. We have
(Fig. 2.1.78)

Iao π R4
rR = ρa2o = =
π R2 4 π R2
so that r = π/4. 

Problem 2.1.15 You should determine the inertia centroidal kernel of the T-section
of Fig. 2.1.79.

Solution It is easy to verify that the requested inertia centroidal ellipse and the
requested inertia centroidal kernel are those of Fig. 2.1.80. 

Problem 2.1.16 You should determine the inertia centroidal kernel of the system of
concentrated areas of Fig. 2.1.81.

Fig. 2.1.79

Fig. 2.1.80
2.1 Geometry of Areas 163

Fig. 2.1.81

Fig. 2.1.82

Solution Evidently the requested inertia centroidal kernel is a quadrilateral whose

vertexes B1 , B2 , B3 , B4 are the antipoles of the tangents b1 , b2 , b3 , b4 (Fig. 2.1.82).
We denote with P1 [resp. P2 ] [resp. P3 ] [resp. P4 ] the point in which the area
A1 = 100 cm2 [resp. A2 = 200 cm2 ] [resp. A3 = 300 cm2 ] [resp. A4 = 400 cm2 ] is
concentrated. We arbitrarily orientate b1 , b2 , b3 , b4 (Fig. 2.1.82). To determine B1 ,
with reference to the axes b1 , b2 we have

Ib1 = Sb1 hB1 , Ib1 b2 = Sb1 kB1

from which

hB1 = 30 cm, kB1 = 22.9 cm

so that B1 is found (Fig. 2.1.83). About B2 we have

Fig. 2.1.83
164 2 The Problem of Saint Venant

hB2 = 40 cm, kB2 = −6 cm,

about B3 we have

hB3 = 30 cm, kB3 = −26.7 cm,

about B4 we have

hB4 = 40 cm, kB4 = −18 cm.

So the the requested inertia centroidal kernel is the quadrilateral of Fig. 2.1.83. 

2.2 The Problem of Saint Venant

2.2.1 Introduction

We consider in a plane α an area A that we call cross section. We denote with G

its centroid and with z the centroidal line normal to α. We call z axis of the beam
(Fig. 2.2.1). We denote with V the volume described from A when we translate G
along a segment of z of length l, maintaining A normal to z. We call a solid having
such volume V beam of Saint Venant2.2.1 and l length of the beam (Fig. 2.2.2).
To study the behavior under load of the beam of Saint Venant, we adopt a carte-
sian orthogonal frame of reference G, x, y, z. We choose as its origin the centroid G

Fig. 2.2.1

Fig. 2.2.2

2.2.1 Adhémar Jean Claude Barré, count of Saint Venant, Filliers en Brie 1797 – Saint Ouen 1886.
2.2 The Problem of Saint Venant 165

Fig. 2.2.3
G

Fig. 2.2.4
G

of the terminal cross section of left and as axis z the axis of the beam (Fig. 2.2.3).
As we will see, for some load conditions, the axes x, y must be principal diameters
of the inertia centroidal ellipse of the cross section (Fig. 2.2.4).
The material of which the solid is made is supposed homogeneous, isotropic and
linearly elastic. The deformations are, for hypothesis, small.
We call bases of the beam of Saint Venant the terminal cross sections left and
right and lateral surface of the beam the boundary of V deprived of the bases
(Fig. 2.2.3).
The solid is not constrained, so that S = Sp .
About the load, we suppose the volumetric load X, Y, Z identically zero and the
superficial load only distributed.2.2.2
We suppose that the superficial load px , py , pz distributed on the lateral surface
is identically zero. So the load applied on the solid is a distributed superficial load
applied only on the bases of the beam. We suppose that such load is a system of
forces equivalent to zero.
Such problem is an elastic equilibrium problem which we call problem of Saint
Venant.
The Saint Venant found the exact solution of six of such problems.
The first problem of Saint Venant is the centroidal axial load. The load px , py ,
pz distributed on the right base is equivalent to the system constituted by an only

2.2.2 The presence of a concentrated load imposes the employment of the Functional analysis. We
notice that a concentrated load can be approximate with a distributed load with precision high as
we want. We also notice that a concentrated load is a model useful for the calculation but not
verifiable in the physical reality.
166 2 The Problem of Saint Venant

Fig. 2.2.5

force N having for line of action the axis z (Fig. 2.2.5). The load applied on the left
base constitutes, together with that applied on the right base, a system equivalent
to zero.
The second [resp. third] problem of Saint Venant is the right bending of axis x
[resp. y]. The load px , py , pz distributed on the right base is equivalent to the system
constituted by an only couple Mx [resp. My ] having axis2.2.3 x [resp. y] (Fig. 2.2.6
[resp. 2.2.7]). The load applied on the base of left constitutes, together with that
applied on the base of right, a system equivalent to zero.
The fourth problem of Saint Venant is the torsion. The load px , py , pz distributed
on the right base is equivalent to the system constituted by an only couple Mt having
axis z (Fig. 2.2.8). The load applied on the base of left constitutes, together with that
applied on the base of right, a system equivalent to zero.

Fig. 2.2.6

Fig. 2.2.7

Fig. 2.2.8

2.2.3 The axis of a couple is the line normal to the plane on which the couple acts.
2.2 The Problem of Saint Venant 167

Fig. 2.2.9
G

The fifth [resp. sixth] problem of Saint Venant is the shear of axis x [resp. y]. The
load px , py , pz distributed on the right base is equivalent to the system constituted
by an only force Tx [resp. Ty ] having for line of action the axis x [resp. y] (Fig. 2.2.9
[resp. 2.2.10]). The load applied on the base of left constitutes, together with that
applied on the base of right, a system equivalent to zero.
We assume the following postulate of Saint Venant.
[2.2.1] We consider two problems, constituted by a solid of Saint Venant submitted
on the bases to two different load conditions. If the superficial loads applied on the
right base in the two problems are equivalent, then both the problems admit the
same solution everywhere except in a small region near to the bases. We call such
region zone of extinction2.2.4 (Fig. 2.2.11). 

Fig. 2.2.10

Fig. 2.2.11

2.2.4 Precisely, the distance d is supposed small with respect to the diameter of the smallest circle
containing the cross section.
168 2 The Problem of Saint Venant

Remark 2.2.1 In the hypothesis of the postulate of Saint Venant both the problems
admit the same solution u,v,w practically everywhere. They are also consequently
equal the partial derivatives of u,v,w. As a consequence, since the (1.1.28), both the
problems admit the same state of strain εx, εy, εz, γxy, γxz, γyz, practically everywhere.
As a consequence, since the (1.4.5), both the problems admit the same state of stress
σx, σy, σz, τxy, τxz, τyz, practically everywhere. 

Let us consider now a problem of Saint Venant, which we call problem S.

Obviously the superficial load px , py , pz applied on the right base z = l is equivalent
to the system constituted by the three forces2.2.5 (Fig. 2.2.12)
  
N= pz dA, Tx = px dA, Ty = py dA,
A A A

and by the three couples2.2.6

  
 
Mx = pz y dA, My = − pz x dA, Mt = py x − px y dA.
A A A

In such system N [resp. Tx ] [resp. Ty ] has line of action z [resp. x] [resp. y] and
Mx [resp. My ] [resp. Mt ] has axis x [resp. y] [resp. z] (Fig. 2.2.12).
Since the postulate of Saint Venant [2.2.1], the solution of the problem of Saint
Venant S coincides practically everywhere with the solution of the problem of Saint
Venant of Fig. 2.2.14.
Since the principle of superposition [1.5.4], the solution of the problem of Saint
Venant of Fig. 2.2.14 is a linear combination of the solutions of the first prob-
lem of Saint Venant (Fig. 2.2.5), of the second one (Fig. 2.2.6), of the third one
(Fig. 2.2.7), of the fourth one (Fig. 2.2.8), of the fifth one (Fig. 2.2.9) and of the
sixth one (Fig. 2.2.10). This way, the solutions of the six fundamental problems of

Fig. 2.2.12

2.2.5 A denotes the area of the cross section of the beam.

2.2.6 We assume positive the anticlockwise moments. About Mt , we refer to Fig. 2.2.13.
2.2 The Problem of Saint Venant 169

Fig. 2.2.13

Fig. 2.2.14

Saint Venant give immediately the solution of any problem S of Saint Venant by
linear combination, everywhere except in the zone of extinction.

2.2.2 State of Stress

We will see that in the six fundamental problems of Saint Venant it always results
identically in V

σx = 0, σy = 0, τxy = 0. (2.2.1)

From this some properties of the state of stress follow that are valid for each one
of the six fundamental problems of Saint Venant and then, as previously seen, for
each problem of Saint Venant.
Let P be any point of the solid of Saint Venant and α be any plane containing the
line α passing through P and parallel to the axis z (Fig. 2.2.16). We denote with n the
170 2 The Problem of Saint Venant

Fig. 2.2.15
G

normal to α passing through P and with nx , ny , nz its direction cosines (Fig. 2.2.15).
It is obvious that nz = 0. We denote with m the normal to n and z passing through
P and with mx , my , mz its direction cosines (Fig. 2.2.15). It is obvious that mz = 0.
Insofar

σn = 0, τnm = 0.

Consequently on α it results (Fig. 2.2.16)

tn = τnz .

Another property of the state of stress implicated from (2.2.1) is the following.
Let P be an arbitrary point of the volume V occupied from the solid of Saint Venant.
It follows from (2.2.1)

σx (P) τxy (P) τxz (P)

τyx (P) σy (P) τyz (P) = 0
τzx (P) τzy (P) σz (P)

and as consequence the state of stress in P is plane.

To determine the plane of stress is enough, since the [1.2.11], to individualize
an unloaded plane passing through P. We denote with z the line parallel to z and
passing through P. We consider the cross section passing through P, the relative
stress vector

tz (P)

Fig. 2.2.16
2.2 The Problem of Saint Venant 171

and its component in the plane of the cross section

τz (P) .

We denote with r the line of action of such component and with s the normal to
r and z (Fig. 2.2.17). Obviously

τz s = 0. (2.2.2)

It follows that the plane β passing through P and individualized by the lines
z and r is unloaded. In fact the normal one to β is s. Moreover, since β contains
the parallel one to z passing through P, ts is parallel to z. Reasoning for absurd, we
suppose ts = 0. Well, since ts lies in β and has line of action z , on β a τsz = 0 exists.
Because of the symmetry of the τ , on the plane of normal z a τz exists having
component on s equal to τsz . Insofar

τz s = τsz = 0.

Since the (2.2.2) this is absurd. Then the plane β is unloaded. Insofar the plane
of stress in P is the plane passing through P, parallel to z and individualized
by τ z (P).
In the problem of Saint Venant the first two of the (1.2.3) are identically zero.
The third one, since the (2.2.1), furnishes

∂τzx ∂τzy
= 0, = 0.
∂z ∂z

P
G

Fig. 2.2.17
172 2 The Problem of Saint Venant

As a consequence the functions τzx , τzy are constant with respect to z. Then the
distribution of the shearing stresses is the same one in every cross section of the
beam.2.2.7
A last important property common to the six problems of Saint Venant is the
following. Let (Fig. 2.2.18)

– Q be a point of the lateral surface of the beam,

– α be the plane passing through Q and tangent to the beam,
– n be the normal one to α passing through Q and orientated going out from the
material,
– nx , ny , nz the direction cosines of n,
– n the unit vector of n.

Clearly nz = 0. Insofar, keeping in mind of the hypothesis that the lateral surface
is unloaded, the third of the (1.2.5) becomes

τzx nx + τzy ny = 0.
As a consequence
τz × n = 0 (2.2.3)

and this implicates (Fig. 2.2.19) that τ z is orthogonal to n. Insofar in every problem
of Saint Venant in any point Q of the lateral surface of the beam the line of action
of the τ z is the line t passing through Q and tangent to the boundary of the cross
section (Fig. 2.2.19).

Fig. 2.2.18

Fig. 2.2.19

2.2.7 This implicates that the points of a parallel to z have the same plane of stress.
2.3 Bending 173

2.3 Bending

2.3.1 Right Bending of Axis x

We analyze the second problem of Saint Venant, the right bending of axis x
(Fig. 2.2.6). The axes x, y of the frame of reference must rigorously be inertia prin-
cipal axes (Fig. 2.3.1). The cross section of the beam can be both monoconnected
and multiconnected. We denote with V the volume occupied by the solid of Saint
Venant and consider the functions

σx = 0 in V

σy = 0 in V

Mx
σz = y in V
Ix
(2.3.1)
τxy = 0 in V

τxz = 0 in V

τyz = 0 in V

where the symbol Mx denotes a non zero real number and the symbol Ix denotes the
moment of inertia of the cross section A with respect to axis x (Fig. 2.3.1).
To prove that the (2.3.1) is the unique solution of the second problem of Saint
Venant, as seen in Sect. 1.5 we must verify that such state of stress satisfies

• the inside compatibility, that is the (1.1.29),

• the outside compatibility, that is the conditions that the constraints impose to the
points of Su ,
• the inside equilibrium, that is the (1.2.3),
• the boundary equilibrium, that is the (1.2.5) in the points of Sp .

Fig. 2.3.1
174 2 The Problem of Saint Venant

Since the (1.4.1), the state of strain associated to (2.3.1) is

νMx
εx = − y in V
EIx
νMx
εy = − y in V
EIx
Mx (2.3.2)
εz = y in V
EIx
γxy = 0 in V
γxz = 0 in V
γyz = 0 in V.

Insofar the functions εx , εy , εz , γ xy , γ xz , γ yz are constant or linear in V. This

implicates that their second order partial derivatives are all identically zero in V.
The Eq. (1.1.29) of the inside compatibility are therefore banally satisfied.
About the outside compatibility there are no conditions to satisfy because for
hypothesis the beam is not constrained.
About the inside equilibrium we notice that

– for hypothesis X = Y = Z = 0 in V,
– since the functions σ x , σ y , τ xy , τ xz , τ yz are constant, their first order partial
derivatives are all identically zero in V,
– since the function σ z is constant with respect to z, its first order partial derivative
with respect to z is identically zero in V.

Insofar the (1.2.3) are banally satisfied.

At this point the six functions (2.3.1) constitute a state of stress physically
possible. We find the load to apply on the beam to produce it.
About the volumetric load, for hypothesis X = Y = Z = 0 in V.
About the superficial load, let (Fig. 2.2.18)

– Q be a point of the lateral surface,

– α the plane passing through Q and tangent to the lateral surface,
– n the line passing through Q, normal to α and directed going out from the material,
– nx , ny , nz the direction cosines of n.

Evidently nz = 0. As a consequence, from the (2.3.1) and (1.2.5) it follows

px (Q) = 0, py (Q) = 0, pz (Q) = 0.

Insofar the lateral surface is unloaded.

So the state of stress (2.3.1) is physically possible and to establish it we must
apply a load only on the bases of the beam. Insofar the problem that we are analyzing
is a problem of Saint Venant.
2.3 Bending 175

We now see that a load must be applied on the right base z = l to establish in
the solid the state of stress (2.3.1). Let Q be any point of the right base z = l. We
observe that the plane passing through Q and tangent to the surface of the beam is
obviously the base z = l. Moreover the line n passing through Q, normal to such
plane and directed going out from the material is clearly parallel and equiverse to z.
Then its direction cosines nx , ny , nz are

nx = 0, ny = 0, nz = 1.

As a consequence, from the (2.3.1) and (1.2.5) it follows

Mx
px (Q) = τxz = 0, py (Q) = τyz = 0, px (Q) = σz = y.
Ix
Insofar the superficial load to apply on the right base is linearly distributed, par-
allel to the axis z and equiverse to z in the points of the right base having positive
coordinate y. Clearly this load is equivalent to a system constituted to the most by
a force N having line of action z, by a couple Mx having axis x and by a couple My
having axis y. Since the static moment with respect to a centroidal axis is zero, we
easily obtain (Fig. 2.3.1)

  
Mx Mx Mx
N= pz dA = y dA = y dA = Sx = 0.
A A Ix Ix A Ix

The value of the couple Mx having axis x is equal to (Fig. 2.3.2)

   
Mx 2 Mx
Mx = dMx = (pz dA) (y) = y dA = y2 dA
A A A Ix Ix A

Mx
= Ix = Mx
Ix

so that the constant Mx fixed in the (2.3.1) is exactly the intensity of the couple of
axis x to apply on the right base (Fig. 2.3.3).

Fig. 2.3.2
176 2 The Problem of Saint Venant

Fig. 2.3.3
G

About My , we observe that x, y are conjugate diameters of the inertia centroidal

ellipse of the cross section A. Insofar, since the [2.1.13], the centrifugal moment Ixy
is zero. Insofar (Fig. 2.3.2)
   
Mx Mx
My = dMy = (pz dA) (x) = y x dA = x y dA
A A A Ix Ix A

Mx
= Ixy = 0.
Ix

So we have found that the distributed superficial load to apply on the right base
is equivalent to a system constituted only by a couple of axis x and of intensity Mx
(Fig. 2.3.3). Such load produces in the beam the state of stress (2.3.1).
We find now the load to apply on the left base z = 0. Let Q be a point of the
base z = 0, n be the normal to the plane tangent in Q to the surface of the beam.
Obviously

nx = 0, ny = 0, nz = −1.

Insofar

Mx
px (Q) = −τxz = 0, py (Q) = −τyz = 0, pz (Q) = −σz = − y.
Ix
Proceeding as for the right base, we found that such load is equivalent to the sys-
tem constituted only by a couple of axis x and of intensity Mx . Its verse is opposite
to one of the couple applied on the right base (Fig. 2.3.4).
In conclusion the (2.3.1) is the unique solution of the second problem of Saint
Venant (Fig. 2.2.6). In particular we call the third of the (2.3.1),

Mx
σz = y, (2.3.3)
Ix
2.3 Bending 177

Fig. 2.3.4
G

formula of Navier for the right bending of axis x. Its graphic representation, that we
call diagram of the σ z , is given in Fig. 2.3.5. Since σ z is constant with respect to z,
such diagram of the σ z repeats itself in all the cross sections.
We call Mx bending moment of axis x.
We call neutral axis of the bending, and denote with the symbol n, the set of the
points of the cross section such that σ z = 0. Clearly in the problem in examination
n = x. Obviously all the points of a parallel AB to n have σ z constant, with value
represented by the segment HK of the diagram (Fig. 2.3.5).
Let us see now how the beam deforms it (Fig. 2.3.6). Evidently, to obtain the field
of the displacements u, v, w we must integrate the (1.1.28) and (2.3.2). Clearly, the
hypothesis that the beam is free in the space implies that we can determine u, v, w
apart from a rigid translation and a rigid rotation.
Nevertheless to be clearer we prefer to do such analysis supposing the beam
constrained with a punctual rigid joint in the centroid G of the left base z = 0, i.e.
subject to the conditions

Fig. 2.3.5

Fig. 2.3.6
178 2 The Problem of Saint Venant

u (0, 0, 0) = 0, v (0, 0, 0) = 0, w (0, 0, 0) = 0,

∂u ∂v ∂v ∂u (2.3.4)
(0, 0, 0) = 0, (0, 0, 0) = 0, − (0, 0, 0) = 0.
∂z ∂z ∂x ∂y

First of all, from the (1.1.28) and (2.3.2) we have

∂u νMx
=− y, (2.3.5)
∂x EIx
∂v νMx
=− y, (2.3.6)
∂y EIx
∂w Mx
= y, (2.3.7)
∂z EIx
∂u ∂v
+ = 0, (2.3.8)
∂y ∂x
∂u ∂w
+ = 0, (2.3.9)
∂z ∂x
∂v ∂w
+ = 0. (2.3.10)
∂z ∂y

From the (2.3.6), by integration, we obtain

νMx 2
v (x, y, z) = − y + f (x, z) . (2.3.11)
2EIx

From the (2.3.11) and (2.3.8) we obtain

∂u ∂v ∂f
=− = − (x, z) . (2.3.12)
∂y ∂x ∂x

From the (2.3.6) and (2.3.10) we obtain

∂w ∂v ∂f
=− = − (x, z) . (2.3.13)
∂y ∂z ∂z

From the (2.3.7) by integration we obtain

∂f
u (x, y, z) = −y (x, z) + g (x, z) . (2.3.14)
∂x
From the (2.3.13) by integration we obtain

∂f
w (x, y, z) = −y (x, z) + h (x, z) . (2.3.15)
∂z
2.3 Bending 179

From the (2.3.14) and (2.3.5) we obtain

∂ 2f ∂g νMx
−y (x, z) + (x, z) = − y
∂x 2 ∂x EIx
so that the functions f, g must be such that

∂ 2f νMx ∂g
− (x, z) + y+ (x, z) = 0 ∀ (x, y, z) ∈ V.
∂x 2 EIx ∂x

As a consequence it must result

∂ 2f νMx
− (x, z) + =0 (2.3.16)
∂x2 EIx

and
∂g
(x, z) = 0. (2.3.17)
∂x
From the (2.3.15) and (2.3.7) we obtain

∂ 2f ∂h Mx
−y (x, z) + (x, z) = y
∂z2 ∂z EIx
so that the functions f, h must be such that

∂ 2f Mx ∂h
− (x, z) − y+ (x, z) = 0 ∀ (x, y, z) ∈ V.
∂z2 EIx ∂z

As a consequence it must result

∂ 2f Mx
(x, z) + =0 (2.3.18)
∂z2 EIx

and
∂h
(x, z) = 0. (2.3.19)
∂z
From the (2.3.17) it follows

g (x, z) = g1 (z) (2.3.20)

and from the (2.3.19) it follows

h (x, z) = h1 (x) . (2.3.21)

180 2 The Problem of Saint Venant

The (2.3.9), (2.3.14), (2.3.15), (2.3.20), and (2.3.21) furnish

∂u ∂w ∂ 2f ∂g ∂ 2f
0= + = −y (x, z) + (x, z) − y (x, z)
∂z ∂x ∂x∂z ∂z ∂x∂z
∂h ∂ 2f dg1 dh1
+ (x, z) = −2y (x, z) + (z) + (x) .
∂x ∂x∂z dz dx

Consequently ∀ (x, y, z) ∈ V

∂ 2f
(x, z) = 0 (2.3.22)
∂x∂z
dg1 dh1
(z) + (x) = 0. (2.3.23)
dz dx
The (2.3.23) implicates that a real number A exists such that

dg1
(z) = A (2.3.24)
dz
and

dh1
(x) = −A. (2.3.25)
dx
From the (2.3.24) [resp. (2.3.25)] it follows that a real number B [resp. C] exists
such that

g1 (z) = Az + B (2.3.26)

h1 (x) = −Ax + C. (2.3.27)

From the (2.3.20) and (2.3.26) it follows

g (x, z) = Az + B. (2.3.28)

From the (2.3.21) and (2.3.27) it follows

h (x, z) = −Az + C. (2.3.29)

From the (2.3.16) by integration we obtain

∂f νMx
= x + f2 (z)
∂x EIx
2.3 Bending 181

from which by integration we obtain

νMx 2
f (x, z) = x + xf1 (z) + f2 (z) . (2.3.30)
2EIx
From the (2.3.30), (2.3.18), and (2.3.22) it follows

Mx ∂ 2f d 2 f1 d 2 f2
= − 2 (x, z) = −x 2 (z) − 2 (z)
EIx ∂z dz dz
and

∂ 2f df1
0= (x, z) = (z)
∂x∂z dz
so that

d 2 f2 Mx
(z) = −
dz2 EIx
and
df1
(z) = 0.
dz
From this by integration we get that three real number D, G, L exist such that

Mx 2
f2 (z) = − z + Dz + G (2.3.31)
2EIx
and
f1 (z) = L. (2.3.32)

From the (2.3.14), (2.3.30), (2.3.32), and (2.3.28) it follows

νMx
u (x, y, z) = −y x + L + Az + B. (2.3.33)
EIx
From the (2.3.11), (2.3.30), (2.3.32), and (2.3.31) it follows

νMx 2 νMx 2
v (x, y, z) = − y + x + xL
2EIx 2EIx
(2.3.34)
Mx 2
− z + D z + G.
2EIx
From the (2.3.15), (2.3.30), (2.3.31), (2.3.32), and (2.3.29) it follows

Mx
w (x, y, z) = −y − z + D − Ax + C. (2.3.35)
EIx
182 2 The Problem of Saint Venant

Imposing the conditions (2.3.4) we have

0 = u (0, 0, 0) = B, 0 = v (0, 0, 0) = G, 0 = w (0, 0, 0) = C.

∂u ∂v ∂v ∂u
0= (0, 0, 0) = A, 0 = (0, 0, 0) = D, 0 = − (0, 0, 0)
∂z ∂z ∂x ∂y

νMx νMx
= − x+L+ x + L (0, 0, 0) = 2L.
EIx EIx

From this and from the (2.3.33), (2.3.34), and (2.3.35) it follows

νMx
u =− xy
EIx

Mx  2 
v = νx − νy2 − z2 (2.3.36)
2EIx

Mx
w= yz .
EIx

The (2.3.36) allows to understand how the beam deforms it. We easily can verify
that

– during the deformation the cross sections preserve its plane,

– the points of the beam that were found on the axis z before the application of the
load are found, to happened deformation, on a arc of parabola (Fig. 2.3.6),
– since the hypothesis of small deformations, such arc of parabola can be approxi-
mate with an arc of circumference,  
– such arc of circumference lies in the plane y, z and has center Q = 0, − EI x
Mx , 0 .

Obviously such arc of circumference has radius (Fig. 2.3.7)

1 Mx
= . (2.3.37)
r EIx

We call flexural rigidity of the beam the positive real number EIx . Clearly the
flexural rigidity is measured in kg/cm2 . We see from the (2.3.37) that greater the
flexural rigidity is, less the beam deforms it.
We also underline that, since the (2.3.36)

∂ 2v Mx
=− . (2.3.38)
∂z2 EIx
2.3 Bending 183

Fig. 2.3.7

Moreover with the (2.3.36) we easily can verify that (Fig. 2.3.7)

– the lines of the plane

 y, z parallel
 to axis z turn into arcs of circumference that still
EIx
have center Q = 0, − Mx , 0 ,
– the lines of the plane y, z parallel to axis y turn into lines passing through Q.

Let us consider now the deformation of the generic cross section. Using the
(2.3.36) we can easily verify that the lines of the plane
 x, z parallel
 to axis x turn into
EIx
arcs of circumference that always have center T = 0, − νMx , 0 . Moreover the lines
of the plane x, z parallel to axis y turn into lines all passing through T (Fig. 2.3.8).
In the problem of bending we call

Fig. 2.3.8
184 2 The Problem of Saint Venant

– load plane the centroidal plane where lie the couples acting on the bases of the
beam,
– load axis s the intersection between load plane and cross section,
– flexural plane the plane where lies the axis of the beam to happened deformation,
– flexural axis f the intersection between flexural plane and cross section.

Clearly in the problem in examination, which is the right bending of axis x, we

have that

– the load plane is the plane y, z,

– the load axis s is the axis y,
– the flexural plane is the plane y, z,
– the flexural axis f is the axis y.

We notice that in the problem in examination the load axis s and the neutral axis
n are a couple of conjugate diameters of the inertia centroidal ellipse of the cross
section. Precisely, they are the couple of principal diameters.
We notice that in the problem in examination the load plane and the flexural plane
coincide.2.3.1

Remark 2.3.1 The (2.3.3) shows that the σ z is maximum in the points more distant
from the neutral axis. This result suggests the optimal shape for the cross section
subject to bending. Evidently the material needs to be disposed as far away as possi-
ble from the neutral axis. In fact let us consider the I-cross section of Fig. 2.3.9 and
the rectangular cross section of Fig. 2.3.10. We suppose that two sections have equal
area and equal height H. In Fig. 2.3.9 [resp. 2.3.10] the diagram of σ z is shown. We
denote with Īx , M̄x the moment of inertia and the bending moment of the I-cross
section. We denote with I x , M x the moment of inertia and the bending moment of
the rectangular cross section. Making to work in both cases the material with the
same value of the σ zmax , we have

Fig. 2.3.9

2.3.1 It is for this motive that in the problem in examination the bending is said right.
2.3 Bending 185

Fig. 2.3.10

Mx H MxH
=
Ix 2 Ix2
from which

Ix
Mx = Mx.
Ix

Insofar, since I x is much greater than I x , the beam of Fig. 2.3.9 can support a bending
moment much greater than what the beam of Fig. 2.3.10 can support. 
About the structural safety, since the Remark 1.7.3, denoting with s the safety
factor, the safety condition in any point of the beam is

σo
σz ≤ .
s
So the beam of Saint Venant is structural safety if

σ0
σzmax ≤
s
that is, since the (2.3.3), if

Mx σ0
ymax ≤ .
Ix s
We call section modulus and denote with the symbol Wx , the greatness

Ix
Wx = .
ymax
This way the safety condition becomes

Mx
≤ σa . (2.3.39)
Wx

Clearly the section modulus is a geometric characteristic of the cross section. In

the case of standard iron beam, the section modulus is directly furnished as tabulated
data from the steelmaking factories.
186 2 The Problem of Saint Venant

Fig. 2.3.11

Fig. 2.3.12

Problem 2.3.1 The beam of Fig. 2.3.11 is a standard channel iron with h = 18 cm
and σ a = 1,600 kg/cm2 (Fig. 2.3.12) . The load is Mx = 2 tm (Fig. 2.3.12). You
shall verify the structural safety.
Solution Clearly the supports of Fig. 2.3.11 have reaction zero. Then the beam is
loaded only on the bases by bending moments of axis x. Since the axis x is principal
of inertia, it treats of right bending of axis x.
The structural safety verification is performed verifying that

σzmax ≤ σa .

Since (Fig. 2.3.13)

Mx 2 · 105 kg
ymax = · 9 = 1329 2 ≤ 1, 600 = σa
Ix 1, 354 cm
the beam is in structural safety. 

Problem 2.3.2 You shall design the beam of Fig. 2.3.14 choosing a standard iron
with σa = 1, 400 kg/cm2 (Fig. 2.3.15). The load is Mx = 9 tm (Fig. 2.3.15).
Solution Clearly the constraint of Fig. 2.3.14 has reaction −Mx . Then the beam is
loaded only on the bases by bending moments of axis x. Since the axis x is principal
of inertia, it treats of right bending of axis x.

Fig. 2.3.13
2.3 Bending 187

Fig. 2.3.14

Fig. 2.3.15

Since the case is of standard iron beam, we can choose the cross section with the
tabulated data from the steelmaking factories. Since the beam must be in structural
safety, we need of a cross section such that s σzmax ≤ σ0 . So we choose the cross
section HE 240 A (Fig. 2.3.15), having

900, 000
Wx ≥ = 642.9 cm3 .
1, 400

In Fig. 2.3.16 we draw the diagram of σz , where

Mx kg
σzmax = = 1, 333 2 . 
Wx cm

Fig. 2.3.16

2.3.2 Right Bending of Axis y

We analyze the right bending of axis y, the third problem of Saint Venant (Fig. 2.2.7).
The axes x, y of the frame of reference must rigorously be inertia principal
axes (Fig. 2.3.1). The cross section of the beam can be both monoconnected and
multiconnected. We consider the functions
188 2 The Problem of Saint Venant

σx = 0
σy = 0
My
σz = − x
Iy (2.3.40)
τxy = 0
τxz = 0
τyz = 0

where the symbol My denotes a non zero real number and the symbol Iy denotes the
moment of inertia of the cross section A with respect to the axis y (Fig. 2.3.1).
Proceeding exactly as in Sect. 2.3.1, we prove that

– the (2.3.40) is the unique solution of the third problem of Saint Venant,
– the distributed superficial load applied on the right base is equivalent to a system
constituted only by a couple of axis y and of intensity My (Figs. 2.3.17 and 2.3.18),
– the state of strain is

νMy νMy My
εx = x, εy = x, εz = − x, γxy = 0, γxz = 0, γyz = 0.
EIy EIy EIy

We call the third of the (2.3.40),

My
σz = − x, (2.3.41)
Iy

formula of Navier for the right bending of axis y. Its graphic representation, that we
call diagram of the σ z , is given in Fig. 2.3.19. Since σ z is constant with respect to z,
such diagram of the σ z repeats itself in all the cross sections.
We call My bending moment of axis y.

Fig. 2.3.17

Fig. 2.3.18
2.3 Bending 189

Fig. 2.3.19

About the field of the displacements u, v, w we obtain, supposing the beam

constrained with a punctual rigid joint in the centroid G of the left base z = 0

My
u=− (νy2 − νx2 − z2 )
2EIy
νMy
v= xy (2.3.42)
EIy
My
w=− xz.
EIy

From the (2.3.42) we easily get that

– during the deformation the cross section preserve itself plane,

– the deformation of the beam is perfectly analogous to that seen in Sect. 2.3.1.

We call flexural rigidity of the beam the positive real number EIy .
Finally, we easily verify that in the problem in examination, i.e. the right bending
of axis y, we have

– the load plane is the plane x, z,

– the load axis s is the axis x,
– the neutral axis n is the axis y,
– the flexural plane is the plane x, z,
– the flexural axis f is the axis x.

We notice that in the problem in examination the load axis s and the neutral axis
n are a couple of conjugate diameters of the inertia centroidal ellipse of the cross
section. Precisely, they are the couple of principal diameters.
We notice that in the problem in examination the load plane and the flexural plane
coincide.
We call section modulus and denote with the symbol Wy , the greatness

Iy
Wy = .
xmax
190 2 The Problem of Saint Venant

2.3.3 Deviated Bending

We consider the following problem of Saint Venant. The superficial load distributed
on the right base is equivalent to the system constituted only by a couple of intensity
M and acting in a load plane s, z (Fig. 2.3.20). We suppose that the load axis s is not
a principal direction of inertia2.3.2 (Fig. 2.3.21). The superficial load distributed on
the left base is such that the total load applied on the beam is a system equivalent to
zero.
The axes x, y of the frame of reference must rigorously be inertia principal
axes (Fig. 2.3.21). The cross section of the beam can be both monoconnected and
multiconnected.
To analyze this problem we use the principle of superposition and the postulate
of Saint Venant.
Clearly the vector M lies in the plane of the right base and is orthogonal to s
(Fig. 2.3.21). We consider its orthogonal component Mx [resp. My ] on the axis x
[resp. y] (Fig 2.3.22). We easily verify that the system constituted by the subsystem
Mx and by the subsystem My is equivalent to the system only constituted by the
vector M.

Fig. 2.3.20

Fig. 2.3.21

2.3.2 Otherwise the deviate bending is the already examined right bending.
2.3 Bending 191

Fig. 2.3.22

As a consequence, because of the principle of superposition and of the postulate

of Saint Venant, anywhere except in the zone of extinction, the unique solution of
the problem in examination is the sum of the (2.3.1) and of the (2.3.40).
So the deviated bending is the sum of two right bending. In particular, we have
the binomial formula of the σ z in a point P = (x, y) of A (Fig. 2.3.23)

Mx My
σz = y− x. (2.3.43)
Ix Iy

In the Sect. 2.3.1 we have called neutral axis of the bending, and denotate with
the symbol n, the set of the points of the cross section such that σ z = 0. Clearly,
since the (2.3.43), in the problem in examination the neutral axis n is the line of
equation (Fig. 2.3.24)

Mx My
y− x = 0. (2.3.44)
Ix Iy

We expressly observe that also in the deviated bending, as it already happened

in the right bending, the neutral axis n is centroidal.

Fig. 2.3.23

Fig. 2.3.24
192 2 The Problem of Saint Venant

About the structural safety, since the Remark 1.7.3, denoting with s the safety
factor, the safety condition in any point of the beam is
σ0
σz ≤ = σa .
s
So we must verify that the maximum of the function (2.3.43) is not greater than
σ a.
To easily calculate such maximum we remember from the Analytic geometry the
following property.
Let

– O, x, y be an orthogonal cartesian frame of reference (Fig. 2.3.25),

– a, b, c be three real numbers such that a2 + b2 > 0,
– r be the line of equation a x + b y + c = 0,
– α, β be the two half plane in which r divides the plane (Fig. 2.3.25).

It is known that the possible cases are the two following

 
1 ∀ (xα , yα ) ∈ α axα + byα + c > 0 and ∀ xβ , yβ  ∈ β axβ + byβ + c < 0
2 ∀ (xα , yα ) ∈ α axα + byα + c < 0 and ∀ xβ , yβ ∈ β axβ + byβ + c > 0.

We call positive the half plane in which ax + by + c assumes positive values.

Let now

– Q = (x̄, ȳ) be any point of the plane (Fig. 2.3.26),

– n be the normal to r passing through Q,
– I be the point common to n and r,
– dr = dist(Q, r) = dis(Q, I) [resp. −dis(Q, r) = −dist(Q, I)] if Q belongs [resp.
belongs not] to the positive half plane.

Fig. 2.3.25

Fig. 2.3.26
2.3 Bending 193

With such notations we easily have

[2.3.1] A real number k2.3.3 exists such that

∀P = (x, y) ∈ 2 ax + by + c = kdr . 

From the [2.3.1], (2.3.43) it follows that a real number k exists such that for every
point P = (x, y) of the cross section it results

σz = kdn (2.3.45)

where dn denotes the orthogonal distance of P from the neutral axis n (Fig. 2.3.27).
To find an operational expression of k, we observe that the system of the σ z dis-
tributed on the generic cross section must be equivalent2.3.4 to the system constituted
by the alone vector M. We denote

– with Ms the orthogonal component of M on the load axis s,

– with Mn the orthogonal component of M on the neutral axis n,
– with ds the orthogonal distance of P from the load axis s.

We can easily prove that the superficial load px , py , pz distributed on the right
base of the beam has moment Ms [resp. Mn ] with respect to axis s [resp. n].
Moreover, since by definition M is normal to axis s, we have Ms = 0.
As a consequence, it must be

Mn = (σz dA) (dn )
A

0 = Ms = (σz dA) (ds )
A

from which

Fig. 2.3.27

2.3.3 We easily have k = (a2 + b2 )1/2 .

2.3.4 It is enough to consider the equilibrium of the part [z, l] of the beam.
194 2 The Problem of Saint Venant
  
Mn = σz dn dA = kdn2 dA =k dn2 dA = kIn (2.3.46)
A A A

  
0= σz ds dA = kdn ds dA = k dn ds dA = kIns (2.3.47)
A A A

where In [resp. Ins ] is the inertia [resp. centrifugal] moment of the area A of the cross
section with respect to the axis n [resp. n and s].
From the (2.3.45) and (2.3.46) the formula of Navier of the deviated bending
follows

Mn
σz = dn . (2.3.48)
In

Its graphic representation, that we call diagram of the σ z , is given in Fig. 2.3.28.
We see that the points of any parallel to n always have the same value of σ z .
Moreover, since σ z is constant with respect to z, such diagram of the σ z repeats
itself in all the cross sections. We underline that in the points of the neutral axis n it
results σ z = 0.
We see from the (2.3.48) that the maximum value of σ z is assumed in the most
distant points from the neutral axis n.
Obviously we suppose M = 0. Then, since the (2.3.45), k = 0. Then, since the
(2.3.47)

Ins = 0.

From this, since the [2.1.14], we get that in the deviated bending the neutral axis n
and the load axis s are conjugate diameters of the inertia centroidal ellipse of the
cross section.

Fig. 2.3.28
2.4 Axial Load 195

Fig. 2.3.29

About the deformation of the beam, we easily can verify that

– during the deformation any cross section preserve itself plane,

– the points of the beam that were found on the axis z before the application of the
load are found, to happened deformation, on a arc of circumference,
– such arc of circumference lies in the flexural plane f, z, where the flexural axis f is
normal to n and z (Fig. 2.3.29),
– the flexural plane is distinct from the load plane.

Remark 2.3.2 In a deviated bending problem the data is the cross section and the
bending moment M. To verify the structural safety, first of all we find any couple
of conjugate diameters of the inertia centroidal ellipse of the cross section. Then we
draw the load axis s as the centroidal line normal to M. Then we draw the neutral
axis n as conjugate diameter of s, by using the found couple of conjugate diame-
ters. Then we use the Navier’s formula (2.3.48) or we draw the diagram of the σ z
(Fig. 2.3.28). 

2.4 Axial Load

2.4.1 Centroidal Axial Load

We analyze the centroidal axial load, the first problem of Saint Venant (Fig. 2.2.5).
The frame of reference O, x, y is any cartesian orthogonal. The cross section of
the beam can be both monoconnected and multiconnected. We denote with V the
volume occupied by the solid of Saint Venant and consider the real functions defined
in V

σx = 0
σy = 0
N
σz =
A (2.4.1)
τxy = 0
τxz = 0
τyz = 0.
196 2 The Problem of Saint Venant

Fig. 2.4.1

where the symbol N denotes a non zero real number and the symbol A denotes the
area of the cross section (Fig. 2.4.1).
To prove that the (2.4.1) is the unique solution of the first problem of Saint
Venant, as already seen in Sect. 1.5, we must verify that such state of stress satisfies

– the inside compatibility, that is the (1.1.29),

– the outside compatibility, that is the conditions that the constraints impose to the
points of Su ,
– the inside equilibrium, that is the (1.2.3),
– the boundary equilibrium, that is the (1.2.5) in the points of Sp .

Since the (1.4.1), the state of strain associated to (2.4.1) is

νN
εx = −
EA
νN
εy =−
EA
N
εz = (2.4.2)
EA
γxy =0
γxz =0
γyz = 0.

Insofar the functions εx , εy , εz , γ xy , γ xz , γ yz are constant in V. This implicates

that their second order partial derivatives are all identically zero in V. The Eq.
(1.1.29) of the inside compatibility are therefore banally satisfied.
About the outside compatibility there are no conditions to satisfy because by
hypothesis the beam is not constrained.
About the inside equilibrium we notice that

– by hypothesis X = Y = Z = 0 in V,
– since the functions σ x , σ y , σ z , τ xy , τ xz , τ yz are constant, their first order partial
derivatives are all identically zero in V.

Insofar the (1.2.3) are banally satisfied.

At this point the six functions (2.4.1) constitute a state of stress physically
possible. We find now that a load must be applied on the beam to produce it.
2.4 Axial Load 197

Fig. 2.4.2

About the volumetric load, for hypothesis X = Y = Z = 0 in V.

About the superficial load, let (Fig. 2.4.2)

– Q be a point of the lateral surface,

– α the plane passing through Q and tangent to the lateral surface,
– n the line passing through Q, normal to α and directed going out from the material,
– nx , ny , nz the direction cosines of n.

Evidently nz = 0. As a consequence, from the (2.3.1) and (1.2.5) it follows

px (Q) = σx nx + τxy ny + τxz nz = 0

py (Q) = τyx nx + σy ny + τyz nz = 0
pz (Q) = τzx nx + τzy ny + σz nz = 0.

Insofar the lateral surface is unloaded.

So the state of stress (2.4.1) is physically possible and to establish it we must
apply a load only on the bases of the beam. Insofar the problem that we are analyzing
is a problem of Saint Venant.
We now see that a load must be applied on the right base z = l to establish in
the solid the state of stress (2.4.1). Let Q any point of the right base z = l. We
observe that the plane passing through Q and tangent to the surface of the beam is
obviously the base z = l. Moreover the line n passing through Q, normal to such
plane and directed going out from the material is clearly parallel and equiverse to z
(Fig. 2.4.3). Then its direction cosines nx , ny , nz are

nx = 0, ny = 0, nz = 1.

As a consequence, from the (2.4.1) and (1.2.5) it follows

N
px (Q) = τxz = 0, py (Q) = τyz = 0, pz (Q) = σz = .
A

Fig. 2.4.3
198 2 The Problem of Saint Venant

Fig. 2.4.4

Insofar the superficial load applied on the right base is uniformly distributed and
is parallel and equiverse to axis z. Consequently, it is equivalent to the system con-
stituted by an unique centroidal force, parallel and equiverse to z, having intensity
(Fig. 2.4.4)
  
N N N
pz dA = dA = dA = . A = N.
A A A A A A

Such load N2.4.1 produces in the beam the state of stress (2.4.1).
Proceeding as for the right base, we verify that the load applied on the left base
z = 0 is equivalent to the system constituted by an unique centroidal force, parallel
and opposite to z, having intensity N (Fig. 2.4.5).
In conclusion the (2.4.1) is the unique solution of the first problem of Saint
Venant (Fig. 2.2.5). Particularly important it is the third of the (2.4.1), that is

N
σz = . (2.4.3)
A
Let us see now how the beam deforms itself. Evidently, to obtain the field of the
displacements u, v, w we must integrate the (1.1.28) and (2.4.2). Clearly, being for
hypothesis the beam free in the space, we can determine u, v, w to less than a rigid
translation and a rigid rotation.
Nevertheless to be clearer we prefer to do such analysis supposing the beam
constrained with a punctual rigid joint in the centroid G of the left base z = 0, that
is subject to the conditions

u(0, 0, 0) = 0, v(0, 0, 0) = 0, w(0, 0, 0) = 0.

∂u ∂v ∂v ∂u (2.4.4)
(0, 0, 0) = 0, (0, 0, 0) = 0, − (0, 0, 0) = 0.
∂z ∂z ∂x ∂y

Fig. 2.4.5

2.4.1 A positive N is of traction.

2.4 Axial Load 199

We easily obtain
νN
u=− x
EA
νN
v=− y (2.4.5)
EA
N
w= z.
EA
The (2.4.5) allows to understand as the beam deforms itself. We easily can verify
that

– during the deformation the cross sections preserve him plane and normal to z,
– the points of the beam that were found on the axis z before the application of the
load are still found, to happened deformation, on the axis z,
– the points of the beam that were found on a parallel to axis z before the application
of the load are found, to happened deformation, on a parallel to axis z,
– we have
   
I I σz I N N I Nl
εz dz = dz = dz = dz = ,
0 0 E 0 EA EA 0 EA

so that (Fig. 2.4.6), denoting with l [resp. l ] the length of the beam before [after]
the deformation and putting l = l − l

Nl
l = , (2.4.6)
EA

– if N > 0 the cross section contracts itself. Precisely any point P of any cross section
suffers a displacement according the line passing through P and G, of intensity
(Fig. 2.4.7)

νN
dist(P, G)
EA

and as a consequence the cross section contracts itself preserving the form
(Fig. 2.4.8).

Fig. 2.4.6
200 2 The Problem of Saint Venant

Fig. 2.4.7

Fig. 2.4.8

Remark 2.4.1 We can calculate Δl also from the (2.4.5). In fact we have

Nl
l = w(0, 0, l) = .
EA

About the structural safety, since the Remark 1.7.3 and the (2.4.3), denoting with
s the safety factor, the safety condition in any point of the beam is

N σ0
≤ = σa . 
A s
Problem 2.4.1 An iron thread of 6 mm of diameter and of 10 m of length suffers a
tensile centroidal load of 400 kg. You shall determine the modulus of Young knowing
that the thread elongation is 6.8 mm.
Solution We denote with z the axis of the thread. Since

0.68 cm l 1   σz 1 N 1 400 kg
= = εz = σz − ν σx + σy = = · = ·
1, 000 cm l E E E A E π · (0.3)2 cm2
kg
it results E = 2, 078, 800 cm 2. 

Problem 2.4.2 You shall calculate the necessary area for the member AB of the
kg
structure of Fig. 2.4.9, assuming σa = 1, 600 cm 2.

Solution From the equilibrium triangle of the hinge B it immediately follows that
the beam AB is in traction and bears 5 t. From the safety condition

N
≤ σa
A
we have A = 3.12 cm2 . 
2.4 Axial Load 201

Fig. 2.4.9

2.4.2 Non Centroidal Axial Load

We consider now the following problem of Saint Venant. The superficial load dis-
tributed on the right base is equivalent to the system constituted by an unique
non centroidal force N, parallel to axis z (Fig. 2.4.10). We call such problem non
centroidal axial load.
We call center of pressure and denote with the symbol C the point common to
the line of action of N and to right base (Fig. 2.4.11). The frame of reference is
now rigorously principal of inertia (Fig. 2.4.12). The cross section of the beam can

Fig. 2.4.10

Fig. 2.4.11

Fig. 2.4.12
202 2 The Problem of Saint Venant

Fig. 2.4.13

be both monoconnected and multiconnected. We call load axis and denote with the
symbol s the line passing through C and G (Fig. 2.4.12).
To analyze this problem we make use of the principle of superposition and of the
postulate of Saint Venant. Preliminarily we put (Fig. 2.4.13)

e = dist(C, G) = (x02 + y20 )1/2 ,

where x0 , y0 are the coordinates of C (Fig. 2.4.12), and introduce the point A =
(x0 ,0) of Fig. 2.4.14. Then we consider the system Σ 2 constituted by (Fig. 2.4.15)
– the force N of Fig. 2.4.11 applied in C,
– two forces both parallel to z and applied in A, having equal intensity N and
opposite directions,
– two forces both parallel to z and applied in G, having equal intensity N and
opposite directions.

Obviously such system Σ 2 of Fig. 2.4.15 is equivalent to the system Σ 1 of

Fig. 2.4.11. Now we observe that in the system Σ 2 the force N applied in C and
the force N applied in A and having opposite direction constitute a couple of axis x

Fig. 2.4.14

Fig. 2.4.15
2.4 Axial Load 203

Fig. 2.4.16

Fig. 2.4.17

and intensity2.4.2 Mx = N y0 (Fig. 2.4.16). Moreover in the system Σ 2 the remained

force N applied in A and the force N applied in G and having opposite direction
constitute a couple of axis y and intensity2.4.3 My = – N x0 (Fig. 2.4.17).
In such way the system Σ 2 constituted by the force N applied in G, by the bend-
ing moment of axis x Mx = N y0 and by the bending moment of axis y My = – N x0
(Fig. 2.4.17) is equivalent to the system Σ 1 constituted by the unique force N applied
in C (Fig. 2.4.11).
As a consequence, because of the principle of superposition and of the postulate
of Saint Venant, anywhere except in the zone of extinction, the unique solution of the
problem in examination is the sum of a centroidal axial load problem (Fig. 2.4.18),

Fig. 2.4.18

2.4.2 Such couple is anticlockwise for an observer that places himself along the axis x.
2.4.3 Such couple is clockwise for an observer that places himself along the axis y.
204 2 The Problem of Saint Venant

Fig. 2.4.19

Fig. 2.4.20

of a right bending of axis x problem (Fig. 2.4.19) and of a right bending of axis y
problem (Fig. 2.4.20).
So in the problem in examination the state of stress is sum of the (2.4.1), (2.3.1),
and (2.3.40). In particular, we have the trinomial formula of the σ z in a point P =
(x, y) of the cross section (Fig. 2.4.21)

N Ny0 Nx0
σz = + y+ x (2.4.7)
A Ix Iy

where A is the area of the cross section, Ix [resp. Iy ] is the moment of inertia of the
area A with respect to axis x [resp. y].
We call neutral axis of the non centroidal axial load, and denotate with the sym-
bol n, the set of the points of the cross section such that σ z = 0. Clearly, since
the (2.4.7), in the problem in examination the neutral axis n is the line of equation
(Fig. 2.4.22)

Fig. 2.4.21
2.4 Axial Load 205

Fig. 2.4.22

N Ny0 Nx0
+ y+ x = 0. (2.4.8)
A Ix Iy

We expressly observe that, since G = (0,0) doesn’t satisfy the (2.4.8), in the non
centroidal axial load the neutral axis n is not centroidal.
Remark 2.4.2 We notice that, since by hypothesis N = 0, from the (2.4.8) it follows
that C does not belong to n. 

About the structural safety, since the Remark 1.7.3, denoting with s the safety
factor, the safety condition in any point of the beam is
σ0
σz ≤ = σa .
s
So we must verify that the maximum of the function (2.4.7) is not greater than
σ a . As in the deviated bending problem, we easily calculate such maximum using
the Analytic geometry property [2.3.1]. So from the [2.3.1], (2.4.7) it follows that a
real number k exists such that for every point P = (x, y) of the cross section it results

σz = kdn , (2.4.9)

where dn denotes the orthogonal distance of P from the neutral axis n (Fig. 2.4.22).
To find an operational expression of k, we observe that the system of the σ z
distributed on the generic cross section must balance2.4.4 the system constituted by
the alone vector N passing through C.
Insofar, it must be2.4.5
  
N = σz dA = kdn dA = k dn dA = kSn
A A A
where Sn denotes the static moment of the area A of the cross section with respect
to the neutral axis n. From this, since Sn = 0 because n is non centroidal, we get

2.4.4 Such property is obtained by imposing the equilibrium of the generic part [z, l] of the beam.
2.4.5 We are imposing the equilibrium of the part [z, l] of the beam to the translation according z.
206 2 The Problem of Saint Venant

N
k= . (2.4.10)
Sn
From (2.4.9) and (2.4.10) we have the first monomial formula of the non
centroidal axial load
N
σz = dn . (2.4.11)
Sn

Moreover the system constituted by the alone vector N passing through C and the
system of the σ z distributed on the generic cross section must have equal moment
with respect to axis n.2.4.6 Insofar
  
N·d = (σz dA)(dn ) = σz dn dA = kdn2 dA
A A A

=k dn2 dA = kIn
A

where d = dist(C, n) (Fig. 2.4.23) and In denotes the inertia moment of the whole
cross section with respect to neuter axis n. From this, since In = 0,2.4.7 we get

Nd
k= . (2.4.12)
In

From (2.4.9) and (2.4.12), putting Mn = N d, we have the formula of Navier of

the non centroidal axial load2.4.8
Mn
σz = dn . (2.4.13)
In

Fig. 2.4.23

2.4.6 We are imposing the equilibrium of the part [z, l] of the beam to the rotation around the axis n.
2.4.7 Otherwise we have N = 0, against the hypothesis.
2.4.8 The (2.4.8) is also called second monomial formula of the non centroidal axial load.
2.4 Axial Load 207

Fig. 2.4.24

The graphic representation of the (2.4.13), that we call diagram of the σ z , is given
in Fig. 2.4.24. We see that the points of any parallel to n always have the same value
of σ z (Fig. 2.4.25). Moreover, since σ z is constant with respect to z, such diagram
of the σ z repeats itself in all the cross sections. We underline that in the points of
the neuter axis n it results σ z = 0. We see from the (2.3.48) that the maximum value
of σ z is assumed in the most distant points from the neuter axis n. We see from the
trinomial formula (2.4.7) that

N
σz (G) = σz (0, 0) = . (2.4.14)
A

Remark 2.4.3 The conditions σ z = 0 on n and σ z = N/A in G allow us of immediately

draw the diagram of the σ z of Fig. 2.4.25. 

We now prove that

[2.3.1] In the non centroidal axial load the center of pressure C and the neutral axis
n are antipole and antipolar with respect to the inertia centroidal ellipse of the cross
section.

Fig. 2.4.25
208 2 The Problem of Saint Venant

Fig. 2.4.26

Proof Clearly the system constituted by the alone vector N passing through C and
the system of the σ z distributed on the generic cross section must have equal moment
with respect to load axis s.2.4.9 So, denoting with ds the real number dist(P, s) [resp.
−dist(P, s)] if P belongs [resp. belongs not] to the positive half plane individualized
from s (Fig. 2.4.26), we have
  
0=N·0= (σz dA)(ds ) = σz ds dA = kdn ds dA
A A A
=k dn ds dA = kIns .
A
From this, since from the (2.4.10) we have k = 0, it follows Ins = 0, so that

Ins = Sn dist(C, s). (2.4.15)

On the other hand, from the (2.4.10) and (2.4.12) we get In = Sn · d, so that

In = Sn dist(C, n). (2.4.16)

From the [2.1.5], (2.4.15), and (2.4.16) the thesis follows. 

Remark 2.4.4 In a non centroidal axial load problem the data is the cross section and
the center of pressure C. To perform the design, or to verify the structural safety,
first of all we find any couple of conjugate diameters of the inertia centroidal ellipse
of the cross section. Then, using the Remark 2.1.5, we draw the neutral axis n as
antipolar of C. Then, using the Remark 2.4.3, we draw the diagram of the σ z . 

Remark 2.4.5 In a non centroidal axial load problem the data is the cross section
and the center of pressure C. We consider the inertia centroidal kernel of the cross
section. The possible cases are the three following

2.4.9 We are imposing the equilibrium of the part [z, l] of the beam to the rotation around the axis s.
2.4 Axial Load 209

1 C is on the boundary of the kernel (Fig. 2.4.27),

2 C is intern to the kernel (Fig. 2.4.28),
3 C is external to the kernel (Fig. 2.4.29).

Since the [2.4.1], the neutral axis n is the antipolar of C. Then

– in the case 1 n is tangent to the cross section, so that this one is all distended or all
compressed (Fig. 2.4.27),
– in the case 2 we consider the load axis s passing through C and G. Then we
consider the intersection W between s and the boundary of the kernel. As already
seen, the antipolar r of W is tangent to the cross section. Since the [2.1.7], n is
parallel to r. Since the [2.1.12], n is more distant from G than r. Insofar the cross
section is all distended or all compressed (Fig. 2.4.28),
– in the case 3, reasoning as in the case 2, we see that the cross section is partly
distended and partly compressed (Fig. 2.4.29). 

Remark 2.4.6 We observe that in the non centroidal axial load problem it is possible
to calculate the σ z employing a binomial formula.

Fig. 2.4.27

Fig. 2.4.28
210 2 The Problem of Saint Venant

Fig. 2.4.29

Fig. 2.4.30

In fact let us consider the system Σ 3 constituted by (Fig. 2.4.30)

– the force N of Fig. 2.4.11 applied in C,

– two forces both parallel to z and applied in G, having equal intensity N and
opposite directions.

Obviously such system Σ 3 of Fig. 2.4.30 is equivalent to the system Σ 1 of

Fig. 2.4.11.
Now we observe that in the system Σ 3 the force N applied in C and the force
N applied in G and having opposite direction constitute a couple M of intensity
M = N · e and of axis normal to s and z (Fig. 2.4.31). In such way the system Σ 3

Fig. 2.4.31
2.4 Axial Load 211

Fig. 2.4.32

constituted by the force N applied in G and by the bending moment M (Fig. 2.4.32)
is equivalent to the system Σ 1 constituted by the unique force N applied in C
Fig. 2.4.11.
As a consequence, because of the principle of superposition and of the postulate
of Saint Venant, anywhere except in the zone of extinction, the unique solution of the
problem in examination is the sum of a centroidal axial load problem (Fig. 2.4.18)
and of a deviated bending having s as load axis (Fig. 2.4.33). Clearly, in the deviated
bending the neutral axis n0 is the centroidal line parallel to n (2.4.34).
So in the problem in examination we have the binomial formula of the σ z in a
point P = (x, y) of the cross section (Fig. 2.4.34)

N Mn0
σz = + dno (2.4.17)
A In0

Fig. 2.4.33

Fig. 2.4.34
212 2 The Problem of Saint Venant

where

– A is the area of the cross section,

– Mn0 is the orthogonal component of M on n0 ,
– In0 is the moment of inertia of the area A with respect to axis n0 ,
– dn0 = dist(P, no ). 

About the deformation of the beam, we easily can verify that in the non centroidal
axial load problem

– during the deformation the cross sections preserve its plane,

– the points of the beam that were found on the axis z before the application of the
load are found, to happened deformation, on a arc of circumference,
– such arc of circumference lies in the flexural plane f, z, where the flexural axis f is
normal to n and z.

Problem 2.4.3 We consider the steel column of Fig. 2.4.35, having the cross section
of Fig. 2.4.36. On the top a vertical superficial load whose resultant P pass through
the point C of Fig. 2.4.9 is distributed. You shall point out the procedure to verify
the structural safety of the column.
Solution Evidently this is a problem of Saint Venant of non centroidal axial load.
To verify the structural safety we need the diagram of σ z . To this aim we indeed
determine, with the procedure already seen in Problem 2.1.10, a couple of conjugate
diameters of the inertia centroidal ellipse of the cross section (Fig. 2.4.37). Then we
draw the neutral axis n, as antipolar of C, with the procedure already seen in Remark
2.1.5 (Fig. 2.4.38).

Fig. 2.4.35

Fig. 2.4.36
2.4 Axial Load 213

Fig. 2.4.37

Fig. 2.4.38

Fig. 2.4.39

Finally, using the procedure already seen in Remark 2.4.3, we easily draw the
diagram of the σ z (Fig. 2.3.39). Since the beam is compressed, it is necessary to
also verify the stability safety. 

Problem 2.4.4 We consider the steel column of Fig. 2.4.40, having the cross section
of Fig. 2.4.41. On the top is distributed a vertical superficial load (whose resultant
P=120 t pass through the point C of Fig. 2.4.41) is distributed. We know that the
thickness of the wall is 508 mm, that dist(C, G)=15 cm, that σ 0 /s = 1,600 kg/cm2 .
You shall verify the structural safety of the column.
Solution Evidently this is a problem of Saint Venant of non centroidal axial load. To
verify the structural safety we need the diagram of σ z . Since the inertia centroidal
ellipse of the cross section is a circle, the neutral axis n is normal to the load axis
s passing through C and G (Fig. 2.4.41). So the position of n is given from the
214 2 The Problem of Saint Venant

Fig. 2.4.40

Fig. 2.4.41

condition

ρn20 = dist (G, n)dist (G, C).

We easily have

17.62 = 15 dist(G, n)

from which

dist(G, n) = 20.65 cm.

Insofar the diagram of σ z is that of Fig. 2.4.42 and

 
P P d
σz max = dn max = dist(G, n) +
Sn A dist (G, n) 2
120 · 000 kg
= [20.65 + 25.4] = 1, 556 2 .
172 · 20.65 cm

Fig. 2.4.42
2.4 Axial Load 215

Since

kg kg
σz max = 1, 556 2
≤ 1, 600 2 = σa
cm cm
the column is in structural safety. Since the beam is compressed, it is necessary to
also verify the stability safety. 

2.4.3 Material Non Resistant to Traction

The theory of the elasticity developed in this book is mathematically valid only for
material resistant both to traction and to compression. Such material, as the steel, at
the tensile breaking test shows a diagram load-deformation that practically is polarly
symmetrical (Fig. 2.4.43). There are nevertheless some materials largely used in the
past and still today used in some types of construction that substantially are not
resistant to traction. They are the masonries and the concrete. The mathematical
model for such materials has been proposed only toward the 1990. It is not of simple
use. In fact the condition σz ≤ 0 imposes the use of the Theory of variational
inequalities of the modern Mathematical analysis.
Nevertheless the constructions performed with such materials and still in exercise
are so numerous that a technical theory was debugging and largely used for such
materials. It deals with civil constructions and the interested structures work under
non centroidal axial load.
We consider the solid of Saint Venant, in material not resistant to traction, loaded
on the two bases with a non centroidal axial load. As seen in Remark 2.4.5, if N
is of compression and the center of pressure C is inside or on the contour to the
inertia centroidal kernel of the cross section, the solution of Saint Venant satisfies
anywhere the condition σz ≤ 0. Then we can exactly calculate the state of stress
with the (2.4.13).
But if N is of compression and C is external to the inertia centroidal kernel, the
(2.4.13) is not applicable to the case of the material non resistant to traction. In fact
it furnishes σ z of compression that don’t balance the applied load N.
Nevertheless for a practical purpose, if N is of compression and C is external to
the inertia centroidal kernel, a technical theory that has been largely employed is
available. It consists to determine, by attempts, the position of the neutral axis n.
Let’s report us to the case, frequent in the practice, that (Fig. 2.4.44)

Fig. 2.4.43
216 2 The Problem of Saint Venant

Fig. 2.4.44

– the cross section admits an axis s of symmetry with respect to a direction r,

– the center of pressure C is a point of s and is external to the inertia centroidal
kernel.

Such technical theory makes the following conjectures

– the neutral axis n is parallel to r,

– the reacting section Ax is the part of cross section individualized by n and C
(Fig. 2.4.45),
– in the reacting section the σ z vary linearly with the distance from the neutral
axis n,
– the σ z are zero in the points of n,
– the σ z acting on the reacting section balance the load N.

We choose an axis n of attempt by fixing a value of attempt x = dist(C, n)

(Fig. 2.4.45). For every point P of Ax we put dn = dn (P) = dist(P, n) (Fig. 2.4.46).
We assume besides that it exists a real number k such that in every point P of Ax it
results

σz = kdn . (2.4.18)

Fig. 2.4.45

Fig. 2.4.46
2.4 Axial Load 217

Fig. 2.4.47

Since the σ z must balance the external load N, it must be

  
N= σz dA = kdn dA = k dn dA.
Ax Ax Ax

From this, denoting with Sx the static moment of Ax with respect to axis n, we
have N = k Sx , from which, taking into account the (2.4.18), it follows
N
σz = dn . (2.4.19)
Sn

Denoting with Gx the centroid of the reacting section Ax , we obviously have

(Fig. 2.4.46)
N N N
σz (Gx ) = dn (Gx ) = dn (Gx ) = .
Sx Ax dn (Gx ) Ax

We now denote with n0 the parallel to n passing through C. Since the σ z acting
on the reacting section Ax must balance the load N, the two system must have equal
moment with respect to n0 . So (Fig. 2.4.46)
 
0=N · 0= dMn0 = (σz dA)(x − dn )
Ax Ax
    
N N
= dn (x − dn )dA = x dn dA − 2
dn dA
Ax Sx Sx Ax Ax

from which, denoting with Ix the moment of inertia of Ax with respect to n, it follows

N
0= (x Sx − Ix ).
Sx
As a consequence, since N < 0, it must be

x Sx − Ix = 0.

We calculate Sx and Ix . If the preceding condition is satisfied, we assume the axis

n of attempt of Fig. 2.4.45 as real neutral axis of the problem. If otherwise it results

x Sx − Ix = 0
218 2 The Problem of Saint Venant

we put f(x) = x Sx – Ix and we perform another attempt choosing for x a new value.
Making a diagram of f, it is easy to get with few attempts the value x0 in which f is
zero (Fig. 2.4.47).

Remark 2.4.7 We can easily prove that in the this technical theory the neutral axis
n and the center of pressure C are antipolar and antipole with respect to the inertia
centroidal ellipse of the reacting section Ax . We can also prove that C is a point of
the contour of the inertia centroidal kernel of the reacting section Ax . 

Problem 2.4.5 A column in masonry have rectangular cross section (Fig. 2.4.48) and
is submitted to a non centroidal axial load of compression. The center of pressure
C is on the axis of symmetry s (Fig. 2.4.48) and is external to the inertia centroidal
kernel. The masonry has σ a = 4 kg/cm2 . You shall verify the structural safety.

Solution We observe indeed that, since the Problem 2.1.11, for the rectangle of
Fig. 2.4.48 the inertia centroidal kernel is the one of Fig. 2.4.49. As a consequence,
since the Remark 2.4.7, the neutral axis n has the position of Fig. 2.4.50. Then, from
the (2.4.19), we have

Fig. 2.4.48

Fig. 2.4.49

Fig. 2.4.50
2.5 Torsion 219

N N3a 2 N
σz max = dn max =  = · .
Sx 3a
b3a 2 3 ab

To this point the verification of the structural safety is immediate. 

2.5 Torsion

2.5.1 The Exact Solution

We study the fourth problem of Saint Venant (Fig. 2.2.8). We assume any orthogonal
cartesian reference frame O, x, y, z satisfying the conditions that z is parallel to the
axis of the beam and that O belongs to the left base. The cross section A of the
beam can be both monoconnected (Fig. 2.5.1) and multiconnected (Fig. 2.5.2). We
denote

– with ∂A the frontier of A, supposed regular,

– in every point Q of A, with n the line normal to A, passing through Q and going
out of A,
– with nx , ny , nz the direction cosines of n,
– ϑany nonzero real number.

We consider the boundary problem

[2.5.1] Find a real function ϕ(x, y) defined on A, having in A at least all the second
order partial derivatives and such that

∂ 2ϕ ∂ 2ϕ
(x, y) + (x, y) = 0 ∀(x, y) ∈ A (2.5.1)
∂x2 ∂y2

Fig. 2.5.1

Fig. 2.5.2
220 2 The Problem of Saint Venant

dϕ
(x, y) = ynx − xny ∀(x, y) ∈ ∂A.  (2.5.2)
dn
Remark 2.5.1 The (2.5.1) is called equation of Laplace.2.5.1 Every function that
∂2 ∂2
satisfies the (2.5.1) is said harmonic in A. The operator ∂x 2 + ∂y2 is called Laplace
operator. 

Remark 2.5.2 We call problem of Neumann2.5.2 every boundary problem in which a

function, having the transformed by the Laplace operator assigned on the open and
having the directional derivative along n assigned on the frontier, needs to be found.
So the problem [2.5.1] is a problem of Neumann. 

Remark 2.5.3 It is known in Functional analysis that every problem of Neumann

is equivalent to a problem, called of Dirichlet,2.5.3 in which another function χ ,
harmonic in the open and assuming assigned values on the frontier, needs to be
found. In the case in examination χ = x +y
2 2
2 + const. 

Since
  
  ∂y ∂x
−ynx + xny ds = − + dA = 0 dA = 0,
∂A A ∂x ∂y

a known theorem of Functional analysis assures that the problem [2.5.1] admits at
least a solution and that, said ϕ 1 and ϕ 2 any two solutions of such problem, a real
number c exists such that

ϕ2 = ϕ1 + c in A.

After that we can denote with ϕ a solution of the problem [2.5.1] such that
ϕ(0,0) = 0 and consider the three functions

u = −ϑzy
v = ϑzx (2.5.3)
w = ϑϕ.

We will prove that the (2.5.3) is the unique solution of the fourth problem of Saint
Venant. As already seen in Sect. 1.5, we must verify that such field of displacements
satisfies

– the inside compatibility, that is the (1.1.29) in V,

– the outside compatibility, that is the conditions that the constraints impose to the
points of Su ,

2.5.1 Pierre Simon Laplace, Beaumont en Auge 1749 – Paris 1827.

2.5.2 Karl Gottfried Neumann, Konigsberg 1832 – Leipzig 1925.
2.5.3 Peter Gustav Lejeune Dirichlet, Duren 1805 – Gottingen 1859.
2.5 Torsion 221

– the inside equilibrium, that is the (1.2.3) in V,

– the boundary equilibrium, that is the (1.2.5) in the points of Sp .

Since u, v, w are regular, the inside compatibility (1.1.29) is banally satisfied.

About the inside equilibrium (1.2.3), since the (1.1.28) the state of strain
associated to (2.5.3) is

∂
εx = ( − ϑzy) = 0
∂x
∂
εy = (ϑzx) = 0
∂y
∂
εz = (ϑϕ(x, y)) = 0
∂z
(2.5.4)
∂ ∂
γxy = ( − ϑzy) + (ϑzx) = −ϑz + ϑz = 0
∂y ∂x
∂ ∂ ∂ϕ
γxz = ( − ϑzy) + (ϑϕ) = −ϑy + ϑ
∂z ∂x ∂x
∂ ∂ ∂ϕ
γyz = (ϑzx) + (ϑϕ) = ϑx + ϑ .
∂z ∂y ∂y
Since the (1.4.5), the state of stress associated to the state of strain (2.5.4) is

σx = 2Gεx + λ(εx + εy + εz ) = 0
σy = 2Gεy + λ(εx + εy + εz ) = 0
σz = 2Gεz + λ(εx + εy + εz ) = 0
τxy = Gγxy = 0
(2.5.5)
∂ϕ
τxz = Gγxz = Gϑ −y +
∂x
∂ϕ
τyz = Gγyz = Gϑ x + .
∂y
We recall that by hypothesis it is X = Y = Z = 0 in V. Moreover since the
functions σ x , σ y , σ z , τ xy are constant, they all have the first order partial derivatives
identically zero in V. On the other hand, since the functions τ xz , τ yz are constant with
respect to z, they have the first order partial derivatives with respect to z identically
zero in V. So the first and the second of the (1.2.3) are banally satisfied. About the
third of the (1.2.3), since ϕ satisfies the (2.5.1), we have
   
∂τzx ∂τzy ∂σz ∂ ∂ϕ ∂ ∂ϕ
+ + = Gϑ −y + + Gϑ x +
∂x ∂y ∂z ∂x ∂x ∂y ∂y
∂ 2ϕ ∂ 2ϕ
= Gϑ + = 0.
∂x2 ∂y2
222 2 The Problem of Saint Venant

Fig. 2.5.3

Fig. 2.5.4

Such way we have seen that, if the volumetric load is identically zero, the (2.5.3)
is a field of displacements that can physically exist for the beam of Saint Venant.
Let us see now what load is necessary to apply on the beam to deform it according
the law (2.5.3). Let (Figs. 2.5.3 and 2.5.4)

– Q be every point of the lateral surface of the beam,

– α 0 be the tangent plane passing through Q,
– n be the line normal to α Q , passing through Q and going out of the material,
– nx , ny , nz be the direction cosines of n.

Since nz = 0, from the (1.2.5) and (2.5.2) it follows

px = σx nx + τxy ny + τxz nz = 0 nx + 0 ny + τxz 0 = 0

py = τyx nx + σy ny + τyz nz = 0 nx + 0 ny + τyz 0 = 0
∂ϕ ∂ϕ
pz = τzx nx + τzy ny + σz nz = Gϑ −y + nx + Gϑ x + ny
∂x ∂y
∂ϕ ∂ϕ
= Gϑ −ynx + xny + nx + ny
∂x ∂y
dϕ
= Gϑ −ynx + xny + = 0.
dn

Insofar the lateral surface is unloaded and then the (2.5.3) is the solution of a
problem of Saint Venant.
Let us see now what load is necessary to be applied on the right base to deform
the beam according the law (2.5.3). Let
2.5 Torsion 223

– Q be an every point of the right base of the beam,

– n be the line normal to the right base, passing through Q and going out of the
material,
– nx , ny , nz be the direction cosines of n.

Since nx = 0, ny = 0, nz = 1, from the (1.2.5) and (2.5.2) it follows

px = σx nx + τxy ny + τxz nz = 0 nx + 0 ny + τxz · 1

∂ϕ
= τzx = Gϑ −y +
∂x
py = τyx nx + σy ny + τyz nz = 0 nx + 0 ny + τyz · 1 (2.5.6)
∂ϕ
= τzy = Gϑ x +
∂y
pz = τzx nx + τzy ny + σz nz = τzx 0 + τzy 0 + 0nz = 0.

As a consequence, on the right base we must not apply axial load. In fact
 
N= pz dA = 0 dA = 0.
Z A

Besides on the right base we must not apply bending moment of axis x. In fact
 
Mx = (pz dA)(y) = 0 · y dA = 0.
A A

Besides on the right base we must not apply bending moment of axis y. In fact
 
My = − (pz dA)(x) = − 0 · x dA = 0.
A A

Besides on the right base we must not apply shear of axis x. In fact, since the
(2.5.5) and (2.5.1) we have

∂ ∂ ∂τzx ∂x ∂τzy ∂x
(xτzx ) + (xτzy ) = x + τzx + x + τzy
∂x ∂y ∂x ∂x ∂y ∂y

∂τzx ∂τzy ∂ 2ϕ ∂ 2ϕ
=x + τzx + x = τzx + xGϑ + 2 = τzx ;
∂x ∂y ∂x 2 ∂y

so that, taking into account the (2.5.6), the formulas of Gauss and the (2.5.5), it
results
224 2 The Problem of Saint Venant
    
∂ ∂
ITx = px dA = τzx dA = (xτzx ) + (xτzy ) dA
A ∂x ∂y
A A

 
= (xτzx nx + xτzy ny )ds = x τzx nx + τzy ny ds
∂A ∂A
∂ϕ ∂ϕ
= xGϑ −ynx + xny + nx + ny ds
∂A ∂x ∂y

= xGϑ · 0 ds = 0.
∂A

Analogously we get

Ty = py dA = 0.
A

Remark 2.5.4 Another way to prove that Tx = 0 and Ty = 0 is the following. By

absurd, let us suppose Tx = 0 on the right base. Then the equilibrium of the beam
to rotation around y imposes the existence of a bending moment My of axis y on
the left base (Fig. 2.5.5). This is absurd since the pz are anywhere zero on the left
base. 

In conclusion on the right base to the more there is a twisting moment Mt . Since
the (2.5.6), assuming positive the moment counterclockwise
 
 
Mt = (py dA)(x) − (px dA)(y) = (py x − px y)dA
A A

∂ϕ ∂ϕ
= Gϑ x2 + y2 + x −y dA
A ∂y ∂x

from which
Mt
ϑ=  . (2.5.7)
∂ϕ ∂ϕ
G x2 + y2 + x −y dA
A ∂y ∂x

We call torsional rigidity of the beam the real number


∂ϕ ∂ϕ
G x2 + y2 + x −y dA.
A ∂y ∂x

Fig. 2.5.5
2.5 Torsion 225

We can prove that the torsional rigidity is always positive.

Insofar the (2.5.3) is the solution of the fourth problem of Saint Venant. About
the deformation of the beam, from the (2.5.3) it obviously follows that after the
deformation the points of the axis of the beam still stay on the axis z. Besides, since
εz = 0 anywhere, the axis of the beam stays of length l, meaning it doesn’t lengthen
and it doesn’t shorten. From considerations of Analytical geometry the lines parallel
to the axis z but distinguished by the axis z turn into lines skew2.5.4 with respect to
the axis z.
We now consider the cross section of the beam individualized by the coordinate
z ∈ [0, l]. It follows from (2.5.3) that the cross section hunches, meaning the points
of the cross section move, to happened deformation, to a surface S of the space that,
in the general case, is not a plane. Then in the torsion the cross sections don’t stay
plane. Besides it follows from the (2.5.3) that, if we project on the cross section indi-
vidualized by the coordinate z the points of S, we get in the plane of the cross section
a figure that with good approximation is2.5.5 that gotten rigidly rotating around the
axis z of ϑz radians the cross section of the undeformed beam (Fig. 2.5.7).

Fig. 2.5.6

Fig. 2.5.7

2.5.4 It deals with a linearization consequent to the hypothesis of small deformations. In the (non-
linear) field of the great deformations, in the problem of the torsion the lines parallel to z but
distinguished by the axis z turn into cylindrical helixes of the space.
2.5.5 We consider the cross section z = k, with 0 ≤ k ≥ l, a frame of reference 0, x, y and a generic
point P0 = (x0 , y0 ) of the cross section (Fig. 2.5.6). When we project on the plane z = k the
point of S in which P0 is moved to happened deformation, evidently we obtain the point P2 = (x0
+ u, y0 + v). We consider the displacement s1 = P0 P2 = (u, v) = ( − ϑky0 , ϑkx0 ) Since s1 ×
OP0 = u x0 + v y0 = 0, P0 is moved orthogonally to the line passing through O and P0 . Since
|s1 | = (u2 + v2 ) 12 = ϑk dist(O, P0 ), the length of the displacement s1 is equal to the length of the
arc P0 P1 of the circumference of center O and radius dist(o, p0 ), obtained rotating the cross section
1
of α = ϑk radians around O. Since the hypothesis of small displacements, (u2 + v2 ) 2 is small, so
that α = ϑk is small, so that we can approximate the segment P0 P2 with the arc P0 P1 , so that we
can approximate P2 with P1 . In other words, in the hypothesis of small displacements the arc P0 P1
can be linearized with the segment P0 P2 .
226 2 The Problem of Saint Venant

Fig. 2.5.8

This way, ϑ has dimension L–1 and measure the angle from which it rotates, in
the sense just specified, the cross section at distance of 1 m, if this is the unity of
measure, from the left base. Two cross sections of the beam having distance of 1 m
among them, rotate among them of

ϑz2 − ϑz1 = ϑ(z2 − z1 ) = ϑ.

Then the constant ϑ that appears in the (2.5.3) is the angle from which they rotate,
in the sense before specified, two cross sections of the beam at unitary distance
among them. For this motive ϑ is said unitary angle of torsion.
Finally, about the outside compatibility, it is easy to verify that, since we have
chosen ϕ such that ϕ(0,0) = 0, the (2.5.3) satisfies the conditions imposed by a
punctual rigid joint in the origin O = A of the left base z = 0, that is (Fig. 2.5.8)

u(0, 0, 0) = 0, v(0, 0, 0) = 0, w(0, 0, 0) = 0,

∂u ∂v ∂v ∂u (2.5.8)
(0, 0, 0) = 0, (0, 0, 0) = 0, + (0, 0, 0) = 0.
∂z ∂z ∂x ∂y

Remark 2.5.5 It is important to underline that the solution of the problem of the tor-
sion found by Saint Venant is valid only if the beam is bound in the alone point O of
the base of left. If the base of left is not free to hunch, dangerous secondary stresses
arise, which can provoke the breakup of the beam. This case will be examined in the
following. 

2.5.2 The Circular Cross Section

The case in which the cross section of the beam is a circle or an annulus is of great
practical importance. In fact in cars and ships the driving shafts that transmit the
rotatory motion from the motor to the propeller organs have a circle or an annulus
as cross section. Moreover the circle and the annulus are the only cross sections that
do not hunch. Therefore we treat this case with great attention.
We consider the circular cross section of Fig. 2.5.9 and denote with r0 its radius.
We assume the center of the circle as origin O of the reference frame (Fig. 2.5.9).

Fig. 2.5.9
2.5 Torsion 227

First of all we prove that the function ϕ identically zero is solution of the problem
of Neumann [2.5.1] related to the circle A of Fig. 2.5.9.
Since ϕ = 0 is a constant function, it admits partial derivatives of every order and
all these coincide with the function identically zero. Insofar

∂ 2ϕ ∂ 2ϕ
= 0, = 0,
∂x2 ∂y2
so that ϕ is harmonic in A. Besides, for the same motive, in every point (x, y) of ∂A
the directional derivative of ϕ along n is zero. Then, in order that ϕ = 0 is solution
of the problem of Neumann [2.5.1], it is enough to verify that in every point (x, y)
of ∂A it results
ynx − xny = 0.

To this aim let us observe that the normal n to ∂A in (x, y) is the line passing
through O and (x, y). Denoting with α the angle between n and the axis x, we have
(Fig. 2.5.9)

x = r0 cos α, y = r0 sinα

and
nx = cos α, ny = sinα,

so that

x = r0 nx , y = r0 ny

so that

ynx − xny = ro nx ny − r0 nx ny = 0.

Insofar ϕ = 0 is solution of the problem of Neumann [2.5.1] related to the circle

A of Fig. 2.5.9. As a consequence, since the (2.5.3), w = 0, so that the circular cross
section do not hunch.
From the (2.5.5) and (2.5.7) it follows

τzx = −Gϑy, τzy = Gϑx. (2.5.9)


Mt = Gϑ (x2 + y2 )dA. (2.5.10)
A

We call polar inertia moment, and denote with the symbol Ip , the nonnegative
real number2.5.6

- - - - r04
2.5.6 Obviously Ip = A (x
2 + y2 )dA = Ax
2 dA+
Ay
2 dA =2 Ay
2 dA = 2Ix = π 2.
228 2 The Problem of Saint Venant

(x2 + y2 )dA.
A

Insofar from the (2.5.10) we have

Mt
ϑ= (2.5.11)
GIp

so that, taking into account the (2.5.9)

Mt
τzx = − y
Ip
(2.5.12)
Mt
τzy = x.
Ip

We see from the (2.5.11) that the torsional rigidity of the circular cross section is
G Ip .
We now prove that the shear stresses distributed on the cross section have the
following important properties
(1) in every point P of the circle the total shear stress τ z is normal to the line passing
through O and P (Fig. 2.5.10),
(2) the total shear stress τ z varies linearly with the distance from the center O
(Fig. 2.5.11).
As for as the first property, it is obviously enough to prove that

r × τz = 0

Fig. 2.5.10

Fig. 2.5.11
2.5 Torsion 229

Fig. 2.5.12

where r is the oriented segment having first end O and second end P. Putting
P=(x, y), it results r = (rx , ry ) = (x, y), so that from the (2.5.12) we immediately
have (Fig. 2.5.12)
Mt Mt
r × τ z = rx τzx + ry τzy = xτzx + yτzy = − xy + xy = 0.
Ip Ip
1
About the second property, since r = dist(P, 0) = (x2 + y2 ) 2 , it results

!1
2 2 2
2 21 Mt Mt Mt 2 1
τz = 2
(τzx + τzy ) = − y + x = (x + y2 ) 2
Ip Ip Ip

so that
Mt
τz = r. (2.5.13)
Ip

Remark 2.5.6 The diagram of the τ z is that of Fig. 2.5.11. Since the (2.5.13) the τ z
are zero in the center, vary linearly with the distance from the center, are constant
on every concentric circumference, have the maximum value on the boundary of the
circle. Moreover on every concentric circumference the τ z are in every point tangent
to the circumference (Fig. 2.5.10). 

Remark 2.5.7 The analytical solution of the problem of the torsion is only known
for very few cross sections. Here we treat only the case of the elliptic cross section
(Fig. 2.5.13).
We consider the function
a2 − b2
ϕ=− xy.
a2 + b2

Fig. 2.5.13
230 2 The Problem of Saint Venant

Evidently ϕ is harmonic. Let now consider a point (x̄, ȳ) ∈ ∂A. Since the ellipse
has equation
x2 y2
2
+ 2 =1
a b
the tangent t to ellipse in (x̄, ȳ) has equation
xx̄ yȳ
2
+ 2 = 1,
a b
the line n normal to t and passing through (x̄, ȳ) has equation

ȳ x̄
(x − x̄) 2
− (y − ȳ) 2 = 0,
b a
the direction cosines of n are
x̄ 1 ȳ 1
nx = 1 , ny = 1 ,
a2  x̄2 ȳ2 2 b2  x̄2 ȳ2 2
a4
+ b4 a4
+ b4
we have
∂ϕ ∂ϕ
(x̄, ȳ)nx + (x̄, ȳ)ny − ȳnx + x̄ny
∂x ∂y
x̄ 1 a2 − b2 ȳ 1 a2 − b2
= 2  − ȳ − ȳ + 2  1 − x̄ + x̄
a x̄2 ȳ2 2
1
a +b
2 2 b x̄2 ȳ2 2 a2 + b2
a 4 + b 4 a 4 + b 4

2 x̄ȳ x̄ȳ
= 1 − + 2 = 0.
x̄2 2
ȳ 2 a2 +b 2 a + b2
a4
+ b4

From the (2.5.7), since

π ab3 π a3 b
Ix = , Iy = ,
4 4
we obtain
Mt
ϕ= -  
G A x2 + y2 + x ∂ϕ
∂y − y ∂ϕ
∂x dA
Mt
= -  a2 −b2 a2 −b2 2

G A x2 + y2 − a2 +b2
x2 + a2 +b2
y dA
 
Mt Mt a2 + b2
=     = .
G 1 − aa2 −b
2 2
I y + 1 + a2 −b2
I x
π Ga3 b3
+b 2 a +b
2 2
2.5 Torsion 231

From the (2.5.5) we have

2Mt 2Mt
τzx = − y, τzy = x,
π ab3 π a3 b
from which
1
2Mt x2 y2 2
τz = + 4 .
π ab a4 b

Let us consider now any diameter d of the ellipse (Fig. 2.5.14). Since for every
P = (x, y) of d it results

τzx a2 y
=− 2
τzy b x

we know that in every point P of d the τ z have always the same direction. As a
consequence, since on the boundary the τ z are tangent to the boundary, in every
point of d the τ z are parallel to the tangent in P0 to the boundary (Fig. 2.5.14).
Banally in the center O of the ellipse it results

1
2Mt 02 02 2
τz = + = 0.
π ab a4 b4

Moreover, since d is a line passing through O, a real number k exists such that
for every (x, y) of d it results y = kx. As a consequence, for every P=(x, y) of d, we
have
1
2Mt x2 (kx)2 2
τz (P) = τz (x, y) = +
π ab a4 b4
⎡ 1
⎤
2Mt 1 k 2 2
=⎣ + 4 cos (arctgk)⎦ dist(P, O).
π ab a4 b

So in the points of d the τ z vary linearly (Fig. 2.5.14). Then the maximum of the
τ z arises on the boundary of the ellipse. Denoting with t the tangent in P0 to the
ellipse (Fig. 2.5.14), we easily verify that

Fig. 2.5.14
232 2 The Problem of Saint Venant

 2
x02 y20 1
+ =
a4 b4 dist(O, t)

so that the maximum of the τ z arises in the end points of the shorter diameter of the
ellipse. In the case of Fig. 2.5.13 such maximum arises in the points H1 and H2 . It
results

2Mt
τz max = . 
π a2 b

2.5.3 The Stress Concentration

When a beam is submitted to torsion, in some types of cross section the dangerous
phenomenon of the stress concentration is verified. This problem can be analyzed
in various ways. Here we examine the method of the hydrodynamic analogy, which
offers the advantage of the physical evidence.
We denote with A the cross section of the beam and consider the problem2.5.7
[2.5.2] Find a vectorial function τ z : A → 2 such that

div τ z = 0 in A
curl τ z = 2 Gϑ in A
τ z × n = 0 in ∂A. 

We have
[2.5.3] The solution (2.5.5) of the fourth problem of Saint Venant is solution of the
problem [2.5.2].
Proof We consider the solution (2.5.5) of the fourth problem of Saint Venant. We
easily have in every point of A

∂τzx ∂τzy ∂ 2ϕ ∂ 2ϕ
div τ z = + = Gϕ +
∂x ∂y ∂x2 ∂y2
∂τzy ∂τzx ∂ 2ϕ ∂ 2ϕ
curl τ z = − = Gϕ 1 + +1− = 2Gϑ.
∂x ∂y ∂x∂y ∂x∂y

 
2.5.7 Let v = vx , vy , vz : ⊆ 3 → 3 . We call divergence of v and denote with the symbol
∂vx ∂vy ∂vz
div v the real function, defined in Ω, div v = ∂x + ∂y + ∂z .
We call curl (or rotation) of v
 
∂vy
and we denote with the symbol curl v (or rot v) the vectorial function curl v = ∂v ∂y − ∂z x +
z
   
∂vx ∂vz ∂vy ∂vz
∂z − ∂x y + ∂x − ∂y z, defined in Ω and having values in  . In the bidimensional case
3
 
∂vy ∂v
we adopt the analogous definitions v = (vx , vy ), div v = ∂x + ∂y , curl v = ∂xy − ∂v
∂vx x
∂y z, and we
∂vy ∂vz
also write, more concisely, curl v = ∂x − ∂y .
2.5 Torsion 233

The condition τ z ×n = 0 in ∂A is satisfied since in every problem of Saint Venant

in any boundary point τ z is tangent to the boundary. The thesis follows. 
Remark 2.5.8 In the problem [2.5.2] the equation div τ z = 0 is a condition of
equilibrium. In fact from the (2.5.5) it follows

∂σz
= 0.
∂z
From this, since in the problem of Saint Venant the volumetric load is identically
zero, it results

∂τzx ∂τzy ∂σz

div τ z = + + + Z,
∂x ∂y ∂z
so that the equation div τ z = 0 coincides with one of the equilibrium conditions
(1.2.3).
Moreover in the problem [2.5.2] the equation curl τ z = 2Gϑ is a condition
of compatibility. In fact from the (2.5.4) and from the fourth and fifth of the
compatibility conditions (1.1.29) it follows

∂ ∂γyz ∂γzx 1 ∂
0= − + =− curl τ z
∂x ∂x ∂y G ∂x
∂ ∂γyz ∂γzx 1 ∂
0= − = curl τ z
∂y ∂x ∂y G ∂y

from which

curl τ z = const in A.

Vice versa this last condition, together with the (2.5.4), implicates that the fourth
and fifth of the compatibility conditions (1.1.29) are true. Insofar the equation
curl τ z = 2Gϑ is equivalent to two equations of compatibility. 
The problem [2.5.2] is also the mathematical model of a simple problem of
Hydraulics. An analogy between such problem and that of the torsion is established.
Such analogy, that we call hydrodynamic analogy, furnishes in a simple way a lot
of important indications about the torsion.
We consider a tumbler with the base equal to the one of the solid of Saint Venant.
We fill this container with water and, with an opportune device, we rotate it around
the vertical axis z with constant angular speed c. At the instant t0 we arrest the
motion of rotation. If c is sufficiently small, in the first instants following t0 the
water motion can be regarded as steady,2.5.8 without friction, independent from z
and simulated by the mathematical model2.5.9
[2.5.4] Find a vectorial function v: A → 2 such that

2.5.8 The motion is said steady when the speed of the water is independent from the time.
2.5.9 We denote with v the velocity of the water.
234 2 The Problem of Saint Venant

div v = 0 in A
curl v = 2 Gϑ in A
v × n = 0 in ∂A. 

Remark 2.5.9 Obviously in the problem [2.5.4] the boundary condition expresses
the fact that the walls of the container are impermeable. 

Comparing the problems [2.5.2] and [2.5.4], we see that only the symbol with
which the unknown function is denoted changes, so that they are mathematically
coincident. With this an analogy among the precedent hydrodynamic problem and
the problem of the torsion is established. By virtue of such analogy the distribution
of the τ z in the torsion coincides with the distribution of the speeds v of the particles
of water. Insofar if we know the path γ of a particle of water (Fig. 2.5.15), then2.5.10
in every point of γ the τ z are tangent to γ .

Remark 2.5.10 When we use the hydrodynamic analogy it is very useful to employ
the property of the Hydraulics called constant flow.2.5.11 Such property, that in
Mathematical physics is called theorem of the divergence, can be obtained in the

Fig. 2.5.15

2.5.10 We call streamline every curve γ such that, whatever a point P of γ is chosen, the velocity
v (P, t) of P at the instant t is tangent to γ (Fig. 2.5.16). In general γ doesn’t coincide with the path
β of the fluid particle, that is situated in the point A at the instant tA (Fig. 2.5.17). Clearly, if the
flow is steady, the path β of the fluid particle is also a streamline (Fig. 2.5.18).
2.5.11 We consider a vectorial function v :  ⊆ 2 → 2 and a regular plane curve δ having first
end Q1 and second end Q2 , such that δ ⊆  (Fig. 2.5.19). We denote in every point Q of δ with t
[resp. n] the tangent [resp. normal] line to δ. We call flow of v through δ the real number

 Q1 ,Q2  Q1 ,Q2
v × n ds = (vx nx + vy ny )ds.
δ δ

If v is the velocity of a fluid, clearly such curvilinear integral is the area of fluid that goes through
δ throughout an unitary time interval.
2.5 Torsion 235

Fig. 2.5.16

Fig. 2.5.17

Fig. 2.5.18

Fig. 2.5.19

following way. We suppose plane the motion of the water and we consider two
streamlines γ 1 , γ 2 relative to the instant t. We call stream tube the portion of
plane individualized by γ 1 and γ 2 . We denote with A, D [resp. B, C] two points
of γ 1 [resp. γ 2 ] and we consider the portion W of stream tube individualized by
the segments AB and CD (Fig. 2.5.20). Since the water is a fluid incompressible,
the principle of conservation of mass assures that in every point of the fluid results
div v = 0. Then, since the Gauss formulas

Fig. 2.5.20
236 2 The Problem of Saint Venant
  
∂vx ∂vy  
0= div v dW = + dW = vx nx + vy ny ds
W W ∂x ∂y ∂W
  A,B  B,C  C,D
= v × n ds = v × n ds + v × n ds + v × n ds
∂W ∂W ∂W ∂W
 D,A
+ v × n ds.
∂W
Since in every point of γ 1 [resp. γ 2 ] v is tangent to γ 1 [resp. γ 2 ], we have
v × n = 0, so that
 A,B  C,D
0= v × n ds + v × n ds.
∂W ∂W

Insofar
 C,D  B,A
v × n ds = v × n ds.
∂W ∂W

In conclusion, if the fluid is incompressible, in a stream tube the flow is

constant. 

Remark 2.5.11 We consider two streamlines γ 1 , γ 2 so near that we can approxi-

mate the stream tube with a regular curve γ that we call mean curve. We consider
every point P of γ , the tangent ξ to γ in P, the normal η to ξ passing through P
(Fig. 2.5.21). We call chord the segment c = AB individualized by η, γ 1 , γ 2 . Since
γ 1 , γ 2 are very near, in Hydraulics we assume that in c the velocity v is constant,
parallel to ξ and equiverse (Fig. 2.5.22). In reality the fluid viscosity imposes the
condition v(A) = v(B) = 0, so that the real velocity profile along c is the one of
Fig. 2.5.23. Nevertheless in the practical calculations such real velocity profile can
be quietly approximate with the constant diagram of Fig. 2.5.22. As a consequence,
the flow in P is
  
v × n ds = v ds = v ds = v c. 
A,B A,B A,B

Fig. 2.5.21

Fig. 2.5.22
2.5 Torsion 237

Fig. 2.5.23

Fig. 2.5.24

Remark 2.5.12 We consider two streamlines γ 1 , γ 2 very near and two chords c1 ,
c2 of their stream tube (Fig. 2.5.24). As seen in the Remarks 2.5.10, 2.5.11, it
results v1 c1 = v2 c2 . So the velocity increases where the stream tube gets narrower
(Fig. 2.5.24). 

The hydrodynamic analogy furnishes, for beams submitted to torsion, very

important indications about the stress concentration.
This dangerous phenomenon arises when the cross section has a reentrant angle,
as in the points A, B, C, D, E, F, G, H, I of Fig. 2.5.25. It is well known in Fluid
dynamics that in correspondence of every of these points, the streamlines became
infinitely nearer, so that, as seen in Remark 2.5.12, the speed is theoretically infinite
in these points (Fig. 2.5.25). Then in every of these points the τ z would be theoret-
ically infinite and obviously the material would collapse. To avoid such dangerous
phenomenon, in the cross sections the reentering angles are always replaced by
convenient fitting (Fig. 2.5.26).

Fig. 2.5.25

Fig. 2.5.26
238 2 The Problem of Saint Venant

Instead, in correspondence of the points A1 , B1 , C1 , D1 , E1 , F1 , H1 , I1 , J1 , K1 ,

L1 , M1 , N1 of Fig. 2.5.26, the edges of the cross section can be left sharp. In fact
it is well known in Fluid dynamics that in proximity of every of these points the
streamlines are outdistanced among them so that the fluid velocity is practically
zero. Such zones are called of stagnation (Fig. 2.5.26). In every of these points it is
v = 0, so that τ z = 0.

Problem 2.5.1 You shall verify the structural safety of a beam of Saint Venant
submitted to torsion.

Solution By virtue of (2.5.5) the state of stress in the beam is independent from z.
Insofar we consider every point P of every cross section.
To calculate the principal stresses in P, we introduce an orthogonal cartesian
reference frame P, x , y , z such that z is parallel to z and x is parallel to τ z
(Fig. 2.5.27).

From the (2.5.5) we have

 1
2 2
σz = σz = 0, τz x = τz = τzx
2
+ τzy , τz y = 0

and, taking into account the (1.2.8) and (1.2.9)

σx = 0, σy = 0, τx y = 0.

Then τy = 0. Then y is a principal direction of stress. Then a principal stress is

ση = σy = 0 and the other two principal stress σξ , σζ are furnished by the circle of
Mohr for the set of planes of support y (Fig. 2.5.28). This way σξ = τz , σζ = −τz

Fig. 2.5.27

Fig. 2.5.28
2.5 Torsion 239

Fig. 2.5.29

(Fig. 2.5.29). Since the maximum shearing stress criterion, the beam is in structural
safety if
σ0 σa
τz max ≤ ≤ . 
2s 2

Problem 2.5.2 You shall determine the angle of rotation among the end cross sec-
tions of a beam of Saint Venant submitted to torsion. The cross section is a circle
with F = 3 cm. The beam has length 1.5 m (Fig. 2.5.30) and is made with a steel
having E = 2.1 · 106 kg/cm2 , ν = 0.3. Moreover τ zmax = 800 kg/cm2 .
Solution From the (2.5.13) it follows

Mt 2Mt
800 = τz max = R=
Ip π (1.5)3
so that Mt = 2,826 kg cm = 0.028 tm. Since

E
G=
2(1 + ν)
we have G = 807,000 kg/cm2 . Since

Mt
ϑ= ,
GIp
we have
2, 826 · 2
ϑ= = 0.00044 cm−1 .
807, 000 · π · (1.5)4

As a consequence, the angle of rotation among the end sections of the beam is

0.00044 · 1, 500 = 0.66 radians = 37.83◦ . 

Fig. 2.5.30
240 2 The Problem of Saint Venant

2.5.4 Closed Thin Walled Cross Section

We call thin walled beam a beam of Saint Venant whose cross section A is such
that

– in its plane it exists a regular curve γ, that we call middle curve of the cross section,
open (Fig. 2.5.31) or closed (Fig. 2.5.32),
– for every point P of γ the intersection between n2.5.12 and A is a segment c, that
we call chord, whose middle point is P,
– in every point P of γ the chord c is negligible in comparison to the diameter d of
the smallest circle containing the cross section, and d is negligible in comparison
to l,
– in every point P of γ the tangents to ∂A in the end points A and B of the chord
individualized by P are practically parallel to t (Figs. 2.5.31 and 2.5.32).

In all the branches of the Structural engineering the one-dimensional structural

elements are frequently realized with thin walled beams.The thin walled cross sec-
tion allows in fact to exploit at the most the possibilities of resistance of the material,
in the sense that a notable saving in weight of material is realized.
Remark 2.5.13 Sometimes, in the most important engineering applications, the
thickness of the wall must be very small. In such case it becomes necessary, for
matters of stability of the equilibrium, to stiffen with additional elements (called
ribs) some cross sections of the beam. 

Fig. 2.5.31

Fig. 2.5.32

2.5.12 We denote with n the normal to the tangent t to γ in P (Figs. 2.5.31 and 2.5.32).
2.5 Torsion 241

When a thin walled beam, submitted to torsion, is constrained with the punctual
rigid joint (2.5.8), in the field of small displacements we can use the results of Saint
Venant. Then, applying the hydrodynamic analogy to this particular cross section,
it is possible to build some technical theory2.5.13 that allow to approximate with
simple formulas the state of stress of the beam.
An important class of closed thin walled cross sections is when the area A is
two-connected.2.5.14 We denote with γ 1 , γ 2 the closed regular curves that constitute
the frontier ∂A of A (Fig. 2.5.33). For such type of cross section the hydrodynamic
analogy shows that in every point P of every chord c (Fig. 2.5.34)

– τ z has constant intensity,

– τ z is parallel to the tangent to γ ,
– τ z is oriented according to Mt .

As a consequence, since the principle of constant flow (Remark 2.5.12), a real

number k exists such that

τz c = k

for all the points of γ . Such way

k
τz = . (2.5.14)
c

Fig. 2.5.33

Fig. 2.5.34

2.5.13 We call technical theory a procedure of calculation not supported by a mathematical model,
that is not simulated by a mathematical problem.
2.5.14 This is a definition of Mathematical analysis. In practice an area is two-connected if has a
cavity and only one.
242 2 The Problem of Saint Venant

Fig. 2.5.35

To obtain an operating expression for k, we impose the equilibrium to the rotation

around an axis z of the trunk of the beam individualized by every cross section K
and the right base (Fig. 2.5.35). We choose the axis z parallel to z and passing
through every point O of the plane K.
Insofar the τ z distributed on the area A of the cross section K must have moment
with reference to axis z equal to the applied twisting moment Mt .
We introduce on γ a curvilinear abscissa s, that is (Fig. 2.5.36) we choose a
point W of γ as origin and a verse to go along γ ; moreover to every point P
- W,P
of γ we associate the real number2.5.15 s = γ ds, taken with the sign plus
[resp. minus] if P follows [resp. precedes] W in the chosen verse to go along γ .
Then (Fig. 2.5.36)

– we consider two very near points P, Q of γ ,

– we put ds = s(Q) – s(P),
– we denote with c the chord in P and consider the chord in Q,

so that we individualize an elementary area dA of A.

Since τ z is constant in c and ds is very small, we can admit that τ z is constant in
dA. It is obvious then that the system of the τ z distributed on dA is equivalent to the
system constituted by the only force dF = τ z dA, equiverse to τ z and having as line
of action the tangent in P to γ (Fig. 2.5.36).
Insofar, denoting with h the arm of dF with respect to O (Fig. 2.5.36), we have
 
Mt = (dF)(h) = τz h dA. (2.5.15)
A A

Fig. 2.5.36

2.5.15 Inthe Integration theory we prove that such curvilinear integral measures the length of the
arc of γ individualized by the points W, P.
2.5 Torsion 243

Since τ z h is constant in c, from known theorems of the Integration theory it

follows that in the (2.5.15) we can substitute dA with c ds and the double inte-
gral double with a curvilinear integral along γ . This way, taking into account the
(2.5.14), we get
  
Mt = τz h dA = τz h c ds = k h ds. (2.5.16)
A γ γ

But h ds/2 is just the area dΩ of the triangle of vertexes 0, P, Q (Fig. 2.5.35),
so that h ds = 2 dΩ. As a consequence, denoting with Ω the area contained in the
middle curve γ (Fig. 2.5.37), we have
 
Mt = k h ds = k 2d = 2 k.
γ 

From this, taking into account the (2.5.14), we get the first formula of Bredt2.5.16

Mt
τz = . (2.5.17)
2c

After that, to calculate the unitary angle of torsion ϑ, we observe that the right
base of the beam is rotated, to happened deformation, of the angle ϑl. Then the
[1.6.5] furnishes for the strain work the expression

1
L= Mt ϑl. (2.5.18)
2

Fig. 2.5.37

2.5.16 R. Bredt (1896).

244 2 The Problem of Saint Venant

Because of the (1.6.13), (1.4.1), and (2.5.5) the internal energy E of the beam is

E= ϕdV
V
1 
= σx εx + σy εy + σz εz + τxy γxy + τxz γxz + τyz γyz dV
V 2
    
1   1 1
= τxz γxz + τyz γyz dV = τzx
2
+ τzy2
dV = τ 2 dV.
2 V 2G V 2G V z

Insofar, because of the Fubini2.5.17 theorem, we get

  I 
1 1
E= τz2 dV = τz2 dA dz.
2G V 2G 0 A

Since in the torsion the state of stress is independent from z, the double integral
is constant with respect to z. Consequently
  I 
1 l
E= τz2 dA dz = τz2 dA.
2G A 0 2G A

from which, since the (2.5.17), we have

   2
1 l l Mt
E= τz2 dA = τz2 c ds = cds
2G A 2G γ 2G γ 2c

lMt2 1
= ds.
8G2 γ c

So, taking into account the (1.6.8), we get


1 lMt2 1
Mt ϑl = L = E = ds
2 8G2 γ c

from which

Mt 1
ϑ= ds. (2.5.19)
4G2 γ c

The (2.5.19) is called second formula of Bredt.

Remark 2.5.14 In the type of cross section in examination, the result τ z c = k can
also be gotten with simple considerations of equilibrium. We consider in fact two
cross sections H, K of the beam (Fig. 2.5.35). We consider then the part of the beam

2.5.17 Guido Fubini, Venezia 1879 – New York 1943.

2.5 Torsion 245

Fig. 2.5.38

individualized by H, K and by two every chords c1 , c2 (Fig. 2.5.38). For such part
we impose the equilibrium to the translation along the axis z.
Since the principle of dissection, we apply therefore2.5.18 (Fig. 2.5.38)

– the stresses σ z , τ z on the individualized portions of the cross sections H, K,

– the stresses ση1 , τη1 ξ1 , τη1 z [resp. ση2 , τη2 ξ2 , τη2 z ] on the rectangle ABA B [resp.
CDC D ].

When we impose the equilibrium to the translation along the axis z, clearly only
the τη1 z , τη2 z give non-zero contributions. Moreover, since in the torsion the state
of stress is independent from z and the hydrodynamic analogy guarantees that τ z is
constant along every chord, the τη1 z [resp. τη2 z ] are constant on the rectangle ABA B
[resp. CDC D ]. Then, putting τ1 = τη1 z [resp. τ2 = τη2 z ], we have (Fig. 2.5.38)

τ1 c1 dist(H, K) − τ2 c2 dist(H, K) = 0

from which τ 1 c1 = τ 2 c2 . 

In the Structural engineering we meet a lot of closed thin walled cross sec-
tions in which the area A has a degree of connection greater than two. For them
an approximate but simple method of calculation can be employed that elaborates
the reasoning made to obtain the formulas of Bredt.
We exemplify such general method for a closed thin walled three-connected cross
section (Fig. 2.5.39). The middle curve of the section is constituted by three arcs of
curve γ 1 , γ 2 , γ 3 , having end points H, K (Fig. 2.5.40). We orientate these arcs of
curve so that we can go counterclockwise along the closed curves γ 1 , γ 2 and γ 1 ,
γ 3 (Fig. 2.5.40).

2.5.18 We denote with A, B [resp. C, D] the end points of the chord c1 [resp. c2 ]; with ξ 1 [resp. ξ 2 ]
the line passing through A and B [resp. C and D]; with η1 [resp. η2 ] a line normal to ξ 1 [resp. ξ 2 ]
and z.
246 2 The Problem of Saint Venant

Fig. 2.5.39

Fig. 2.5.40

Fig. 2.5.41

For such type of cross section the hydrodynamic analogy shows that in every
chord c (Fig. 2.5.41)

– τ z has constant intensity,

– τ z is parallel to the tangent to the middle curve,
– τ z is oriented according to Mt .

As a consequence, since the principle of constant flow (Remark 2.5.12), three

real number t1 , t2 , t3 exist such that

τz c = t1 in γ1
τz c = t2 in γ2 (2.5.20)
τz c = t3 in γ3 .

We assume as unknown of the problem the four real numbers t1 , t2 , t3 ,ϑ, where ϑ
denotes the unitary angle of torsion. We will now write a linear algebraic system
of four equations in the four unknown t1 , t2 , t3 ,ϑ, with non singular matrix of the
coefficients. Its unique solution resolves simply and with satisfactory approximation
the problem of the torsion for the closed thin walled three-connected cross section
of Fig. 2.5.39. In fact with the (2.5.20) we can calculate the τ z in any point of the
cross section.
The first equation of the resolving system is gotten applying the theorem of the
divergence (Remark 2.5.10) to the portion W of A individualized by a chord c1 of
2.5 Torsion 247

Fig. 2.5.42

γ 1 , by a chord c2 of γ 2 , by a chord c3 of γ 3 , all chosen very near to H (Fig. 2.5.42).

Because of the theorem of the divergence, the flow of τ z going out from ∂W must be
zero. Insofar, keeping in mind of the Remark 2.5.11 and of the (2.5.20) and (2.2.3),
we have (Figs. 2.5.41 and 2.5.42)

t1 − t2 − t3 = 0. (2.5.21)

As seen in the Remark 2.5.14, the (2.5.21) is an equilibrium condition.

The second equation of the resolving system is gotten imposing the equilib-
rium to the rotation around an axis z of the trunk of the beam individualized by
every cross section K and the right base (Fig. 2.5.37). We choose the axis z par-
allel to z and passing through every point O of the plane K (Fig. 2.5.43). Insofar
the τ z distributed on the area A of the cross section K must have moment with
reference to axis z equal to the applied twisting moment Mt . We denote with
A1 the part of A individualized from the middle curve γ 1 and with Ω 1 the area
individualized from γ 1 and the points O, H, K (Fig. 2.5.43). We assume the coun-
terclockwise moments positive and calculate the moment M1 that the τ z distributed
on the area A1 have with reference to axis z , that is with respect to O. The same
procedure that has conducted to the formulas of Bredt furnishes, keeping in mind
of the (2.5.20)
  
M1 = dM = (dF)(h) = (τz dA)(h)
A1 A1 A1
  H,K  H,K 
= τz h dA = τz h c ds = t1 h ds = t1 2d = 2t1 1.
A1 γ1 γ1 1
(2.5.22)

Analogously, we denote with A2 the part of A individualized from the middle

curve γ 2 and with Ω 2 the area individualized from γ 2 and the points O, H, K

Fig. 2.5.43
248 2 The Problem of Saint Venant

Fig. 2.5.44

(Fig. 2.5.44). Then the moment M2 that the τ z distributed on the area A2 have with
respect to O is
  
M2 = dM = − (dF)(h) = − τz h dA
A2 A2 A2
 H,K  H,K  (2.5.23)
=− τz h c ds = −t2 h ds = −t2 2d = −2t2 2.
γ2 γ2 2

Analogously, we denote with A3 the part of A individualized from the middle

curve γ 3 and with Ω 3 the area individualized from γ 3 and the points O, H, K
(Fig. 2.5.45). Then the moment M3 that the τ z distributed on the area A3 have with
respect to O is
  
M2 = dM = − (dF)(h) = τz h dA
A3 A3 A3
 H,K  H,K  (2.5.24)
= τz h c ds = t3 h ds = t3 2d = 2t3 3 .
γ3 γ3 3

As a consequence, we have the condition of equilibrium

21 t1 − 22 t2 + 23 t3 = Mt . (2.5.25)

The third and fourth equation of the resolving system is gotten using the theorem
of Stokes.2.5.19 First of all, because of the [2.5.3], we observe that

Fig. 2.5.45

2.5.19 SirGeorge Gabriel Stokes, Skreen 1819 – Cambridge 1903. The theorem of Stokes decrees
that: Let D be a portion of the three-dimensional space, S be a regular surface contained in D and
2.5 Torsion 249

Fig. 2.5.46

curl τ z = 2G ϑ in A.

Now we denote with S12 [resp. S13 ] the portion of the cross section individualized
from the closed curve constituted by γ 1 and γ 2 [resp. γ 1 and γ 3 ] and oriented as
in Fig. 2.5.47 [resp. 2.5.48]. After that we prolong on S13 the function τ z so that it
results

curl τ z = 2G ϑ in S13 . (2.5.26)

Then we apply the theorem of Stokes to the regular surface S12 bounded
from the closed curve constituted by γ 1 and γ 2 . Taking into account the direc-
tion of τ z (Fig. 2.5.41), the fact that everywhere curl τ z is normal to S12 , the
(2.5.20) and (2.5.26), the fact that τ z is tangent to γ 1 and γ 2 in every point, we
have (Fig. 2.5.47)
  
2GϑS13 = 2Gϑ dσ = 2G ϑdσ = curl τ z dσ
S12 S12 S12
  H,K  K,H
= (curl τ z × n)dσ = (τ z × t)ds + (τ z × t)ds
S12 γ1 γ2
 H,K  K,H  H,K  K,H
t1 t2
= τz ds + τz ds = ds + ds
γ1 γ2 γ1 c γ2 c

Fig. 2.5.47

bounded by a closed curve γ , ν be a vectorial function defined in D. Then the circulation of ν

along γ is equal to the flow of curl ν through S (Fig. 2.5.46).
250 2 The Problem of Saint Venant

Fig. 2.5.48

from which
 H,K  K,H
1 1
t1 ds + t2 ds − 2G S12 ϑ = 0. (2.5.27)
γ1 c γ2 c

Finally we apply the theorem of Stokes to the regular surface S13 bounded from
the closed curve constituted by γ 1 and γ 3 . Taking into account the direction of
τ z (Fig. 2.5.41), the fact that everywhere curl τ z is normal to S13 , the (2.5.20) and
(2.5.26), the fact that τ z is tangent to γ 1 and γ 3 in every point, we have (Fig. 2.5.48)

  
2GϑS13 = 2Gϑ dσ = 2Gϑdσ = curl τ z dσ
S13 S13 S13
  H,K  K,H
= (curl τ z × n)dσ = (τ z × t)ds + (τ z × t)ds
S13 γ1 γ3
 H,K  K,H  H,K  K,H
t1 t3
= τz ds + τz ds = ds + ds
γ1 γ3 γ1 c γ3 c

from which
 H,K  K,H
1 1
t1 ds + t3 ds − 2GS13 ϑ = 0. (2.5.28)
γ1 c γ3 c

The (2.5.21), (2.5.25), (2.5.27), and (2.5.28), constitute the resolving system for
the closed thin walled three-connected cross section of Fig. 2.5.39. We can verify
that it is a linear algebraic system of four equations in the four unknown t1 , t2 , t3 , ϑ,
with non singular matrix of the coefficients.

2.5.5 Open Thin Walled Cross Section

We will afterwards see that the optimal cross section for a beam submitted to torsion
is really the closed thin walled cross section. Nevertheless it is possible that a beam
with an open thin walled cross section is forced to bear a torsional load. In such case
2.5 Torsion 251

Fig. 2.5.49

too it is also possible to calculate the state of stress in an approximate way but of
simple application.
We begin such analysis studying the torsion of a beam having for cross section
a lengthened rectangle (Fig. 2.5.49). The length of the long [resp. short] side is 2a
[resp. 2b], where a » b. We will now draw, employing the hydrodynamic analogy,
same expressions for τ z and ϑ that are as precise as a is greater than b.
It is known from the Dynamics of fluids that for the cross section A of Fig. 2.5.49,
introducing the reference frame of Fig. 2.5.50, if a » b then

– the paths of the fluid particles, and therefore the streamlines, are practically
parallel to axis y,
– the velocity of the fluid is independent from y,
– in the center of the rectangle the speed of the fluid is zero.

Since in every point P of A the velocity is tangent to the streamline passing

through P, as a consequence, if the rectangle is very lengthened, we have practically
everywhere τ z parallel to the long side of the rectangle. Insofar we have, with good
approximation (Fig. 2.5.50)

τzx = 0 in A. (2.5.29)

From the [2.5.3], (2.5.29) we get

∂τzy ∂τzx ∂τzy

2Gϑ = curl τ z = − = ,
∂x ∂y ∂x
from which, since τ z (0) = 0, it follows

τzy = 2Gϑx in A. (2.5.30)

To gain a more operating expression for τ z , we impose the equilibrium to the

rotation around an axis z of the trunk of the beam individualized by every cross
section K and the right base (Fig. 2.5.37). We choose the axis z parallel to z and

Fig. 2.5.50
252 2 The Problem of Saint Venant

Fig. 2.5.51

passing through the point O of the plane K (Fig. 2.5.51). Insofar the τ z distributed
on the area A of the cross section K must have moment with respect to axis z equal
to applied twisting moment Mt . We now consider the stream tube individualized by
the streamline x ∈ ]0, b[ and by the streamline x + dx, where dx denotes a positive
number as small as we like (Fig. 2.5.51). We denote

– with dA the infinitesimal area of the stream tube,

– with ∂dA the boundary of dA,
– with dM the moment of the τ z distributed on dA, with respect to O.

Clearly, since dA ⊆ A, we have

div τ z = 0 in dA
curl τ z = 2Gϑ in dA.
Moreover, since ∂dA is constituted by two streamlines, it results

τz × n = 0 in ∂dA.

Insofar the system of the τ z distributed on dA coincides with the system of the τ z
that arises in the closed thin walled cross section dA when loaded with the twisting
moment dM (Fig. 2.5.52). As a consequence, since the (2.5.17) and (2.5.29) we get

dM
τzy = τz = . (2.5.31)
2dx

Since the rectangle is very lengthened, the area Ω contained in the middle curve
of the stream tube of Fig. 2.5.52 is equal, with good approximation, to the area of

Fig. 2.5.52
2.5 Torsion 253

the rectangle having base 2a and height 2x. Consequently, taking into account the
(2.5.30), from the (2.5.31) we get

dM = τzy 2dx = 16Gϑax2 dx.

So
  b b3
Mt = dM = 16Gϑax2 dx = 16Gϑa
0 3

from which

3 Mt
ϑ= (2.5.32)
16 Gab3

from which, because of the (2.5.30)

3 Mt
τzy = x, (2.5.33)
8 ab3

whose diagram is given in Fig. 2.5.53.

Remark 2.5.15 It is obvious that the (2.5.33) approximates the state of stress in the
cross section anywhere except in the ends (Fig. 2.5.54). 

Remark 2.5.16 It is possible to get the (2.5.32) from the (2.5.33) imposing that the
strain work L is equal to the internal energy. In fact, proceeding in the way followed
for the two-connected closed thin walled cross section, we have

Fig. 2.5.53

Fig. 2.5.54
254 2 The Problem of Saint Venant

1 1
Mt ϑl = L = E = (σx εx + σy εy + σz εz + τxy γxy + τxz γxz + τyz γyz )dV
2 V 2
   I 
1 1 1
= τzy γzy dV = τ dV =
2
τzy
2
dA dz
2 V 2G V zy 2G 0 A
 !  I
1 3 Mt 2 9 l Mt2 3 l Mt2
= 3
x dA dz = 2 3
Iy =
2G A 8 ab 0 128G a b 32G ab3

from which

3Mt
ϑ= . 
16G ab3

The structural steel is a type of beam with open thin walled cross section that
sometimes is forced to bear a torsional load. For such type of beam an approximate
but simple method of calculation can be useed that employs the results gotten for
the beam having for cross section a lengthened rectangle.
We exemplify such method for an H-beam (Fig. 2.5.55). The Fluid dynamics
shows that in such case the paths of the fluid particles, and then the streamlines,
and the velocity diagrams are those of (Fig. 2.5.56). We choose a point O of the
cross section and denote with M1 [resp. M2 ] [resp. M3 ] the moment with respect to
O of the τ z distributed on the rectangle ABCD [resp. EFMN] [resp. LGHI]. Now
we dissect the beam of Fig. 2.5.55 with the plane individualized by the axis z and

Fig. 2.5.55

Fig. 2.5.56
2.5 Torsion 255

Fig. 2.5.57

by the chord EN and with the plane individualized by the axis z and by the chord
FM (Fig. 2.5.56). This way we obtain the three cantilevers of Fig. 2.5.57, whose
cross sections are lengthened rectangles. We apply on the upper [resp. middle] [resp.
lower] cantilever the twisting moment M1 [resp. M2 ] [resp. M3 ] (Fig. 2.5.57). We
apply in the cantilevers of Fig. 2.5.57 on the planes of cut the σ and the τ present in
the beam of Fig. 2.5.55. Then, since the principle of dissection [1.2.1], the cantilever
of Fig. 2.5.55 and those of Fig. 2.5.57 have the same state of strain, and then the
same state of stress.
From the (2.5.5) we easily get that in Fig. 2.5.55 the σ acting on the planes of
cut are zero everywhere. On the contrary, since the symmetry of τ , in Fig. 2.5.55
the τ acting on the planes of cut are non-zero. However we observe that, using the
hydrodynamic analogy, in the cantilevers of Fig. 2.5.57, disregarding such τ and
applying only the twisting moments M1 , M2 , M3 , we obtain a distribution of τ z
(Fig.2.5.58) whose streamlines and diagram are practically equal to those present in
the assigned beam (Fig. 2.5.56). This result holds everywhere, except in proximity
of the chords EN, FM.
For this reason we apply on the cantilevers of Fig. 2.5.57 only the twisting
moments M1 , M2 , M3 and we approximate with the solution of such problem the
solution of the torsion of the H-beam of Fig. 2.5.55. Insofar, to determinate such
solution, it is enough to calculate M1 , M2 , M3 . In fact in Fig. 2.5.57 we have the
torsion of beams having for cross section a lengthened rectangle, already examined
in precedence.

Fig. 2.5.58
256 2 The Problem of Saint Venant

We will now write a linear algebraic system of three equations in the three
unknown M1 , M2 , M3 , with non singular matrix of the coefficients.
The first equation of the resolving system is gotten imposing the equilibrium to
the rotation around an axis z of the trunk of the beam individualized by every cross
section K and the right base (Fig. 2.5.37). We choose the axis z parallel to z and
passing through the point O of the plane K. Insofar the τ z distributed on the area
A of the cross section K must have moment with respect to axis z equal to applied
twisting moment Mt . So, with our notations, we must have

M1 + M2 + M3 = Mt . (2.5.34)

The second and third equations are compatibility equations. We denote with ϑ the
unitary angle of torsion of the cantilever of Fig. 2.5.55 and with ϑ1 [resp. ϑ2 ] [resp.
ϑ3 ] the unitary angle of torsion of the upper [resp. middle] [resp. lower] cantilever
of Fig. 2.5.57. Because we have substantially satisfied the principle of dissection
[1.2.1], the generic cross section at the abscissa z must rotate practically of the same
angle in both problems of Figs. 2.5.55 and 2.5.57. So we must have (Fig. 2.5.59)

ϑz = ϑ1 z = ϑ2 z = ϑ3 z,

from which

ϑ1 = ϑ2 , ϑ1 = ϑ3 ,

from which, since the (2.5.32)

M1 M2
=
a1 b31 a2 b32
(2.5.35)
M1 M3
= .
a1 b31 a3 b33

The (2.5.34) and (2.5.35) constitute a linear algebraic system with non singular
matrix of the coefficients. Since the (2.5.33), its unique solution M1 , M2 , M3 allows
to calculate all the stresses in the H-beam of Fig. 2.5.55.

Fig. 2.5.59
2.5 Torsion 257

Remark 2.5.17 In the previous analysis the compatibility condition ϑ1 = ϑ2

clearly guarantees that there are neither laceration nor overlapping of material in
correspondence of the chord EN. 

Remark 2.5.18 Obviously in the parts of cross section broken in Fig. 2.5.60 the
preceding technical theory doesn’t furnish reliable values of the state of stress.
We notice that usually in such parts there are reentering angles where we have a
convenient fitting. 

The technique explained for the H-beam of Fig. 2.5.55 can be used to any open
thin walled cross section composed by lengthened rectangles. For some sections
of frequent employment (Fig. 2.5.61) we give in Fig. 2.5.62 the streamlines, that
suggest the correct decomposition in lengthened rectangles (Fig. 2.5.63).

Fig. 2.5.60

Fig. 2.5.61

Fig. 2.5.62
258 2 The Problem of Saint Venant

Fig. 2.5.63

Fig. 2.5.64

Fig. 2.5.65

Fig. 2.5.66

Such technique can be adapted to other thin walled cross section. In fact for the
beam of Fig. 2.5.64, that has a thin walled cross section (Fig. 2.5.65), the Fluid
dynamics gives the streamlines and the speed diagrams of Fig. 2.5.66.
Reasoning as for the H-beam of Fig. 2.5.55, we decompose the cross section of
Fig. 2.5.65 in the lengthened rectangle ABCL, in the closed thin walled cross section
EFMN and in the lengthened rectangle DEFG (Fig. 2.5.67). It is very important to
observe that we are able to operate the decomposition in such way only because
the new sections have the streamlines of Fig. 2.5.68, which are practically equal

Fig. 2.5.67

Fig. 2.5.68
2.5 Torsion 259

Fig. 2.5.69

to the ones of Fig 2.5.66. In fact they substantially only differ in correspondence
of the chords CL, DG. So we choose a point O of the cross section and denote
with M1 [resp. M2 ] [resp. M3 ] the moment with respect to O of the τ z distributed
on the portion of cross section ABCL [resp. EFMN] [resp. DEFG] (Fig. 2.5.65).
Then, proceeding as for the H-beam of Fig. 2.5.55, we obtain the three cantilevers
of Fig. 2.5.69, on which we apply the twisting moments M1 , M2 , M3 . Then, since the
principle of dissection [1.2.1], with good approximation the cantilever of Fig. 2.5.64
and those of Fig. 2.5.69 have the same state of strain, and then the same state of
stress.
Reasoning as for the H-beam of Fig. 2.5.55, we obtain the linear algebraic system
of three equations in the three unknown M1 , M2 , M3 , with non singular matrix of
the coefficients:

M1 + M2 + M3 = Mt

3 M1 M2 1
= ds
16G a1 b31 4G2 γ c
M1 M3
3
= .
a1 b1 a3 b33

Its unique solution furnishes, via the (2.5.33) and (2.5.17), the state of stress in
the cantilever of Fig. 2.5.64. This approximate analysis is satisfactory everywhere,
except for the parts of cross section broken in Fig. 2.5.70. But usually in such parts
there are reentering angles where we have a convenient fitting of material.

Remark 2.5.19 When, using the hydrodynamic analogy, we decompose a thin walled
cross section in closed parts and open parts, we can simplify the previous procedure
by observing that the open parts absorb an aliquot of the twisting moment substan-
tially negligible. We can easily verify such property for the section of Fig. 2.5.71.
Let us denote with Ma [resp. Mb ] the aliquot of Mt absorbed by the closed [resp.
open] part of the cross section and with Ca [resp. Cb ] the torsional rigidity of the

Fig. 2.5.70
260 2 The Problem of Saint Venant

Fig. 2.5.71

closed [resp. open] part of the cross section. Clearly Ca is much greater than Cb .
Then from the compatibility condition

Ma Mb
ϑ= =
Ca Mb
and the thesis follows. 

Remark 2.5.20 In the previous procedure of calculation when we decompose the

cross section the respect of the hydrodynamic analogy is obligatory. In fact if we
design the structure of Fig. 2.5.64 with the decomposition of Fig. 2.5.72, the collapse
of the structure assuredly happens. 

Remark 2.5.21 In the open thin walled cross section of Fig. 2.5.73 the material is
arranged in close proximity of a middle curve γ (Fig. 2.5.74). For this cross section
the hydrodynamic analogy shows that

Fig. 2.5.72

Fig. 2.5.73

Fig. 2.5.74
2.5 Torsion 261

Fig. 2.5.75

– the streamlines and the speed diagram are those of Fig. 2.5.75,
– if the chord c is constant along γ , then the diagram of speed is substantially
constant along γ , except for the ends,
– if the chord c is constant along γ , the diagram of velocity is practically equal to
the velocity which is verified in the lengthened rectangular cross section having
base equal to the length lγ of γ and height c (Fig. 2.5.76).
Insofar in the problem of the torsion the opened thin walled cross section of
Fig. 2.5.73 can be approximate, if the chord is constant, with the lengthened
rectangular cross section of Fig. 2.5.76. 

We now perform a simple analysis that furnishes important indications on the

way to optimize the form of a cross section submitted to torsion. We consider the
cantilever of Fig. 2.5.77, submitted to the twisting moment Mt and having the closed
thin walled cross section of Fig. 2.5.78. We specify in Fig. 2.5.79 its dimensions,

Fig. 2.5.76

Fig. 2.5.77

Fig. 2.5.78
262 2 The Problem of Saint Venant

Fig. 2.5.79

Fig. 2.5.80

where d is supposed really smaller than a. We compare such cantilever with the
cantilever of Fig. 2.5.80, submitted to the same twisting moment Mt and having the
open thin walled cross section of Fig. 2.5.81. This beam is gotten from the one of
Fig. 2.5.77 performing a cut according the plane individualized from the axis of the
beam and from the chord HH , a cut according the plane individualized from the
axis of the beam and from the chord KK (Fig. 2.5.79) and removing the resulting
thin strip of material.
Evidently the two beams of Fig. 2.5.77 and of Fig. 2.5.80, that are submitted to
the same twisting moment Mt , have practically the same weight and the same form.
Nevertheless the beam with open cross section is submitted to great stresses and
deforms itself even more than the beam with closed cross section.
In fact let us denote with τ a [resp. τ c ] the maximum shearing stress in the open
[resp. closed] thin walled cross section of Fig. 2.5.82 [resp. 2.5.78]. Clearly the

Fig. 2.5.81

Fig. 2.5.82
2.5 Torsion 263

Fig. 2.5.83

middle curve γ of such open thin walled cross section has length lγ = 4a. Then,
since the Remark 2.5.21 we can calculate τ a in Fig. 2.5.83. So, since the (2.5.33), it
results

3 Mt 3Mt
τa =  2 = . (2.5.36)
8 4ad2
(2a) d2

Moreover in Fig. 2.5.79 the area contained in the middle curve is practically Ω
= a2 . So with the formula of Bredt (2.5.17) we get

Mt
τc = . (2.5.37)
2a2 d
From the (2.5.36) and (2.5.37) we get

3Mt
τa 4ad2 a
= = 1.5 .
τc Mt d
2a2 d
Insofar

τa  τc .

Let us denote now with ϑa [resp. ϑc ] the unitary angle of torsion of the open [resp.
closed] thin walled cross section of Fig. 2.5.81 [resp. 2.5.78]. Since the (2.5.32) and
(2.5.19) we have
 
3 Mt
 3
3  a 2
16 G
ϑa (2a) d2
=   =
ϑc 4 d
Mt  
4 G(a2 )2 4da

so that the beam with open cross section deforms itself even more than the beam
with closed cross section.
This simple analysis shows the closed thin walled cross section is very more able
to absorb the torsion than the open thin walled cross section.
264 2 The Problem of Saint Venant

Fig. 2.5.84

Problem 2.5.3 Two steel cantilevers of length l are both submitted to a twisting
moment Mt . They have as cross section a circle of radius R and a lengthened
rectangle of base π R and height R (Fig. 2.5.84), so that the beams have equal
weight. We denote with with τ 1 [resp. τ 2 ] the maximum shearing stress in the rect-
angular [resp. circular] cross section and with ϑ1 [resp. ϑ2 ] the unitary angle
of torsion of the rectangular [resp. circular] cross section. You shall determine
τ 1 /τ 2 and ϑ1 /ϑ2 .
Solution We easily get, using the previous theoretical results

3Mt R/2 3Mt R/2

τ1 (π R/2)(R/2)3 (π R/2)(R/2)3 3
= = = ,
τ2 Mt R 2Mt R 2
Ip π R4
3 Mt 3 Mt
ϑ1 16G  π R  R 3 16 G  π R  R 3 3
=  2
2 =  2
2 = .
ϑ2 Mt 2Mt 2 
GIp G π R4

2.5.6 Non Uniform Torsion

In constructive practice it can happen that in a cantilever, submitted to twisting

moment at the free end, the rigid jointed end is prevented to hunch. This is usu-
ally verified when the rigid joint of a steel cantilever is realized by welding the base
on a thick steel plate. Another cause that prevents the hunch is the symmetry, as it
happens for the middle cross section in Fig. 2.5.85. In such cases it is experimen-
tally verified that the preceding analyses of the torsional problem furnish incorrect
values of stress. Such problem is called non uniform torsion.
For thin walled cross section we can analyze the non uniform torsion with a
technical theory according to which

Fig. 2.5.85
2.5 Torsion 265

Fig. 2.5.86

– the real state of stress is sum of the τ̂z that arise when the beam is free to hunch,
and of non negligible secondary stresses σz, τ̃z ,
– the unitary angle of torsion is not constant2.5.20 along the axis z of the beam.

In such technical theory we denote with (Fig. 2.5.86)

– G, x, y, z a reference frame in which G is the centroid of the left base, x, y

rigorously are inertia principal axes, z is the axis of the beam,
– ρ the middle curve of the thin walled cross section,
– t the tangent to ρ in the generic point of ρ,
– A the area of the cross section,
– ∂A the boundary of A.

Moreover we assume that

– the only load applied on the beam is a twisting couple Mt , acting on the right base,
– the cross section of abscissa z ∈ ]0, l[ hunches and rigidly rotates around the axis
z of the angle ϑ(z),
– the beam is constrained from a punctual rigid joint in the centroid G of the left
base,
– the left base is prevented to hunch, i.e.

w(x, y, 0) = 0 ∀(x, y) ∈ A, (2.5.38)

– in every point of ρ it results

γzt = 0, (2.5.39)

– everywhere

σz = Eεz . (2.5.40)

2.5.20 This is the motive for which such problem is called non uniform torsion.
266 2 The Problem of Saint Venant

Remark 2.5.22 For hypothesis, the cross section of abscissa z [resp. z+dz] rigidly
rotates of ϑ(z) [resp. ϑ(z + dz)]. Approximating ϑ(z + dz) with ϑ(z) + ∂ϑ
∂z (z)dz, the
∂ϑ
unitary angle of torsion is ∂z (z). 

By hypothesis, in this technical theory the cross section of abscissa z ∈ ]0, l[

rigidly rotates around the axis z of the angle ϑ(z). As a consequence, since the
hypothesis of small deformations, in its points it results, as seen in the footnote
2.5.5

u = −ϑy, v = ϑx, (2.5.41)

so that

∂u ∂w dϑ ∂w
τzx = G + = G −y (z) +
∂z ∂x dz ∂x
∂v ∂w dϑ ∂w
τzy = G + = G x (z) +
∂z ∂y dz ∂y
so that

∂τzy ∂τzx dϑ
curl τ z = − = 2G (z). (2.5.42)
∂x ∂y dz

From the third one of the (1.2.3) we get

∂τzx ∂τzy ∂σz

div τ z = + =− . (2.5.43)
∂x ∂y ∂z

So the τ z that arise in the cross section are such that

dϑ
curl τ z = 2G (z) in A
dz
∂σz (2.5.44)
div τ z = − in A
∂z
τ z × n = 0 in ∂A.

It is mathematically obvious that the solution τz of the problem (2.5.44) is sum

of the solution τ̂z of the problem

dϑ
curl τ z = 2G (z) in A
dz
(2.5.45)
div τ z = 0 in A
τ z × n = 0 in ∂A
2.5 Torsion 267

and of the solution τ̃z of the problem

curl τ z = 0 in A
∂σz
div τ z = − in A (2.5.46)
∂z
τ z × n = 0 in ∂A.

About the problem (2.5.45), since in A

dϑ
(z) = const,
dz
taking into account the results obtained for the problem [2.5.2], we obviously get
that

– the problem (2.5.45) has a unique solution, that we denote τ̂z ,

– such unique solution is just the solution of the problem of the beam, supposed free
to hunch, submitted to the twisting moment

dϑ
M̂t = C (z) (2.5.47)
dz
where C denotes the torsional rigidity of the cross section,
– we can determinate such unique solution exactly with the formulas of Saint Venant
or approximatively with the previous technical theories given for the closed or
open thin walled cross section.

About the problem (2.5.46), first of all we prove that in every chord it results
τ.z = .
τzt = const. To this aim we consider a point P of the middle curve ρ, the
tangent t to ρ, the normal r to t passing through P. From the first of (2.5.46) it
follows
∂ τ̃zt ∂ τ̃zr
− = curl τ̃ z = 0. (2.5.48)
∂r ∂t
Since the third of (2.5.46) τ̃z = τ̃zt in the points of ∂A. Then, since the chord is
short, approximatively it results τ̃zr = 0 along the chord. So the (2.5.46) furnishes

∂ τ̃zt
= 0,
∂r
from which we get that in every chord it results τ̃z = τ̃zt = const. As a consequence,
it is enough to determine the value of τ.z only in the points of ρ.
Let P be a point of ρ. As seen in footnote 2.5.5, since the (2.5.41) P suffers a
displacement s = ϑ(z) dist(G, P) normal to the segment GP (Fig. 2.5.87). Denoting
with st the orthogonal component of s on t and putting h = dist(G, t), clearly we
have (Fig. 2.5.87)
268 2 The Problem of Saint Venant

Fig. 2.5.87

st = s sinα = ϑ(z) dist(O, P) sinα = h ϑ(z). (2.5.49)

From the (2.5.39) it follows

∂st ∂w
0= +
∂z ∂t
so that, since the (2.5.49)

∂w ∂st dϑ
=− = −h . (2.5.50)
∂t ∂z dz
Denoting with P0 a point of ρ (Fig. 2.5.87), from the (2.5.50) we get
 
P0 ,P ∂w P0, P dϑ
ds = −h ds;
ρ ∂t ρ dz

so that
 
P0, P ∂w dϑ P0, P
w(P) − w(P0 ) = ds = − h ds. (2.5.51)
ρ ∂t dz ρ

We denote with dΩ the area 2hds of the triangle of Fig. 2.5.88, having base ds
and height h, and with Ω(P) the area individualized by ρ, G, P, P0 (Fig. 2.5.89), that
we call sectoral area. Then from the (2.5.51) it follows
 P0, P
dϑ dϑ
w(P) + w(P0 ) − d = w(P0 ) − 2(P) . (2.5.52)
dz ρ dz

Fig. 2.5.88
2.5 Torsion 269

Fig. 2.5.89

Since Ω(P) is constant with respect to z, from the (2.5.40) and (2.5.52) we get

∂w d2 ϑ
σz = E (P0 ) − 2E(P) 2 . (2.5.53)
∂z dz
Since the external load is only a twisting moment, we must have
  
∂w d2 ϑ
0= σz dA = E (P0 ) dA − 2E  dA;
A ∂z A dz2 A

so that

∂w 2 d2 ϑ
(P0 ) =  dA
∂z A dz2 A

from which, taking into account the (2.5.53)

  
d2 ϑ 1
σz = σz (P) = −2E 2 (P) −  dA . (2.5.54)
dz A A

From this, denoting with c(P) the length of the chord and putting

1
b= c(Q)(Q)ds (2.5.55)
A ρ

we have

d2 ϑ
σz = −2E [(P) − b]. (2.5.56)
dzz

Let us now turn our attention to the problem of calculating the secondary stresses
τ̃ z . We decompose the middle curve ρ in the branches ρ 1 , ρ 2 , ρ 3 (Fig. 2.5.90). We
call closed [resp. open] a branch that belongs [resp. not belongs] to a closed curve.
In a branch ρ i we denote with Pi an end point and with P a point. We denote with
Ai the part of cross section individuated by Pi and P and with ∂Ai its boundary (Fig.
2.5.91). Applying the divergence theorem we get
270 2 The Problem of Saint Venant

Fig. 2.5.90

Fig. 2.5.91

 
∂ τ̃zx ∂ τ̃zy
div τ̃ z dA = + dA
Ai Ai ∂x ∂y
 
= (τ̃zx nx + τ̃zy nx )ds = (τ̃ z × n)ds.
∂Ai ∂Ai

From this, taking into account the second and the third of the (2.5.46), since with
good approximation τ̃z is constant on every chord and normal to it, we obtain
  
∂σz
− dA = div τ̃ z dA = (τ̃ z × n)ds
Ai ∂z Ai ∂Ai
=.
τz (P)c(P) − .
τz (Pi )c(Pi ).

As a consequence, taking into account the (2.5.55)


d3 ϑ
.
τz (P)c(P) = .
τz (Pi )c(Pi ) + 2E[(P) − b]dA
Ai dz3
 (2.5.57)
d3 ϑ Pi ,P
=.
τz (Pi )c(Pi ) + 2E 3 c(Q)[(Q) − b]ds.
dz ρi

If ρ i is an open branch, choosing as Pi an end point of ρ we easily have

.
τz (Pi ) = 0. (2.5.58)

If ρ i is a closed branch, we calculate the value of .

τz (Pi ) applying the divergence
theorem in the branch points and the theorem of Stokes to closed curves of ρ.
For instance in the branch ρ i of Fig. 2.5.92 we choose P1 =P2 =P3 =A and apply
the divergence theorem in the branch point A (Fig. 2.5.93). Since the area W of
Fig. 2.5.93 is infinitesimal, we approximatively have
2.5 Torsion 271

Fig. 2.5.92

Fig. 2.5.93

 
0= τ z dW =
div. τ z × n)ds
(.
W ∂W (2.5.59)
=.
τz (P1 )c(P1 ) − .
τz (P2 )c(P2 ) − .
τz (P3 )c(P3 ).

Applying the theorem of Stokes to plane surface S12 bounded from the closed
curve constituted by ρ 1 and ρ 2 (Fig. 2.5.94), we get
 
τ z × n)dS =
(curl. τ z × t)ds.
(.
S12 ρ12

From this, taking into account the (2.5.46), we get (Fig. 2.5.93)

   A,B  B,A
0= (curl .
τ z × n)dS = .
τz ds = .
τz ds + .
τz ds
S12 ρ12 ρ1 ρ2
 A,B  B,A
1 1
= .
τz (P)c(P) ds + .
τz (P)c(P) ds.
ρ1 c(P) ρ2 c(P)

Fig. 2.5.94
272 2 The Problem of Saint Venant

From this, taking into account the (2.5.57), we get

 A,B 1
0= {.τz (P1 )c(P1 )
ρ1 c(P)
 "  B,A
d3 ϑ A,P 1
+2E 3 c(Q)[(Q) − b]dr ds + {.τz (P2 )c(P2 )
dz ρ1 ρ2 c(P)
 "  A,B
d3 ϑ A,p 1
+2E 3 c(Q)[(Q) − b]dr ds = . τz (P1 )c(P1 ) ds
dz ρ2 ρ1 c(P)
 #  A,P " (2.5.60)
d3 ϑ A,B 1
+2E 3 c(Q)[(Q) − b]dr ds
dz ρ1 c(P) ρ1
 B,A
1
+.τz (P2 )c(P2 ) ds
ρ2 c(P)
 #  A,P "
d3 ϑ B,A 1
+ 2E 3 c(Q)[(Q) − b]dr ds.
dz ρ2 c(P) ρ2

Applying the theorem of Stokes to plane surface S13 bounded from the closed
curve constituted by ρ 1 and ρ 3 (Fig. 2.5.94), we get an expression analogous to
(2.5.60). Such expression, together with the (2.5.60) and (2.5.59), constitutes a lin-
ear algebraic system of three equations in the three unknown . τz (P1 ), .
τz (P2 ), .
τz (P3 ),
with non singular matrix of the coefficients.
Such way in the non uniform torsion of a thin walled cross section, if we known
the function ϑ(z), then the primary [resp. secondary] stresses τ̂z [resp. . τz, σz ] are
completely calculated from the (2.5.45) [resp. (2.5.56) and (2.5.57)].
To determine ϑ(z) we observe that for an open cross section, since the (2.5.57),
(2.5.58), and (2.5.55) it results

 
.t (z) =
M τz dA)(h) =
(. .
τz ch ds
A ρ
 
d3 ϑ Pi ,P
= 2hE 3 c(Q)[(Q) − b]dr ds
ρ dz ρi
 
d3 ϑ Pi, P
= 4E 3 c(Q)[(Q) − b]dr d
dz ρ ρi
#    "
d3 ϑ Pi ,P
= 4E 3  c(Q)[(Q) − b]dr − c( − b)ds
dz ρi ∂ρ ρ
2.5 Torsion 273

 
d3 ϑ d3 ϑ
= − 4E c( − b)ds = −4E c( − b)2 ds
dz3 ρ dz3 ρ
 
d3 ϑ d3 ϑ
− 4E b c( − b)ds = −4E c( − b)2 ds.
dz3 ρ dz3 ρ

We call rigidity to hunch the constant


.
C = 4E c( − b)2 ds. (2.5.61)
ρ

Insofar

d3 ϑ
.t (z) = −.
M C . (2.5.62)
dz3

If the thin walled cross section is closed, we can verify that the (2.5.62) still sub-
sists. However the rigidity to the hunch has a different expression from the (2.5.61)
and this is because the (2.5.58) doesn’t subsist. As usual, we must employ the theo-
rem of Stokes a number of times which depends on the degree of connection of the
cross section.
Since the solution of the problem in examination (2.5.44) is just the sum of the
solution / τz of the problem (2.5.45) and the solution τ˜z of the problem (2.5.46), then
in every cross section of the cantilever it must result

/t (z) + M̃t (z) = Mt .

M

From this and from the (2.5.47) and (2.5.62) we get the resolving differential
equation of the non uniform torsion

dϑ . d3 ϑ
C − C 3 = Mt in ]0, l[. (2.5.63)
dz dz

The boundary conditions are

ϑ(0) = 0
dϑ
(0) = 0
dz
σz (x, y, l) = 0 ∀(x, y) ∈ A

or, since the (2.5.56)

274 2 The Problem of Saint Venant

ϑ(0) = 0
dϑ
(0) = 0
dz (2.5.64)
d2 ϑ
(l) = 0.
dz2

Since C, .
C ∈ ]0, +∞[, we put

1
C 2
α= .
.
C

The (2.5.63) is a third order ordinary differential equation with constant coeffi-
cients. We easily verify that its general integral is

Mt
ϑ(z) = d1 sinh(az) + d2 cosh(az) + z + d1
α 2.
C
and that the solution of the boundary problem (2.5.63) and (2.5.64) is

Mt
ϑ(z) = {αz − sinh(αz) + tgh(αl)[ cosh(αz) − 1]} . (2.5.65)
αC
From the (2.5.65) and (2.5.62) it follows

.t (z) = Mt [cosh(αz) − tgh(αl)sinh(αz)]

M (2.5.66)

so that

Mt
.t (0) = Mt , M
M .t (l) =
cosh(αl)
so that the twisting moment Mt is almost exclusively absorbed by the secondary
stresses τ.z in proximity of the joint. However the (2.5.66) shows that the perturba-
tion, caused by the fact that the constrained cross section is not free to hunch, is not
located in proximity of the joint, but in the big part of the beam (Fig. 2.5.95).

Fig. 2.5.95
2.5 Torsion 275

Fig. 2.5.96

Fig. 2.5.97

Problem 2.5.4 You shall determine the primary and secondary stresses in the
nonuniform torsion of the cantilever of Fig. 2.5.96. The cross section is the thin
walled H-section of Fig. 2.5.97 and the constrained section is prevented to hunch.

Proof First of all the function ϑ (z) needs to be determined. About the torsional
rigidity C of the cross section of Fig. 2.5.97, we suppose that every cross section
of the cantilever is free to hunch and that on the free end a twisting moment Mt
is applied. We denote with γ the unitary angle of torsion. From the (2.5.34) and
(2.5.35) we easily get

M1 M2
=
dc3 (h − c) c31
M1 Mt − M1 − M2
3
=
dc dc3
so that

Mt dc3
M1 =
(h − c) c31 + 2dc3

Fig. 2.5.98
276 2 The Problem of Saint Venant

so that

3 M1 3Mt
γ = γ1 = =  
16G dc3 16G (h − c) c31 + 2dc3

so that

16  
C= G (h − c) c31 + 2dc3 . (2.5.67)
3

About the rigidity to hunch C̃, in first place we determine the function sectorial
area Ω. To such purpose we assume the centroidal inertia reference frame G, x, y
of Fig. 2.5.98, G as pole of the sectorial areas and P0 as initial point (Fig. 2.5.98).
As a consequence, we obviously get the diagram of Ω of Fig. 2.5.99, which is
symmetrical with respect to both axes x and y. From this and from the (2.5.55) we
get b = 0. Insofar, since the (2.5.61), we have

E c h2 d3
C̃ = . (2.5.68)
24
The (2.5.67) and (2.5.68) furnish the value of α, from which, using the (2.5.65),
we obtain the rotation ϑ (z) of any cross section of the cantilever.
Let us consider now a cross section of abscissa z ∈ ]0, l[. The primary stresses
τ̂z are easily calculated resolving the problem of Fig. 2.5.100, in which every cross
section of the cantilever is free to hunch and, since the (2.5.66), on the free end a
twisting couple of intensity
 
M̂t (Z) = Mt − M̃t (z) = Mt 1 − cosh (αz) + tgh(al) sinh (αl)

acts.

Fig. 2.5.99

Fig. 2.5.100
2.5 Torsion 277

Fig. 2.5.101

About the secondary stresses σz, τ̃z , from the (2.5.74) it follows

d2 ϑ Mt α  
2
(z) = − sinh (αz) + tgh (αl) cosh (αz)
dz C
which, taking into account the (2.5.56) and the Fig. 2.5.99, furnishes the diagram
of σz . Such diagram is linear and also symmetrical with respect to axes x, y
(Fig. 2.5.101). 
About the τ̃z , from the (2.5.63) it follows

d3 ϑ Mt α 2  
(z) = − cosh (αz) + tgh (αl) sinh (αz)
dz3 C
which, together with the (2.5.57) and the Fig. 2.5.99, furnishes the diagram of
the τ̃z . Such diagram is parabolic and also symmetrical with respect to axes x, y
(Fig. 2.5.102). 

Remark 2.5.23 From the Fig. 2.5.101 it results that every wing of the H-section
is submitted to a bending moment (Fig. 2.5.103). For this motive, in the nonuni-
form torsion of thin walled beams, we can say that the torsion is partly absorbed by
differential bending. 

Fig. 2.5.102

Fig. 2.5.103
278 2 The Problem of Saint Venant

2.6 Shear

2.6.1 The Exact Solution

We now study the shear of axis y, that is the sixth problem of Saint Venant
(Fig. 2.2.10). We assume an orthogonal cartesian reference frame O, x, y, z such
that

– z is parallel to the axis of the beam,

– x, y rigorously are inertia principal axes of the cross section of the beam. So in
particular O = G.

The cross section A can be both monoconnected and multiconnected. We denote

with ∂A its boundary. Moreover we denote with

– l ∈ ]0, +∞[ the length of the beam,

– Ix the moment of inertia of the area A with respect to axis x,
– E [resp. ν] the modulus of Young [resp. Poisson] of the material,
– Ty any non-zero real number.

We consider the problem of Neumann

[2.6.1] Find a real function ψ defined in A, having in A all the first and second order
partial derivatives and such that

∂ 2ψ ∂ 2ψ
(x, y) + (x, y) = −2y ∀(x, y) ∈ A (2.6.1)
∂x2 ∂y2

dψ ν
(x, y) = ν x y nx − (x2 − y2 )ny ∀(x, y) ∈ ∂A.  (2.6.2)
dn 2
Since the axis x is centroidal, it results
   
ν
ν x y nx − (x2 − y2 )ny ds − −2y dA
∂A 2 A

∂ ∂ ν 2 
= (ν x y) − (x − y2 ) + 2y dA = 2(1 + ν)Sx = 0
A ∂x ∂y 2

and as consequence, because of a known theorem of Functional analysis

– the problem [2.6.1] admits at least a solution,

– for every couple of solutions ψ 1 , ψ 2 a real number c exists such that ψ 2 = ψ 1 +
c in A.
2.6 Shear 279

After that we can consider any solution ψ of the problem of Neumann [2.6.1],
denote with ϕ every solution of the problem of Neumann [2.5.1] and put
  
∂ψ ∂ψ ν 2
y −x − x(x − y ) dA
2
Ty A ∂x ∂y 2
β=  . (2.6.3)
Elx ∂ ϕ ∂ϕ
x +y +x
2 2
−y dA
A ∂y ∂x

Then we consider the real functions defined in A × [0, l]

νTy
u(x, y, z) = (l − z)xy − βzy
EIx
Ty  z 2 νTy
v(x, y, z) = l− z + (l − z)(y2 − x2 ) + βzx (2.6.4)
2EIx 3 2Elx
Ty  z Ty
w (x, y, z) = − l− yz + ψ (x, y) + βϕ (x, y) .
EIx 2 EIx

We will prove that the (2.6.4) is the unique solution of the sixth problem of Saint
Venant. As already seen in Sect. 1.5, we must verify that such field of displacements
satisfies

– the inside compatibility, that is the (1.1.29) in V,

– the outside compatibility, that is the conditions that the constraints impose to the
points of Su ,
– the inside equilibrium, that is the (1.2.3) in V,
– the boundary equilibrium, that is the (1.2.5) in the points of Sp .

Since u, v, w are regular, the inside compatibility (1.1.29) is banally satisfied.

About the inside equilibrium (1.2.3), first of all we observe that, since the
(1.1.28), the state of strain produced by (2.6.4) is

νTy
εx = (l − z) y
EIx
νTy
εy = (l − z) y
EIx
Ty
εz =− (l − z) y
EIx (2.6.5)
γxy =0
 
Ty ∂ψ ∂ϕ
γxz = −νxy + + β −y +
EIx ∂x ∂x
  ∂ψ 
Ty ν  2 ∂ϕ
γyz = x − y2 + +β x+ .
EIx 2 ∂y ∂x
280 2 The Problem of Saint Venant

From the (2.6.5) and (1.4.5), taking into account that

(2v − 1) Ty
εx + εy + εz = (l − z) y ,
EIx

the state of stress produced by (2.6.4) is

σx = 0
σy = 0
Ty (l − z)
σz = − y
Ix
τxy = 0 (2.6.6)
Ty ∂ψ ∂ϕ
τzx = −νxy + + Gβ −y +
2 (1 + ν) Ix ∂x ∂x
Ty 
ν 2  ∂ψ ∂ϕ
τzy = x − y2 + + Gβ x + .
2 (1 + ν) Ix 2 ∂y ∂y

We recall that by hypothesis X = Y = Z = 0 in V. Moreover since the functions

σx , σy , τxy are constant, they all have the first order partial derivatives identi-
cally zero in V. On the other hand, since the functions τzx , τzy are constant with
respect to z, they have the first order partial derivatives with respect to z iden-
tically zero in V. So the first and the second of the (1.2.3) are banally satisfied.
About the third of the (1.2.3), since ϕ [resp. ψ] satisfies the (2.5.1) [resp. (2.6.1)],
we have

∂τzx ∂τzy ∂σz Ty ∂ 2ψ ∂ 2ψ

+ + +Z = + + 2y
∂x ∂y ∂z 2 (1 + ν) Ix ∂x2 ∂y2
∂ 2ϕ ∂ 2ϕ
+ Gβ + = 0.
∂x2 ∂y2

Such way we have seen that, if the volumetric load is identically zero, the (2.6.4)
are a field of displacements that can physically exist for the beam of Saint Venant.
Let us see now what load needs to be applied on the beam to deform it according
the law (2.6.4). Let (Figs. 2.5.3 and 2.5.4)

– Q be a point of the lateral surface of the beam,

– α 0 be the tangent plane passing through Q,
– n be the line normal to α 0 , passing through Q and going out of the material,
– nx , ny , nz be the direction cosines of n.
2.6 Shear 281

Since nz = 0, from the (1.2.5), (2.5.2), and (2.6.2) it follows

px = σx nx + τxy ny + τxz nz = 0
py = τyx nx + σy ny + τyz nz = 0
dϕ
pz = τzx nx + τzy ny + σz nz = Gβ − ynx + xny
dn
 
Ty dψ ν
+ − νxynx + (x2 − y2 )ny = 0.
2(1 + ν)Ix dn 2
Insofar the lateral surface is unloaded and then the (2.6.4) is the solution of a
problem of Saint Venant.
Let us see now what load needs to be applied on the right base to deform the
beam according the law (2.6.4). Let

– Q be every point of the right base of the beam,

– n be the line normal to the right base, passing through Q and going out of the
material,
– nx , ny , nz be the direction cosines of n.

Since nx = 0, ny = 0, nz = 1, from the (1.2.5) and (2.6.6) it follows

px = σx nx + τxy ny + τxz nz = τxz

Ty ∂ψ ∂ϕ
= −νxy + + Gβ −y +
2(1 + ν)Ix ∂x ∂x
py = τyx nx + σy ny + τyz nz = τyz
 
Ty ν 2 ∂ψ ∂ϕ (2.6.7)
= (x − y2 ) + + Gβ x +
2(1 + ν)Ix 2 ∂y ∂y
pz = τzx nx + τzy ny + σz nz = σz
Ty
= − (l − l)y = 0.
Ix
Analogously, on the left base we get

px = σx nx + τxy ny + τxz nz = −τxz

Ty ∂ψ ∂ϕ
=− −νxy + − Gβ −y +
2(1 + ν)Ix ∂x ∂x
py = τyx nx + σy ny + τyz nz = −τyz
 
Ty ν 2 ∂ψ ∂ϕ (2.6.8)
=− (x − y ) +
2
− Gβ x +
2(1 + ν)Ix 2 ∂y ∂y
pz = τzx nx + τzy ny + σz nz = −σz
Ty Ty l
= (l − 0)y = y.
Ix Ix
282 2 The Problem of Saint Venant

As a consequence, on the right base we must not apply axial load. In fact
 
N = pz dA = 0 dA = 0.
A A

Moreover on the right base we must not apply bending moment of axis x. In fact
 
Mx = (pz dA)(y) = 0 dA = 0.
A A

Moreover on the right base we must not apply bending moment of axis y. In fact
 
My = − (pz dA)(x) = − 0 dA = 0.
A A

We now see if we must apply on the right base a shear of axis x. Since the (2.6.6),
(2.5.1), and (2.6.1) we have

∂ ∂ ∂τzx ∂τzy
(xτzx ) + (xτzy ) = τzx + x +x
∂x ∂y ∂x ∂y
Ty ∂ 2ψ ∂ 2ψ
= τzx + x + − 2νy
2(1 + ν)Ix ∂x2 ∂y2
∂ 2ϕ ∂ 2ϕ Ty
+ Gβx + = τzx − xy;
∂x 2 ∂y 2 Ix

as a consequence, taking into account the (2.6.7), (2.6.6), and (2.6.2) and the
formulas of Gauss
 
Tx = px dA = τxz dA
A  A

∂ ∂ Ty
= (xτzx ) + (xτzy ) + xy dA
∂x ∂y Ix
A 
Ty
= (xτzx nx + xτzy ny )ds + xy dA
Ix A
∂A  
Ty dψ ν 2 Ty
= x − νxynx + (x − y )ny ds + Ixy
2
∂A 2(1 + ν)I x dn 2 Ix
Ty
= Ixy .
Ix
Since the axes x, y are principal of inertia, they are a couple of conjugate diam-
eters of the inertia centroidal ellipse and then, because of the [2.1.13], Ixy = 0.
Consequently we have

Tx = 0.
2.6 Shear 283

Let us observe now that we must apply on the right base a shear of axis y. In fact,
since the (2.6.6), (2.5.1), and (2.6.1) we have

∂ ∂ ∂τzx ∂τzy
(yτzx ) + (yτzy ) = τzy + y +y
∂x ∂y ∂x ∂y
Ty ∂ 2ψ ∂ 2ψ
= τzy + y + − 2νy
2(1 + ν)Ix ∂x2 ∂y2
∂ 2ϕ ∂ 2ϕ Ty
+ Gβy + = τzy − y2 ;
∂x2 ∂y2 Ix

as a consequence, taking into account the (2.6.7), (2.6.6), and (2.6.2) and the
formulas of Gauss
 
Ty = py dA = τyz dA
A  A

∂ ∂ Ty
= (yτzx ) + (yτzy ) + y2 dA
∂x ∂y Ix
A 
Ty
= (yτzx nx + yτzy ny )ds + y2 dA
∂A Ix A
  
Ty dψ ν 2 Ty
= y − νxynx + (x − y )ny ds + Ix = Ty .
2
∂A 2(1 + ν)I x dn 2 Ix

We now see if we must apply on the right base a twisting moment Mt . Since the
(2.6.7) and (2.6.3) we have
 
Mt = [(py dA)(x) − (px dA)(y)] = (xpy − ypx ) dA
A A
  
Ty ν 2 ∂ψ ∂ψ
= (x − y )x + νxy2 + x
2
−y dA
2(1 + ν)Ix A 2 ∂y ∂x

∂ϕ ∂ϕ
+ Gβ x2 + y2 + x −y dA = 0.
A ∂y ∂x

With an analogous reasoning we understand, from the (2.6.7), that on the left base
the axial load is zero, the shearing load of axis x is zero, the bending moment of axis
y is zero, the twisting moment is zero, the shearing load of axis y has intensity – Ty ,
the bending moment of axis x has intensity l Ty .
This way the (2.6.4) are the unique, unless a rigid translation according to z,
solution of the sixth problem of Saint Venant.
Remark 2.6.1 We notice that, since the (2.6.6), in the problem of the shear of axis y
the diagram of the τ z is identical in every cross section. 

Remark 2.6.2 In the problem of the shear of axis y the diagram of the σ z is different
in every cross section. Precisely in the cross section of abscissa z such diagram
284 2 The Problem of Saint Venant

Fig. 2.6.1

coincides with the one furnished by the formula of Navier for the right bending of
axis x of a beam submitted to the bases to an external moment – Ty (l – z). Clearly
– Ty (l – z) is the bending moment in the section of abscissa z of the cantilever of
Fig. 2.6.1. 

About the deformation of the beam, from the (2.6.4) it follows that

– the axis of the beam, i.e. the axis z, turns in a cubic parabola contained in the plane
y, z,
– the centroid of the right base has a displacement according y given by (Fig. 2.6.2)

Ty l3
v(0, 0, l) = ,
3EIx

– to happened deformation, the length of the axis of the beam still is l,

– in the general case the cross section hunches.

Let us consider now a fibre of the beam, that is an infinitesimal element of

material (Fig. 2.6.3)

– parallel to axis z,
– having end points (x, y, z), (x, y, z + dz),
– having as cross section the elementary rectangle dx, dy.

Fig. 2.6.2

Fig. 2.6.3
2.6 Shear 285

As seen in Sect. 1.1.4, during the deformation an infinitesimal element of mate-

rial containing the point (x, y, z) rigidly rotates around the axis z of the infinitesimal
angle
 
1 ∂u ∂v
(x, y, z) − (x, y, z) .
2 ∂y ∂x

Analogously, during the deformation an infinitesimal element of material con-

taining the point (x, y, z + dz) rigidly rotates around the axis z of the infinitesimal
angle
 
1 ∂u ∂v
(x, y, z + dz) − (x, y, z + dz) .
2 ∂y ∂x

Insofar the two end cross sections of the fibre of Fig. 2.6.3 rotate among
themselves of the angle

 
1 ∂u ∂v
ϑ1 = (x, y, z + dz) − (x, y, z + dz)
2 ∂y ∂x
 
1 ∂u ∂v
− (x, y, z) − (x, y, z) .
2 ∂y ∂x

Since dz is infinitesimal, we can approximate ϑ1 obtaining

 
1 ∂u ∂ ∂u
ϑ1 = (x, y, z) + dz (x, y, z)
2 ∂y ∂z ∂y
   
∂v ∂ ∂v 1 ∂u ∂v
− (x, y, z) + dz (x, y, z) − (x, y, z) − (x, y, z)
∂x ∂z ∂x 2 ∂y ∂x
1 ∂ ∂u ∂v
= dz − .
2 ∂z ∂y ∂x

We call local unitary angle of torsion the number

ϑ1
ϑ0 =
dz
so that in the point (x, y, z) it results

1 ∂ ∂u ∂v
ϑ0 = − . (2.6.9)
2 ∂z ∂y ∂x

We call average unitary angle of torsion the number


1
ϑ= ϑ0 dA. (2.6.10)
A A
286 2 The Problem of Saint Venant

Insofar in the sixth problem of Saint Venant from the (2.6.9) we get

νTy
ϑ0 = − x−β (2.6.11)
EIx
and then, taking into account the (2.6.10) and the fact that y is centroidal


1 νTy
ϑ= − x − β dA
A A EIx
(2.6.12)
1 νTy
= − Sy − βA = −β.
A EIx

The fifth problem of Saint Venant (Fig. 2.2.9) is perfectly analogous to the sixth
problem of Saint Venant. The frame of reference O, x, y, z is principal of inertia. We
denote with Iy the moment of inertia of the area A of the cross section with respect
to the axis y and with Tx any non-zero real number.
We consider the problem of Neumann
[2.6.2] Find a real function χ defined in A, having in A all the first and second order
partial derivatives and such that

∂ 2χ ∂ 2χ
(x, y) + (x, y) = −2x ∀(x, y) ∈ A
∂x2 ∂y2
dχ ν
(x, y) = − (y2 − x2 )nx + νxyny ∀(x, y) ∈ ∂A. 
dn 2

Since the axis y is centroidal, it results

   
ν
νxynx − (y2 − x2 )ny ds − −2x dA
∂A 2 A

∂ ∂ ν 2 
= (νxy) − (y − x ) + 2x dA = 2Sy = 0
2
A ∂x ∂y 2

and as a consequence, because of a known theorem of Functional analysis

– the problem [2.6.2] admits at least a solution,

– for every couple of solutions χ 1 , χ 2 a real number c exists such that χ 2 = χ 1 + c
in A.

After that we can consider any solution χ of the problem of Neumann [2.6.2],
any solution ϕ of the problem of Neumann [2.5.1] and put
2.6 Shear 287
  
∂χ ∂χ ν 2
Tx x −y − y(x + y ) dA
2
A ∂y ∂x 2
α=  . (2.6.13)
∂ϕ ∂ϕ
EIy x +y +x
2 2
−y dA
A ∂y ∂x

Then we consider the real functions defined in A × [0, l]

Tx  z 2 νTx
u(x, y, z) = l− z + (l − z)(x2 − y2 ) − αzy
2EIy 3 2EIy
νTx
v(x, y, z) = (l − z)xy + αzx (2.6.14)
EIy
Tx  z Tx
w(x, y, z) = − l− xz + χ (x, y) + αϕ(x, y).
EIy 2 EIy

Proceeding as for the sixth problem of Saint Venant, from the (2.6.14) we get

νTx
εx = (l − z)x
EIy
νTx
εy = (l − z)x
EIy
Tx
εz = − (l − z)x
EIy (2.6.15)
γxy = 0
 
Tx ν 2 ∂χ ∂ϕ
γxz = (y − x2 ) + + α −y +
EIy 2 ∂x ∂x
 
Ty ∂χ ∂ϕ
γyz = −νxy + +α x+ ,
EIx ∂y ∂y

and

σx = 0
σy = 0
Tx (l − z)
σz = − x
Iy
τxy = 0 (2.6.16)
Ty ν 2 ∂χ ∂ϕ
τzx = (y − x2 ) + + Gα −y +
2(1 + ν)Iy 2 ∂x ∂x
Ty ∂χ ∂ϕ
τzy = −νxy + + Gα x + .
2(1 + ν)Iy ∂y ∂y
288 2 The Problem of Saint Venant

Moreover we verify that the (1.2.3) are satisfied and the lateral surface is
unloaded. Then the (2.6.14) is the solution of a problem of Saint Venant.
We also verify that the load that needs to be applied on the right base to deform
the beam according the law (2.6.14) is
 
Tx ν 2 ∂χ ∂ϕ
px = (y − x ) +
2
+ Gβ −y +
2(1 + ν)Iy 2 ∂x ∂x
 
Tx ∂χ ∂ϕ (2.6.17)
py = −νxy + + Gβ x +
2(1 + ν)Iy ∂y ∂y
pz = 0

and that the load that needs to be applied on the left base to deform the beam
according the law (2.6.14) is
 
Tx ν 2 ∂χ ∂ϕ
px = − (y − x2 ) + − Gβ −y +
2(1 + ν)Iy 2 ∂x ∂x
 
Tx ∂χ ∂ϕ
py = − −νxy + + Gβ x + (2.6.18)
2(1 + ν)Iy ∂y ∂y
Tx l
pz = x.
Iy

Insofar on the right base it results an axial load

N = 0,

a bending moment of axis x

Mx = 0,

a bending moment of axis y

My = 0,

a shear of axis x

Tx = Tx ,

a shear of axis y

Ty = 0,

a twisting moment

Mt = 0.
2.6 Shear 289

Analogously we obtain that on the left base the axial load is zero, the shearing
load of axis y is zero, the bending moment of axis x is zero, the twisting moment is
zero, the shearing load of axis x has intensity – Tx , the bending moment of axis y
has intensity l Tx .
This way the (2.6.14) are the unique, unless a rigid translation according to z,
solution of the fifth problem of Saint Venant.
About the deformation of the beam, from the (2.6.14) it follows that

– the axis of the beam, i.e. the axis z, turns into a cubic parabola contained in the
plane x, z,
– the centroid of the right base has a displacement according y given by

Tx l3
u(0, 0, l) = ,
3EIy

– to happened deformation, the length of the axis of the beam is still l,

– in the general case the cross section hunches.

In the fifth problem of Saint Venant from the (2.6.9) we get

νTx
ϑ0 = y − α. (2.6.19)
EIy

From the (2.6.19), taking into account the (2.6.10) and the fact that x is centroidal,
we get

ϑ = −α. (2.6.20)

Remark 2.6.3 We can verify immediately that in the problem of the torsion the
average unitary angle of torsion (2.6.10) coincides with the unitary angle of
torsion. 

Let us consider now a problem of Saint Venant loaded on the right base only by a
shearing force T having line of action s lying in the right base (Fig. 2.6.4). We call
shear center of the cross section, both monoconnected and multiconnected, a point
Ct of the right base such that if s pass through Ct then the average unitary angle of
torsion (2.6.10) is zero.

Fig. 2.6.4
290 2 The Problem of Saint Venant

We have
[2.6.3] Every cross section admits one and only one shear center, given by the point

  
1 ∂ψ ∂ψ ν 2
C= y −x − x(x + y ) dA,
2
2(1 + ν)Ix A ∂x ∂y 2
  
1 ∂χ ∂χ ν 2
− x −y − y(x + y ) dA .
2
2(1 + ν)Iy A ∂y ∂x 2

Proof We consider a generic cross section, both monoconnected and multicon-

nected, and an inertia principal frame of reference G, x, y, z (Fig. 2.6.4). We denote
with C = (xc , yc ) the point (2.6.21) and suppose that on the right base a shearing
force T acts having a line of action passing through C and components Tx , Ty . We
denote with δ the distance of the centroid G from the line of action of T (Fig. 2.6.4).
Evidently the system of forces of Fig. 2.6.4 is equivalent to the system of forces of
Fig. 2.6.5. Therefore, for the principle of superposition [1.5.4] and for the postulate
of Saint Venant [2.2.1] the problem of Saint Venant of Fig. 2.6.4 is, except in the
zone of extinction, sum

– of the fourth problem of Saint Venant, in which the beam is submitted on the right
base to the twisting moment T·δ,
– of the fifth problem of Saint Venant, in which the beam is submitted on the right
base to a shear given by the component Tx of T,
– of the sixth problem of Saint Venant, in which the beam is submitted on the right
base to a shear given by the component Ty of T.

Insofar the average unitary angle of torsion ϑ of the problem of Fig. 2.6.4 is the
sum of the average unitary angle of torsion of these three problems. Consequently,
keeping in mind of the (2.5.7), (2.6.12), and (2.6.20), it results

Tδ
ϑ= −α−β
- ∂ϕ ∂ϕ
G x2 + y2 +x −y dA
A ∂y ∂x

Fig. 2.6.5
2.6 Shear 291

from which, taking into account the (2.6.13) and (2.6.3), we get

1
ϑ= {2Tδ(1 + ν)Ix Iy
- ∂ϕ ∂ϕ
EIx Iy x 2 + y2 + x − y dA
A ∂y ∂x
  
∂χ ∂χ ν 2 (2.6.22)
− Tx Ix x −y − y(x + y2 ) dA
A ∂y ∂x 2
  
∂ψ ∂ψ ν
− Ty Iy y −x − x(x2 + y2 ) dA.
A ∂x ∂y 2

Obviously (Fig. 2.6.4)

Tδ = −Tx yc + Ty xc .

From this, taking into account the (2.6.22) and (2.6.21), it follows

1
ϑ= {−2Tx yc (1 + ν)Ix Iy
- ∂ϕ ∂ϕ
EIx Iy x2 + y2 + x − y dA
A ∂y ∂x
  
∂χ ∂χ ν
+ 2Ty xc (1 + ν)Ix Iy − Tx Ix x −y − y(x2 + y2 ) dA
A ∂y ∂x 2
  
∂ψ ∂ψ ν
− Ty Iy y −x − x(x2 + y2 ) dA = 0.
A ∂x ∂y 2

So the existence of the shear center is proven.

About the uniqueness, reasoning by absurd, let Ct , C̃t be two different shear
center of a same cross section. We apply on the right base a couple constituted by
two shearing forces having intensity F, passing one for Ct and the other for C̃t ,
both normal to the line passing through Ct and C̃t (Fig. 2.6.6). Therefore the beam
is submitted to the non-zero twisting moment F dist(Ct , C̃t ). As a consequence,
taking into account the Remark 2.6.3, we see that the beam has an average unitary
angle of torsion ϑ given by the (2.5.7) and then non zero. On the other hand, using
the principle of superposition [1.5.4] and the postulate of Saint Venant [2.2.1], the
problem of Saint Venant in examination is, except in the zone of extinction, sum of

Fig. 2.6.6
292 2 The Problem of Saint Venant

two problems. In every of such problems the beam is submitted on the right base to
a shearing force passing through a shear center. Consequently, since the definition
of shear center, the beam has an average unitary angle of torsion ϑ = 0. This absurd
proves the uniqueness of the shear center. 
Remark 2.6.4 From the [2.6.3] it obviously follows that the shear center depends
only from the shape of the cross section. 

Because of the Remark 2.6.4, from the principle of symmetry of the Physics it
follows
[2.6.4] If the cross section has a symmetry axis r, then the shear center belongs
to r. 
Remark 2.6.5 From the [2.6.4] it obviously follows that

– the shear center of a circle is its center,

– the shear center of a rectangle is its center,
– the shear center of a H-cross section is the point of Fig. 2.6.8. 

In the general case in examination, the beam is submitted on the right base to a
shearing force T having a line of action s (Fig. 2.6.7). The cross section indifferently
is both monoconnected and multiconnected.
In this problem we call

– load axis the line s centroidal and parallel to the line of action of T,
– neutral axis the set n of the points of the cross section in which σ z = 0.

Fig. 2.6.7

Fig. 2.6.8
2.6 Shear 293

Fig. 2.6.9

Denoting with Tx , Ty the components of T in a principal inertia frame of reference

(Fig. 2.6.9) and reasoning as in the proof of the [2.6.3], since the (2.6.6), (2.6.16),
and (2.5.5), we get that

Ty Tx
σz = − y− x (2.6.23)
Ix Iy

so that n is the centroidal line of equation

Ty Tx
− y − x = 0. (2.6.24)
Ix Iy

Furthermore, reasoning as in Sect. 2.3.3, from the (2.6.23) we obtain (Fig. 2.6.10)
the following results

[2.6.5] In the general problem of the shear of Fig. 2.6.7 the neutral axis and the load
axis are a couple of conjugate diameters of the inertia centroidal ellipse of the cross
section. 

Fig. 2.6.10
294 2 The Problem of Saint Venant

[2.6.6] In the general problem of the shear of Fig. 2.6.7 it exactly results

Tf (l − z)
σz = − dn .  (2.6.25)
In

[2.6.7] In the general problem of the shear of Fig. 2.6.7 the axis of the beam to
happened deformation turns into an arc of cubic parabola that is contained in the
plane f, z. The points of the generic cross section suffer displacements that are sum
of three aliquots: a displacement on another plane, for the deviate bending that
coexists with the shear; an hunch χ from pure shear, which is generally a negligible
aliquot of the total displacement, due to the force T transported in Ct ; a torsional
hunch ϕ, due to the twisting moment born when T is transported in Ct . 

In the [2.6.6] dn denotes the orthogonal distance from the neutral axis of the
point in which we calculate the σ z and f is a line normal to n, oriented so that
the orthogonal projection Tf of T on f is positive (Fig. 2.6.10). We call f flexu-
ral axis and f, z flexural plane. In Fig. 2.6.10 the diagram of the stresses (2.6.25)
is given.

2.6.2 The Approximate Solution

For the problem of the shear a brilliant technical theory that furnishes an approx-
imate but simple expression of the τ z , eliminating the congenital analytical
difficulties of the problem of Neumann, exists.
In this theory, due to Zhuravskii,2.6.1 we consider a beam of Saint Venant having
a generic monoconnected cross section and submitted on the right base to a shearing
force T passing through the shear center Ct (Fig. 2.6.11). We call chord of the cross
section a segment c that divides it into two parts (Fig. 2.6.12).
Zhuravskii conjectured that if the cross section is monoconnected and T passes
through Ct , then in every chord c the component of τ z orthogonal to c is constant
(Fig. 2.6.12).

Fig. 2.6.11

2.6.1 Dmitrii Ivanovich Zhuravskii, Beloe 1821 – St. Petersburg 1891.

2.6 Shear 295

Fig. 2.6.12

We assume this hypothesis and consider (Fig. 2.6.12)

– the load axis s, that is the centroidal line parallel to the line of action of T,
– the neutral axis n, that is the conjugate diameter of s furnished from the [2.6.5],
– the flexural axis f, that is the line normal to n, oriented so that the orthogonal
projection Tf of T on f is positive.

We isolate from the beam a trunk operating a cut according to the section S
of abscissa z and according to the section S of abscissa z + dz (Fig. 2.6.13). We
consider a chord c of the cross section (Fig. 2.6.14) and we denote with A1 , A2 the
two parts in which c divides the area A of the cross section.

Fig. 2.6.13

Fig. 2.6.14
296 2 The Problem of Saint Venant

As seen in Sect. 2.1.1, it results

0 = Sn (A) = Sn (A1 ) + Sn (A2 ),

so that the static moment Sn (A1 ) of A1 with respect to the neutral axis n and the
static moment Sn (A2 ) of A2 with respect to the neutral axis n are equal in absolute
value and of opposite sign. We denote with A0 the area A1 [resp. A2 ] if Sn (A1 ) >
0 [resp. < 0] (Fig. 2.6.14) and with m the normal one to the chord c (Fig. 2.6.14).
Then on the trunk dz of beam of Fig. 2.6.13 we operate a cut according to the plane
individualized by the chord and the axis z. Now we apply the principle of dissection
[1.2.1] to the portion of trunk individualized by A0 (Fig. 2.6.15).
Since in the problems of Saint Venant the volumetric loads are identically zero,
imposing the equilibrium to the translation according to the axis z of the portion of
beam of Fig. 2.6.15, the sum

– of the resultant one of the σ z that act on the portion A0 of the cross section S of
Fig. 2.6.15,
– of the resultant one of the σ z that act on the portion A0 of the cross section S of
Fig. 2.6.15,
– of the resultant one of the τ mz that act on the rectangle ABCD of Fig. 2.6.15,

must be zero.
The σ z are given from the (2.6.25) so that, if Tf > 0, are of compression in the
points having dn > 0 (Fig. 2.6.10). This way

– the resultant one of the σ z that act on the portion A0 of the cross section S of
Fig. 2.6.15, is parallel to the axis z and has intensity
  
Tf (l − z)
dn dA , (2.6.26)
A0 In

– the resultant one of the σ z that act on the portion A0 of the cross section S of
Fig. 2.6.15 is parallel but opposite to the axis z and has intensity
  
Tf (l − (z + dz))
− dn dA . (2.6.27)
A0 In

Fig. 2.6.15
2.6 Shear 297

Fig. 2.6.16

Fig. 2.6.17

About the resultant one of the τ mz that act on the rectangle ABCD of Fig. 2.6.15,
we first of all observe that, since the Zhuravskii hypothesis, in the points of the
chord c the τ zm is constant (Fig. 2.6.16). Then, for the property of the shearing
stress symmetry [1.2.2], on the side DC of the rectangle ABCD (Fig. 2.6.17) τ mz is
constant.So, since dz is as small as we want, we can admit that τ mz is constant in the
whole rectangle ABCD.
As far as the direction of the τ mz we have
[2.6.8] In every point P of the cross section, denoting with c a chord passing through
P and with m the normal one to c, if f is directed so that Tf > 0 then τ zm is directed
toward the part of cross section having positive static moment with respect to the
neutral axis n (Fig. 2.6.16).
Proof In fact if the shearing load T is such that Tf > 0, obviously in Fig. 2.6.17 the
resultant one of the σ z that act on the portion A0 of the cross section S is greater than
the resultant one of the σ z that act on the portion A0 of the cross section S . Insofar
the resultant one of the τ mz that act on the rectangle ABCD must be opposite to z.
The thesis follows. 
Since the [2.6.8] the resultant one of the τ mz that act on the rectangle ABCD of
Fig. 2.6.17 is opposite to z and has intensity
 
− τmz dB = −τmz dB
ABCD ABCD (2.6.28)
= −τmz meas(ABCD) = −τmz cdz.

Consequently, imposing the equilibrium to the translation according to the axis z

of the portion of beam of Fig. 2.6.15 and taking into account the (2.6.26), (2.6.27),
298 2 The Problem of Saint Venant

and (2.6.28), we get

     
Tf (l − z) Tf (l − (z + dz))
dn dA − dn dA − τmz cdz = 0.
A0 In A0 In

From this it follows

   
Tf dz Tf
0= dn dA − τmz cdz = dz dn dA − τmz c
In In A0
A0

that furnishes, denoting with Sn the static moment of the part A0 of A with respect
to the neutral axis n, the formula of Zhuravskii

Tf Sn
τmz = . (2.6.29)
In c
Remark 2.6.6 Obviously the shorter the chord we have chosen, the more precise the
conjecture of Zhuravskii will be, and therefore his formula (2.6.29). 

Remark 2.6.7 It is possible to obtain from the (2.6.29) an approximate law of varia-
tion of the τ zt along a chord c, where t denotes the direction of c (Fig. 2.6.18). Since
the volumetric loads are identically zero, assuming the frame of reference G, t, m, z
the third of the (1.2.3) becomes

∂τzt ∂τzm ∂σz

+ + = 0.
∂t ∂m ∂z

Making in such relation the partial t-derivative we get

∂ 2 τzt ∂ ∂τzm ∂ 2 σz
+ + = 0. (2.6.30)
∂t2 ∂m ∂t ∂t∂z

Fig. 2.6.18
2.6 Shear 299

Since τ zm is constant along c, we have

∂τzm
= 0,
∂t
so that

∂ ∂τzm
= 0.
∂m ∂t

Because of the [2.3.1] three real numbers al , a2 , a3 exist such that

∀ (t, m) ∈ A dn (t, m) = a1 t + a2 m + a3 ; (2.6.31)

then, since the (2.6.25), we get

∂ 2 σz Tf ∂dn Tf a1
= = . (2.6.32)
∂t∂z In ∂t In
From this and from the (2.6.30) we get

∂ 2 τzt Tf a1
=− .
∂t 2 In

Insofar τ zt varies parabolically along the chord. We notice that the (2.6.29) also
furnishes the value of τ zm in the end points H, K of the chord (Fig. 2.6.18). Then,
since the (2.2.3), we know the value of τ zt in the points H, K.
When the chord is parallel to the neutral axis, in the (2.6.31) it results a1 = 0.
Consequently, since the (2.6.32) and (2.6.30) we get

∂ 2 τzt
= 0,
∂t2
so that τ zt varies linearly along the chord. Obviously in such case the values of τ zt
in the end points H, K of the chord are enough to trace the diagram of the τ zt on the
chord. 

Remark 2.6.8 We consider the points H1 , H2 where the neutral axis n intersects the
boundary ∂A. If the tangents to ∂A in such points are parallel (Fig. 2.6.19), then the
τ zf has its maximum on the neutral axis. In fact from the (2.6.29) we get

∂τzf Tf ∂ Sn Tf 1 ∂Sn Sn ∂c
=− =− − 2 . (2.6.33)
∂f In ∂f c In c ∂f c ∂f
300 2 The Problem of Saint Venant

Fig. 2.6.19

Since by hypothesis the tangents to ∂A in H1 and in H2 are parallel, obviously on

the neutral axis it results
∂c
= 0. (2.6.34)
∂f
Clearly Sn has its maximum on the neutral axis. Insofar

∂Sn
= 0. (2.6.35)
∂f
From the (2.6.33), (2.6.34), and (2.6.35) it follows

∂τzf
= 0.
∂f
The thesis follows (Fig. 2.6.20). 

Remark 2.6.9 The formula of Zhuravskii has been extended with success to the
beams with strongly variable cross sections (Fig. 2.6.20). We call axis of such a
beam and we still denote with the symbol z a normal line to the left base, passing
through its centroid. We still call cross section of the beam the intersection of the
beam with a plane normal to z. In the cross section of abscissa z we call neutral
axis the diameter n of the inertia centroidal ellipse conjugated to the direction of
the shearing load T. We denote with f the normal one to n (Fig. 2.6.16) and we
suppose valid the (2.6.29). We consider in the cross section of abscissa z a chord
c (Fig. 2.6.16). After that we consider the portion of beam individualized by the
cross section of abscissa z, from the cross section of abscissa z + dz and from the
plane c, z. Proceeding as for the beam with constant cross section, we get that the

Fig. 2.6.20
2.6 Shear 301

approximate value of the τ zm in the generic point of the chord c is obtained in the
following way (Fig. 2.6.16)
   
Tf (l − z) Tf (l − z) d Tf (l − z)
dn dA − dn dA + dn dA dz
A(z) In A(z) In dz A(z) In
− τzm c dz = 0
 
from which τzm c = − dz
d
Tf (l − z) SInn and then

Tf Sn Tf (l − z) d Sn
τz m = + . 
In c c dz In

Remark 2.6.10 Let us suppose that a generic monoconnected cross section is sub-
mitted on the right base to a shearing force T having line of action s . Obviously we
can calculate the state of stress

– exactly, except in the zone of extinction, transporting T in the centroid

 G and
adding the state of stress due to the twisting moment T dist G, s to the (2.6.6)
and (2.6.16), in which Tx , Ty are the components of T in a principal of inertia
frame of reference,
– approximatively, transporting T in the  center Ct and adding the state of stress
 shear
due to the twisting moment T dist Ct , s to the (2.6.29). 

Remark 2.6.11 The constructive practice has shown that the state of stress induced
by shearing loads as a rule is not negligible. 

Remark 2.6.12 Let’s appraise the maximum of the τ zf considered in the Remark
2.6.8. To this aim we must consider the chord c0 = H1 ,H2 of Fig. 2.6.19.
The chord c0 divides the cross section A in the two parts Aa and Ab , hav-
ing respectively, with respect to the neutral axis n, static moment Sa , Sb and
moment of inertia Ia , Ib . Denoting with Ca [resp. Cb ] the antipole of n with
respect to the inertia centroidal ellipse of Aa [resp. Ab ], we obviously have Ia =
Sa dist (Ca , n) resp. Ib = Sb dist (Cb , n) .
We denote with the symbol h∗ the distance among Ca and Cb measured according
to the normal one to n. Evidently

Sa + Sb = 0,
In = Ia + Ib = Sa dist (Ca , n) + Sb dist (Cb , n) = Sa h∗ ,
so that
Tf
τzf (G) = .
h∗ c0
Since the (2.6.25) the σ z acting above [resp. below] of the neutral axis have
resultant applied in Ca [resp. Cb ]. It is for this reason that sometimes h∗ is called
arm of the inside couple (Fig. 2.6.21). 
302 2 The Problem of Saint Venant

Fig. 2.6.21

Problem 2.6.1 A cantilever is submitted on the free right base to a vertical centroidal
shearing load T. The cross section, constant, is a rectangle of base b and height h
(Fig. 2.6.22). You shall determine the shearing stresses.

Solution Since the Remark 2.6.5, Ct = G. Consequently we can use the (2.6.29)
to calculate τ z in every point P of the cross section. The line s of Fig. 2.6.23 is
centroidal and parallel to T. So s is the load axis. The line n of Fig. 2.6.23 is clearly
the conjugate diameter of s. So n is the neutral axis. To calculate τ z in any point P of
the cross section, since the Remark 2.6.6 it is convenient to consider the horizontal
chord HK (Fig. 2.6.23).

Fig. 2.6.22

Fig. 2.6.23
2.6 Shear 303

Fig. 2.6.24

Since H, K ∈ ∂A, in H and K τ z must be tangent to the contour and therefore

vertical (Fig. 2.6.23). Since we suppose that the chord HK is short, we have with
good approximation that τ z is vertical in P and its value is given from the (2.6.29).
So we have (Fig. 2.6.24)

Tf Sn TSn
τz (P) = τzf (P) = τzm (P) = τzm = = 3 
In c b h
12 b
    
T h2 − f b f + 12 h2 − f 6 T h2
= = − f2 .
b2 h3 b h3 4
12

Insofar τ z varies parabolically with respect to f and its diagram is Fig. 2.6.25,
with

6T h2 3 T 3T
τz max = τzf (G) = τz (G) = − 02 = = .  (2.6.36)
b h3 4 2 bh 2A

Fig. 2.6.25
304 2 The Problem of Saint Venant

2.6.3 The Circular Cross Section

In the case of circular cross section, since the Remark 2.6.5 we have the shear center
Ct coinciding with the center G of the circle. We suppose that on the right base of
the beam a shearing load T centroidal and vertical acts (Fig. 2.6.26). Since T passes
through Ct we can calculate τ z with the (2.6.29). Obviously in this case the neutral
axis n is centroidal and horizontal (Fig. 2.6.26). We denote with f the normal to n
passing through G.
We consider a point P of the cross section (Fig. 2.6.27). To calculate the value
of τ z in P we draw an horizontal chord passing through P. We denote with c its
length and with H, K its end points (Fig. 2.6.27). Finally we denote with ϕ the angle

Fig. 2.6.26

Fig. 2.6.27

Fig. 2.6.28
2.6 Shear 305

included between f and the line passing through G and H (Fig. 2.6.27). Insofar

c = 2r sin ϕ. (2.6.37)

To use the (2.6.29) we calculate the static moment of the area B below the chord
c with respect to neutral axis n (Fig. 2.6.27). As a preliminary, we observe that ψ
varies from ϕ to 0 when f = r sin ψ varies from r sin ϕ to r (Fig. 2.6.28). So we
have
  
df ∼
Sn = dS = (dB) f + = f dB
2
  0
∼
= f (df 2r sin ψ) = 2r sin ψ (r cos ψ) d (r cos ψ)
ϕ
 ϕ (2.6.38)
= 2r sin ψ r cos ψ r sin ψ dψ
0
 ϕ
2
= 2r3 sin2 ψ cos ψ dψ = r3 sin3 ϕ.
0 3

From the (2.6.29), (2.6.37), and (2.6.38) it follows

 
3 T r sin ϕ
2 3 3
T Sn
τzm (P) = τzf (P) = = 4 
In c πr
2r sin ϕ
4
4 T 4T 2
= sin2 ϕ = sin ϕ.
3 π r2 3A
Insofar the τ zm have the diagram of Fig. 2.6.29 and

4 T 2π 4T
τzm max = τzm (G) = τzm (W) = sin = . (2.6.39)
3A 2 3A

Problem 2.6.2 The cantilever of Fig. 2.6.30 has a circular constant cross section
with F = 8 cm. The cantilever is of steel, has length 1 m and is submitted at the

Fig. 2.6.29
306 2 The Problem of Saint Venant

Fig. 2.6.30

free end to a shearing load T = 0.6 t centroidal and vertical. It is assigned the value
σ a = σ 0 /s = 1,600 kg/cm2 . You shall verify the structural safety.
Solution As seen in Sect. 1.7.2, denoting with σ ζ [resp. σ ξ ] the maximum [resp.
minimum] principal stress in a point P of the cantilever, we are in structural safety
in P if
σ0
σζ − σξ ≤ . (2.6.40)
s
Since the diagram of the τ z is constant with z while the diagram of the σ z has
the maximum at the constrained end, it is enough to verify the structural safety in
the points of the constrained end, that is at cross section z = 0. Evidently in this
problem the neutral axis is the horizontal centroidal line and the diagram of the σ z
is the one of Fig. 2.6.31, where

Tl 4Tl 4 · 600 · 100 kg

σz max = R= = = 1, 194 2 .
In πR3 π4 3 cm

Moreover the diagram of the τ zm (Fig. 2.6.29) is furnished from the (2.6.29),
where
4T 4 · 600 kg
τzm max = = = 16 2 .
3A 3 · π · 42 cm
We observe that the state of stress nearer to the crisis happens in the points A, B,
C of Fig. 2.6.32, so that it is enough verify the structural safety in these points.
We assume τ zmax = τ zmmax . In the point A we have (Fig. 2.6.32)

kg
σx = 0, σy = 0, σz = −1, 194 , τxy = 0, τxz = 0, τyz = 0 .
cm2

Fig. 2.6.31
2.6 Shear 307

Fig. 2.6.32

Insofar the principal stresses are

kg
σξ = 0, ση = 0, σζ = −1, 194 ,
cm2
so that the (2.6.40) is satisfied and then in the point A we are in structural safety.
In the point B we have (Fig. 2.6.32)

kg
σx = 0, σy = 0, σz = 0, τxy = 0, τxz = 16 , τyz = 0 .
cm2
Insofar the principal stresses are

kg kg
σξ = −16 , ση = Q, σζ = 16 ,
cm2 cm2
so that the (2.6.40) is satisfied and then in the point B we are in structural safety.
In the point C we have (Fig. 2.6.32)

0 1
4T 2 4 · 600   4 · 600 3 2
kg
τzx = τzm = sin ϕ = 1 − cos2 ϕ = 1− =7 .
3A 3·π ·4 2 3 · π · 42 4 cm2
Moreover

τzx
= tgϕ ,
τzy
from which
 
3
cosϕ cosϕ 4 kg
τzy = τzx = τzx  1/2 = 7.  2 1/2
= 7.9 .
sinϕ 1 − cos2 ϕ cm2
1− 3
4
308 2 The Problem of Saint Venant

Finally, since the (2.6.25) it results

T l 600 · 100 · 4 · 3 kg
σz = dn = 895.25 2 .
In π · 44 cm
Insofar in the point C we have

kg kg kg
σx = 0, σy = 0, σz = 895.25 2
, τxy = 0, τxz = 7 2 , τyz = 7.9 2 .
cm cm cm
Since
 1/2 kg
τz = τzx
2
+ τzy
2
= 10.55 2 ,
cm
the circle of Mohr for the set of planes having as support the normal to z and τz
furnishes the principal stresses
 1/2
σz  σz 2 kg
σξ = + τz2 = 895.37 2
2 2 cm
ση = 0
  1/2
σz σz 2 kg
σζ = − + τz 2
= −0.12 2 .
2 2 cm

Insofar the (2.6.40) is satisfied and then in the point C we are in structural
safety. 

2.6.4 Open Thin Walled Cross Section

Let us consider now a cantilever having open thin walled monoconnected cross
section A (Fig. 2.6.33). The load is applied only on the right base and is constituted
by a shearing force of intensity T having a line of action generic but passing through
the shear center Ct (Fig. 2.6.33). Let

– ρ be the middle curve of the cross section (Fig. 2.6.33),

– Q be any point of ρ,

Fig. 2.6.33
2.6 Shear 309

Fig. 2.6.34

– t be the tangent to ρ at Q (Fig. 2.6.34),

– n be the normal to t at Q (Fig. 2.6.34),
– H, K be the points in which n intersects the boundary ∂A of the cross section
(Fig. 2.6.34),
– c be the chord individualized from H, K.

Since T pass through Ct we can calculate the τ z with the formula of Zhuravskii
(2.6.29). We notice that, since c is very short, such formula furnishes a practi-
cally exact result. Since in the problems of Saint Venant the τ z in the points of
the boundary ∂A is tangent to ∂A, it results

τzn (H) = 0, τzn (K) = 0 .

As a consequence, taking into account the fact that c is very short, we can assume
that in every point of c it results

τzn = 0. (2.6.41)

On the other hand the (2.6.29) is obtained supposing that the component of τ z
according to the normal one to the generic chord is constant. Consequently we have
that, in the case of open thin walled monoconnected cross section, in every point of
a generic chord c the τ z is constant and normal to c.
Insofar drawing the diagram of the τ z for the middle curve ρ we represent the
whole distribution of the τ z on the area A of the cross section.
As far as the determination of τ z (Q), to employ the (2.6.29) we first of all must
determine the neutral axis n, that is the conjugate diameter of the load axis s with
respect to the inertia centroidal ellipse of the cross section. As seen in Sect. 2.1.3,
to this aim it is enough to determine the antipole S0 of a non-centroidal line s0
parallel to the line of action of T. Then n is the line passing through G and S0
(Fig. 2.6.35). After having individualized the neutral axis n, we consider the cen-
troidal line f normal to n and oriented so that the orthogonal projection Tf of T on
f is positive (Fig. 2.6.36). Then, to calculate the value of τ z in the point Q of ρ,
we trace the shorter chord c passing through Q (Fig. 2.6.37). The normal m to c is
parallel to the tangent t to ρ (Fig. 2.6.37).
310 2 The Problem of Saint Venant

Fig. 2.6.35

Fig. 2.6.36

Fig. 2.6.37

In conclusion from the (2.6.29) we have

Tf Sn
τz (Q) = τzm (Q) = . (2.6.42)
In c
In the (2.6.42) In denotes the moment of inertia of the whole cross section with
respect to the neutral axis n, Sn denotes the static moment with respect to the neutral
axis n of one of the two parts in which the chord c divides the cross section. As for
as the verse of τ z , we use the [2.6.8].
2.6 Shear 311

We have
[2.6.9] In the parts in which the middle curve ρ is straight and the chord is constant,
that is in the parts of the cross section constituted by lengthened rectangles, the
diagram of the τ z is parabolic.
Proof Let’s report us to the Fig. 2.6.33 and to the segment A, B of ρ (Fig. 2.6.38).We
consider a point Q of ρ (Fig. 2.6.39) and we put x = dist(A, Q) (Fig. 2.6.40). We
denote with c the chord passing through Q and normal to ρ (Fig. 2.6.39). We observe
that in the (2.6.42) Tf , In , c are constant along the segment A, B. As for Sn , it is

Fig. 2.6.38

Fig. 2.6.39

Fig. 2.6.40
312 2 The Problem of Saint Venant

the static moment with respect to neutral axis n of the broken area of Fig. 2.6.39.
As underlined in footnote 2.1.2, Sn is equal to the sum of the static moment a
with respect to neutral axis n of the rectangle P1 P2 P3 P5 and of the static moment
Sn (P4 P5 P6 P7 ) with respect to neutral axis n of the rectangle P4 P5 P6 P7

Sn = a + Sn (P4 P5 P6 P7 ). (2.6.43)

It is obvious that, varying Q, a is constant. Instead, denoting with G̃ the centroid

of the rectangle P4 P5 P6 P7 , it results (Fig. 2.6.40)

Sn (P4 P5 P6 P7 ) = meas2 (P4 P5 P6 P7 ) dist(G̃, n) = c dist(G̃, n). (2.6.44)

We denote with D the point intersection between n and ρ (Fig. 2.6.40), with d
the distance between A and D, with α the angle between n and ρ (Fig. 2.6.40).It
obviously results (Fig. 2.6.40)
 x
dist(G̃, n) = [dist(A, D) − dist(A, G̃)]sinα = d − sin α,
2
from which, taking into account the (2.6.43) and (2.6.44), we get
 x
Sn = a + c x d − sin α
 c  2 (2.6.45)
= − sin α x2 + (c d sin α)x + a.
2
Insofar Sn varies parabolically with x and this, for the (2.6.42), implicates that τ z
varies parabolically on ρ. 
We also have
[2.6.10] We suppose that a part p̂ of the middle curve ρ is straight and parallel to
the neutral axis n. If in p̂ the chord is constant then in p̂ the diagram of the τ z is
linear.
Proof Let’s us report to the case of Fig. 2.6.41, in which the part p̂ of the mid-
dle curve ρ is a segment A, B parallel to the neutral axis n. Reasoning as for the
Fig. 2.6.40, evidently for every Q it results dist(G̃, n) = p = const. It follows of it,
since the (2.6.43) and (2.6.44), that

Sn = a + cxp. (2.6.46)

Insofar the parabola (2.6.45) degenerates in the line (2.6.46). 

We also have
[2.6.11] We suppose that a part p̂ of the middle curve ρ is straight and non parallel
to the neutral axis n. We suppose that in p̂ the chord is constant and we denote with
D the point intersection between ρ and n. Then D is a point of local extreme for the
parabolic diagram of the τ z .
2.6 Shear 313

Fig. 2.6.41

Proof Let’s report to the case of Figs. 2.6.38 and 2.6.40, in which the part p̂ of the
middle curve ρ is a segment A, B non parallel to the neutral axis n. Since the (2.6.42)
and (2.6.45) in p̂ it results

Tf  c  
τz = − sin α x2 + (c d sin α)x + a . (2.6.47)
In c 2
It is known from the Mathematical analysis that the points of local extreme of a
real function τ z defined and differentiable in ] − ∞, +∞[ must be searched among
the zeros of its derivative. So we must search the points of the line r (Fig. 2.6.42) in
which
dτz
= 0.
dx

From this, since the (2.6.47) and taking into account the fact that α = 0, we have
the condition

− x + d = 0.

Fig. 2.6.42
314 2 The Problem of Saint Venant

Since at x = d it results

dτz d2 τz Tf
(d) = 0, 2 (d) = − sin α ,
dx dx In c
we conclude that D is a point of local extreme for τ z . 
Remark 2.6.13 We can prove the [2.6.11] in another way. Let us refer to case of
Fig. 2.6.37, consider the segment H, K of the middle curve ρ and denote with E
the point intersection between the segment H, K and n. Well E is a point of local
extreme. In fact we can consider some points E1 , E2 , E3 very near to E (Fig. 2.6.43).
Since the (2.6.42), denoting with Sn (A1 ) the static moment of the area A1 with
respect to the neutral axis n, we get (Fig. 2.6.43)

Tf Sn (A1 )
τz (E1 ) = .
In c
Analogously (Fig. 2.6.44)
Tf Sn (A2 )
τz (E2 ) = .
In c

Fig. 2.6.43

Fig. 2.6.44
2.6 Shear 315

Nevertheless A2 = A1 ∪ A12 and meas2 (A1 ∩ A12 ) = 0, so that

Sn (A2 ) = Sn (A1 ) + Sn (A12 ).

Consequently

Tf Sn (A1 ) Tf Sn (A12 ) Tf Sn (A12 )

τz (E2 ) = + = τz (E1 ) + . (2.6.48)
In c In c In c

Since the centroid of the area A1 and that of the area Al2 are both in the positive
half-plane (Figs. 2.6.43 and 2.6.45), we have that Sn (A1 ) and Sn (A12 ) have both
positive sign. Of here and from the (2.6.48) it follows that τ z (E2 ) is in modulus
greater than τ z (E1 ). With the same reasoning it is gotten that τ z (E) is in modulus
greater than τ z (E2 ).
We now appraise τ z (E3 ). To this aim we denote with AE the area individualized
by the chord passing through E (Fig. 2.6.46), with A3 the area individualized by
the chord passing through E3 (Fig. 2.6.47) and with AE3 the elementary area indi-
vidualized by the chord passing through E and by the chord passing through E3

Fig. 2.6.45

Fig. 2.6.46
316 2 The Problem of Saint Venant

Fig. 2.6.47

Fig. 2.6.48

(Fig. 2.6.48). Evidently the union of AE and AE3 is just A3 and the intersection of
AE and AE3 has two-dimensional measure zero. Then Sn (A3 ) = Sn (AE ) + Sn (AE3 ).
However the centroid of AE lies in the positive half-plane while the centroid of AE3
lies in the negative half-plane. Then Sn (AE ) is negative while Sn (AE3 ) is positive. As
a consequence, since

Tf Sn (A3 ) Tf Sn (AE ) Tf Sn (AE3 ) Tf Sn (AE3 )

τz (E3 ) = = + = τz (E) + ,
In c In c In c In c
we have that τ z (E) is in modulus greater than τ z (E3 ).
This way E is a point of local extreme for the diagram of τ z . 

We easily prove that if the middle curve has not points of ramification then the
diagram of the τ z has a number of inversions equal to the number of times in which
the neutral axis cuts the middle curve minus the unity.
Another approximate but important property is that of the practical immutability
of the flow of τ z in the shear. We have
2.6 Shear 317

[2.6.12] In every junction of the cross section the outgoing flow of τz is practically
zero.
Proof Let’s us report to the cross section of Fig. 2.6.36 and consider the junction
BHK. We denote with F the flow of τ z going out from such junction. We observe
that the part HK of the boundary of the junction belongs to ∂A. Consequently τ z
is tangent to the boundary of the junction and of here it follows that through the
segment HK the flow of τ z is zero. Now we denote with c(H) [resp. c(K)] the length
of the chord passing through H [resp. K]. We observe that τ z is normal to c(H) [resp.
c(K)] and constant along c(H) [resp. c(K)]. We also observe that in H [resp. K] τ z is
right oriented. So the flow going out from the junction through the chord c(H) [resp.
c(K)] is – τ z (H) c(H) [resp. τ z (K) c(K)]. Likewise, denoting with c(B) the length of
the chord passing through B and observing that τ z (B) is direct downward, we have
that the flow going out from the junction through the chord c(B) is – τ z (B) c(B).
Insofar

 = −τz (H) c(H) + τz (K) c(K) − τz (B) c(B).

Applying the divergence theorem to the junction BHK we draw


div τ z dB = .
BHK

On the other hand from the third of the (1.2.3) on from the (2.6.25) we get

∂τzx ∂τzy ∂σz Tf

div τ z = + =− = − dn .
∂x ∂y ∂z In
Insofar, denoting with Sn (BHK) the static moment of the junction BHK with
respect to the neutral axis n, we have

Tf Tf
= − dn dB = − Sn (BHK).
BHK I n In
But we have before seen that F is a sum of addenda. In every addenda a static
moment appears with respect to n of an area really greater than BHK. Therefore
approximatively F = 0. 
Problem 2.6.3 You shall trace the diagram of τ z for the shear problem of Fig. 2.6.49.
Solution Since T passes through Ct , we can use the (2.6.29) to calculate the τ z that
arise in the cross section. Indeed in the first place the neutral axis n needs to be
determined. Since the load axis is the centroidal line parallel to T and since the
inertia centroidal ellipse is Fig. 2.6.50, it is obvious that in this problem the neutral
axis n is the horizontal centroidal line. We trace the line f normal to n, passing
through G and oriented so that the orthogonal projection Tf of T on f is positive
(Fig. 2.6.51). Evidently in this case Tf = T. We start to trace the diagram of the τ z
for the segment A, B of the middle curve ρ (Fig. 2.6.51). Since the [2.6.10] such
318 2 The Problem of Saint Venant

Fig. 2.6.49

Fig. 2.6.50

Fig. 2.6.51

diagram is linear. Moreover obviously Sn = 0 in the point A and then τ z (A) = 0.

Finally we calculate τ z (B) with the (2.6.42), obtaining

Tf Sn (AAB )
τz (B) = . (2.6.49)
In c

So we have the diagram of Fig. 2.6.52. Now in the same scale we trace the dia-
gram of the τ z for the segment C, D of the middle curve ρ (Fig. 2.6.51). First of all
2.6 Shear 319

Fig. 2.6.52

Fig. 2.6.53

we calculate τ z (C). From the (2.6.42), denoting with AAC the area individualized by
the chord c1 passing through C (Fig. 2.6.53), we get

Tf Sn (AAC )
τz (C) = . (2.6.50)
In c1

From the (2.6.48) and (2.6.49) it follows (Fig. 2.6.54)

Tf
τz (C)c1 − τz (B)c = [Sn (AAc ) − Sn (AAB )]. (2.6.51)
In

We denote with ABC the area of the junction BC, that is the area individualized by
the chord c passing through B and by the chord c1 passing through C (Fig. 2.6.55).
Since AAC – AAB = ABC , from the (2.6.51) it follows
320 2 The Problem of Saint Venant

Fig. 2.6.54

Fig. 2.6.55

Tf
τz (C)c1 − τz (B)c = Sn (ABC ). (2.6.52)
In

Since the cross section is thin walled, ABC is negligible with respect to AAB
and to AAC . Insofar Sn (ABC ) is negligible with respect to Sn (AAB ) and to Sn (AAC ).
Consequently, because of the (2.6.49), (2.6.50), and (2.6.52), we get

τz (C)c1 − τz (B)c ∼
= 0. (2.6.53)

The (2.6.53) is the flow immutability assured from the [2.6.12] and furnishes

c
τz (C) = τz (B).
c1
So we have obtained the point Z1 of the diagram of the τ z on the segment C, D
of the middle curve ρ (Fig. 2.6.56). Since the [2.6.9] such diagram is parabolic and
2.6 Shear 321

Fig. 2.6.56

since the [2.6.11] such diagram assumes a local extreme in the point O in which n
intersects ρ.
We can trace such parabola as envelope of tangents.2.6.2 In fact using the (2.6.42)
we calculate the value of τ z in O, so that we have the point Z2 of the parabola
(Fig. 2.6.56). Moreover, since the [2.6.11], in O the tangent to the parabola is
vertical. Consequently2.6.3 the tangent to the parabola in C is the line Z1 , Z2 of
Fig. 2.6.57.
Reasoning in the same way we immediately know the tangent to the parabola in
D, and consequently τ z (D). In fact let’s now report to the whole interval C, D. As
clarifyed in the footnote 2.6.3, the tangents to the parabola in C and in D have point
of intersection on the vertical one to the segment C, D passing through its middle
point. Consequently the tangent to the parabola in D must pass through the point
Z4 (Fig. 2.6.58). Well, applying the same reasoning to the interval O, D, the tangent

Fig. 2.6.57

2.6.2 To perform such construction it is necessary to know two points of the parabola and the
tangents to the parabola in such points.
2.6.3 It is well known that the tangents to a parabola in two every points of abscissa x and x
1 2
intersect in a point of abscissa (xl + x2 )/2.
322 2 The Problem of Saint Venant

Fig. 2.6.58

Fig. 2.6.59

Fig. 2.6.60

to the parabola in D must pass through the point Z5 (Fig. 2.6.59). In conclusion the
tangent to the parabola in D is individualized from the points Z4 and Z5 (Fig. 2.6.60)
and now we know the value of τ z in D.Then, as elucidated in the footnote 2.6.2, we
can trace the parabola of Fig. 2.6.60.
2.6 Shear 323

Fig. 2.6.61

Now in the same scale we trace the diagram of the τ z for the segment E, F of
the middle curve ρ (Fig. 2.6.51). To calculate τ z (C), applying the [2.6.12] to the
junction DE, we immediately get (Fig. 2.6.61)

c1
τz (E) = τz (D).
c
Since τ z (F) = 0, taking into account the [2.6.10] we can complete the required
diagram (Fig. 2.6.61). About the verse of the τ z , we remember that τ z in every point
P is directed towards the part of the cross section, individualized from the chord,
having positive static moment. We can immediately recognize that the verses of the
τ z are those data in Fig. 2.6.62. 

Fig. 2.6.62
324 2 The Problem of Saint Venant

Remark 2.6.14 In the junctions of an open thin walled cross section (Fig. 2.6.63)
the (2.6.29) doesn’t furnish reliable results. So in Fig. 2.6.62 the values of τ z in
the junctions are not furnished. Nevertheless as a rule in the junctions considerable
fittings exist to prevent the phenomenon of the torsional stress concentration. For a
more precise analysis it is necessary to use the exact theory of Saint Venant. 

Problem 2.6.4 You shall trace the diagram of τ z for the shear problem of Fig. 2.6.64.
Solution Since T passes through Ct , we can use the (2.6.29) to calculate the τ z
that arise in the cross section. Indeed in the first place the neutral axis n needs
to be determined. Since the load axis s is the horizontal centroidal line and since
the inertia centroidal ellipse is Fig. 2.6.50, it is obvious that the neutral axis n is
the vertical centroidal line (Fig. 2.6.65). We trace the line f normal to n, passing
through G and oriented so that the orthogonal projection Tf of T on f is positive
(Fig. 2.6.65). Evidently in this case Tf = T. We start to trace the diagram of the

Fig. 2.6.63

Fig. 2.6.64

Fig. 2.6.65
2.6 Shear 325

τ z for the segment A, B of the middle curve ρ (Fig. 2.6.65). Since the [2.6.9] such
diagram is parabolic. Moreover Sn (A) = 0 and then τ z (A) = 0. Denoting with O1 ,
O2 the points intersection between n and ρ (Fig. 2.6.65), since the [2.6.11] O1 is a
point of local extreme for the parabola and then we known the tangent in O1 to the
parabola. Finally we calculate τ z (O1 ) with the (2.6.42). So the diagram of the τ z
for the segment A, B must pass through the points Z1 , Z2 and we known the tangent
in O1 (Fig. 2.6.66). So, proceeding as in Problem 2.6.3, we can trace such diagram
(Fig. 2.6.67).
Now in the same scale we trace the diagram of the τ z for the segment C, D of the
middle curve ρ (Fig. 2.6.65). First of all, employing the approximate flow constancy,
we calculate τ z (C) (Fig. 2.6.68). After that we observe that, because of the [2.6.10],
the diagram of the τ z for the segment C, D is linear.
Now we denote with O3 the point intersection between f and ρ (Fig. 2.6.68) and
observe that τ z (O3 ) = 0. In fact, being f axis of geometric symmetry for the cross
section (Fig. 2.6.65), the areas individualized by the chord passing through O3 must
have the respective centroids GAO3 , GFO3 in symmetrical position with respect to the
axis f (Fig. 2.6.69). Moreover it is known from the Geometry of areas (Sect. 2.1) that
the centroid of the total area is a point of the segment GAO3 , GFO3 . Consequently
GAO3 , GFO3 are points of n. It follows that the areas individualized by the chord

Fig. 2.6.66

Fig. 2.6.67
326 2 The Problem of Saint Venant

Fig. 2.6.68

Fig. 2.6.69

Fig. 2.6.70

passing through O3 (Fig. 2.6.69) have static moment zero with respect to n. This
implicates that τ z (O3 ) = 0. This way, on the segment C, D the diagram of the τ z is
individualized (Fig. 2.6.70). We can now complete the diagram of the τ z observing
that f is axis of symmetry for the cross section with respect to the geometry and to
the load (Fig. 2.6.65). So the complete diagram of the τ z is Fig. 2.6.70.
As for as the verse of the τ z we observe that for the point W of Fig. 2.6.71 the
part of cross section emphasized in Fig. 2.6.71 has centroid with positive abscissa
2.6 Shear 327

Fig. 2.6.71

Fig. 2.6.72

and has positive static moment with respect to n. Then in W the τ z is directed
towards such area (Fig. 2.6.72). In the point O1 the same reasoning involves that
τ z is oriented as shown in Fig. 2.6.72.
Let us now consider a generic point Q of the segment CO3 (Fig. 2.6.73) and
denote with AAQ the area individualized from A and Q (Fig. 2.6.73). It is clear that
as soon as Q draws near to O3 the centroid of the area AAQ moves it toward left,
up to n, when Q coincides with O3 , as we have already seen. Then if Q is a point
of the segment CO3 , the centroid of the area AAQ lies in the positive half-plane and
consequently the area AAQ has positive static moment with respect to n. This way,
in the points of the segment CO3 τ z is directed upward (Fig. 2.6.74). The condition
that f must be axis of symmetry for the diagram of the τ z furnishes the verses of the

Fig. 2.6.73
328 2 The Problem of Saint Venant

Fig. 2.6.74

τ z in the remaining part of cross section (Fig. 2.6.74). We observe that in the point
O3 the diagram introduces an inversion of the verse of the τ z . 

Problem 2.6.5 You shall trace the diagram of τ z for the shear problem of Fig.2.6.75.
Solution Since T passes through Ct , we can use the (2.6.29) to calculate the τ z that
arise in the cross section. Then, using the principle of superposition, the diagram of
the τ z is sum of the diagram of Fig. 2.6.62 and of the diagram of Fig. 2.6.70. 

Problem 2.6.6 You shall trace the diagram of τ z for the shear problem of Fig. 2.6.76.
Solution Since T passes through Ct , we can use the (2.6.29) to calculate the τ z that
arise in the cross section. Indeed in the first place the neutral axis n needs to be

Fig. 2.6.75

Fig. 2.6.76
2.6 Shear 329

determined. Since the load axis is the centroidal line parallel to T and since the
inertia centroidal ellipse is Fig. 2.6.77, it is obvious that the neutral axis n is the
horizontal centroidal line (Fig. 2.6.78). We trace the line f normal to n, passing
through G and oriented so that the orthogonal projection Tf of T on f is positive
(Fig. 2.6.78). Evidently in this case Tf = T. We start to trace the diagram of the τ z
for the segment A, B of the middle curve ρ (Fig. 2.6.78). Since the [2.6.10] such
diagram is linear. Moreover obviously Sn = 0 in the point A and then τ z (A) = 0.
Now we calculate τ z (B). The left area individualized by the chord c passing
through B has static moment with respect to n (Fig. 2.6.79)
   
b c1
Sn = c − dist G̃, n ,
2 2

where G̃ denotes his centroid. Then, because of the (2.6.42)

    T  
Tf b c1 f b c1
τz (B) = c − dist G̃, n = − dist G̃, n .
In c 2 2 In 2 2

Fig. 2.6.77

Fig. 2.6.78

Fig. 2.6.79
330 2 The Problem of Saint Venant

Fig. 2.6.80

So the diagram of the τ z for the segment A, B of the middle curve ρ is Fig. 2.6.80.
We now observe that f is axis of symmetry for the cross section with respect to
the geometry and to the load (Fig. 2.6.99). So the diagram of the τ z for the segment
C, D of the middle curve ρ is Fig. 2.6.80. As far as the verse of the τ z , reasoning
as in the problem 2.6.4 we easily get the result of Fig. 2.6.80. Finally, we trace the
diagram of the τ z for the segment E, F of the middle curve ρ. Since the [2.6.9] such
diagram is parabolic and since the [2.6.11] such diagram assumes a local extreme
in the point G in which n intersects ρ.
We observe that the [2.6.12] furnishes (Fig. 2.6.80)

τz (E)c1 = τz (B)c + τz (C)c = 2cτz (B)

from which
2c
τz (E) = τz (B).
c1
In conclusion the diagram of τz that arises in the shear problem of Fig. 2.6.76 is
Fig. 2.6.81. 

Problem 2.6.7 You shall trace the diagram of τ z for the shear problem of Fig. 2.6.82.

Fig. 2.6.81
2.6 Shear 331

Fig. 2.6.82

Solution Since T passes through Ct , we can use the (2.6.29) to calculate the τ z that
arise in the cross section. To use the (2.6.29), in the first place the neutral axis n
needs to be determined. Since the load axis is the centroidal line parallel to T and
since the inertia centroidal ellipse is Fig. 2.6.83, it is obvious that the neutral axis n is
the horizontal centroidal line (Fig. 2.6.84). Proceeding as in the preceding problems,
it is easily verified that the diagram of the τz is Fig. 2.6.85. It results symmetric with
respect to axis n and with respect to axis f, so that in modulus (Fig. 2.6.84)

τz (A) = τz (B) = τz (C) = τz (D). 

Fig. 2.6.83

Fig. 2.6.84
332 2 The Problem of Saint Venant

Fig. 2.6.85

Fig. 2.6.86

Problem 2.6.8 You shall trace the diagram of τ z for the shear problem of Fig.2.6.86.
Solution Since T passes through Ct , we can use the (2.6.29) to calculate the τ z that
arise in the cross section. To use the (2.6.29) in the first place the neutral axis n
needs to be determined. Since the load axis is the centroidal line parallel to T and
since the inertia centroidal ellipse is Fig. 2.6.83, it is obvious that the neutral axis n
is the vertical centroidal line (Fig. 2.6.87).

Fig. 2.6.87
2.6 Shear 333

Fig. 2.6.88

We trace the diagram of the τ z for the segment E, F of the middle curve ρ
(Fig. 2.6.87). Let Q be every point of such segment. Clearly the area individual-
ized from the chord passing through Q is symmetric with respect to n and then has
centroid on n (Fig. 2.6.88). Consequently its static moment with respect to n is zero,
so that τ z (Q) = 0. Proceeding as in the preceding problems, it is easily verified that
the diagram of the τz is Fig. 2.6.85. It is symmetric with respect to axis n and with
respect to axis f (Fig. 2.6.89). 

Problem 2.6.9 You shall trace the diagram of τ z for the shear problem of Fig. 2.6.90.
Solution Since T passes through Ct , we can use the (2.6.29) to calculate the τ z that
arise in the cross section. To use the (2.6.29), in the first place the neutral axis n
needs to be determined. Proceeding as seen in Sect. 2.1, we determine the iner-
tia centroidal ellipse (Fig. 2.6.91) and successively the neutral axis n as conjugate
diameter of the load axis, i.e. of the centroidal line parallel to T (Fig. 2.6.92). We

Fig. 2.6.89
334 2 The Problem of Saint Venant

Fig. 2.6.90

Fig. 2.6.91

Fig. 2.6.92

trace the line f normal to n, passing through G and oriented so that the orthogonal
projection Tf of T on f is positive (Fig. 2.6.92).

Now we trace the diagram of the τ z for the segments A, B and C, D constituting
the middle curve ρ (Fig. 2.6.92). Since the [2.6.9] such diagrams are parabolic and
since the [2.6.11] such diagrams assume a local extreme in the points O1 , O2 in
which n intersects ρ (Fig. 2.6.92). Then the required diagram of τ z is Fig. 2.6.93.
2.6 Shear 335

Fig. 2.6.93

Fig. 2.6.94

We notice that the τ z distributed on the part C, D [resp. A, B] of the cross section
(Fig. 2.6.94) [resp. (Fig. 2.6.95)] are vertical [resp. horizontal]. As a consequence
the τ z distributed on the part C, D of the cross section have T as resultant and the τ z
distributed on the part A, B of the cross section have resultant zero. Consequently the
τ z distributed on the part A, B of the cross section have a point of inversion E, where
obviously τ z (E) = 0 (Fig. 2.6.93). Clearly E is such that dist(E, O1 ) = dist(A, O1 )
(Fig. 2.6.96).
Furthermore it results

 B E B
0= τz dA = τz c dx =c τz dx + c τz dx
AB
A A E

Fig. 2.6.95
336 2 The Problem of Saint Venant

Fig. 2.6.96

so that

E B
τz dx = − τz dx. 

A E

Problem 2.6.10 You shall trace the diagram of τ z for the shear problem of
Fig. 2.6.33.

Solution Since T passes through Ct , we can use the (2.6.29) to calculate the τ z that
arise in the cross section. For this problem we have already determined the neutral
axis n (Fig. 2.6.36). We trace the line f normal to n, passing through G and ori-
ented so that the orthogonal projection Tf of T on f is positive (Fig. 2.6.97). Now
we trace the diagram of the τ z that, since the [2.6.9], is constituted from parabo-
las (Fig. 2.6.98). We notice that, since the [2.6.11], such parabolas assume a local
extreme in the points O1 , O2 , O3 in which n intersects ρ (Fig. 2.6.97). In particular
we can observe that the parabola concavity for the segment A, B is really that of
Fig. 2.6.98 (Fig. 2.6.99). In fact, since the position of O3 , the opposite concavity is
impossible (Fig. 2.6.100). 

Fig. 2.6.97
2.6 Shear 337

Fig. 2.6.98

Fig. 2.6.99

We have seen in Sect. 2.6.1 that the determination of the shear center requires
the solution of two boundary problems. Nevertheless, since the [2.6.4], if the cross
section has an axis of symmetry then the shear center must belong to it. As example,
in the I-section of Fig. 2.6.83 it results Ct = G.
In the case of an open thin walled cross section the determination of the shear
center can be done by an approximate but simple way. This advantageous technique,
achieved from the formula of Zhuravskii (2.6.29), furnishes for such type of cross
section practically exact results.
We have seen in Sect. 2.6.2 that the system constituted by the τ z furnished by the
formula of Zhuravskii is equivalent to the system constituted by the unique shearing
force T passing through the shear center Ct . Then it deals with assigning by the
338 2 The Problem of Saint Venant

Fig. 2.6.100

(2.6.29) a first distribution of τ z and with determining its central axis ŝ1 , to which
Ct must belong. Enough now to perform a second time this procedure applying by
the (2.6.29) a second distribution of τ z and determining its central axis ŝ2 , to which
Ct must belong. This way Ct is determined as point of intersection of ŝ1 and ŝ2 .
When this procedure is practically performed, to build a distribution of τ z with
the (2.6.29), rather than to assign a shearing force Ti declaring its intensity, direction
and verse, it is worthwhile to choose a neutral axis ni in the most convenient way for
the following tracing of the diagram of the τ z and to suppose that Tifi = 1, where fi
is the centroidal line normal to ni , oriented as we like. This obviously is equivalent
to assign Ti in intensity, direction and verse. In fact when we assign the neutral axis
ni , the direction of Ti is achieved parallel to the diameter of the inertia centroidal
ellipse conjugated to n. Moreover the condition Tifi = 1 obviously determines the
intensity and the verse of Ti .

Problem 2.6.11 You shall determine the shear center of a cross section constituted
by two rectangles lengthened but not parallel (Fig. 2.6.101).

Solution The determination of the shear center Ct in this case is particularly sim-
ple. In fact let us calculate with the (2.6.29) the distribution of τ z produced from a
shearing force T1 passing through Ct. We denote with A, B the two rectangles that
constitute the cross section (Fig. 2.6.102). We observe that, since τ z is constant on
every chord, the τ z distributed on the rectangle A [resp. B] admit the line a [resp. b]

Fig. 2.6.101
2.6 Shear 339

Fig. 2.6.102

as axis of symmetry (Fig. 2.6.102). So the distribution of τ z produced from a shear-

ing force T1 passing through Ct is equivalent to a system of two forces, respectively
having for line of action a and b. Since the lines a, b have in common the point
O (Fig. 2.6.102), we conclude that the distribution of τ z produced from a shearing
force T1 passing through Ct is equivalent to a system constituted by a unique force
having line of action passing through O. So the central axis ŝ1 passes through O.
Applying again the same reasoning to another shearing force T2 passing through
Ct , the central axis ŝ2 also passes through O. Insofar, since Ct is the intersection
between ŝ1 and ŝ2 , we get that Ct = O (Fig. 2.6.103). 

Problem 2.6.12 You shall determine the shear center of the cross sections shown in
Fig. 2.6.104.

Solution From the Problem 2.6.11 it immediately follows that the shear centers of
the cross sections shown in Fig. 2.6.104 are those shown in Fig. 2.6.105. 

Problem 2.6.13 You shall determine the shear center Ct of the symmetrical cross
section shown in Fig. 2.6.106.

Fig. 2.6.103

Fig. 2.6.104

Fig. 2.6.105
340 2 The Problem of Saint Venant

Fig. 2.6.106

Fig. 2.6.107

Solution Since the cross section in examination admits the horizontal centroidal line
a as axis of symmetry (Fig. 2.6.107), because of the [2.6.4] the shear center Ct must
belong to a. This premised, we consider the distribution of τ z produced from a shear-
ing force T passing through Ct and such that the neutral axis n is a (Fig. 2.6.107). We
denote with f the centroidal line normal to n and oriented downward (Fig. 2.6.107).
Finally we suppose that Tf = 1 kg. It is clear that (Problem 2.6.3) T is vertical, has
intensity T = Tf = 1 kg and is directed downward. Clearly we can determine the
line of action of T by determining the central axis of the τ z produced from T. Using
the (2.6.29), we easily get the diagram of such τ z (Fig. 2.6.108).
We denote with b, c, d the axes individualized by the middle curve ρ
(Fig. 2.6.109). Evidently, for reason of symmetry, the τ z distributed on the rect-
angle individualized by b [resp. c] [resp. d] admit as central axis b [resp. c] [resp.
d] and then are equivalent to the alone force Rb [resp. Rc ] [resp. Rd ] of Fig. 2.6.109.
Furthermore, since the diagram of Fig. 2.6.108 is symmetrical with respect to axis
n, we also have Rb = Rd . Then the system constituted by the forces Rb , Rc, Rd
of Fig. 2.6.109 is equivalent to the system only constituted by the Rc applied at a
distance e from c such that (Fig. 2.6.110)

e Rc = hRb .
2.6 Shear 341

Fig. 2.6.108

Fig. 2.6.109

Fig. 2.6.110
342 2 The Problem of Saint Venant

Fig. 2.6.111

Clearly such line of action of Rc is the central axis of the τ z produced from the
shearing force T passing through Ct . So we obtain Ct as intersection between such
central axis and the symmetry axis n (Fig. 2.6.110). Since e > 0, Ct is found on the
left with respect to c. 

Problem 2.6.14 You shall determine the shear center Ct of the cross section shown
in Fig. 2.6.111.
Solution We apply the standard approximate procedure for the open thin walled
cross section, determining two lines ŝ1 , ŝ2 to which Ct must belong. To determine
ŝ1 we suppose to have applied on the cross section a shearing force T1 passing
through Ct . We calculate with the formula of Zhuravskii the τ z produced from T1 .
After that we determine ŝ1 as central axis of such distribution of τ z . To simplify the
calculations we choose T1 such that the neutral axis n1 is the horizontal centroidal
line and such that T1f1 = 1 kg, where f1 is the centroidal line normal to n1 and
directed downward (Fig. 2.6.112). Clearly T1 has the direction of the load axis s1 ,
that is of the diameter conjugated to n1 with respect to the inertia centroidal ellipse
(Fig. 2.6.112). Equally clear is the condition T1f1 = 1 kg which individualizes the
intensity and the verse of T1 (Fig. 2.6.112).

Fig. 2.6.112
2.6 Shear 343

Fig. 2.6.113

To individualize the line of action ŝ1 of T1 , we determine with the (2.6.29) the
distribution of the τ z produced from T1 . Proceeding as in the precedent problems
we easily verify that such diagram is Fig. 2.6.113. We now deal with finding the
central axis of the distribution of the τ z just calculated. To this aim we decompose
first of all the cross section in the rectangles A, B, C1 , C2 and denote with a [resp.
b] [resp. c] the part of the middle curve ρ individualized by A [resp. B] [resp. C1 ,
C2 ] (Fig. 2.6.114). We observe that the τ z insisting on the rectangle A are a system
of distributed forces parallel, equiverse and symmetric with respect to a. Insofar
the system of the τ z insisting on the rectangle A is equivalent to the system consti-
tuted by the alone force Ra , parallel and equiverse to the τ z , having line of action
a (Fig. 2.6.115). Moreover, denoting with ca [resp. la ] the height [resp. base] of the
rectangle A, obviously it results
  la  la
Ra = τz dA = τz ca dx = ca τz dx
A 0 0

so that Ra can be obtained multiplying ca for the area of the triangle of Fig. 2.6.113
relative to A.

Fig. 2.6.114
344 2 The Problem of Saint Venant

Fig. 2.6.115

Analogously the τ z insisting on the rectangle B are a system of distributed forces

parallel, equiverse and symmetric with respect to b. Insofar the system of the τ z
insisting on the rectangle B is equivalent to the system constituted by the alone
force Rb , parallel and equiverse to such τ z , having line of action b (Fig. 2.6.115).
Moreover, denoting with cb [resp. lb ] the base [resp. height] of the rectangle B,
obviously it results
  lb  lb
Rb = τz dA = τz cb dx = cb τz dx
B 0 0

so that Rb can be obtained multiplying cb for the area of the parabola of Fig. 2.6.113.
In the same way the system of the τ z insisting on the rectangles C1 , C2 is equiv-
alent to the system constituted by the alone force Rc , parallel to such τ z , having line
of action c (Fig. 2.6.115), of intensity
 
Rc = τz dC − τz dC.
C2 C1

Such way the distribution of the τ z produced from T1 is equivalent to the system
constituted by the three forces Ra , Rb , Rc having respectively line of action a, b, c
(Fig. 2.6.116). Insofar we can determinate the central axis ŝ1 of the distribution of
the τ z produced from T1 individualizing a system constituted by an unique force
and equivalent to Fig. 2.6.116. To this aim we verify that the three forces Ra , Rb , Rc
have resultant T1 (Fig. 2.6.117). After that we denote with O the point intersection

Fig. 2.6.116
2.6 Shear 345

Fig. 2.6.117

Fig. 2.6.118

between b and c and consider a line ŝ1 parallel to T1 (Fig. 2.6.118). The condi-
tion that the system of Fig. 2.6.116 and the system constituted by the alone T1 are
equivalent furnish

Ra dist(O, a) − T1 dist(O, ŝ1 ) = 0.

So the position of the central axis ŝ1 is determined (Fig. 2.6.119).

Now we determine the central axis ŝ2 . With the same reasoning, we suppose
to have applied on the cross section a shearing force T2 passing through Ct . We
calculate by the formula of Zhuravskii the τ z produced from T2 . To simplify the
calculations we choose T2 such that the neutral axis n2 is the vertical centroidal
line and such that T2f2 = 1 kg, where f2 is the centroidal line normal to n2 and
directed toward right (Fig. 2.6.120). Clearly T2 has the direction of the load axis s2 ,
that is of the diameter conjugated to n2 with respect to the inertia centroidal ellipse

Fig. 2.6.119
346 2 The Problem of Saint Venant

Fig. 2.6.120

Fig. 2.6.121

(Fig. 2.6.121). Equally clear is that the condition T2f2 = 1 kg individualizes the
intensity and the verse of T2 (Fig. 2.6.121).
To individualize the line of action ŝ2 of T2 , we determine with the (2.6.29) the
distribution of the τ z produced from T2 . Proceeding as in the precedent problems we
easyly verify that such diagram is Fig. 2.6.122. We now deal with finding the central
axis of the distribution of the just calculated τ z . Proceeding like in the precedent
case we get that the distribution of the τ z produced from T2 is equivalent to the
system constituted by three forces Sa , Sb , Sc having respectively line of action a, b,
c (Fig. 2.6.123). Insofar we can determinate the central axis ŝ2 of the distribution
of the τ z produced from T2 individualizing a system constituted by a unique force
and equivalent to that of Fig. 2.6.123. To this aim we verify that the three forces
Sa , Sb , Sc have resultant T2 (Fig. 2.6.124). After that we consider a line ŝ2 parallel
to T2 (Fig. 2.6.125). The condition that the system of Fig. 2.6.123 and the system
constituted by the alone T2 applied on ŝ2 are equivalent furnishes the position of
the central axis ŝ2 (Fig. 2.6.126). Intersecting ŝ1 and ŝ2 we get the shear center Ct
(Fig. 2.6.127). 
2.6 Shear 347

Fig. 2.6.122

Fig. 2.6.123

Fig. 2.6.124

Fig. 2.6.125
348 2 The Problem of Saint Venant

Fig. 2.6.126

Fig. 2.6.127

Fig. 2.6.128

Problem 2.6.15 You shall verify the structural safety for the cantilever of Fig.
2.6.128. At the free end a shearing force T is applied. The cross section, shown
in Fig. 2.6.129, is constant, open and thin walled. To the constrained cross section
hunch is allowed.
Solution Since T doesn’t pass for Ct , the beam operates not only on shear and bend-
ing but also on torsion. In fact since the postulate of Saint Venant [2.2.1] we can
replace the applied load of Fig. 2.6.129 with the equivalent system of forces of
Fig. 2.6.130, where d = (b – c) /2. Then for the principle of superposition [1.5.4]
the problem in examination is sum of the problem of Fig. 2.6.131 and of the prob-
lem of Fig. 2.6.132. We assume the reference frame of Fig. 2.6.131. In the problem
of shear and bending of Fig. 2.6.131 the stresses σ z , τ z only arise. The σ z are fur-
nished by the (2.6.25) and then assume their maximum at the jointed cross section z
= 0. The τ z are constant with z. In the problem of Fig. 2.6.132 the stresses τ z only
2.6 Shear 349

Fig. 2.6.129

Fig. 2.6.130

Fig. 2.6.131

arise. Such τ z are constant with z. Insofar we must verify the structural safety only
at the jointed cross section z = 0. In the problem of torsion of Fig. 2.6.131 at the
jointed cross section z = 0 we easily get the neutral axis n and then the diagram
of the σ z (Fig. 2.6.133). Furthermore in the same problem of Fig. 2.6.131, since T
passes through Ct , we can calculate the τ z by the formula of Zhuravskii (2.6.29).
350 2 The Problem of Saint Venant

Fig. 2.6.132

Fig. 2.6.133

Fig. 2.6.134

This way, proceeding as in the preceding problems, we trace the diagram of τ z of

Fig. 2.6.134.
About the problem of Fig. 2.6.132, using the approximate technique developed
in Sect. 2.5.5, we easily get the diagram of the τ z of Fig. 2.6.135. So in the problem
in examination we must verify the structural safety only at the jointed cross section
2.6 Shear 351

Fig. 2.6.135

z = 0, where the component of stress σ z , τ z only act. The σ z have the diagram
shown in Fig. 2.6.133. The τ z have a diagram that is vectorial sum of the diagram
of Fig. 2.6.134 and of the diagram of Fig. 2.6.135.
Theoretically it is necessary to verify the structural safety in every point of the
jointed cross section z = 0. In practice it is enough to perform the procedure only
in few points, where the state of stress appears onerous to a greater extent. To such
intention we explicitly notice, although this is evident, that in the point Q1 [resp.
Q2 ] of Fig. 2.6.133 the τ z originated from the shear and from the torsion must be
added [resp. subtract]. 

2.6.5 Closed Thin Walled Cross Section

The formula of Zhuravskii (2.6.29) is valid if the cross section is monoconnected.

However it is easy to extend it to the case of the closed thin walled cross section,
that obviously is multiconnected. Without loss of generality, we take in examination
a cantilever constrained by a punctual rigid joint, having the closed thin walled cross
section of Fig. 2.6.136 and submitted in the free extreme to a shearing force T whose
line of action s̃ pass through the shear center Ct (Fig. 2.6.136).
From the exact solution of the shear problem we get that for the problem in
examination (Fig. 2.6.137) the (2.6.25) is true. In such formula n denotes the neutral
axis, that is the diameter conjugated to the direction of T with respect to the inertia
centroidal ellipse of the cross section. Moreover f denotes a centroidal line normal
to n and directed so that the orthogonal projection of T on f is positive (Fig. 2.6.138).

Fig. 2.6.136
352 2 The Problem of Saint Venant

Fig. 2.6.137

Fig. 2.6.138

Still remembering the exact solution of the shear problem, we translate the load
T in G, assume the inertia principal frame of reference G, x, y of Fig. 2.6.137 and
denote with Tx [resp. Ty ] the orthogonal projection of T on the axis x [resp. y]. Then
from the (2.6.6) and (2.6.16) it follows

∂τzy ∂τzx Tx ∂ 2χ
curl τ z = − = −νy +
∂x ∂y 2(1 + ν)Iy ∂x∂y
Tx ∂ 2χ Ty ∂ 2ψ
− νy + + νx +
2(1 + ν)Iy ∂x∂y 2(1 + ν)Ix ∂x∂y
Ty ∂ 2ψ Tx ν Ty ν
− −νx + =− y+ x
2(1 + ν)Ix ∂x∂y (1 + ν)Iy (1 + ν)Ix

so that the mean value of curl τ z on the area A of the cross section is

-  
A curl τ z dA Tx ν Ty ν
= − y dA + x dA
A A(1 + ν)Iy A A(1 + ν)Ix A
(2.6.54)
Tx ν Ty ν
= − Sx + Sy = 0.
A(1 + ν)Iy A(1 + ν)Ix
2.6 Shear 353

After that we consider a generic cross section of the beam and a generic point
P of the middle curve ρ of the cross section (Fig. 2.6.138). We call chord2.6.4 at P
of the cross section the segment H, K normal to the tangent in P to ρ, as shown
in Fig. 2.6.138. Supposing that s̃ pass through Ct , we assume valid the hypothe-
sis of Zhuravskii that in every chord of the cross section the component of the τ z
according to the normal one to the chord is constant. Well we assume as unknown
of the problem the values of τ z in two convenient points Q1 , Q2 of the cross section
(Fig. 2.6.139). We consider a point P of ρ (Fig. 2.6.139). Applying the procedure of
Zhuravskii, we isolate from the beam the element shown in Fig. 2.6.140.
The equilibrium to the horizontal translation of such element (Fig. 2.6.140) it
furnishes
 
Tf (l − z) Tf [l − (z + dz)]
dn dA − dn dA
A0 In A0 In
− τz (P)c(P) dz + τz (Q1 )c(Q1 )dz = 0,

from which

Tf Sn c(Q1 )
τz (P) = + τz (Q1 ). (2.6.55)
In c(P) c(P)

It is evident (Fig. 2.6.139) that whatever P is chosen on ρ it is possible to express

τ z (P) with the (2.6.55) or with the formula

Fig. 2.6.139

Fig. 2.6.140

2.6.4 Insofar in the closed thin walled cross section a chord doesn t generally divide the cross section
into two parts.
354 2 The Problem of Saint Venant

Tf Sn c(Q1 ) c(Q2 )
τz (P) = + τz (Q1 ) − τz (Q2 )
In c(P) c(P) c(P)

or with the (2.6.29). As easy consequence, the [2.6.9], [2.6.10], [2.6.11], and
[2.6.12] are true for the closed thin walled cross section too. Insofar the diagram
of the τ z for the problem of Fig. 2.6.136 is determined, to less than τ z (Q1 ), τ z (Q2 ).
To calculate τ z (Q1 ), τ z (Q2 ) we write a linear algebraic system of two equations
with two unknown with non-singular matrix of the coefficients. In such system we
apply twice the theorem of Stokes, and then solve completely the problem. We notice
that the portion ρ 1 [resp. ρ 2 ] of the middle curve ρ of the cross section (Fig. 2.6.141)
[resp. (Fig. 2.6.142)] is a closed curve. We denote with S1 [resp. S2 ] the surface
constituted by the portion of plane delimited by ρ 1 [resp. ρ 2 ] (Fig. 2.6.141) [resp.
(Fig. 2.6.142)]. Suitably prolonging the function τ z on S1 and on S2 , the theorem of
Stokes furnishes the two equations
 
τ z × t ds = curl τ z × n dσ ,
ρ1 S1
 
τ z × t ds = curl τ z × n dσ ,
ρ2 S2

from which
 
τz ds = curl τ z dσ , (2.6.56)
ρ1 S1
 
τz ds = curl τ z dσ . (2.6.57)
ρ2 S2

Since the (2.6.54), we have that in the solving system (2.6.56) and (2.6.57) the
surface integrals are respectively equal to 2Gϑmeas2 (S1 ), 2Gϑmeas2 (S2 ), where ϑ
is the unitary angle of torsion of the cross section of Fig. 2.6.136 submitted to the
twisting moment T dist (G, Ct ).

Fig. 2.6.141

Fig. 2.6.142
2.6 Shear 355

Remark 2.6.15 In a closed thin walled cross section, if the middle curve ρ has m
closed meshes, it needs to assume as unknown of the problem the values of τ z in m
convenient points of ρ. 

It is easy to extend to the closed thin walled cross section the approximate tech-
nique that allows to determine the shear center of an open thin walled cross section.
In fact to determine the shear center Ct of the section of Fig. 2.6.136 it is enough
to apply on the cross section a shearing force T1 [resp. T2 ] passing for Ct . Exactly
as in the case of the open thin walled cross section, the force T1 [resp. T2 ] is cho-
sen assigning a convenient neutral axis n1 [resp. n2 ] and supposing T1f1 = 1 [resp.
T2f2 = 1]. In fact when we assign the neutral axis ni the direction of Ti is achieved,
which is parallel to the diameter of the inertia centroidal ellipse conjugated to ni .
Moreover the condition Tifi = 1 obviously determines the intensity and the verse
of Ti . Then it deals with calculating the two distributions of τ z by the previous for-
mulas and with determining their central axes ŝ1 , ŝ2 , to which Ct must belong. This
way Ct is determined as point of intersection of ŝ1 and ŝ2 .
Remark 2.6.16 We notice that the position of the shear center strongly changes if
a closed cross section becomes open. In fact, since the [2.6.4], the cross section of
Fig. 2.6.143 has shear center Ct coincident with the centroid G. Well if we operate
a thin cut as shown in Fig. 2.6.144, then the shear center Ct suffers an important
displacement (Fig. 2.6.150). Clearly the cross section of Fig. 2.6.144 and that of
Fig. 2.6.143 have the same centroid. Moreover, since the horizontal centroidal line
a (Fig. 2.6.145) is axis of symmetry, because of the [2.6.4] Ct must belong to a.

Fig. 2.6.143

Fig. 2.6.144
356 2 The Problem of Saint Venant

Fig. 2.6.145

Fig. 2.6.146

Then to individualize Ct , we apply on the section of Fig. 2.6.146 a shearing force

T passing through Ct and such that its neutral axis is a (Fig. 2.6.146). Besides we
denote with f the centroidal line normal to n, oriented as in Fig. 2.6.146, and we
suppose Tf = 1. It is obvious that for the cross section of Fig. 2.6.144 the axes n
and f are inertia principal axes. It immediately follows that T is vertical and has
intensity 1 kg.
We easily verify that the verses of the τ z generate by T are those of
Fig. 2.6.147. Moreover, since T is vertical, the horizontal τ z must have zero resul-
tant. Consequently the resultant one of the τ z insisting on the superior part of the
cross section is equal and opposite to the resultant one of the τ z insisting on the
inferior part of the cross section (Fig. 2.6.148). Now we impose that the system of

Fig. 2.6.147

Fig. 2.6.148
2.6 Shear 357

Fig. 2.6.149

Fig. 2.6.150

forces F, – F, F1 , F2 of Fig. 2.6.149 and the only shearing force T passing through
Ct have equal moment with respect to the point Q of Fig. 2.6.148. This way we
determine the central axis of the τ z , i.e. the line of action ŝ of T (Fig. 2.6.149). From
this the position of Ct follows (Fig. 2.6.150). 
Chapter 3
The Two-Dimensional Problems

3.1 Panels

3.1.1 The Problem of the Panel

Let V be the portion of the three-dimensional space occupied by a body and O, x,

y, z a cartesian orthogonal frame of reference. We say that the body is in a state of
plane stress with plane of stress x, y when ∀ (x, y, z1 ), (x, y, z2 ) ∈ V

σx (x, y, z1 ) = σx (x, y, z2 )
σy (x, y, z1 ) = σy (x, y, z2 )
σz (x, y, z1 ) = 0
(3.1.1)
τxy (x, y, z1 ) = τxy (x, y, z2 )
τxz (x, y, z1 ) = 0
τyz (x, y, z1 ) = 0.

Remark 3.1.1 Evidently if a body is in a state of plane stress then in every point of
the body the state of stress is plane. 

Taking into account the fact that the coefficient of Poisson ν is different from zero
(Sect. 1.4.2), it is easy to verify that the sextuple one of functions (3.1.1) cannot be
a solution of a problem of the elastic equilibrium. Then the problems of plane stress
don t exist in the physical reality.
Nevertheless a problem of notable practical interest exists, that of the plane panel,
whose solution satisfies with good approximation the (3.1.1).
We call panel a cylinder having the bases parallel to the plane x, y and height h
very small in comparison to the middle dimensions of the bases (Fig. 3.1.1). The
material of the panel is supposed homogeneous, isotropic and linearly elastic. The
deformations are, for hypothesis, small. We denote with A the generic cross section
of the cylinder, both monoconnected and multiconnected, with ∂A its boundary and,
∀ (x, y) ∈ ∂A, with nx , ny the direction cosines of the normal one to the tangent in
(x, y) to ∂A, directed going out from ∂A (Fig. 3.1.2).
The load applied on the panel consists of a volumetric load X̃, Ỹ, Z̃ such that
∀ (x, y, z1 ) , (x, y, z2 ) ∈ V

A. Maceri, Theory of Elasticity, DOI 10.1007/978-3-642-11392-5_3, 359


C Springer-Verlag Berlin Heidelberg 2010
360 3 The Two-Dimensional Problems

Fig. 3.1.1

Fig. 3.1.2

X̃ (x, y, z1 ) = X̃ (x, y, z2 )
Ỹ (x, y, z1 ) = Ỹ (x, y, z2 ) (3.1.2)
Z̃ (x, y, z1 ) = 0

and in a superficial load p̃x , p̃y , p̃z distributed on the lateral surface S of the cylinder,
normal to z and constant with z, i.e. (Fig. 3.1.1) ∀ (x, y, z1 ), (x, y, z2 ) ∈ S

p̃x (x, y, z1 ) = p̃x (x, y, z2 )

p̃y (x, y, z1 ) = p̃y (x, y, z2 ) (3.1.3)
p̃z (x, y, z1 ) = 0.

The two bases are unloaded, so that ∀ (x, y) ∈ A

     
p̃x x, y, − h2 = p̃y x, y, − h2 = p̃z x, y, − h2 = 0
     
p̃x x, y, h2 = p̃y x, y, h2 = p̃z x, y, h2 = 0.

Then the plane x, y is evidently a plane of symmetry both geometric and of

load for the panel. Insofar, for the principle of symmetry, every point of the panel
belonging to the plane x, y must have a displacement according to z equal to zero:

w (x, y, 0) = 0 ∀ (x, y) ∈ A.

In other terms, every point of the middle plane of the panel stays, during the
deformation, in the middle plane of the panel.
3.1 Panels 361

It has been always verified in the Laboratories material tests that every panel
substantially operates in a state of plane stress (of plane x, y). Precisely it has always
been verified that the unique solution of such problem of the elastic equilibrium with
good approximation ∀ (x, y, z1 ) , (x, y, z2 ) ∈ V satisfy the condition

σ̃x (x, y, z1 ) = σ̃x (x, y, z2 )

σ̃y (x, y, z1 ) = σ̃y (x, y, z2 )
σ̃z (x, y, z1 ) = 0
(3.1.4)
τ̃xy (x, y, z1 ) = τ̃xy (x, y, z2 )
τ̃xz (x, y, z1 ) = 0
τ̃yz (x, y, z1 ) = 0.

Remark 3.1.2 We can easily justify the third of the (3.1.4). In fact, observing that
the faces of the panel are unloaded, from the (1.2.5) we easily get ∀ (x, y) ∈ A

h h h
τ̃zx x, y, − = τ̃zy x, y, − = σ̃z x, y, − =0 (3.1.5)
2 2 2
h h h
τ̃zx x, y, = τ̃zy x, y, = σ̃z x, y, = 0. (3.1.6)
2 2 2

Besides, from the third of the (1.2.3) and from the (3.1.2) we draws ∀ (x, y, z) ∈ V

∂ τ̃zx ∂ τ̃zy ∂ σ̃z

(x, y, z) + (x, y, z) + (x, y, z) = 0. (3.1.7)
∂x ∂x ∂z
From the (3.1.5) and (3.1.7) it follows ∀ (x, y) ∈ A

∂ σ̃z h
x, y, − =0 (3.1.8)
∂z 2

and from the (3.1.6) and (3.1.7) we have ∀ (x, y) ∈ A

∂ σ̃z h
x, y, = 0. (3.1.9)
∂z 2

Since for hypothesis h is small, from the (3.1.5), (3.1.7), (3.1.6), and (3.1.9) it
follows with good approximation the third of the (3.1.4). 
Let us consider now the problem

[3.1.1] Find three real functions σx (x, y) , σy (x, y) , τxy (x, y) such that3.1.1

∂σx ∂τxy
+ +X =0 on A (3.1.10)
∂x ∂y

   
3.1.1 We put ∀((x, y), z) ∈ ∂A × −h 2, h 2 px (x, y) = p̃x (x, y, z), py (x, y) = p̃y (x, y, z) ;∀ (x, y, z)
∈ V X (x, y) = X̃ (x, y, z), Y (x, y) = Ỹ (x, y, z).
362 3 The Two-Dimensional Problems

∂τyx ∂σy
+ +Y =0 on A (3.1.11)
∂x ∂y
   
∂ 2 σx + σy ∂ 2 σx + σy ∂X ∂Y
+ = − (1 + ν) + on A (3.1.12)
∂x 2 ∂y2 ∂x ∂y
σx nx + τxy ny = px on ∂A (3.1.13)

τyx nx + σy ny = py on ∂A.  (3.1.14)

Employing the tools of the Functional analysis we can prove that, in convenient
hypotheses on the open A and on the known terms X, Y, px , py , the problem [3.1.1]
admits an unique solution σ x , σ y , τ xy .
 if σ
Well we can easily prove that, x , σ 
y , τxy are the unique solution of the problem
[3.1.1] and ∀ ((x, y) , z) ∈ A× − h 2, h 2

σ̃x (x, y, z) = σx (x, y)

σ̃y (x, y, z) = σy (x, y)
σ̃z (x, y, z) = 0
(3.1.15)
τ̃xy (x, y, z) = τxy (x, y)
τ̃xz (x, y, z) = 0
τ̃yz (x, y, z) = 0,

then the functions (3.1.15) satisfy the condition (3.1.4) and are a sextuple of stresses
balanced and approximately compatible for the problem of the elastic equilibrium
of the panel.
In fact the (3.1.15) banally satisfy the (3.1.4). Furthermore the fifth of the (3.1.15)
and the (3.1.10) assure that on V

∂ σ̃x ∂ τ̃xy ∂ τ̃xz

+ + + X̃ = 0;
∂x ∂y ∂z
the sixth of the (3.1.15) and the (3.1.11) assure that on V

∂ τ̃yx ∂ σ̃y ∂ τ̃yz

+ + + Ỹ = 0;
∂x ∂y ∂z
the (3.1.15) and the (3.1.2) assure that on V

∂ τ̃zx ∂ τ̃zy ∂ σ̃z

+ + + Z̃ = 0;
∂x ∂y ∂z
the (3.1.13) and the (3.1.15) assure that on S

σ̃x nx + τ̃xy ny + τ̃xz nz = p̃x ;

the (3.1.14) and the (3.1.15) assure that on S

τ̃yx nx + σ̃y ny + τ̃yz nz = p̃y ;

3.1 Panels 363

the (3.1.15) and the (3.1.3) assure that on S

τ̃zx nx + τ̃zy ny + σ̃z nz = p̃z ;

the (3.1.15) and the hypothesis that the faces of the panel are unloaded assure that
on the face z = −h/2[resp. z = h/2]

σ̃x nx + τ̃xy ny + τ̃xz nz = p̃x

τ̃yx nx + σ̃y ny + τ̃yz nz = p̃y
τ̃zx nx + τ̃zy ny + σ̃z nz = p̃z .

Insofar the sextuple one of stresses (3.1.15) is balanced (internally and at the
boundary).
With regard to the compatibility, from the (3.1.15) we get that ∀ (x, y, z) ∈ V

1  
ε̃x (x, y, z) = σx (x, y) − νσy (x, y)
E
1  
ε̃y (x, y, z) = σy (x, y) − νσx (x, y)
E
v 
ε̃z (x, y, z) = − σx (x, y) + σy (x, y)
E
2 (1 + ν)
γ̃xy (x, y, z) = τxy (x, y)
E
γ̃xz (x, y, z) = 0
γ̃yz (x, y, z) = 0,

from which
 
∂ 2 ε̃x ∂ 2 ε̃y 1 ∂ 2 σx ∂ 2 σy ∂ 2 σy ∂ 2 σx
+ = + −ν 2 −ν 2
∂y 2 ∂x2 E ∂y 2 ∂x 2 ∂y ∂x
from which, taking account of (3.1.10), (3.1.11), and (3.1.12), we have
 
∂ 2 γ̃xy 2 (1 + ν) ∂ 2 τxy 1+ν ∂ 2 σx ∂X ∂ 2 σy ∂Y
= = − 2 − − −
∂x∂y E ∂x∂y E ∂x ∂x ∂y2 ∂y
   
1 + ν ∂ σx 2 ∂ σy
2 1 ∂ σx2 ∂ σy
2 ∂ σx
2 ∂ 2 σy
= + + + + +
E ∂x2 ∂y2 E ∂x2 ∂x2 ∂y2 ∂y2
 
1 ∂ σx2 ∂ σy
2 ∂ σy
2 ∂ σx
2
= + −ν 2 −ν 2
E ∂y 2 ∂x 2 ∂y ∂x
so that the first compatibility equation is satisfied;
 
∂ 2 ε̃y ∂ 2 ε̃z ν ∂ 2 σx ∂ 2 σy ∂ 2 γ̃yz
+ =− + , =0
∂z2 ∂y2 E ∂y2 ∂y2 ∂y∂z
364 3 The Two-Dimensional Problems

so that the second compatibility equation is not satisfied;

 
∂ 2 ε̃z ∂ 2 ε̃x ν ∂ 2 σx ∂ 2 σy ∂ 2 γ̃zx
+ 2 =− + , =0
∂x 2 ∂z E ∂x 2 ∂x2 ∂z∂x
so that the third compatibility equation is not satisfied;

∂ 2 ε̃x ∂ ∂ γ̃xy ∂ γ̃xz ∂ γ̃zy

2 = 0, + − =0
∂y∂z ∂x ∂z ∂y ∂x
so that the fourth compatibility equation is satisfied;

∂ 2 ε̃y ∂ ∂ γ̃yz ∂ γ̃yx ∂ γ̃xz

2 = 0, + − =0
∂z∂x ∂y ∂x ∂z ∂y

so that the fifth compatibility equation is satisfied;

 
∂ 2 ε̃z ν ∂ 2 σx ∂ 2 σy ∂ ∂ γ̃zx ∂ γ̃zy ∂ γ̃yx
2 =− + , + − =0
∂x∂y E ∂x∂y ∂x∂y ∂z ∂y ∂x ∂z

so that the sixth compatibility equation is not satisfied.

It has always been verified in the Laboratories material tests that for every panel
substantially the second, third and sixth compatibility equation are satisfied with
good approximation. So the problem of the panel is, with good approximation, a
problem of plane stress. To determine its solution σ̃x , σ̃y , σ̃z , τ̃xy , τ̃xz , τ̃yz is enough,
because of the (3.1.15), to resolve the bidimensional problem [3.1.1].

3.1.2 Rectangular Panels

To solve the bidimensional boundary problem [3.1.1] we can employ the Airy3.1.2
function.
We consider the problem (Fig. 3.1.2)
[3.1.2] Find a real function Φ(x, y) such that

∂ 4 ∂ 4 ∂ 4
+ 2 + =0 on A (3.1.16)
∂x4 ∂x2 ∂y2 ∂y4
∂ 2 ∂ 2
nx − ny = px on ∂A (3.1.17)
∂y2 ∂x∂y
∂ 2 ∂ 2
− nx + ny = py on ∂A.  (3.1.18)
∂x∂y ∂x2

3.1.2 Sir George Biddel Airy, Alnwich 1801 – Greenwich 1892.

3.1 Panels 365

We call function of Airy every solution Φ 3.1.3 of the problem [3.1.2]. It is easy to
verify that every Airy function Φ individualizes the solution of the problem [3.1.1].
In fact if Φ is a solution of the problem [3.1.2], putting

∂ 2
σx = (3.1.19)
∂y2

∂ 2
σy = (3.1.20)
∂x2

∂ 2
τxy = − (3.1.21)
∂x∂y

we immediately verify that the triplet σ x , σ y , τ xy is solution of the problem [3.1.1].

Remark 3.1.3 Every polynomial in x and y of order non greater than three is an Airy
function. In fact all its derivatives of fourth order are identically zero. 
Let us consider now a rectangular panel having thickness b, base l and height h.
We assume the orthogonal cartesian reference O, x, y of Fig. 3.1.3 and we denote
with N a real number. We consider the Airy function (Remark 3.1.3)

N 2
= x . (3.1.22)
2bh

Since the (3.1.19), (3.1.20), and (3.1.21) the Airy function (3.1.22) individualizes
the state of stress
N
σx = 0, σy = , τxy = 0. (3.1.23)
bh
Since the (3.1.13) and (3.1.14), to induce in the panel the state of stress (3.1.23)
we must apply on the edge ∂A at y = 0 (where nx = 0, ny = −1) the load

px = σx nx + τxy ny = 0
py = τyx nx + σy ny = − bNh ;

Fig. 3.1.3

3.1.3 Every function satisfying the (3.1.16) is called biharmonic on A.

366 3 The Two-Dimensional Problems

at x = −h/2 (where nx = −1, ny = 0) the load

px = σx nx + τxy ny = 0
py = τyx nx + σy ny = 0;
at y = l (where nx = 0, ny = 1) the load
px = σx nx + τxy ny = 0
py = τyx nx + σy ny = bNh ;
at x = h/2 (where nx = 1, ny = 0) the load
px = σx nx + τxy ny = 0
py = τyx nx + σy ny = 0.
As a consequence the load to be applied on the panel is that of Fig. 3.1.4 (panel
submitted to axial load).
Remark 3.1.4 Evidently the load to apply on the edge y = l of the panel is equiv-
alent to a concentrated force having the axis x as line of action and intensity
(Fig. 3.1.5)

 h /2  h /2 
N N h/2
b py dx = b dx = dx = N.
−h/2 −h/2 bh h −h/2

If l is much greater than h, since the postulate of Saint Venant the state of stress
of the panels of Figs. 3.1.4 and 3.1.5 are anywhere equal except near the loaded
edges. Furthermore, if we denote with A the area of the cross section of the panel,
obviously
N
σy = . 
A
Let us examine now another load condition for the rectangular panel having
thickness b, base l and height h. We assume the orthogonal cartesian reference
O, x, y of Fig. 3.1.3 and we denote with M a real number. We consider the Airy

Fig. 3.1.4

Fig. 3.1.5
3.1 Panels 367

function (Remark 3.1.3)

2M 3
= x . (3.1.24)
b h3
Since the (3.1.19), (3.1.20), and (3.1.21) the Airy function (3.1.24) individualizes
the state of stress

12 M
σx = 0, σy = x, τxy = 0. (3.1.25)
b h3
Since the (3.1.13) and (3.1.14), to induce in the panel the state of stress (3.1.25)
we must apply on the edge ∂A at y = 0 (where nx = 0, ny = −1) the load

12 M
px = 0, py = − x;
b h3
at x = −h/2 (where nx = −1, ny = 0) the load

px = 0, py = 0;

at y = l (where nx = 0, ny = 1) the load

12 M
px = 0, py = x;
b h3
at x = h/2 (where nx = 1, ny = 0) the load

px = 0, py = 0.

As a consequence the load to be applied on the panel is that of Fig. 3.1.6 (panel
submitted to bending).
Remark 3.1.5 Evidently the load to apply on the edge y = l of the panel has resultant
having the axis y as line of action and intensity
 h/2  h/2
12 M
b py dx = x dx = 0
−h/2 h3 −h/2

Fig. 3.1.6
368 3 The Two-Dimensional Problems

Fig. 3.1.7

and resultant moment having axis z (orthogonal both to axis x and to the axis y) and
intensity (Fig. 3.1.7)
 h/2  h /2
12 M
b py x dx = x2 dx = M.
−h/2 h3 −h/2

If l is much greater than h, since the postulate of Saint Venant the state of stress of
the panels of Figs. 3.1.6 and 3.1.7 are anywhere equal except near the loaded edges.
Furthermore, since the moment of inertia with respect to z of the cross section of the
panel is (Fig. 3.1.8)

b h3
I= ,
12
we obviously have (formula of Navier)

M
σy = x. 
I
Let us examine now a third load condition for the rectangular panel having thick-
ness b, base l and height h. We assume the orthogonal cartesian reference O, x, y of
Fig. 3.1.3 and we denote with T a real number. We consider the function

2T 3 3T
= x y− xy, (3.1.26)
b h3 2bh
that clearly satisfies the (3.1.16). Then we can assume the (3.1.26) as Airy func-
tion for the panel. Since the (3.1.19), (3.1.20), and (3.1.21) such Airy function
individualizes the state of stress
12 T 3T 6T 2
σx = 0, σy = xy, τxy = − x . (3.1.27)
b h3 2 b h b h3
Since the (3.1.13) and (3.1.14), to induce in the panel the state of stress (3.1.27)
we must apply on the edge ∂A at y = 0 (where nx = 0, ny = −1) the load

Fig. 3.1.8
3.1 Panels 369

3T 6T 2
px = − + x , py = 0;
2 b h b h3
at x = −h/2 (where nx = −1, ny = 0) the load

3T 6 T h2
px = 0, py = − = 0;
2 b h b h3 4
at y = l (where nx = 0, ny = 1) the load

3T 6T 2 12 l T
px = − x , py = x;
2 b h b h3 b h3
at x = h/2 (where nx = 1, ny = 0) the load
3T 6 T h2
px = 0, py = − = 0.
2 b h b h3 4
As a consequence the load to be applied on the panel is that of Fig. 3.1.9 (panel
submitted to shear).
Remark 3.1.6 Evidently the load to apply on the edge y = l of the panel has a
resultant of intensity T and parallel to axis x (because its component on the axis x is
 h/2  h/2  h/2
3T 6T 3T T
b px dx = dx − x2 dx = − =T
−h/2 2h −h/2 h3 −h/2 2 2

and its component on the axis y is

 h/2  h /2
12 l T
b py dx = x dx = 0),
−h/2 h3 −h/2

and resultant moment having axis z (orthogonal both to axis x and to the axis y) and
intensity (Fig. 3.1.10)
 h/2  h/2
12 l T
b py x dx = x2 dx = T l.
−h/2 h3 −h/2

If l is much greater than h, since the postulate of Saint Venant the state of stress of
the panels of Figs. 3.1.9 and 3.1.10 are anywhere equal except near the loaded edges.

Fig. 3.1.9
370 3 The Two-Dimensional Problems

Fig. 3.1.10

Furthermore, observing that the bending moment in the cross section of abscissa y
is (Fig. 3.1.10) M (y) = T y and that

b h3
I= ,
12
we obviously have (formula of Navier)

M (y)
σy = x.
I
We also notice that τ xy is constant with y and that its maximum value is on the
T
axis y and is equal to 1.5 times bh . 

3.1.3 Circular Panels

The circular panels are more easily treated employing the polar coordinates. We
consider an orthogonal cartesian reference O, x, y and every point P(x, y) of the
plane (Fig. 3.1.11). We consider the straight line r passing through O and P. Then
we rotate around O the axes x, y in sense counterclockwise up to overlap the axis
x with r. The axes r, t that we obtained respectively form with the axes x and y an
angle of ϑ radiant (Fig. 3.1.12). Putting r = dist (P, O), we call r [resp. ϑ] first
[resp. second] polar coordinate of P and we use the notation

P = ((r, ϑ)) .

We have (Fig. 3.1.12)

x = r cosϑ, y = r sinϑ (3.1.28)

Fig. 3.1.11

Fig. 3.1.12
3.1 Panels 371

and
 1 y
2
r = x2 + y2 , ϑ = arctg , (3.1.29)
x
so that

∂r x
= = cosϑ
∂x r
(3.1.30)
∂r y
= = sinϑ,
∂y r
∂ϑ y sinϑ
=− 2 =−
∂x r r
(3.1.31)
∂ϑ x cosϑ
= 2 = .
∂y r r

Furthermore, since the unit vector x [resp. y] has in the frame of reference O, r, t
the direction cosines xr = cosϑ, xt = −sinϑ [resp. yr = sinϑ, yt = cosϑ], from the
(1.2.8) and (1.2.9) we get

σx = σr xr2 + σt xt2 + 2τrt xr xt

= σr cos2 ϑ + σt sin2 ϑ − τrt sin2ϑ
σy = σr y2r + σt y2t + 2τrt yr yt
(3.1.32)
= σr sin2 ϑ + σt cos t2 ϑ + τrt sin 2ϑ
τxy = σr xr yr + σt xt yt + τrt (xr yt + xt yr )
= σ2r sin 2ϑ − σ2t sin 2ϑ + τrt cos 2ϑ.

Let us consider now the problem

[3.1.3] Find three real functions σr (r, ϑ) , σt (r, ϑ) , τrt (r, ϑ) such that on A

∂σr 1 ∂τrt σr − σt
+ + =0 (3.1.33)
∂r r ∂ϑ r

1 ∂σt ∂τrt τrt

+ +2 =0 (3.1.34)
r ∂ϑ ∂r r

∂2 1 ∂2 1 ∂
(σr + σt ) + (σr + σt ) + (σr + σt ) = 0 (3.1.35)
∂r2 r2 ∂ϑ 2 r ∂r

and on ∂A3.1.4

every point P = (x, y) = ((r, t)) of ∂A we denote with nr , nt , the direction cosines (in the
3.1.4 In
frame of reference O, r, t of Fig. 3.1.12) of the outgoing normal n to the tangent in P to ∂A. Besides
we denote with pr , pt the components on the axes r, t of the load p (so that px + py = pr + pt ).
372 3 The Two-Dimensional Problems

σr nr + τrt nt = pr (3.1.36)

τrt nr + σt nt = pt .  (3.1.37)

It is easy to be prove that problems [3.1.1] and [3.1.3] are equivalent. Let in fact
σ x , σ y , τ xy a solution of the problem [3.1.1]. For every point P of A we denote with
σ r , σ t , τ rt the components of stress according to the axes r, t of Fig. 3.1.12, so that
the (3.1.32) are true. Well from the (3.1.10) we get

∂σx ∂τxy ∂  
0= + = σr cos2 ϑ + σt sin2 ϑ − τrt sin2ϑ
∂x ∂y ∂x

∂ σr σt 
+ sin2ϑ − sin2ϑ + τrt cos2ϑ
∂y 2 2
 
∂   ∂r
= σr cos2 ϑ + σt sin2 ϑ − τrt sin2ϑ
∂r ∂x
 
∂   ∂ϑ
+ σr cos2 ϑ + σt sin2 ϑ − τrt sin2ϑ
∂ϑ ∂x
   
∂ σr σt ∂r
+ sin2ϑ − sin2ϑ + τrt cos2ϑ
∂r 2 2 ∂y
  
∂ σr σt ∂ϑ
+ sin2ϑ − sin2ϑ + τrt cos2ϑ
∂ϑ 2 2 ∂y

that furnishes, taking account of (3.1.30) and (3.1.31)

∂σr ∂σt 2 ∂τrt

0 = cosϑ cos2 ϑ + sin ϑ − sin2ϑ
∂r ∂r ∂r

sinϑ ∂σr ∂σt 2
− (σt − σr ) sin2ϑ − 2τrt cos2ϑ + cos2 ϑ + sin ϑ
r ∂ϑ ∂ϑ
  
∂τrt ∂σr ∂σt sin2ϑ ∂τrt
− sin2ϑ + sinϑ − + cos2ϑ
∂ϑ ∂r ∂r 2 ∂r
cosϑ
+ [(σr − σt ) cos2ϑ − 2τrt sin2ϑ
r

∂σr ∂σt sin2ϑ ∂τrt
+ − + cos2ϑ
∂ϑ ∂ϑ 2 ∂ϑ
∂σr ∂σt sinϑ ∂τrt ∂τrt cosϑ
= cosϑ − + 2τrt − sinϑ + + σr − σt .
∂r ∂ϑ r ∂r ∂ϑ r
(3.1.38)
3.1 Panels 373

From the (3.1.11) we get

∂τyx ∂σy ∂  σr σt 
0= + = sin2ϑ − sin2ϑ + τrt cos2ϑ
∂x ∂y ∂x 2 2
∂  
+ σr sin2 ϑ + σt cos2 ϑ + τrt sin2ϑ
∂y
   ∂r
∂ σr σt
= sin2ϑ − sin2ϑ + τrt cos2ϑ
∂r 2 2 ∂x
   
∂ σr σt ∂ϑ
+ sin2ϑ − sin2ϑ + τrt cos2ϑ
∂ϑ 2 2 ∂x
 
∂   ∂r
+ σr sin2 ϑ + σt cos2 ϑ + τrt sin2ϑ
∂r ∂y
 
∂   ∂ϑ
+ σr sin2 ϑ + σt cos2 ϑ + τrt sin2ϑ
∂ϑ ∂y

that furnishes, taking account of (3.1.30) and (3.1.31)

 
∂σr ∂σt sin2ϑ ∂τrt
0 = cosϑ − + cos2ϑ
∂r ∂r 2 ∂r

sinϑ ∂σr ∂σt sin2ϑ ∂τrt
− − + cos2ϑ + (σr − σt ) cos2ϑ
r ∂ϑ ∂ϑ 2 ∂ϑ
  
∂σr ∂σt ∂τrt
−2τrt sin2ϑ + sinϑ sin ϑ +
2 cos ϑ +
2 sin2ϑ
∂r ∂r ∂r
 (3.1.39)
cosϑ ∂σr ∂σt
+ (σr − σt ) sin2ϑ + 2τrt cos2ϑ + sin2 ϑ cos2 ϑ
r ∂ϑ ∂ϑ

∂τrt ∂σr ∂σt cosϑ
+ sin2ϑ = sinϑ + + 2τrt
∂ϑ ∂r ∂ϑ r
∂τrt ∂τrt sinϑ
+ cosϑ + + σr − σt .
∂r ∂ϑ r

Multiplying for cosϑ [resp. sinϑ] the first one and the second member of the
(3.1.38) [resp. (3.1.39)] and adding, we obtain

∂σr ∂τrt 1
0= + + σr − σt
∂r ∂ϑ r

so that the (3.1.33) is true.

Multiplying for −sinϑ [resp. cosϑ] the first one and the second member of the
(3.1.38) [resp. (3.1.39)] and adding, we obtain
374 3 The Two-Dimensional Problems

∂τrt ∂σt 1
0= + + 2τrt
∂r ∂ϑ r

so that the (3.1.34) is true.

From the (3.1.12), taking account of (3.1.30), (3.1.31), and (3.1.32) we get
   
∂ ∂   ∂ ∂  
0= σx + σy + σx + σy
∂x ∂x ∂y ∂y
 
∂ ∂   sinϑ ∂  
= cosϑ σx + σy − σx + σy
∂x ∂r r ∂ϑ
 
∂ ∂ cosϑ ∂
+ sinϑ (σr + σt ) + (σr + σt )
∂y ∂r r ∂ϑ
 
∂ ∂ sinϑ ∂
= cosϑ (σr + σt ) − (σr + σt )
∂x ∂r r ∂ϑ
 
∂ ∂ cosϑ ∂
+ sinϑ (σr + σt ) + (σr + σt )
∂y ∂r r ∂ϑ
∂2 1 ∂2 1 ∂
= 2 (σr + σt ) + 2 (σr + σt ) + (σr + σt )
∂r r ∂ϑ 2 r ∂r

so that the (3.1.35) is true.

We now observe that the axis r [resp. t] has in the reference frame O, x, y
(Fig. 3.1.12) direction cosines rx = cosϑ [resp. tx = −sinϑ], ry = sinϑ [resp.
ty = cosϑ]. Multiplying first and second member of the (3.1.13) [resp. (3.1.14)] for
cosϑ [resp. sinϑ], adding and taking into account the (3.1.32) and the properties of
the scalar product we get

pr = px rx + py ry = px cosϑ + py sinϑ
= σx nx cosϑ + τxy ny cosϑ + τxy nx sinϑ + σy ny sinϑ
= σr cos3 ϑ nx + σt sin2 ϑ cosϑnx − τrt sin2ϑ cosϑ nx
σr − σt σr − σt
+ sin2ϑ cosϑ ny + τrt cos2ϑ cosϑ ny + sin2ϑ sinϑ nx
2 2
+τrt cos2ϑ sinϑ nx + σt sin ϑ ny + σt cos ϑ sinϑ ny
3 2

+τrt sin2ϑ sinϑ ny = σr nr + τrt nt

so that the (3.1.36) is true. Analogously, multiplying first and second member of the
(3.1.13) [resp. (3.1.14)] for −sinϑ[resp. cosϑ], adding and taking into account the
(3.1.32) and the properties of the scalar product we get

pt = px tx + py ty = −px sinϑ + py cosϑ

= −σx nx sinϑ − τxy ny sinϑ + τxy nx cosϑ + σy ny cosϑ
= τrt nr + σt nt

so that the (3.1.37) is true.

3.1 Panels 375

With analogous reasoning we prove that, if (σr , σt , τrt ) is solution of problem

[3.1.3], then the components of stress according to the axes x, y (Fig. 3.1.12)
furnished from the (3.1.32) are solution of the problem [3.1.1].
Remark 3.1.7 The (3.1.33) [resp. (3.1.34)] expresses the equilibrium of an infinites-
imal element of panel (containing the point P = ((r, t))) to the translation according
to the axis r [resp. t] (Fig. 3.1.13). We consider in fact the infinitesimal element
of panel of Fig. 3.1.13. The component according to r of the forces applied on the
edges AD and BC is equal to

∂σr
− σr r dϑ b + σr + dr (r + dr) dϑb
∂r
or, neglecting the infinitesimal ones of superior order

∂σr
r dr dϑ b + σr dr dϑ b.
∂r

The component according to r of the forces applied on the edges AB and CD is

equal to

∂τrt dϑ
− τrt dr b + τrt + dϑ dr b − 2σt dr b sin
∂ϑ 2
or, neglecting the infinitesimal ones of superior order

∂τrt
dr dϑ b − σt dr dϑ b.
∂ϑ

As a consequence, since the loads of mass are zero for hypothesis, the equilib-
rium of the element of Fig. 3.1.13 to the translation according to the axis r furnishes

∂σr ∂τrt
r dr dϑ b + σr dr dϑ b + dr dϑ b − σt dr dϑ b = 0
∂r ∂ϑ
from which the (3.1.33) immediately follows.

Fig. 3.1.13
376 3 The Two-Dimensional Problems

The component according to t of the forces applied on the edges AD and BC is

equal to
∂τrt
− τrt r dϑ b + τrt + dr (r + dr) dϑ b
∂r

or, neglecting the infinitesimal ones of superior order

∂τrt
r dr dϑ b + τrt dr dϑ b.
∂r
The component according to t of the forces applied on the edges AB and CD is
equal to
dϑ ∂σt
2 τrt dr b sin − σt dr b + σt + dϑ dr b
2 ∂ϑ
or, neglecting the infinitesimal ones of superior order
∂σt
τrt dr dϑ b + dr dϑ b.
∂ϑ
As a consequence, since the loads of mass are zero for hypothesis, the equilib-
rium of the element of Fig. 3.1.13 to the translation according to the axis t furnishes

∂τrt ∂σt
r dr dϑ b + 2 τrt dr dϑ b + dr dϑ b = 0
∂r ∂ϑ

from which the (3.1.34) immediately follows. 

The bidimensional boundary problem [3.1.3] can be resolved employing the
function of Airy in polar coordinates. We call function of Airy in polar coordinates
a solution Φ of the problem
[3.1.4] Find a real function Φ(r, t) such that in A

∂ 4Φ 2 ∂ 4Φ 1 ∂ 4Φ 2 ∂ 3Φ 2 ∂ 3Φ
+ 2 2 2+ 4 + − 3
∂r 4 r ∂r ∂ϑ r ∂ϑ 4 r ∂r 3 r ∂r∂ϑ 2
(3.1.40)
1 ∂ 2ϑ 4 ∂ 2Φ 1 ∂Φ
− 2 2 + 4 + 3 =0
r ∂r r ∂ϑ 2 r ∂r
and in ∂A

nr ∂Φ nr ∂ 2 Φ nt ∂Φ nt ∂ 2 Φ
+ 2 + − = pr (3.1.41)
r ∂r r ∂ϑ 2 r2 ∂ϑ r ∂r∂ϑ
nr ∂Φ nr ∂ 2 Φ ∂ 2Φ
− + nt = pt .  (3.1.42)
r2 ∂ϑ r ∂r∂ϑ ∂r2
It is immediate to verify that every solution Φ of the problem [3.1.4] is solution
of the problem [3.1.3]. In fact, if Φ is a solution of the problem [3.1.4], putting
3.1 Panels 377

1 ∂Φ 1 ∂ 2Φ
σr = + 2
r ∂r r ∂ϑ 2
∂ 2Φ (3.1.43)
σt =
∂r2
1 ∂Φ 1 ∂ 2Φ
τrt = 2 − ,
r ∂ϑ r ∂r∂ϑ
we easily verify that the triplet (σr , σt , τrt ) satisfies the (3.1.33), (3.1.34), (3.1.36),
and (3.1.37). Furthermore, since the (3.1.43) we have

∂2 1 ∂2 1 ∂
(σr + σt ) + (σr + σt ) + (σr + σt )
∂r 2 r ∂ϑ
2 2 r ∂r
∂2 ∂ 2Φ 1 ∂ 2Φ 1 ∂Φ 1 ∂2 ∂ 2Φ 1 ∂ 2Φ 1 ∂Φ
= + + + + +
∂r2 ∂r2 r2 ∂ϑ 2 r ∂r r2 ∂ϑ 2 ∂r2 r2 ∂ϑ 2 r ∂r
1 ∂ ∂ 2Φ 1 ∂ 2Φ 1 ∂Φ ∂ 4Φ ∂ 1 ∂ 3Φ 2 ∂ 2Φ
+ + + = + −
r ∂r ∂r2 r2 ∂ϑ 2 r ∂r ∂r4 ∂r r2 ∂r∂ϑ 2 r3 ∂ϑ 2
∂ 1 ∂ 2Φ 1 ∂Φ 1 ∂ 4Φ 1 ∂ 4Φ 1 ∂ 3Φ 1 ∂ 3Φ
+ − 2 + 2 2 2+ 4 + 3 +
∂r r ∂r
2 2 r ∂r r ∂r ∂ϑ r ∂ϑ 4 r ∂r∂ϑ 2 r ∂r3
1 1 ∂ 3Φ 2 ∂ 2Φ 1 1 ∂Φ 1 ∂ 2Φ ∂ 4Φ
+ − + − 2 + =
r r2 ∂r∂ϑ 2 r3 ∂ϑ 2 r r ∂r r ∂r2 ∂r4
2 ∂ 4Φ 1 ∂ 4Φ 2 ∂ 3Φ 2 ∂ 3Φ 1 ∂ 2Φ 4 ∂ 2Φ 1 ∂Φ
+ + + − − + + 3 .
r ∂r ∂ϑ
2 2 2 r ∂ϑ
4 4 r ∂r 3 r ∂r∂ϑ
3 2 r ∂r
2 2 r ∂ϑ
4 2 r ∂r

So, since the (3.1.40), the (3.1.35) is also it true.

After that we can consider the problem of the circular panel (Fig. 3.1.14) [resp.
pipe (Fig. 3.1.15)]. The cross section is a circle [resp. annulus]. We assume as origin
of the reference frame the center O of the circle [resp. annulus]. Besides, we denote
with pi and pe two real numbers and suppose that on the edge of the panel a load
pr = −pe is applied (Fig. 3.1.14) [resp. a load pr = −pe on the external edge and a
load pr = pi on the internal edge (Fig. 3.1.15)]. Such problem has a polar symmetry
because every line passing through O is axis of geometrical and loading symmetry.
So, since the principle of symmetry of the Physics, all the involved quantity are

Fig. 3.1.14
378 3 The Two-Dimensional Problems

Fig. 3.1.15

independent from ϑ and then all their derivatives with respect to ϑ are zero. As a
consequence, we must determine a function F such that on A

d4 Φ 2 d3 Φ 1 d2 Φ 1 dΦ
4
+ 3
− 2 2
+ 3 = 0. (3.1.44)
dr r dr r dr r dr

To this aim we put t = loge r and observe that

dΦ dΦ dt 1 dΦ dΦ
= = = e−t
dr dt dr r dt dt

d2 Φ d dΦ d2 Φ dΦ
= e−t e−t = e−2t − e−2t
dr2 dt dt dt2 dt
d3 Φ −3t d Φ − 3e−3t d Φ + 2e−3t dΦ
3 2
= e
dr3 dt3 dt2 dt
d Φ
4
−4t d Φ − 6e−4t d Φ + 11e−4t d Φ − 6e−4t dΦ .
4 3 2
= e
dr4 dt4 dt3 dt2 dt

This way the (3.1.44) becomes

d4 Φ −4t d Φ + 11e−4t d Φ − 6e−4t dΦ
3 2
e−4t − 6e
dt4 dt3 dt2 dt
d Φ
3 dΦ d Φ
2 dΦ
+ 2e−4t 3 + 4e−4t − e−4t 2 + e−4t =0
dt dt dt dt

that is
d4 Φ d3 Φ d2 Φ
− 4 + 4 = 0. (3.1.45)
dt4 dt3 dt2

Putting

d2 Φ
f =
dt2
3.1 Panels 379

the (3.1.45) becomes the ordinary differential equation with constant coefficients

d2 f df
2
− 4 + 4f = 0 (3.1.46)
dt dt
whose characteristic equation

k2 − 4 k + 4 = 0

has the double root 2. Thus the general integral of the (3.1.46) is

f = (c1 + c2 t) e2t

and as a consequence

d2 Φ
= (c1 + c2 t) e2t
dt2
from which

dΦ c c2 c2 
1
= − + t e2t + c3
dt 2 4 2
from which
c c2 c2  e2t c2 e2t
1
Φ (t) = − + t − + c3 t + c4
2 4 2 2 2 4
= C e2t + B t e2t + A t + D
from which

Φ (r) = Φ (t (r)) = Ce2logr + B e2logr logr + A logr + D

= A logr + B r2 logr + Cr2 + D. (3.1.47)

For every real number A, B, C, D the (3.1.47) is solution of the (3.1.44) and then
of the (3.1.40).
In the case of Fig. 3.1.14, assuming A = B = 0, the (3.1.42) is banally satisfied
and the (3.1.41) furnishes
1
2Cr = pr = −pe
r
so that C = − p2e . From this and from the (3.1.47) and (3.1.43) we get

σr = −pe , σt = −pe , τrt = 0.

380 3 The Two-Dimensional Problems

In the case of Fig. 3.1.15, assuming B = 0 and

a2 c2 (pe − pi ) pi a2 − pe c2
A= , C=  ,
c2 − a2 2 c2 − a2
on the external edge of the annulus it results

nr ∂Φ nr ∂ 2 Φ nt ∂Φ nr ∂ 2 Φ 1 ∂Φ 1 A
+ 2 + 2 − = = +2C c
r ∂r r ∂ϑ 2 r ∂ϑ r ∂r∂ϑ c ∂r c c

a2 (pe − pi ) pi a2 − pe c2
= + = −pe = pr
c2 − a2 c2 − a2

so that the (3.1.41) is satisfied, and

nr ∂Φ nr ∂ 2 Φ ∂ 2Φ ∂ 2Φ
− + nt = 0 · =0
r2 ∂ϑ r ∂r∂ϑ ∂r2 ∂r2
so that the (3.1.42) is satisfied; on the internal edge of the annulus it results

nr ∂Φ nr ∂ 2 Φ nt ∂Φ nr ∂ 2 Φ 1 ∂Φ 1 A
+ 2 + − =− =− +2C a
r ∂r r ∂ϑ 2 r2 ∂ϑ r ∂r∂ϑ a ∂r a a
c2 (pe − pi ) pi a2 − pe c2
=− − = pi = pr
c2 − a2 c2 − a2
so that the (3.1.41) is satisfied, and

nr ∂Φ nr ∂ 2 Φ ∂ 2Φ ∂ 2Φ
− + nt 2 = 0 · =0
r ∂ϑ
2 r ∂r∂ϑ ∂r ∂r2
so that the (3.1.42) is satisfied. This way, in the case of Fig. 3.1.15, from the (3.1.43)
we get

a2 c2 (pe − pi ) 1 pi a2 − pe c2
σr = · +
c2 − a2 r2 c2 − a2
a2 c2 (pe − pi ) 1 pi a2 − pe c2
σt = − · 2+
c −a
2 2 r c2 − a2
τrt = 0.

In particular in pipes under pressure, in which pe = 0, it results

a2 pi c2 a2 pi c2
σr = 1− , σt = 1+ , τrt = 0
c2 − a2 r2 c2 − a2 r2
3.1 Panels 381

so that σ r is always negative (compression) and σ t is always positive (traction).

The maximum one of σ t is for r = a (to the inside edge of the pipe) and is
equal to

a2 + c2
σtmax = pi
c2 − a2
that is greater than pt (whatever is the thickness of the pipe).

Remark 3.1.8 If the thickness s of the pipe is very small in comparison to the inside
ray a, σtmax is (with good approximation) equal to

a2 + (a + s)2 a
σtmax = pi ∼
= pi . 
(a + s) 2
− a2 s

Let us study now the problem of the circular panel loaded by two diame-
tral forces. To do this we preliminarily have to study the problem of Boussinesq.
Therefore we consider the boundless panel of Fig. 3.1.16 and denote with b its
thickness. The panel is delimited from a plane, so that the area A is an half-plane
delimited by a line that we assume as axis y (Fig. 3.1.17). On the segment OO of
the edge of the panel (Fig. 3.1.16) an uniformly distributed load F (tons/meter) is
applied, normal to the edge of the panel.
We assume the orthogonal cartesian frame of reference O, x, y of Fig. 3.1.17 and
we consider the function

F
Φ (r, ϑ) = − r ϑ sinϑ. (3.1.48)
π

Fig. 3.1.16

Fig. 3.1.17
382 3 The Two-Dimensional Problems

Since

∂Φ F ∂ 2Φ ∂ 3Φ ∂ 4Φ
= − ϑ sin ϑ, = 0, = 0, = 0,
∂r π ∂r 2 ∂r 3 ∂r4
∂ 4Φ ∂Φ F
= 0, = − r (sin ϑ + ϑ cos ϑ) ,
∂r2 ∂ϑ 2 ∂ϑ π
∂ 2Φ F ∂ 2Φ F
= − (sin ϑ + ϑ cos ϑ) , = − r (2 cos ϑ − ϑ sin ϑ) ,
∂r∂ϑ π ∂ϑ 2 π
∂ 3Φ F ∂ 3Φ F
= − (2 cos ϑ − ϑ sin ϑ) , = − r (−3 sin ϑ − ϑ sin ϑ) ,
∂r∂ϑ 2 π ∂ϑ 3 π
∂ 4Φ F
= − r (−4 cos ϑ + ϑ sin ϑ) ,
∂ϑ 4 π
we have for every P (r, ϑ) ∈ A (Fig. 3.1.17)

∂ 4Φ 2 ∂ 4Φ 1 ∂ 4Φ 2 ∂ 3Φ 2 ∂ 3Φ 1 ∂ 2Φ 4 ∂ 2Φ
+ + + − − +
∂r4 r2 ∂r2 ∂ϑ 2 r4 ∂ϑ 4 r ∂r3 r3 ∂r∂ϑ 2 r2 ∂r2 r4 ∂ϑ 2
1 ∂Φ
+ 3 =0
r ∂r
so that the (3.1.48) satisfies the (3.1.40). Let now Q (r, ϑ)be a point of ∂A (that is of
the axis y of Fig. 3.1.17) distinct from O, so that

pr = 0, pt = 0. (3.1.49)

Since nr = 0 and nt = −1, we have

nr ∂Φ nr ∂ 2 Φ nt ∂Φ nr ∂ 2 Φ
+ 2 + − =0 (3.1.50)
r ∂r r ∂ϑ 2 r2 ∂ϑ r ∂r∂ϑ
nr ∂Φ nr ∂ 2 Φ ∂ 2Φ
− + nt = 0. (3.1.51)
r2 ∂ϑ r ∂r∂ϑ ∂r2
The (3.1.49), (3.1.50), and (3.1.51) assure that the (3.1.48) is the function of Airy
of the problem of Fig. 3.1.16, that we call problem of Boussinesq.3.1.5
Hence, because of the (3.1.43), the stress in the panel is given by

2 F cos ϑ
σr = − , σt = 0, τrt = 0. (3.1.52)
π r
Remark 3.1.9 The point O of Fig. 3.1.18 is clearly singular in the respects of the
boundary conditions of the problem [3.1.4]. About such singularity we can develop
the following reasoning. We isolate by the remaining panel a semicircular portion
having ray r and center in O (Fig. 3.1.18). Well we load such portion with the stress

3.1.5 Joseph Valentin Boussinesq, Saint André de Sangonis 1842 – Paris 1929.
3.1 Panels 383

Fig. 3.1.18

(3.1.52) carried on by the remaining panel. Clearly the resultant one of all these
interactions is vertical and has intensity
 π  π
2 4F 2
2 σr cosϑ r dϑ = − cos2 ϑ dϑ = −F.
0 π 0

Evidently the obtained result is independent from the choice of the ray r, which
can be as small as we want, provided that positive. 
The (3.1.52) involve that in proximity of the point in which the concentrated force
F is applied elevated stresses are verified. Precisely in a point P of a line parallel to
the axis y and distant d from it (Fig. 3.1.19) it results

2F
σx = − cos4 ϑ,
πd
2F
σy = − sin2 ϑ cos2 ϑ,
πd
2F
τxy = − sinϑ cos3 ϑ,
πd

so that we easily verify that the stress σ x quickly goes decreasing when P goes away
from the vertical one passing through O (Fig. 3.1.20).
Remark 3.1.10 A second problem of Boussinesq is gotten when the concentrated
force F is direct according to the axis y (Fig. 3.1.21). In such problem we can verify

Fig. 3.1.19

Fig. 3.1.20
384 3 The Two-Dimensional Problems

Fig. 3.1.21

that the function of Airy is

F
Φ= r ϑ cosϑ
π
and that the state of stress is
2 F sinϑ
σr = − , σt = 0, τrt = 0. 
π r
Using the results of the problem of Boussinesq we can now determine the state
of stress in the circular panel loaded by two diametral forces. Precisely we consider
a disk of diameter d and of thickness b (Fig. 3.1.22), submitted to two diametral
distributed forces of intensity F/b (tons/meter). So we must solve the bidimensional
boundary problem of Fig. 3.1.23. Well we determine its solution employing the prin-
ciple of superposition, the solution of the problem of Boussinesq and the solution
of the circular panel of Fig. 3.1.14. In fact in the problem of Boussinesq in all the
points P of a circumference tangent to the axis y, with the center on the axis x and of
diameter d (Fig. 3.1.24), it results r = d cosϑ. It follows of it that σ r has constant
intensity and

2F
σr = − , σt = 0, τrt = 0.
πd

Fig. 3.1.22

Fig. 3.1.23
3.1 Panels 385

Fig. 3.1.24

Fig. 3.1.25

Fig. 3.1.26

This way, for the principle of dissection, the state of stress in the disk of
Fig. 3.1.25 (where on the boundary the state of stress is shown) is given by the
(3.1.52).
Let us consider now the problem of Boussinesq of Fig. 3.1.26. It is obvious that
in the point P of the circumference of Fig. 3.1.26 it results

2F
σr = − , σt = 0, τrt = 0.
πd

This way, for the principle of dissection the solution of Boussinesq for the bound-
less panel of Fig. 3.1.26 also resolves the problem of the disk of Fig. 3.1.27 (where

Fig. 3.1.27
386 3 The Two-Dimensional Problems

Fig. 3.1.28

on the boundary the state of stress is shown). Then, since the principle of super-
position, we know also the solution of the problem of the disk of Fig. 3.1.28.
The load applied on the disk of Fig. 3.1.28 (where on the boundary the state
of stress is known) is precisely that of Fig. 3.1.30: radial, uniform of compression
and of intensity 2F/(π d). In fact in the generic boundary point P of the disk of
Fig. 3.1.28 the state of stress is obviously3.1.6 hydrostatic of compression and of
intensity 2F/(πd) (Fig. 3.1.29).
For the principle of superposition, then, the problem of Fig. 3.1.23 has for
solution the sum of the solution of the problem of Fig. 3.1.30 (that is sum of
two solutions of Boussinesq) and of thesolution of the problem of Fig. 3.1.14,
in which the applied load is pe = −2F (π d). It is easy to verify that it results
(Fig. 3.1.31)

Fig. 3.1.29

Fig. 3.1.30

Fig. 3.1.29 the line passing through P and O and the line passing through P and O are
3.1.6 In
obviously orthogonal.
3.1 Panels 387

Fig. 3.1.31

2F 1 cos4 ϑ cos4 ϑ1
σx = − − ,
π d h d−h

2F 1 sin2 ϑcos2 ϑ sin2 ϑ1 cos2 ϑ1

σy = − − ,
π d h d−h

2F sinϑ cos3 ϑ sinϑ1 cos3 ϑ1

τxy = − + .
π h d−h

Particularly for the points of the axis x is gotten

2F 1 1 1 2F
σx = − − , σy = , τxy = 0,
π d h d−h πd
so that for them the σx (of compression) assume elevated values in proximity of the
points of application of the forces; the σy (of traction) result constant. In the points
of the axis y the σx reaches its maximum in the center of the disk, and it results

6F
σxmax = − .
πd

3.1.4 Effect of a Hole

The panels are often linked among them through riveting. Next to the holes that
we practice in the panel some concentrations of tension, that we must appraise, are
verified. To play this analysis we consider the rectangular panel of Fig. 3.1.32, of
thickness b. In the panel a hole having radius a and center coincident with the center
O of the rectangle (Fig. 3.1.32) is present.

Fig. 3.1.32
388 3 The Two-Dimensional Problems

Fig. 3.1.33

The load applied on the panel is uniformly

  distributed on two opposite edges
(Fig. 3.1.32) and has intensity p t m2 . We assume the frame of reference
O, x, y of Fig. 3.1.33. The presence of a hole produces an alteration of the inside
state of stress, that, since the postulate of Saint Venant, has effect only in an area
near to the hole. Insofar, choosing a radius c suitably greater than the radius of the
hole (Fig. 3.1.33), we can assume that in the points of the panel external to the circle
of center O and radius c the state of stress is the same one that the one of the panel
without hole of Fig. 3.1.4, i.e.

σx = 0, σy = p, τxy = 0. (3.1.53)

This way what remains is to determine the state of stress in the points of the
annulus of center O and radius a and c. To such purpose, we observe indeed that for
the (3.1.53) in the points of the circumference of center O and radius c it results
p p
σr = σy sin2 ϑ = (1 − cos2ϑ) , τrt = σy sinϑcosϑ = sin2ϑ.
2 2
Insofar, for the principle of dissection, the state of stress in the portion of panel
hatched in Fig. 3.1.34 coincides with the state of stress in the annulus of Fig. 3.1.35,
unloaded on the inside edge and submitted on the external edge to the distributed
forces
p p
pr = (1 − cos2ϑ) , pt = sin2ϑ.
2 2

For the principle of superposition, the solution of the problem of Fig. 3.1.35 is
sum of the solution of the problem of Fig. 3.1.36 and of the solution of the problem
of Fig. 3.1.37. The problem of Fig. 3.1.36 has polar symmetry and its solution

Fig. 3.1.34

Fig. 3.1.35
3.1 Panels 389

Fig. 3.1.36

Fig. 3.1.37

has already been determined (Fig. 3.1.15). Precisely in the problem of Fig. 3.1.36
(whose load is pi = 0, pe = −p/2) in every point P = (r, ϑ) it results

p a2 c2 1 p c2
σr = −  2 · 2+  
2 c − a2 r 2 c2 − a2
p a2 c2 1 p c2 (3.1.54)
σt =  · 2+  
2 c2 − a2 r 2 c2 − a2
τrt = 0.

About the problem of Fig. 3.1.37, we consider the function

p 2 a4 p 1 a2
Φ (r, ϑ) = r + 2
− p cos2ϑ. (3.1.55)
4 4 r 2

Since

∂Φ p a4 p 1
= r− cos2ϑ,
∂r 2 2 r3
∂Φ p 2 a4 p 1
= − r + − a2 p sin2ϑ,
∂ϑ 2 2 r4
390 3 The Two-Dimensional Problems

∂ 2Φ 1 ∂ 2Φ p 3a4 p 1
= − pr − a4 p 3 sin2ϑ, = + cos2ϑ,
∂r∂ϑ r ∂r2 2 2 r4
∂ 2Φ 1
= − pr2 + a4 p 4 − 2a2 p cos2ϑ,
∂ϑ 2 r
∂ 3Φ 1
= − 2pr − 2a4 p 3 cos2ϑ,
∂r∂ϑ 2 r
∂ 3Φ 1 ∂ 4Φ 1
= − 6a4 p 5 cos2ϑ, = − 2p + 6a4 p 4 cos2ϑ,
∂r 3 r ∂r ∂ϑ
2 2 r
∂ 4Φ 1
= 30a4 p 6 cos2ϑ,
∂r4 r
∂ 3Φ 1
= 2pr2 + 2a4 p 4 − 2a2 p sin2ϑ,
∂ϑ 3 r
∂ 4Φ 1
= 4pr2 + 4a4 p − 4a2 p cos2ϑ,
∂ϑ 4 r4

it results in A

∂ 4Φ 2 ∂ 4Φ 1 ∂ 4Φ 2 ∂ 3Φ 2 ∂ 3Φ
+ + + −
∂r4 r2 ∂r2 ∂ϑ 2 r4 ∂ϑ 4 r ∂r3 r3 ∂r∂ϑ 2
1 ∂ Φ
2 4 ∂ Φ
2 1 ∂Φ
− 2 2 + 4 + 3 =0
r ∂r r ∂ϑ 2 r ∂r
so that the (3.1.40) is satisfied.
Besides on the inside edge of A it results

nr ∂Φ nr ∂ 2 Φ nt ∂Φ nr ∂ 2 Φ
+ + 2 − = 0,
a ∂r a ∂ϑ
2 2 a ∂ϑ a ∂r∂ϑ
nr ∂Φ nr ∂ 2 Φ ∂ 2Φ
− + nt 2 = 0,
a2 ∂ϑ a ∂r∂ϑ ∂r

so that, because pr = 0, pt = 0, the (3.1.41) and (3.1.42) are satisfied.

Finally on the external edge of A it results3.1.7

nr ∂Φ nr ∂ 2 Φ nt ∂Φ nr ∂ 2 Φ p
+ + − = − cos2ϑ,
c ∂r c ∂ϑ
2 2 c ∂ϑ
2 c ∂r∂ϑ 2
nr ∂Φ nr ∂ 2 Φ ∂ 2Φ p
− + nt 2 = sin2ϑ,
c2 ∂ϑ c ∂r∂ϑ ∂r 2
p p
so that, because pr = − cos2ϑ, pt = sin2ϑ, the (3.1.41) and (3.1.42) are
2 2
satisfied.

3.1.7 By hypothesis a/c ∼

= 0.
3.1 Panels 391

Then the problem of Fig. 3.1.37 admits the (3.1.55) as function of Airy. Since the
(3.1.43), its state of stress is

p a4 a2
σr = − 1+3 4 −4 2 cos2ϑ
2 r r
p a4
σt = 1+3 4 cos2ϑ
2 r
p a4 a2
τrt = 1−3 4 +2 2 sin2ϑ.
2 r r

The state of stress (3.1.54) of the problem of Fig. 3.1.36 can be simplified taking
into account the hypothesis ac ∼= 0. We obtain

p a2 p a2
σr = 1− 2 , σt = 1+ 2 , τrt = 0.
2 r 2 r

As a consequence, the state of stress in the problem of Fig. 3.1.35 (and then in
the portion of panel hatched in Fig. 3.1.34) is

p a2 p a4 a2
σr = 1− 2 − 1+3 4 −4 2 cos2ϑ
2 r 2 r r

p a2 p a4
σt = 1+ 2 + 1+3 4 cos2ϑ
2 r 2 r
p a4 a2
τrt = − 1−3 4 +2 2 sin2ϑ.
2 r r

At the edge of the hole (where r = a) we get

σr = 0, σt = p (1 + 2 cos2ϑ) , τrt = 0,

so that the maximum of σ t is reached at π = 0 (i.e. in the points H, H of

Fig. 3.1.38). It results

σtmax = 3p.

Then the effect of the hole is to triple the stress in the material (Fig. 3.1.38).

Fig. 3.1.38
392 3 The Two-Dimensional Problems

Fig. 3.1.39

Remark 3.1.11 The solution of the problem of the panel with hole of elliptic form,
with the smaller semidiameter 2a parallel to the load of traction (Fig. 3.1.39) is also
known. The maximum value of the σ y is in the points H and H and is
 a
σymax = p 1 + 2 .
c

It grows dangerously to the heigth of a/c. This explains the propagation (in the
fragile materials) of lesions (which are assimilable to crushed ellipsis) normal to the
direction of the stress. To avoid such phenomenon we use some circular holes to the
extremities of the lesion. Such way the σymax is limited to 3 p. 

3.2 Plates

3.2.1 Small Deflections of Thin Plates

The plate of great technical interest is the thin plate with small deflection. Let us
consider a plate having constant thickness s. We suppose that the area A occupied
by the middle plane of the plate is monoconnected and of any shape (Fig. 3.2.1). We
suppose the plate thin, i.e. we suppose that the thickness of the plate is very small
compared with its other dimensions. We assume an orthogonal cartesian frame of
reference O, x, y, z having the axes x, y in the middle plane of the plate (Fig. 3.2.1).
We denote with u, v, w the components according the axes x, y, z of the displacement
of a point of the plate. We will say that the plate is in the field of the small deflections
if w is anywhere negligible with respect to s and the first order partial derivatives of
w are anywhere in modulus near to zero.

Fig. 3.2.1
3.2 Plates 393

For the thin plate in the field of small deflection Kirchhoff proposed an approxi-
mate but mathematically correct model, that has always furnished practically exact
values. The hypotheses of the Kirchhoff’s theory are:

1. the plate is thin,

2. the load applied on the plate is a distributed load q normal to the plate,
3. the load q is such that the plate deform itself remaining in the field of the small
deflections,
4. the displacement of any point of the middle plane is parallel to z (Fig. 3.2.2),
5. the points of the plate lying initially on a normal to the middle plane of the plate
remain on the normal to the middle surface of the plate after bending (Fig. 3.2.2),
6. in every point of the plate the stress σ z is in modulus very near to zero.

We will see that, using these assumptions, all stress components can be expressed
by the function w (x, y), which we call elastic surface of the plate. We also will
see that in these hypotheses the mathematical model that simulates the behavior of
the plate is a boundary problem in the unknown w, constituted by a linear partial
differential equation and by linear boundary conditions. This boundary problem
has unique solution and is analytically linear. Thus its unique solution gives all
necessary information for calculating stresses at any point of the plate.
Remark 3.2.1 Since the hypothesis 4 the middle plane of the plate is not deformed
and the reactive forces at the edges are normal to the plate. 

Remark 3.2.2 The hypothesis 5 is equivalent to the disregard of the effect of shear
forces on the deflection of plates. Such approximation has always been proved
acceptable in the previous hypotheses when A is monoconnected. 

From the hypothesis 5 it follows that for every point (x, y, z) of the plate it results
(Fig. 3.2.2)

∂w ∂w
u (x, y, z) = −z , v (x, y, z) = −z
∂x ∂y
and then, since the (1.1.28)

∂ 2w ∂ 2w ∂ 2w
εx = −z , εy = −z , γxy = −2 z . (3.2.1)
∂x2 ∂y2 ∂x∂y
Remark 3.2.3 The (3.2.1) points out, correctly, that if the concavity is downward
then the inferior fibers shorten. 

Fig. 3.2.2
394 3 The Two-Dimensional Problems

Fig. 3.2.3

From the (1.4.5) and the hypothesis 6 it follows that

2Gεz + λ (εz + εz + εz ) = 0

from which, taking into account the (1.4.5), (1.4.2), (1.4.4), and (3.2.1), we get
(Fig. 3.2.3)

Ez ∂ 2w ∂ 2w
σx = − + ν (3.2.2)
1 − ν2 ∂x2 ∂y2
Ez ∂ w
2 ∂ 2w
σy = − + ν (3.2.3)
1 − v2 ∂y2 ∂x2
Ez ∂ w2
τxy =− . (3.2.4)
1 + ν ∂x∂y

As a consequence on the area s dy [resp. s dx] of Fig. 3.2.4 are then applied:
– the bending moment (positive if stretchs the inferior fibers)
 s/2  s/2
M x = (σx dydz) (z) = dy z σx dz
−s/2 −s/2
 s/2  s/2 !
 
resp. M y = σy dxdz (z) = dx z σy dz ,
−s/2 −s/2

– the twisting moment (positive if anticlockwise)

 
s/2   s/2
M xy =− τxy dydz (z) = −dy z τxy dz
−s/2 −s/2
  !
s/2   s/2
resp. M yx = τyx dxdz (z) = dx z τyx dz ,
−s/2 −s/2

– the shear (positive if downwards)

 s/2  s/2
Qx = τxz dydz = dy τxz dz
−s/2 −s/2
  !
s/2 s/2
resp. Qy = τyz dxdz = dx τyz dz .
−s/2 −s/2
3.2 Plates 395

Fig. 3.2.4

So on the element of plate of Fig. 3.2.4, putting3.2.1

Es3
D=   (3.2.5)
12 1 − ν 2

and taking into account the (3.2.2), (3.2.3), and (3.2.4), on the face parallel to y, z
there is:

– an unitary bending moment Mx having axis parallel to y, verse such that the
inferior fibers are stretched and intensity (t)

M x ∂ 2w ∂ 2w
Mx = = −D + ν , (3.2.6)
dy ∂x2 ∂y2

– an unitary twisting moment Mxy having axis parallel to x, verse anticlockwise (if
positive) and intensity (t)

M xy ∂ 2w
Mxy = = D(1 − v) , (3.2.7)
dy ∂x∂y

– an unitary shear Qx having axis parallel to z, verse downward (if positive) and
intensity (t/m)

Qx
Qx = . (3.2.8)
dy

In the same way, on the face parallel to x, z there is:

– an unitary bending moment My having axis parallel to x, verse such that the
inferior fibers are stretched and intensity (t)

M y ∂ 2w ∂ 2w
My = = −D + ν , (3.2.9)
dx ∂y2 ∂x2

– an unitary twisting moment Myx having axis parallel to y, verse anticlockwise (if
positive) and intensity (t)

3.2.1 We call D flexural rigidity of the plate.

396 3 The Two-Dimensional Problems

M yx ∂ 2w
Myx = = −D(1 − ν) , (3.2.10)
dx ∂x∂y

– an unitary shear Qy having axis parallel to z, verse downward (if positive) and
intensity (t/m)

Qy
Qy = . (3.2.11)
dx

Let us observe now that from the (3.2.6), (3.2.9), and (3.2.7) it banally follows

∂ 2w Mx − ν My
=−  , (3.2.12)
∂x2 D 1 − ν2

∂ 2w My − ν Mx
=−  , (3.2.13)
∂y 2 D 1 − ν2

∂ 2w Mxy
= . (3.2.14)
∂x∂y D (1 − ν)

Such relations allow to link the deformation of the plate to its bending and
twisting moments. In fact we know from the Differential geometry that in every
point P = (x, y, w(x, y)) of the elastic surface (that is a regular surface of 3 ) the
curve obtained intersecting the elastic surface with the plane parallel to plane x, z
[resp. y, z] passing through p is a regular curve of 3 and has curvature (that we call
curvature of the surface according x [resp. y]) exactly equal to

∂ 2w
1 ∂x2
= 
rx 2 3/2
∂w
1+
∂x
⎡ ⎤
⎢ ∂ 2w ⎥
⎢ ⎥
⎢ 1 ∂y2 ⎥
⎢resp. = 3/2 ⎥ .
⎢ ry 2 ⎥
⎣ ∂w ⎦
1+
∂y

Then, since the hypothesis that the plate is in the field of the small deflections,
from the (3.2.12) and (3.2.13) we get the approximate expressions

1 ∂ 2w Mx − νMy
= 2 =−  , (3.2.15)
rx ∂x D 1 − ν2
3.2 Plates 397

1 ∂ 2w My − vMx
= 2 =−  . (3.2.16)
ry ∂y D 1 − ν2

Analogously we get for the torsion of the surface according x and y the
expression

1 ∂ 2w Mxy
= = . (3.2.17)
rxy ∂x∂y D (1 − ν)

After these preliminaries we can establish the differential equation of the math-
ematical model of the plate. To this aim we consider any point P = (x, y) of A. We
isolate from the plate an infinitesimal element by cutting the plate with planes par-
allel to x, z and y, z and passing through P and (x + dx, y + dy), where dx, dy are
positive numbers as small as we want (Fig. 3.2.5). Then we apply on this element
the load q (downward if positive) and the previous bending and twisting moments
(Fig. 3.2.6).
Imposing the equilibrium of the element of Fig. 3.2.5 to the translation according
z, we get

∂Qx ∂Qy
dx dy + dx dy + q dx dy = 0
∂x ∂y
from which

∂Qx ∂Qy
+ + q = 0. (3.2.18)
∂x ∂y

Fig. 3.2.5

Fig. 3.2.6
398 3 The Two-Dimensional Problems

Imposing the equilibrium of the element of Fig. 3.2.5 to the rotation around the
axis parallel to x and passing through P, we get

∂My ∂Mxy
Qy dx dy − dy dx + dx dy = 0
∂y ∂x
from which

∂My ∂Mxy
Qy = − (3.2.19)
∂y ∂x

from which, taking into account the (3.2.9) and (3.2.7)

∂ 3w ∂ 3w
Qy = −D + . (3.2.20)
∂y3 ∂x2 ∂y

Imposing the equilibrium of the element of Fig. 3.2.5 to the rotation around the
axis parallel to y and passing through P, we get

∂Mx ∂Myx
− Qx dy dx + dx dy + dy dx = 0
∂x ∂y
from which
∂Mx ∂Myx
Qx = − (3.2.21)
∂x ∂y

from which, taking into account the (3.2.6) and (3.2.10)

∂ 3w ∂ 3w
Qx = −D + . (3.2.22)
∂x 3 ∂x∂y2

So from the (3.2.18), (3.2.20), and (3.2.22) we immediately have

∂ 4w ∂ 4w ∂ 4w q
+2 2 2 + 4 = on A. (3.2.23)
∂x 4 ∂x ∂y ∂x D

To formulate the boundary conditions, we first of all establish the expressions of

the bending and twisting moments that act in a plane parallel to z. So we consider a
point P of A and a plane α n parallel to z and passing through P. Then we denote with
n the normal to α n (Fig. 3.2.7), with t the intersection between α n and the middle
plane x, y of the plate, with nx , ny [resp. tx , ty ] the direction cosines of n [resp. t].3.2.2
We isolate from the plate the infinitesimal element of Fig. 3.2.7, obtained by cutting
with the plane α n and planes parallel to x, z and y, z.

3.2.2 Obviously tx = –ny , ty = nx .

3.2 Plates 399

Fig. 3.2.7

We denote, as usual, with Mn the unitary bending moment (positive if stretches

the inferior fibers), with Mnt the unitary twisting moment (positive if anticlock-
wise) and with Qn the unitary shear (positive if downward) acting on the face α n
(Fig. 3.2.8). Employing the (1.2.8) and (1.2.9) we easily have
 s/2  s/2  
Mn = (σn dz)(z) = σx z n2x + σy z n2y + 2 τxy z nx ny dz
−s/2 −s/2
 s/2
Mnt = (τnt dz)(z)
−s/2
 s/2
  
= σx z nx tx + σy z ny txy + τxy z nx ty + ny tx dz
−s/2

and then, since the (3.2.2), (3.2.3), (3.2.4)

  ∂ 2w   2 
∂ 2w 2 + ν n2 ∂ w
Mn = −D n2x + ν n2y − 2(1 − ν) + ny x (3.2.24).
∂x2 ∂x∂y ∂y2

∂ w
2
Mnt = −D (ν − 1)nx ny
∂w2
  2  (3.2.25)
2 ∂ w ∂ 2w
+(ν − 1) −nx + ny2
+ (1 − ν)nx ny 2 .
∂x∂y ∂y

Furthermore, about the shear Qn , imposing the equilibrium to the translation

according z of the element of Fig. 3.2.7, we easily get

Qn = Qx nx + Qy ny

from which, since the (3.2.20), (3.2.22)

 
∂ 3w ∂ 3w ∂ 3w ∂ 3w
Qn = −D nx 3 + nx + ny + ny . (3.2.26)
∂x ∂x∂y2 ∂x2 ∂y ∂y3

Fig. 3.2.8
400 3 The Two-Dimensional Problems

The shearing stresses τnz can now be determined by assuming that they are dis-
tributed across the thickness of the plate according to a parabolic law (in the way
seen in problem 2.6.1).
Let us now consider the boundary ∂A and its normal n, that we suppose directed
going out the open A and of direction cosines nx , ny . Well the boundary conditions
on each part of the edge ∂A are:

– part built-in (deflection = 0, rotation = 0)

w = 0,
dw (3.2.27)
= 0;
dn
– part simply supported (deflection = 0, Mn = 0)

w = 0,
  ∂ 2w ∂ 2w   ∂ 2w (3.2.28)
n2x + ν n2y − 2(1 − ν) + n2y + ν n2x = 0;
∂x2 ∂x∂y ∂y2

dMnt
– part free (Mn = 0, Qn = Qn − = 03.2.3 )
dt
  ∂ 2w ∂ 2w   2
n2x + ν n2y − 2(1 − ν) + n2 + ν n2 ∂ w = 0
y x
∂x2 ∂x∂y ∂y2
  ∂ 3w   ∂ 3w
nx 1 + (ν − 1)n2y + ny 1 + (ν − 1)( − 2n 2 + n2 )
x y
∂x3 ∂x2 ∂y
  ∂ 3w   3
+nx 1 + (ν − 1)(n2x − 2n2y ) + n 1 + (ν − 1)n2 ∂ w = 0.
y x
∂x∂y2 ∂y3
(3.2.29)

So we can assume as a mathematical model of the thin plate with small

deformations the boundary problem:

[3.2.1] In the hypothesis in which the known term q and the open one A are regular,
find a real function w defined in A, equipped in A with partial derivatives at least
up to those of the fourth order and such that satisfies in A the differential equation
(3.2.23) and in ∂A the boundary conditions (3.2.27), (3.2.28), and (3.2.29). 

3.2.3 The condition Qn = 0 was formulated by Kirchhoff taking too into account the twisting
dMnt
moment in a model mathematically well posed. He observed that, since dMnt = dt, the
dt
distribution of twisting moments along the edge of the plate is statically equivalent to a distribution
dMnt
of shearing forces of the intensity . He also observed that, as a consequence of this approxi-
dt
mation, if the edge has a corner then in the corner the mathematical model furnishes a concentrated
reaction.
3.2 Plates 401

For the problem [3.2.1] we can prove the following existence and uniqueness
theorem3.2.4 :
[3.2.2] The problem [3.2.1] admits an unique solution. 
As a rule, in any case the unique solution w of problem [3.2.1] is obtained by
using the Finite element method of the Numerical analysis. In turn, from w, employ-
ing the (3.2.6), (3.2.7), (3.2.9), (3.2.20), and (3.2.22), we determine in all the points
of the plate the bending and twisting moments and the shears. Equally, from w,
employing the (3.2.2), (3.2.3), and (3.2.4), we determine in all the points of the
plate the state of stress.
Evidently the problem [3.2.1] is mathematically linear. So
[3.2.3] For the problem [3.2.1] the principle of superposition is held. 
The problem [3.2.1] admits an equivalent formulation in energetic terms. Let us
consider the infinitesimal element of plate of Fig. 3.2.4. Clearly the strain energy
stored in this element can be obtained by calculating the work done by the moments
Mx dy, Mxy dy, My dx, Myx dx on the element during deformation of the plate.
Since the sides of the element remain plane, the work done by the moments Mx
dy [resp. My dx] is obtained by taking half of the product of the moment and the
angle between the corresponding sides of the element after bending:
 
1 ∂ 2w 1  ∂ 2w
(Mx dy) − 2 dx resp. My dx − 2 dy .
2 ∂x 2 ∂y

Analogously, the work done by the moments Mxy dy [resp. Myx dx] is
 
1  ∂ 2w 1  ∂ 2w
Mxy dy dx resp. Myx dx dy .
2 ∂x∂y 2 ∂x∂y

Obviously, the strain energy E of the entire plate is obtained by integrating on A.

Since the (3.2.6), (3.2.7), (3.2.9), and (3.2.10), we easily obtain
 # 2 2
!"
1 ∂ 2w ∂ 2w ∂ 2w ∂ 2w ∂ 2w
E= D + 2 − 2(1 − ν) − dx dy.
A 2 ∂x2 ∂y ∂x2 ∂y2 ∂x∂y
(3.2.30)

Similarly to the three-dimensional case, the problem of the plate admits global
formulations by employing the variational methods. The more powerful formulation
of this type is the minimum potential energy problem. Let us suppose that the open
set A has a Lipschitz-continuous boundary ∂A. We consider the Sobolev space H 2 (A)
and its subset W constituted by the functions w satisfying the boundary conditions
(3.2.27), (3.2.28), and (3.2.29). We call potential energy functional the functional
Jp that to every w ∈ W associate the real number

3.2.4 In the proof we must use the condition 0 < ν < 1/2 .
402 3 The Two-Dimensional Problems

 # 2
1 ∂ 2w ∂ 2w
Jp = D + 2
A 2 ∂x2 ∂y
2
!" 
∂ 2w ∂ 2w ∂ 2w
− 2(1 − ν) − dx dy − q w dx dy.
∂x2 ∂y2 ∂x∂y A

Let us consider now the minimum potential energy problem:

[3.2.4] Find w ∈ W such that Jp (w) ≤ Jp (w̃) ∀w̃ ∈ W. 
In the Mathematical theory of elasticity we prove that the problem of Kirchhoff
[3.2.1] and the weak problem [3.2.4] have the same solution:
[3.2.5] The problems [3.2.1] and [3.2.4] are equivalent. 
Remark 3.2.4 In the problem of the thin plate with small deformations calculations
show that the forces lying in the middle plane caused by edges immovable in the
plane of the plate can be practically disregarded. 

3.2.2 Thin Plates on Elastic Foundation

In the Civil engineering the thin plate is often used as structure of foundation. In
such case the thin plate usually lies directly on the subgrade and still works in the
field of small deflections. The simplest but effective law that we can assume to
simulate the intensity of the reaction R of the subgrade is due to Winkler.3.2.5 He
proposed the expression

R = kw (3.2.31)

where the positive constant k is called modulus of the foundation and depends
largely on the properties of the subgrade.
Clearly the behavior of such plate is simulated by a problem obtained modifying
the differential equation of the problem [3.2.1]. Precisely the modified equation is:

∂ 4w ∂ 4w ∂ 4w q − kw
+2 2 2 + 4 = on A (3.2.32)
∂x 4 ∂x ∂y ∂y D
and the mathematical model of the thin foundation plate with small deflection is the
boundary problem:
[3.2.6] In the hypothesis in which the known term q and the open one A are regular,
find a real function w defined in A, equipped in A with partial derivatives at least
up to those of the fourth order and such that satisfies in A the differential equation
(3.2.32) and in ∂A the boundary conditions (3.2.27), (3.2.28), and (3.2.29). 

3.2.5 Emil Winkler, Falkenberg 1835 – Berlin 1888.

3.3 Shells 403

For the problem [3.2.6] we can prove the following existence and uniqueness
theorem:
[3.2.7] The problem [3.2.6] admits an unique solution. 
Evidently the problem [3.2.6] is mathematically linear. So
[3.2.8] For the problem [3.2.6] the principle of superposition is held. 
As a rule, in any case the unique solution of problem [3.2.6] is obtained by using
the Finite element method of the Numerical analysis.

3.3 Shells

3.3.1 Membranes
We call membrane a body constituted by a very thin layer of material situated round
a regular surface S of 3 . It evidently concerns a two-dimensional problem. The
soap bubble is a classic example of membrane. The thickness of the membrane is
denoted by h and is always considered very small in comparison with the other
dimensions of the membrane and with its radii of curvature. The surface S bisects the
thickness of the membrane and is called middle surface. By specifying analytically
the surface S and the thickness of the membrane at each point of S, a membrane is
entirely defined geometrically.
In practice the membranes are shells so thin that the bending and twisting
moments are negligible. A case in which such hypothesis is perfectly satisfied is
the pneumostructure. In fact this is a structure, that can also have great dimensions,
constituted by gasproof fabrics. Such material is able to absorb tensile stress but can
not absorb neither bending nor twisting moments, that are rigorously equal to zero.
In the mathematical model of the static behavior of the membrane we suppose
that S is regular in the sense of the Differential geometry and that, because of its
thinness, in every point P of S there are only stresses σ1 , σ2 , τ12 parallel to the plane
tangent at P to S and distributed uniformly over the thickness h of the membrane. So
in a membrane the problem of stress analysis is greatly simplified, since the bending
and twisting moments and the shearing forces vanish. We notice that the membrane
is the optimal way to absorb the external load. In fact, since the stress is constant
along the thickness, such structure utilize the material in the best possible way.
In a membrane the only unknowns are the three quantities N1 = h σ1 , N2 = h σ2 ,
N12 = h τ12 . Those internal forces are determined solving a two-dimensional bound-
ary problem obtained formulating the mathematical model of the membrane. Such
problem is constituted by a system of differential equations that simulates the equi-
librium of an element of the membrane and by boundary conditions that simulate
the constraints. Obviously the constraints must act on the membrane reactions tan-
gent to the surface and with moment with respect to any tangent to boundary equal
to zero.
404 3 The Two-Dimensional Problems

In convenient hypotheses of regularity of the data such problem always admits a

solution. So any membrane can always absorb a load only with the forces N1 , N2 ,
N12 . As a matter of fact, that is easily understandable in the case of axial-symmetric
membrane by the elementary Statics. In reality in such case the strips along the par-
allel circles are the funicular of any constant radial pressure and then they can absorb
it. The value of such pressure is perfectly individualized. In fact the strips along the
parallel circles must act on the strip along a meridian the pressure that makes the
meridian the funicular of the system constituted by the external load and the pres-
sure. In particular, as a consequence, unlike what happens for the arcs in masonry
and for the vaults in masonry, it is not necessary to suit the form for the load.
Remark 3.3.1 A membrane is a very rigid structure. This is confirmed by all the
known results. We can also observe that in a membrane the deformations are
essentially due to normal stress and then are smaller than the deformations due to
bending. 

The problem to determine the state of stress in a membrane is particularly simple

in the axial-symmetric case. In such case S is a surface of revolution and the load
is symmetric with respect to its axis. A surface of revolution is obtained by rotation
of a plane curve about an axis r lying in the plane of the curve. This curve is called
the meridian, its plane is called the meridian plane and r is called the axis of the
surface of revolution. The intersection between the surface of revolution and a plane
normal to r is called parallel circle. The position of a meridian passing through a
point P of S individualizes an angle ϑ measured from some datum meridian plane.
The position of a parallel circle passing through a point P individualizes an angle
ϕ made by the normal to the surface and the axis of rotation. So a point P of S
individualizes the angles ϑ, ϕ (Fig. 3.3.1).
Let us consider a point P of S and denote with y the tangent at P to the meridian
passing through P (Fig. 3.3.1). Well the meridian plane α passing through P and the
plane β passing through P and normal to y are the planes of principal curvature of
S at P. The curve intersection between S and α is regular and has some radius of
curvature r1 (Fig. 3.3.2). The curve intersection between S and β is regular and has
radius of curvature r2 (Fig. 3.3.2). Obviously, if we denote with r0 the radius of the
parallel circle passing through P, we have r0 = r2 sin ϕ.
Membranes that have the form of surfaces of revolution and are loaded symmet-
rically with respect to their axis can be studied in a particularly simple way, since by

Fig. 3.3.1
3.3 Shells 405

Fig. 3.3.2

symmetry any quantity is only function of the variable ϕ. In such study we consider
a point P of S and cut out an element of the membrane by the two adjacent meridians
ϑ, ϑ + dϑ and the two parallel circles ϕ, ϕ + dϕ (Fig. 3.3.1), where dϑ, dϕ denote
positive real numbers as small as we want. Evidently the surface area of the relative
element of S is then r1 r2 sin ϕ dϕ dϑ.
From the assumed symmetry of loading and geometric it can be concluded that
there will be no shearing forces acting on the sides of the element. Then we must
consider only the normal forces per unit length Nϕ , Nϑ (Fig. 3.3.1). The intensity
of the external load, which acts in the meridian plane, in the case of symmetry is
resolved in two components Y and Z parallel to the coordinate axes.
In writing the equations of equilibrium of the element, let us begin with the forces
in the direction of the tangent to the meridian. On the upper side of the element the
force

Nϕ r0 dϑ = Nϕ r2 sinϕ dϑ

is acting. The corresponding force on the lower side of the element is

dNϕ dr0
Nϕ + dϕ r0 + dϕ dϑ.
dϕ dϕ

So, by neglecting a small quantity of second order, the forces acting on the upper
and lower side of the element have a resultant in the y direction equal to

d  
Nϕ r0 dϕ dϑ. (3.3.1)
dϕ
The component of the external force in the same direction is

Y r1 r0 dϕ dϑ. (3.3.2)

The forces acting on the lateral sides of the element are equal to Nϑ r1 dϕ and have
a resultant in the direction of the radius of the parallel circle equal to Nϑ r1 dϕ dϑ.
The component of this force in the y direction is (Fig. 3.3.2)

− Nϑ r1 cosϕ dϕ dϑ. (3.3.3)

Summing up the forces (3.3.1), (3.3.2), and (3.3.3) we obtain the equation of
equilibrium in the direction of the tangent to the meridian
406 3 The Two-Dimensional Problems

d  
Nϕ r0 − Nϑ r1 cosϕ + Y r1 r0 = 0. (3.3.4)
dϕ

The second equation of equilibrium is obtained by summing up the projections

of the forces in the z direction. The forces acting on the upper and lower sides of the
element have a resultant in the z direction equal to

Nϕ r0 dϕ dϑ.

The forces acting on the lateral sides of the element and having the resultant
Nϑ r1 dϕ dϑ in the radial direction of the parallel circle give a component in the z
direction of magnitude

Nϑ r1 sinϕ dϕ dϑ.

The external load acting on the element has in the same direction a component

Z r1 r0 dϕ dϑ.

Summing up such forces, we obtain the second equation of equilibrium

Nϕ r0 + Nϑ r1 sinϕ + Z r1 r0 = 0. (3.3.5)

Since the symmetry of the deformation, a small displacement of a point can be

resolved into a component v in the direction of the tangent to the meridian and into
a component w in the direction of the normal to the middle surface. Considering an
element AB of the meridian (Fig. 3.3.3), we see that the increase of the length of
the element due to tangential displacements v and v + dϕdv
dϕ of its ends is equal to
dv
dϕ dϕ. Because of the radial displacements w of the points A and B the length of the
element decreases by an amount w dϕ. We observe that the change in the length of
the element due to the difference in the radial displacements of the points A and B
can be neglected as a small quantity of higher order. Thus the total change in length
of the element AB due to deformation is dϕ dv
dϕ − w dϕ. Dividing this value by
the initial length r1 dϕ of the element, we find the strain of the membrane in the
direction of the tangent to the meridian to be

Fig. 3.3.3
3.3 Shells 407

1 dv w
εϕ = − . (3.3.6)
r1 dϕ r1

Considering an element of a parallel circle it may be seen (Fig. 3.3.3) that, owing
to displacements v and w, the radius r0 of the circle increases by the amount v cosϕ−
w sinϕ. The circumference of the parallel circle increases in the same proportion as
its radius; hence

v 1 w
εϑ = − . (3.3.7)
r2 tgϕ r2
By using the Navier’s law we get

1  
εϕ = Nϕ − νNϑ (3.3.8)
Eh
1  
εϑ = Nϑ − νNϕ . (3.3.9)
Eh
Substituting the (3.3.6) and (3.3.7) in the Eqs. (3.3.8) and (3.3.9), we have
 
Eh 1 dv ν v
Nϕ = −w + −w . (3.3.10)
1 − ν2 r1 dϕ r2 tgϕ
 
Eh 1 v ν dv
Nϑ = −w + −w . (3.3.11)
1 − ν2 r2 tg ϕ r1 dϕ

Substituting the (3.3.10) and (3.3.11) in the Eqs. (3.3.4) and (3.3.5), we have
 
E d r2 h sinϕ dv v
− w + ν h sinϕ −w
1 − ν dϕ
2 r1 dϕ tgϕ
 
E h r1 cosϕ 1 v ν dv (3.3.12)
− −w + − w + Y r1 r2 sinϕ = 0
1 − ν2 r2 tgϕ r1 dϕ
 
Eh r2 sinϕ dv v
− w + ν sinϕ −w
1 − ν2 r1 dϕ tgϕ
  (3.3.13)
E h r1 sinϕ 1 v ν dv
+ −w + − w + Z r1 r2 sinϕ = 0.
1 − ν2 r2 tgϕ r1 dϕ

In conclusion we reached the following mathematical model of the axial-

symmetric membrane:
[3.3.1] Given the open A ⊆ ]0, π [, the geometry h, r1 , r2 and the load Y, Z of the
membrane, find two real functions v, w differentiable in A at least up to the second
order and such to satisfy in A the (3.3.12) and (3.3.13) and the initial conditions
express in tems of v, w and their derivatives. 
408 3 The Two-Dimensional Problems

The solution of problem [3.3.1] furnishes the state of stress of the membrane by
the (3.3.10) and (3.3.11).
Problem 3.3.1 You shall determine the stresses in a thin spherical membrane having
constant thickness h and radius R. The membrane is full of gas at pressure p.
Solution This is an axial-symmetric problem. Its extreme simplicity allows us
to solve it in a very simple way, without resorting to the problem [3.3.1]. We
assume an axis z passing through the center of the sphere as rotation axis. Then
we consider a parallel circle individualized by the coordinate ϕ. We consider the
equilibrium to the translation according z of the segment of the sphere above
parallel circle. Obviously the pressure p acting on the segment of the sphere
has a resultant according z given by p π r2 , where r = R sinϕ. So the
membrane force Nϕ acting on the edge of the segment of the sphere must be
such that

Nϕ 2 π r sinϕ = p π r2 .

Insofar Nϕ = p R/2. Assuming an axis y orthogonal to z as rotation axis of the

shell, with the same considerations we obviously get Nϑ = p R/2 = Nϕ . 
A particularly simple axial-symmetric problem is the circular cylindrical mem-
brane (Fig. 3.3.4) under axial-symmetric load. We assume that the generator of
the membrane is horizontal and parallel to the x axis. An element is cut from the
membrane by two adjacent generators and two cross sections perpendicular to the
x axis. Its position is defined by the coordinate x and the angle ϕ. On the sides of
the element internal forces act (Figs. 3.3.5 and 3.3.6). In addition a load of com-
ponents X, Y, Z will be distributed over the surface of the element. Considering
the equilibrium of the element and summing up the forces in the x direction,
we obtain
∂Nx ∂Nϕx
r dϕ dx + dϕ dx + X dϕ dx = 0.
∂x ∂ϕ

Fig. 3.3.4

Fig. 3.3.5
3.3 Shells 409

Fig. 3.3.6

Similarly, the forces in the y direction give a corresponding equation of

equilibrium
∂Nxϕ ∂Nϕ
r dϕ dx + dϕ dx + Y r dϕ dx = 0.
∂x ∂ϕ

The forces acting in the z direction normal to the shell give the equation

Nϕ dϕ dx + Z dϕ dx = 0.

After simplification, the three equations of equilibrium can be represented in the

following form:

∂Nx 1 ∂Nxϕ
+ = −X
∂x r ∂ϕ
∂Nxϕ 1 ∂Nϕ
+ = −Y
∂x r ∂ϕ
Nϕ = − Zr.

So the third equation immediately furnishes the value of Nϕ . Substituting this

value in the second of the equations, we then obtain Nxϕ by integration. Using the
value of Nxϕ thus obtained we find Nx by integrating the first equation.

Remark 3.3.2 Let us notice that a mathematical analogy between the torsion problem
and the behavior of a stretched elastic membrane subjected to a uniform excess of
pressure on one side exists. Let a very thin homogeneous membrane, such as a soap
film, be stretched under an uniform tension S per unit length over an opening made in
a rigid plate (Fig. 3.3.7). The opening in the plate is assumed to have the same shape
as the cross section of the beam subjected to torsion, and the membrane is supposed
to be fixed at the edge of the opening (Fig. 3.3.7). If q is the pressure per unit area of
the membrane, and if the membrane is in equilibrium, then the force q dx dy, acting
on an element of area dx dy, must be balanced by the resultant of the vertical compo-
nents of the tensile stresses acting on the boundary of the element of area. Obviously
the resultant of the vertical components of the tensile forces acting on the edge
dy [resp. dx] is
410 3 The Two-Dimensional Problems

Fig. 3.3.7

 
∂ 2w ∂ 2w
S dx dy resp. S 2 dx dy
∂x2 ∂y

where we assumed that the deflection w is small. Hence the equation of equilibrium
of the element is

∂ 2w ∂ 2w
q dx dy + S dx dy + S dx dy = 0
∂x2 ∂y2
from which

∂ 2w ∂ 2w q
+ 2 =− .
∂x2 ∂y S

At the boundary the deflection w of the membrane is zero.

Prandtl observed that this problem coincides with the torsion problem [2.5.1]. In
fact, reasoning as in Remark 2.5.3, we can consider the Dirichlet problem (in the
unknown function χ ) equivalent to the Neumann problem [2.5.1]. Then, putting

q q  2 
ψ= χ− x + y2
S 2S
we see that the torsion problem is identical to the previous membrane problem.
Furthermore, if the deflection surface of the membrane is represented by contour
lines (in which the deflection is constant), we may easily conclude that the shearing
stress at a point B in the twisted bar has the direction of the tangent to the contour
line passing through this point (Fig. 3.3.8). 

Fig. 3.3.8
3.3 Shells 411

3.3.2 Thin Shells

We call thin shell a body constituted by a thin layer of material placed around
a regular surface S of 3 . Evidently it still concerns a two-dimensional problem.
However such problem is more complex than problem of the membrane since, dif-
ferently from the membrane (which in reality is a very thin shell), the thin shell can
absorb bending and twisting moments and as well as the shearing forces. This hap-
pens because the thickness of a thin shell is small but big enough to allow to shell
to develop not negligible values of bending and twisting moments and shearing
forces.
We have seen that in a plate under lateral load the equilibrium of an element
is only possible by action of bending and twisting moments, and often of shear-
ing forces. Well the main suppositions of the theory of thin plates also form the
basis for the usual theory of thin shells. There is, however, a substantial difference
in the behavior of plates and shells under the action of external loading. The static
equilibrium of a plate element under a lateral load is only possible by action of
bending and twisting moments, usually accompanied by shearing forces. Instead
a shell, in general, is able to transmit the surface load in the way of the mem-
brane, i.e. by stresses which act parallel to the tangential plane at a given point
of the middle surface and are distributed uniformly over the thickness of the shell.
This property of shells makes them, as a rule, a much more rigid and a more eco-
nomical structure than a plate would be under the same conditions. In principle,
the membrane forces are independent of bending and are wholly defined by the
conditions of static equilibrium. However, the reactive forces and the deformation
obtained by the use of the membrane theory usually are incompatible with the effec-
tive shell’s boundary conditions. To remove this discrepancy the bending of the
shell in the edge zone has to be considered. This bending, however, usually has a
very localized character at the shell’s boundary and may be calculated on the basis
of the same assumptions which were used in the case of small deflections of thin
plates.
If the shell is thick, that is if has a thickness comparable to its radii of curvature,
then it should be analyzed by the three-dimensional elastic theory (Chap. 1).
The thin shells are largely employed in the Mechanical engineering. In the Civil
engineering the main applications are big structures as tanks or domes,3.3.1 loaded
by gas, liquids, snow or loose materials.
We denote the thickness of the shell by h, this quantity always being considered
small in comparison with the other dimensions of the shell and with its radii of cur-
vature. The surface S bisects the thickness of the shell and is called middle surface.
We suppose that S is regular in the sense of the Differential geometry. By specifying
analytically the surface S and the thickness of the shell at each point of S, a shell is

3.3.1 With domes in reinforced concrete that have alone 5 cm of thickness it is possible and easy to
cover areas of 50 m of diameter.
412 3 The Two-Dimensional Problems

entirely defined geometrically. In mathematically modeling the shell, just as made

for the thin plates, we assume that linear elements which are normal to the middle
surface of the shell, remain straight and become normal to the deformed middle
surface of the shell.
As for the membranes, the problem to determine the state of stress in a thin shell
is particularly simple in the axial-symmetric case. In such case S has the form of a
surface of revolution, which is obtained by rotation of a plane curve about an axis r
lying in the plane of the curve. This curve is called the meridian, its plane is called
the meridian plane and r is called the axis of the surface of revolution. The inter-
section between the surface of revolution and a plane normal to r is called parallel
circle. The position of a meridian passing through a point P of S individualizes an
angle ϑ measured from some datum meridian plane. The position of a parallel circle
passing through the point P individualizes the angle ϕ made by the normal to the
surface and the axis of rotation. So a point P of S individualizes the angles ϑ, ϕ
(Fig. 3.3.1).
Let us consider a point P of S and denote with y the tangent at P to the meridian
passing through P (Fig. 3.3.6). Well the meridian plane α passing through P and the
plane β passing through P and normal to t with y are the planes of principal curva-
ture of S at P. The curve intersection between S and α is regular and has some radius
of curvature r1 with center O1 (Fig. 3.3.6). The curve intersection between S and β
is regular and has radius of curvature r2 with center O2 on the axis r (Fig. 3.3.6).
Obviously, if we denote with r0 the radius of the parallel circle passing through P
(Fig. 3.3.6), we have

r0 = r2 sin ϕ. (3.3.14)

Shells that have the form of surfaces of revolution and are loaded symmetrically
with respect to their axis can be studied in a particularly simple way, since by sym-
metry any quantity is only function of the variable ϕ. In the study we consider a point
P of S and cut out an element of the shell by the two adjacent meridians ϑ, ϑ + dϑ
and the two parallel circles ϕ, ϕ + dϕ (Fig. 3.3.9), where dϑ, dϕ denote positive
real numbers as small as we want. Evidently the surface area of this infinitesimal
element of S is then r1 r2 sin ϕ dϕ dϑ.

Fig. 3.3.9
3.3 Shells 413

From the assumed symmetry of loading and geometrical it can be concluded that
only normal stresses will act on the sides of the element lying in the meridian planes.
Such stresses can be reduced to the resultant force

Nϑ r1 dϕ

and to the resultant moment

Mϑ r1 dϕ,

Nϑ and Mϑ being independent of the angle ϑ which defines the position of the
meridian (Fig. 3.3.9). The side of the element perpendicular to the meridian which
is defined by the angle ϕ is acted upon by normal stresses which result in the
force

Nϕ r2 sin ϕ dϑ = Nϕ r0 dϑ (3.3.15)

and in the moment

Mϕ r2 sin ϕ dϑ = Mϕ r0 dϑ

and by shearing stresses which result in the force

Qϕ r2 sin ϕ dϑ = Qϕ r0 dϑ (3.3.16)

normal to S (Fig. 3.3.9). Analogously, the side of the element perpendicular to the
meridians which is defined by the angle ϕ + dϕ is acted upon by stresses which
result in the force

dNϕ dr0
Nϕ + dϕ r0 + dϕ dϑ, (3.3.17)
dϕ dϕ

in the moment

dMϕ dr0
Mϕ + dϕ r0 + dϕ dϑ,
dϕ dϕ
and in the shearing force

dQϕ dr0
Qϕ + dϕ r0 + dϕ dϑ. (3.3.18)
dϕ dϕ

The external load acting upon the element can be resolved, as before, into two
components

Yr1 r2 sin ϕ dϕ dϑ. (3.3.19)

414 3 The Two-Dimensional Problems

and

Zr1 r2 sin ϕ dϕ dϑ

tangent to the meridians and normal to S, respectively.

We can now impose the equilibrium of the infinitesimal element cut out at P
(Fig. 3.3.10). Let us begin with the equilibrium in the direction of the tangent y
to the meridian (Fig. 3.3.6). From the expressions (3.3.15), (3.3.16), (3.3.17), and
(3.3.18) of the forces acting on the upper and lower side of the element and of
from the expression (3.3.19) of the external force, by neglecting a small quantity
of second order, we find a resultant force in the direction of the tangent y to the
meridian equal to

d(Nϕ r0 )
dϕ dϑ + Yr1 r0 dϕ dϑ − Qϕ r0 dϕ dϑ.
dϕ
On the lateral sides of the element two forces of intensity Nϑ r1 dϕ act together.
Their resultant has an intensity of Nϑ r1 dϕ dϑ, direction PO2 and component in
the y direction (Fig. 3.3.6)

− Nϑ r1 cosϕ dϕ dϑ.

So the equation of equilibrium of the element in the direction of the tangent y to the
meridian becomes
 
d Nϕ r0
− Nϑ r1 cosϕ − Qϕ r0 + Yr1 r0 = 0. (3.3.20)
dϕ

Let us consider now the equilibrium of the element in the z direction normal to
the shell (Fig. 3.3.9). The forces acting on the upper and lower sides of the element
have a resultant in the z direction equal to
 
d Qϕ r0
Nϕ r0 dϕ dϑ + dϕ dϑ. (3.3.21)
dϕ

Fig. 3.3.10
3.3 Shells 415

Fig. 3.3.11

Fig. 3.3.12

The forces acting on the lateral sides of the element and having the resultant
Nϑ r1 dϕ dϑ give a component in the z direction of the magnitude (Fig. 3.3.9)

Nϑ r1 sinϕ dϕ dϑ. (3.3.22)

The external load acting on the element has in the same z direction a component

Zr1 r0 dϕ dϑ. (3.3.23)

Summing up the forces (3.3.21), (3.3.22), and (3.3.23), we obtain the second
equation of equilibrium
 
d Qϕ r0
Nϕ r0 + Nϑ r1 sinϕ + + Zr1 r0 = 0. (3.3.24)
dϕ

The third equation is obtained by considering the equilibrium of the moments

with respect to the tangent to the parallel circle of all the forces acting on the ele-
ment. Observing that the angle between the planes in which the moments Mϑ act is
equal to cos ϕ dϑ, we get (Fig. 3.3.10)

dMϕ dr0
Mϕ + dϕ r0 + dϕ dϑ − Mϕ r0 dϑ
dϕ dϕ
−Mϑ r1 cosϕ dϕ dϑ − Qϕ r2 sinϕ r1 dϕ dϑ = 0
from which

1 d(Mϕ r0 ) Mϑ
Qϕ = − cosϕ. (3.3.25)
r1 r0 dϕ r0
416 3 The Two-Dimensional Problems

In the equations of equilibrium (3.3.20) and (3.3.24) there are four unknown
quantities Nϕ , Nϑ , Mϕ , Mϑ , that can be reduced to two if we express the membrane
forces Nϕ , Nϑ and the moments Mϕ , Mϑ in terms of the components v and w of the
displacement.
In the present case of symmetrical deformation of the shell, a small displacement
of a point has only a component v in the direction of the tangent to the meridian and
a component w in the direction of the normal to the middle surface. Considering an
element AB of the meridian (Fig. 3.3.11), we see that the increase of the length of
the element due to tangential displacements v and v + dϕ dv
dϕ of its ends is equal to
dv
dϕ dϕ. Because of the radial displacements w of the points A and B, the length of
the element decreases by an amount w dϕ. We observe that the change in the length
of the element due to the difference in the radial displacements of the points A and B
can be neglected as a small quantity of higher order. Thus the total change in length
of the element AB due to deformation is dϕ dv
dϕ − w dϕ. Dividing this value by the
initial length r1 dϕ of the element, we find the strain of the shell in the direction of
the tangent to the meridian to be

1 dv w
εϕ = − . (3.3.26)
r1 dϕ r1
Considering an element of a parallel circle it may be seen (Fig. 3.3.9) that, as a
consequence of the displacements v and w, the radius r0 of the circle increases by
the amount v cos ϕ − w sinϕ. The circumference of the parallel circle increases in
the same proportion as its radius; hence

v 1 w
εϑ = − . (3.3.27)
r2 tgϕ r2
By using the Navier’s law we get (Fig. 3.3.12)

1  
εϕ = Nϕ − νNϑ (3.3.28)
Eh
1  
εϑ = Nϑ − νNϕ . (3.3.29)
Eh
Substituting the (3.3.28) and (3.3.29) in the Eqs. (3.3.26) and (3.3.27),
we have
 
Eh 1 dv ν v
Nϕ = −w + −w (3.3.30)
1 − ν 2 r1 dϕ r2 tgϕ
 
Eh 1 ν v dv
Nϑ = −w + −w . (3.3.31)
1 − ν 2 r2 tgϕ r1 dϕ
To get similar expressions for the moments Mϕ , Mϑ let us consider the changes
of curvature of the shell element in study (Fig. 3.3.10). Considering the upper and
the lower sides of that element, we see that the initial angle between these two
sides is dϕ. Because of the displacement v along the meridian, the upper side of
3.3 Shells 417

the element rotates with respect to the perpendicular to the meridian plane by the
amount rv1 . As a result of the displacement w, the same side further rotates about
dw
the same axis by the amount r1 dϕ . Hence the total rotation of the upper side of the
element is
v dw
+ .
r1 r1 dϕ
For the lower side of the element the rotation is

v dw d v dw
+ + + dϕ.
r1 r1 dϕ dϕ r1 r1 dϕ
Hence, neglecting the strain of the middle surface, we divide the angular change
by the length r1 dϕ of the arc obtaining for the change of curvature of the meridian

1 d v 1 dw
χϕ = + . (3.3.32)
r1 dϕ r1 r1 dϕ
To find the change of curvature in the plane perpendicular to the meridian, we
observe that, because of the symmetry of the deformation, each of the lateral sides
of the shell element rotates in its meridian plane by the angle rv1 + r1dwdϕ (Fig. 3.3.9).
Since the normal to the right lateral side of the element makes the angle π2 −cosϕ dϑ
with the tangent to the y axis, the rotation of the right side in its own plane has a
component with respect to the y axis equal to − rv1 + r1dwdϕ cosϕ dϑ. This results
in a change of curvature

v 1 dw 1
χϑ = + . (3.3.33)
r1 r1 dϕ r2 tgϕ
From the (3.3.32) and (3.3.33) we then obtain
 
1 d v 1 dw ν v 1 dw
Mϕ = −D + + + (3.3.34)
r1 dϕ r1 r1 dϕ r2 tgϕ r1 r1 dϕ
 
1 v 1 dw ν d v 1 dw
Mϑ = −D + + + (3.3.35)
r2 tgϕ r1 r1 dϕ r1 dϕ r1 r1 dϕ
where

E h3
D= ,
12(1 − ν 2 )

similarly to thin plates. Substituting expressions (3.3.30), (3.3.31), (3.3.34),

(3.3.35), and (3.3.25) into Eqs. (3.3.20) and (3.3.24), we obtain, supposing
h = const.3.3.2

3.3.2 We easily modify the (3.3.36) and (3.3.37) if h is not a constant.

418 3 The Two-Dimensional Problems
   
12 d r2 sinϕ dv 12ν d v
−w + 2 sinϕ −w
h2 dϕ r1 dϕ h dϕ tgϕ
12r1 cosϕ v 12νcosϕ dv
− 2
−w − −w
h r2 tgϕ h2 dϕ
   
1 d r2 sinϕ d v 1 dw ν d cosϕ dw
+ + − v+
r1 dϕ r1 dϕ r1 r1 dϕ r1 dϕ r1 dϕ
1 dw ν cosϕ d v 1 dw Y
+ v+ − + + r1 r2 sinϕ = 0
r1 r22 tg2 ϕ dϕ r1 dϕ r1 r1 dϕ D
(3.3.36)

12r2 sinϕ dv 12ν sinϕ v

−w + −w
h2 r1 dϕ h2 tgϕ
v
12r1 sinϕ 12 ν sinϕ dv
+ −w + −w
h2 r2
tgϕ h2 dϕ
 
d 1 d r2 sinϕ d v 1 dw
− +
dϕ r1 dϕ r1 dϕ r1 r1 dϕ
  !
d 1 d cosϕ dw d 1 dw
−ν v+ + v+
dϕ r1 dϕ r1 dϕ dϕ r1 r22 tg2 ϕ dϕ
 
d cosϕ d v 1 dw Z
+ν + + r1 r2 sinϕ = 0.
dϕ r1 dϕ r1 r1 dϕ D
(3.3.37)
Furthermore from the (3.3.25), (3.3.34), and (3.3.35) we have
 
D d r2 sinϕ d v 1 dw
Qϕ = − +
r1 r2 sinϕ dϕ r1 dϕ r1 r1 dϕ
 
Dν d cosϕ dw D dw
− v+ + v+ (3.3.38)
r1 r2 sinϕ dϕ r1 dϕ r1 r2 tg2 ϕ
2 dϕ
Dν d v 1 dw
+ + .
r1 r2 tgϕ dϕ r1 r1 dϕ

So the mathematical model of the thin axial-symmetric shell is the mono-

dimensional boundary problem
[3.3.2] Given the open A ⊆ ]0, π [, the geometry h, r1 , r2 and the load Y, Z of the
shell, supposed regular, find two real functions v, w differentiable in A at least up
to the third order and such to satisfy in A the (3.3.36) and (3.3.37) and the initial
conditions express in terms of v, w and their derivatives. 
Clearly the solution of problem [3.3.2] furnishes the deformation of the thin
axial-symmetric shell. Then we can calculate the state of stress by the (3.3.30),
(3.3.31), (3.3.34), (3.3.35), and (3.3.38).
3.3 Shells 419

Remark 3.3.3 Clearly for the thin spherical dome having radius R the differen-
tial equations (3.3.26) and (3.3.27) of the problem become simpler since r1 =
r2 = R. 

In the field of Structural engineering we frequently encounter problems in which

a circular cylindrical shell is submitted to the action of forces distributed sym-
metrically with respect to the axis of the cylinder. This particular axial-symmetric
problem admits a simpler formulation.
Let us in fact consider a thin circular cylindrical shell having radius a and the
generator horizontal and parallel to the x axis. An element is cut from the shell by
two adjacent generators and two cross sections perpendicular to the x axis, and its
position is defined by the coordinate x and the angle ϕ (Fig. 3.3.4). We consider the
internal forces acting on the sides of this element (Fig. 3.3.13). It can be concluded
from symmetry that:

– the shearing forces Nxϕ = Nϕx vanish,

– the transverse forces Qϕ vanish,
– the twisting moments Mxϕ = Mϕx vanish,
– the forces Nϕ are constant along the circumference,
– the bending moments Mϕ are constant along the circumference.

Under such conditions of symmetry three of the six equations of equilibrium of

the element are identically satisfied, and we have to consider only the remaining
three equations, that we obtain by projecting the forces on the x and z axes and
by taking the moment of the forces about the y axis. Assuming that the external
forces consist only of a pressure Z normal to the surface, these three equations of
equilibrium are

dNx
a dxdϕ = 0
dx
dQx
a dx dϕ + Nϕ dxdϕ + Z a dx dϕ = 0
dx
dMx
a dx dϕ − Qx a dx dϕ = 0
dx

Fig. 3.3.13
420 3 The Two-Dimensional Problems

from which
dNx
=0 (3.3.39)
dx
dQx
a + Nϕ + Z a = 0 (3.3.40)
dx
dMx
− Qx = 0. (3.3.41)
dx
From the (3.3.39) we get

Nx = c (3.3.42)

where c is a constant.
Let us consider now the displacement of a point of the middle surface of the
shell. From symmetry we conclude that the component v of the displacement in the
circumferential direction vanishes. We thus have to consider only the components u
and w in the x and z directions, respectively. Clearly

∂u
εx =
∂x
and, as particular case of the (3.3.26)
w
εϕ = − .
a
Hence, by applying the Navier’s law, we obtain

Eh ∂u w
Nx = −ν (3.3.43)
1 − ν2 ∂x a
Eh w ∂u
Nϕ = − +ν . (3.3.44)
1 − ν2 a ∂x

From (3.3.39) and (3.3.43) it follows

 
∂u w c 1 − ν2
=ν +
∂x a Eh

from which, because the (3.3.44), we have

Eh
Nϕ = − w + c ν. (3.3.45)
a
Considering the bending moments, we conclude from symmetry that there is no
change in curvature in the circumferential direction. The curvature in the x direction
2
is equal to − ddxw2 . Using the same equations as for plates, we then obtain
3.3 Shells 421

Mϕ = ν Mx
d2 w (3.3.46)
Mx = −D 2
dx

where, as usual,

E h3
D=  
12 1 − ν 2

is the flexural rigidity of the shell.

After that from (3.3.40), (3.3.41), (3.3.45), and (3.3.46) we get

d2 d2 w Eh cν
D + w=Z− . (3.3.47)
dx2 dx2 a2 a

Thus all problems of symmetrical deformation of circular cylindrical shells admit

as mathematical model the problem

[3.3.3] Given the open A ⊆ R, the geometry h, a and the load Z of the shell, supposed
regular, find a real function w differentiable in A at least up to the fourth order and
such to satisfy in A the (3.3.47) and the initial conditions express in terms of w and
its derivatives. 

Remark 3.3.4 The simplest application of this equation is obtained when the
thickness h of the shell is constant. Under such conditions the (3.3.36) becomes

d4 w Z cν
+ 4β 4 w = − (3.3.48)
dx4 D aD

where

1/
Eh 4
β= .
4 D a2

Obviously the (3.3.48) admits the general integral

w = eβx [C1 cos(βx) + C2 sin(βx)] +

+e−βx [C3 cos(βx) + C4 sin(βx)] + f (x)

where f is a particular integral of the (3.3.48) and C1 , C2 , C3 , C4 must be determined

in each particular case from the boundary conditions at the ends of the cylinder. 
422 3 The Two-Dimensional Problems

3.4 Plane Strain Problems

3.4.1 Plane Strain Problems

Let V be the portion of the three-dimensional space occupied by a body and 0, x, y, z
a cartesian orthogonal frame of reference. We say that the body is in state of plane
deformation of plane x, y if ∀(x, y, z1 ), (x, y, z2 ) ∈ V

u(x, y, z1 ) = u(x, y, z2 )
v(x, y, z1 ) = v(x, y, z2 )
w(x, y, z1 ) = 0.

In many important problems of the Civil engineering, among which dams

(Fig. 3.4.1) and tunnels (Fig. 3.4.2), the body is practically in state of plain defor-
mation. In such cases, since the length of the body is practically endless, every cross
section is plane of geometric symmetry and of load. Then, for the principle of sym-
metry of the Physics, all the points of a cross section have longitudinal displacement
equal to zero.
In effect we can prove that every prismatic solid very long and submitted on the
mantle to a transversal load uniform along the axis of the solid is a problem of plane
deformation.
To study this problem, let us consider a right prism of length l that occupies the
volume V of the three-dimensional space. The boundary ∂V is constituted by a side
surface S and the two bases (Fig. 3.4.3). The material is homogeneous, isotropic and
linearly elastic. The deformations are, for hypothesis, small.

Fig. 3.4.1

Fig. 3.4.2
3.4 Plane Strain Problems 423

Fig. 3.4.3

Fig. 3.4.4

The load applied on the prism consists in volumetric forces X̂, Ŷ, Ẑ such that
∀(x, y, z1 ), (x, y, z2 ) ∈ V

X̂(x, y, z1 ) = X̂(x, y, z2 )
Ŷ(x, y, z1 ) = Ŷ(x, y, z2 ) (3.4.1)
Ẑ(x, y, z1 ) = 0

and in superficial forces, applied in S, p̂x , p̂y , p̂z such that ∀(x, y, z1 ), (x, y, z2 ) ∈ S

p̂x (x, y, z1 ) = p̂x (x, y, z2 )

p̂y (x, y, z1 ) = p̂y (x, y, z2 ) (3.4.2)
p̂z (x, y, z1 ) = 0.

On the bases we apply the superficial forces p̂x , p̂y , p̂z furnished by the following
auxiliary problem.
Let us suppose that
– A3.4.1 is a cross section of the prism,
– ∂A is its boundary,
– ∀(x, y) ∈ A, nx , ny are the direction cosines of the normal n to ∂A, oriented going
out from A, (Fig. 3.4.4)
∀ ((x, y), z) ∈ ∂A × {0, l} p̃x (x, y) = px (x, y, z), p̃y (x, y) = −py (x, y, z),
– ∀(x, y, z) ∈ V X̃(x, y) = X(x, y, z), Ỹ(x, y) = Y(x, y, z).

3.4.1 A can be mono-connected or multi-connected.

424 3 The Two-Dimensional Problems

We consider the auxiliary problem (Fig. 3.4.4)

[3.4.1] In the hypothesis of regularity of the data, find three real functions σx , σy , τxy
differentiable in A at least up to second order and such that in A

∂σx ∂τxy
+ + X̃ = 0 (3.4.3)
∂x ∂y
∂τyx ∂σy
+ + Ỹ = 0 (3.4.4)
∂x ∂y
∂ 2 (σx + σy ) ∂ 2 (σx + σy ) 1 ∂X ∂Y
+ =− + (3.4.5)
∂x 2 ∂y2 1−ν ∂x ∂y

and in ∂A

σx nx + τxy ny = p̃x (3.4.6)

τyx nx + σy ny = p̃x .  (3.4.7)

In the Mathematical theory of elasticity it is shown that problem [3.4.1] admits

an unique solution and there are two functions u,v such that in A

∂u 1 
= σx − νσy
∂x E
∂v 1 
= σy − νσx
∂y E
∂u ∂v 1
+ = τxy .
∂y ∂x G

After that we denote with σx (x, y), σy (x, y), τxy (x, y) the unique solution of prob-
lem [3.4.1] and we apply on the left basis of the prism of Fig. 3.4.3 the superficial
forces p̂x , p̂y , p̂z given by

p̂x (x, y, 0) = 0
p̂y (x, y, 0) = 0   (3.4.8)
p̂z (x, y, 0) = −ν σx (x, y) + σy (x, y)

∀(x, y) ∈ A and on the right basis the superficial forces p̂x , p̂y , p̂z given by

p̂x (x, y, l) = 0
p̂y (x, y, l) = 0   (3.4.9)
p̂z (x, y, l) = −ν σx (x, y) + σy (x, y)

∀(x, y) ∈ A. Then the mathematical model of the problem in examination of

Fig. 3.4.3 is
3.4 Plane Strain Problems 425

[3.4.2] In the hypothesis of regularity of the data, find three real functions u, v, w
differentiable in V at least up to second order and solution of problem [1.5.1],
relative to body of Fig. 3.4.3 loaded by the volumetric forces (3.4.1) and by the
superficial forces (3.4.2), (3.4.8), and (3.4.9). 

We have seen (Chap. 1) that the problem [3.4.2] admits an unique3.4.2 solution.
Well the triplet û, v̂, ŵ given by

û(x, y, z) = û(x, y) ∀ ((x, y), z) ∈ A × [0, l]

v̂(x, y, z) = v̂(x, y) ∀ ((x, y), z) ∈ A × [0, l] (3.4.10)
ŵ(x, y, z) = 0 ∀(x, y, z) ∈ V

is just the unique solution of the problem in examination [3.4.2]. As a consequence,

the problem [3.4.2] is a plane strain problem.
To demonstrate that û, v̂, ŵ is the solution of problem [3.4.2], we first of all
observe that from the (3.4.10) it follows

ε̂x (x, y, z) = εx (x, y) ∀ ((x, y), z) ∈ A × [0, l]

ε̂y (x, y, z) = εy (x, y) ∀ ((x, y), z) ∈ A × [0, l]
ε̂z (x, y, z) = 0 ∀(x, y, z) ∈ V
γ̂xy (x, y, z) = γxy (x, y) ∀ ((x, y), z) ∈ A × [0, l]
γ̂xz (x, y, z) = 0 ∀(x, y, z) ∈ V
γ̂yz (x, y, z) = 0 ∀(x, y, z) ∈ V
so that

σ̂x (x, y, z) = σx (x, y) ∀ ((x, y), z) ∈ A × [0, l]

σ̂y (x, y, z) = σy (x, y) ∀ ((x, y), z) ∈ A × [0, l]
 
σ̂z (x, y, z) = −ν σx (x, y) + σy (x, y) ∀ ((x, y), z) ∈ V × [0, l]
(3.4.11)
τ̂xy (x, y, z) = τxy (x, y) ∀ ((x, y), z) ∈ A × [0, l]
τ̂xz (x, y, z) = 0 ∀(x, y, z) ∈ V
τ̂yz (x, y, z) = 0 ∀(x, y, z) ∈ V.

With the (3.4.3), (3.4.4), (3.4.5), and (3.4.11) we easily verify that the (1.2.3)
are satisfied in V. Since nz = 0 in every point of V, with the (3.4.6), (3.4.7),
(3.4.2), and (3.4.11) we easily verify that the (1.2.5) are satisfied in S. Furthermore
in every point of the left basis we have nx = 0, ny = 0, nz = −1. So, tak-
ing into account the (3.4.11) and (3.4.8), the (1.2.5) are satisfied in the left basis.
In the same way we verify that the (1.2.5) are satisfied in the right basis of
the prism. So the (3.4.10) are a balanced solution of the problem [3.4.2]. Since

3.4.2 Unless a polynomial at the most of first degree, which depends from the constraints.
426 3 The Two-Dimensional Problems

the (3.4.10) are a regular triplet of displacement components, they are a com-
patible solution of the problem [3.4.2]. So the (3.4.10) are the solution of the
problem [3.4.2].
Clearly, to know the (3.4.10) we preliminarily must solve the problem [3.4.1].
Remark 3.4.1 We notice that the (3.4.5) is equivalent to the compatibility equation

∂ 2 εx ∂ 2 εy ∂ 2 γxy
+ = . (3.4.12)
∂y2 ∂x2 ∂x∂y
 
In fact let us suppose true the (3.4.12). Since σ̂z = ν σ̂x + σ̂y we have

1    1   
εx = ε̂x = σ̂x − ν σ̂y + σ̂z = 1 − ν 2 σx − ν(1 + ν)σy (3.4.13)
E E
1    1   
εy = ε̂y = σ̂y − ν σ̂x + σ̂z = 1 − ν 2 σy − ν(1 + ν)σy (3.4.14)
E E
2(1 + ν)
γxy = τxy . (3.4.15)
E
Furthermore by differentiating the (3.4.3) with respect to x and the (3.4.4) with
respect to y and summing we have

∂ 2 τxy ∂ 2 σx ∂ 2 σy ∂X ∂Y
2 =− 2 − − − .
∂x∂y ∂x ∂y2 ∂x ∂y
Then from the (3.4.12), (3.4.13), (3.4.14), and (3.4.15) we have

1 ∂ 2   
2 σ − ν (1 + ν) σ + 1 ∂
2   
1 − ν x y 1 − ν 2 σy − ν (1 + ν) σy
E ∂y 2 E ∂x 2
 
2 (1 + ν) ∂ 2 τxy 1 + ν ∂ 2 σx ∂ 2 σy 1 + ν ∂X ∂Y
= =− + − +
E ∂x∂y E ∂x 2 ∂y 2 E ∂x ∂y

from which
   
∂ 2 σx ∂ 2 σy ∂ 2 σx ∂ 2 σy ∂X ∂Y
(1 − ν) + + (1 − ν) + + + =0
∂y 2 ∂x2 ∂x 2 ∂y2 ∂x ∂y
from which the (3.4.5) follows. The vice versa is easily proven. 
Chapter 4
The One-Dimensional Problems

4.1 Ropes

4.1.1 The Funicular

Let us consider a system of n coplanar forces, assigned by the line of action, the
intensity f1 , . . . , fn and the verse (Fig. 4.1.1). We consider the polygon of the forces
012345 obtained drawing, in a convenient scale, from a point 0 the segment 01
parallel and equiverse to f1 , afterwards from the point 1 the segment 12 parallel and
equiverse to f2 , etc. (Fig. 4.1.1). Clearly the oriented segment 0k is the resultant
of the forces f1 , . . . , fk , so that in Fig. 4.1.1 the resultant of the system f1 , . . . , f5
is r = 05.
After that we choose a point P non belonging to the resultant 05 and draw a line a
parallel to P0. Then from the intersection point 1 (Fig. 4.1.1) we draw the line 1 2
parallel to P1, etc. We call funicular polygon the polygon a1 2 3 4 5 b. We easily
prove that the system f1 , . . . , f5 is equivalent to the system constituted by the only
force r applied in the point R intersection of a and b. With equal easiness we prove
that the system f1 , . . . , f5 is equilibrated by the system constituted by the force P0
applied in the first side a of the funicular polygon and the force 5P applied in the
last side b of the funicular polygon.
Remark 4.1.1 Let us suppose that the system f1 , . . . , fn has resultant r = 0. Evidently
in such case the end points 0 and n of the polygon of the forces are coincident and
then the first and last side of the funicular polygon are parallel. Obviously if a = b
the system is a couple. If a = b then f1 , . . . , fn is a null system. 

Remark 4.1.2 When the forces f1 , . . . , fn are parallel, the sides of the polygon of the
forces are parallel (Fig. 4.1.2). As a consequence the tension kP in every trunk of
the cable always has the same horizontal component H (Fig. 4.1.2). 

Remark 4.1.3 When the system is constituted by only two forces, if they have the
same sense then the resultant one is inside (Fig. 4.1.3); if they have opposite sense
then the resultant one is external, on the side and with the sense of the greatest
(Fig. 4.1.4). Obviously the position of the resultant r can also be determined by
calculating the values d1 , d2 for which the resultant r is equivalent to the system. 

A. Maceri, Theory of Elasticity, DOI 10.1007/978-3-642-11392-5_4, 427


C Springer-Verlag Berlin Heidelberg 2010
428 4 The One-Dimensional Problems

Fig. 4.1.1

Fig. 4.1.2

Fig. 4.1.3

Fig. 4.1.4

When the system is constituted by forces q distributed on a curve (Fig. 4.1.5), the
polygon of the forces becomes the curve of the forces (Fig. 4.1.6) and the funicular
polygon becomes the funicular curve or simply the funicular. Obviouly a generic
segment ab in the curve of the forces (Fig. 4.1.6) gives the direction, the sense
and the greatness of the resultant one of the forces distributed on the curve AB
(Fig. 4.1.5). Furthermore such resultant has a line of action passing for the point
4.1 Ropes 429

Fig. 4.1.5

Fig. 4.1.6

intersection of the tangents ta and tb in A and B respectively to the funicular curve.

In the curve of the forces the segments Pa and Pb, parallel to ta and tb respectively
(Fig. 4.1.6), measure the tension in the points A and B of the funicular respectively.
The the case in which the system is constituted by forces q normal to an axis x
and distributed on an its segment (Fig. 4.1.7) is of particular interest. In such case
the curve of the forces degenerates in a line and the polar distance H measures the
horizontal component of the tension in every point of the funicular, component that
in this case is constant (Fig. 4.1.7). Let us consider the tangent to the funicular in
a generic point A and its parallel line Pa passing through P. They form with the
horizontal axis x an angle ϕ such that (Fig. 4.1.7)

dy ac
= = tgϕ
dx H

from which, by differentiating

d2 y d(ac)
dx = . (4.1.1)
dx2 H

Clearly d(ac) is the diminution ab that suffers the segment ac when we pass
from the point A to a very near point B. So d(ac) = −q dx (Fig. 4.1.7) and the
(4.1.1) becomes

Fig. 4.1.7
430 4 The One-Dimensional Problems

d2 y q
2
=− . (4.1.2)
dx H
The (4.1.2) is the differential equation of the funicular of the load q.

4.1.2 The Ropes

The rope is an one-dimensional body deprived of every rigidity except the exten-
sional one.
Let us consider a rope having negligible weight, fixed to the extremities in points
of a plane and submitted to loads f1 , . . . , fn acting in the plane. Clearly the funicular
polygon of the system f1 , . . . , fn is just the configuration of equilibrium of the rope.
Furthermore the tension in every part of the rope is measured in the polygon of the
forces by the segment Pk parallel to the trunk of rope (Fig. 4.1.1).
The problem of the rope is intrinsically complex because it doesn’t have proper
form, meaning doesn’t have an initial configuration.
The rope is only solicited to simple traction, so that the static study reduces it to
that of the determination of the funicular, its configuration of equilibrium.
The elastic extension has on the static regime of the rope a greater influence if
the f/l ratio (Fig. 4.1.8) is smaller. Therefore the elastic extension is considered only
in the tense cables, while in slow cables the length is supposed invariable.
Let us consider now the case of a very tense heavy rope and let us suppose that
the end points A, B are fixed at the same height (Fig. 4.1.8). Evidently in this case in
the funicular configuration of equilibrium the arrow f is small in comparison to the
length l, so that in every point of the rope the inclination of the tangent is small. As a
consequence the unitary weight of the rope measured in the funicular configuration
of equilibrium can be approximated with the proper constant unitary weight q of the
rope. So the (4.1.2) holds, and integrating we have that the funicular is the parabola

q x2
Hy = − + c1 x + c2 .
2
Since the end points A, B are on the axis x, from the similar triangles ACD and
0ca, taking into account that in the parabola the vertex V is the middle point of the
segment CD, we have (Fig. 4.1.8)

Fig. 4.1.8
4.1 Ropes 431

q l2
f = .
8H
Furthermore

q
y= x(l − x)
2H
and the maximum tension is verified in the points of attack A, B and is equal to
 1
2 2
ql
Smax = H 2 + .
2
The length of the rope in the funicular configuration is

 l 2
1
2
dy
L= 1+ dx.
0 dx

The most important case is when the heavy rope is not very tense (Fig. 4.1.9) or
has the end points A, B fixed at very different heights (Figs. 4.1.10 and 4.1.11). In
such cases in many trunks of the rope the tangent has a strong inclination. So the
unitary weight of the rope measured in the funicular configuration of equilibrium
can not be approximated with the proper constant unitary weight q of the rope.
Then the (4.1.2) becomes

d2 y ds
H 2
= −q
dx dx

from which

 1
2 2
d2 y dy
H 2 = −q 1 + . (4.1.3)
dx dx

Fig. 4.1.9
432 4 The One-Dimensional Problems

Fig. 4.1.10

Fig. 4.1.11

The funicular obtained by integrating the (4.1.3) is called the catenary. As we

easily verify, it has the analytical expression4.1.1

H q  
y = c1 − cosh (x − c2 ) − 1 (4.1.4)
q H
where c1 , c2 are the constants of integration.
When the end points A, B are fixed at the same height, with the notations of
Fig. 4.1.9 the (4.1.4) becomes

H q 
y=− cosh x ,
q H
the arrows f is equal to
 
H ql
f = cosh −1 ,
q 2H

the tension S in a generic point of the rope is equal to

S = −qy

4.1.1 We denote with coshα the hyperbolic cosine of α, i.e. coshα = (eα + e−α )/2.
4.2 Beams 433

so that

Smax = H + q,

and the length L of the extended rope is equal to

H ql
L=2 sinh .
q 2H

Remark 4.1.4 The catenary is usually employed in the planning of the lines of dis-
tribution of the electric energy. In such case q can be assumed equal to the proper
unitary weight of the conductor increased of the weight of the muff of ice. We also
must take into account the variation of the temperature, that varies the value of L and
therefore that of the arrow f and as a consequence that of the horizontal component
H of the tension. 

4.2 Beams

4.2.1 The Deflected Beam

The most important one-dimensional problem is the deflected beam. It allows to
completely analyze a large class of structures of great technical interest, plane
or three-dimensional, constituted by some one-dimensional structural elements
assembled by constraints (Fig. 4.2.1).
We will call deflected beam a system constituted by some one-dimensional4.2.1
rectilinear bodies, all disposed on the axis z of a cartesian orthogonal frame of ref-
erence O, x, y, z (Fig. 4.2.2). The bodies are constrained among them and with the
foundation and are all loaded by concentrated and/or distributed external forces nor-
mal to the axis z and by concentrated external couples (Fig. 4.2.3). The axis z pass
through the centroid of the cross section; y, z is plane of geometric and loading sym-
metry, so that the points situated on the axis z have displacement u exactly equal to

Fig. 4.2.1

4.2.1 We call one-dimensional body a three-dimensional body having two very small dimensions
with respect to the third dimension. We call deflected beam a rectilinear one-dimensional body
able to absorb bending moments.
434 4 The One-Dimensional Problems

Fig. 4.2.2

Fig. 4.2.3

zero. The material is homogeneous, isotropic and linearly elastic; the deformations
are small.
Obviously the behavior of a deflected beam can be analyzed employing the
three-dimensional mathematical model. It is however available also an one-
dimensional mathematical model, analytically obviously notably simpler than the
three-dimensional one. However it constitute a least accurate mathematical repre-
sentation of the physical phenomenon. It furnishes smaller information than those
given by the three-dimensional model, but nevertheless enough to perform the
design of the structure or to verify its safety.
As the three-dimensional and two-dimensional models, also that one-
dimensional constitutes a chapter of the Mathematical theory of elasticity, in the
obvious sense that, to pact to use also sophisticated tools of the classical and mod-
ern Mathematical analysis, all the qualitative and quantitative information related
to the mathematical model are obtainable in a practically exact way.
The constraints connect the bodies of the structure among them and with the
foundation. In general they apply on the connected beams some actions that set
a limit to the displacements of the points in which they are installed. The most
important constraint are:

– the fixed end, that prevents every possibility of displacement to the points of the
cross section in which it is installed (Fig. 4.2.3),
– the hinge, that allows the connected cross section to freely rotate around the point
in which it is installed, but forces the connected bodies to have exactly the same
displacement at the point in which it is installed (Fig. 4.2.3),
– the bogie, that allows the constrained cross section to freely rotate around the point
in which it is installed, but allows the point of the constrained body in which it is
installed to have displacements only along the track (Fig. 4.2.3).

To build the mathematical model of a generic deflected beam, we consider its

rectilinear unconstrained trunk AB (Fig. 4.2.4). We introduce the frame of reference
O, z, v (Fig. 4.2.5), denote with p the transversal distributed load applied on the trunk
4.2 Beams 435

Fig. 4.2.4

Fig. 4.2.5

AB and assume p(z) > 0 if in the cross section of abscissa z the external unitary force
p(z) is directed downward (that is in the verse of y).
Let us observe now that the stresses σ and τ practiced by the part AS of beam
on the part SB (Fig. 4.2.5) are a system of forces distributed on the plane surface S
that obviously is equivalent to a system constituted by an only force R whose line of
action is a line of the plane of symmetry y, z. Clearly we can transport the force R
in S. So the previous distribution of σ and τ is equivalent to the system constituted
by the force R applied in S and by a transport moment M. We call such concentrated
couple M bending moment at the cross section S. Furthermore we can decompose
the force R applied in S in a component N according z, that we call normal force at
the cross section S, and in a component T according y, that we call shear at the cross
section S. This way the previous distribution of σ and τ is equivalent to the system
constituted by the concentrated force N applied in S, by the concentrated force T
applied in S and by the couple M concentrated in S (Fig. 4.2.6). In other words, we
can say that the part AS of beam practices on the part SB through the plane surface
S an action constituted by the normal force N, the shear T and the bending moment
M. Obviously, since the principle of action and reaction of the Physics, the part
BS of beam practices on the part AS through the surface S an action constituted by
a normal force N, a shear T and a bending moment M, respectively having equal
intensity to the preceding actions but opposite verse (Fig. 4.2.6).
We will see that the moment M results from a linear distribution of σ z , as shown
in Fig. 4.2.7. So, in particular, the bending moments of Fig. 4.2.6 extends the inferior
fibers (Fig. 4.2.7).
We agree that N is positive if is a tensile force, as it happens in Fig. 4.2.6; that T
is positive if the left part of beam pushes up that of right, as it happens in Fig. 4.2.6;
that M is positive if extends the inferior fibers, as it happens in Fig. 4.2.6.

Fig. 4.2.6
436 4 The One-Dimensional Problems

Fig. 4.2.7

Fig. 4.2.8

Let us cut out now from the beam the infinitesimal element SW, where S is the
cross section of abscissa z and W is the cross section of abscissa z + dz (Fig. 4.2.8).
The load applied on this elements is only constituted by a distributed load p, nor-
mal to z. Assuming the positive real number dz as small as we want, with good
approximation we have (Fig. 4.2.9)

p(t) = p(z) ∀t ∈ [z, z + dz]

N(z + dz) = N(z) + dN(z)
T(z + dz) = T(z) + dT(z)
M(z + dz) = M(z) + dM(z).

Since the principle of dissection, the elementary trunk SW shown in Fig. 4.2.9
is in equilibrium. Then, imposing the equilibrium to the translation according z, we
get

dN
= 0,
dz
so that N = const (Fig. 4.2.9). Imposing the equilibrium to the translation according
y, we get (Fig. 4.2.9)

T − (T + dT) − p dz = 0

Fig. 4.2.9
4.2 Beams 437

from which

dT
= −p. (4.2.1)
dz

Imposing the equilibrium to the rotation around the axis parallel to x and passing
through the centroid of S, we get (Fig. 4.2.9)

dz
M(z) + (p dz) + (T + dT)dz − (M + dM) = 0
2
from which

dM
= T. (4.2.2)
dz

Clearly from the (4.2.1) and (4.2.1) we get4.2.2

d2 M
= −p. (4.2.3)
dz2

When we apply the external load p, it occurs a deformation of the structure and
as a consequence every its point occupies a new position. In particular the points
of the line AB will form an arc of curve what will lie in the plane of symmetry y, z
(Fig. 4.2.10).

Fig. 4.2.10

4.2.2 Because of the hypothesis of small deformations, the deformed configuration of the beam is
very close to the initial one. This allows to obtain the (4.2.3) imposing the equilibrium conditions
on the undeformed beam. Such proceeding way has always received good experimental confirma-
tions in all the problems of the Theory of elasticity, except the problems of equilibrium stability
(Chap. 6).
438 4 The One-Dimensional Problems

Fig. 4.2.11

As the first fundamental hypothesis, we suppose that

during the deformation the displacement of any cross section S of the trunk AB is normal to
the segment AB4.2.3 (Fig. 4.2.10).
(4.2.4)
We call lowering or raising and denote with v(z) the final vertical displacement
of the cross section S of abscissa z (Fig. 4.2.10). We agree that v(z) > 0 if the dis-
placement happens in the verse of the axis v and we call elastic line of the deflected
beam the function v so defined.
Let us denote now with S the position reached by S to happened deformation
and with t the tangent to the elastic line at S (Fig. 4.2.10).
As the second fundamental hypothesis, we suppose that
to happened deformation the points of the cross section S are situated on the plane S normal
to t, so that S is the deformed configuration4.2.4 of the cross section S (Fig. 4.2.10).
(4.2.5)
We call rotation of the cross section S and we denote with the symbol ϕ(S)
or ϕ(z) the angle formed by S and S , that is the angle of which it rotates the
cross section S (Fig. 4.2.11). Evidently ϕ(S) is also the angle that t forms with the
axis of the beam AB (Fig. 4.2.11). As a consequence, if v is differentiable in z,
we have

dv
tg(ϕ(z)) = (z). (4.2.6)
dz

On the other hand the hypothesis that the deformations are small involves that dv
dz
is in modulus everywhere near to zero. So, since the (4.2.6), we approximately have
(Fig. 4.2.12)

tg(ϕ(z)) = sin(ϕ(z)) = ϕ(z). (4.2.7)

4.2.3 Such approximation, valid in the field of small deformations, was suggested by the results
gotten by Saint Venant in the problem of the bending and by the results of the experimental tests
performed in the Laboratories material tests.
4.2.4 Also such approximation is valid in the field of small deformations and was suggested by the
results gotten by Saint Venant in the problem of the bending and by the results of the experimental
tests performed in the Laboratories material tests.
4.2 Beams 439

Fig. 4.2.12

From the (4.2.6) and (4.2.7) we obviously get

dv
ϕ(z) = (z). (4.2.8)
dz
From the (4.2.8) it immediately follows that the rotation ϕ(z) is positive if
clockwise (Fig. 4.2.13).
Let us consider now a generic cross section S and

– denote with E the modulus of Young of the material,

– denote with n the diameter conjugated to the direction y of p with respect to the
inertia centroidal ellipse of the area S,
– call n neutral axis of the cross section S;
– denote with In the moment of inertia of the area S with respect to the neutral axis n,
– denote with m the line normal to the neutral axis n and to axis z,
– suppose that anywhere σx , σy , τxy = 0, as it happens in the Saint Venant’s
problems.

From the hypothesis (4.2.5) it follows (Fig. 4.2.14)

w(n, m, z) = −m ϕ(z).

Fig. 4.2.13

Fig. 4.2.14
440 4 The One-Dimensional Problems

Then, denoting with dn the orthogonal distance from the neutral axis n (so that
dn = m) and taking into account the (4.2.8), we have

∂w d2 v
εz = −dn 2 . (4.2.9)
∂z dz

Consequently εz varies linearly with the orthogonal distance from the neutral axis
n. Furthermore from the (4.2.9) and (1.4.5) it follows that σz also varies linearly with
the distance from the neutral axis, with the law
d2 v
σz = Eεz = −Edn . (4.2.10)
dz2

As a consequence, denoting with Mn the projection of M on n, we have

 
d2 v d2 v
Mn = σz dn dS = −E 2 dn2 dS = −EIn
S dz S dz2

from which4.2.5

d2 v Mn
2
=− (4.2.11)
dz EIn

and, taking into account the (4.2.10)4.2.6

Mn
σz = dn . (4.2.12)
In

In the Differential geometry we prove that if v is a regular curve of 2 assigned

as a cartesian diagram, then the curvature 1r in every point of v is exactly given by

d2 v
1 dz2
= !3
r 2 2
dv
1+
dz

so that, because the hypothesis of small deformations, we can assume

1 d2 v
= 2.
r dz

4.2.5 The (4.2.11) macroscopically simulates (exactly as the Navier’s rules do microscopically) the
answer of the body in terms of deformation when the load is applied.
4.2.6 We notice that the (4.2.12) is just the Navier’s formula of the σ in the Saint Venant’s bending
z
problem.
4.2 Beams 441

As a consequence, since the (4.2.11), we have

1 d2 v Mn
= 2 =− . (4.2.13)
r dz EIn

Finally we notice that from the (4.2.11) and (4.2.3) it follows

d4 v p
4
= .
dz EIn
On the basis of the previous considerations, we assume as mathematical model
of the deflected beam of Fig. 4.2.15, subjected to the external distributed load p, the
problem:

[4.2.1] Supposing regular the data, find a real function v differentiable in ]0, 2 l[ at
least up to fourth order and such that

d4 v p(z)
4
(z) = ∀z ∈ ]0, 2 l[ (4.2.14)
dz EI(z)
v(0) = 0 (4.2.15)

dv
(0) = 0 (4.2.16)
dz
d2 v
(0) = 0 (4.2.17)
dz2
v(2 l) = 0 .  (4.2.18)

Remark 4.2.1 In the problem [4.2.1] the differential equation (4.2.14) is a condi-
tion of equilibrium. Precisely it expresses the equilibrium to the vertical translation
and to the rotation (Fig. 4.2.9) of an elementary trunk dz of beam. On the subject
we notice that from this equilibrium on the purely analytical plan the equilibrium
of any trunk of beam is achieved. The other conditions of the problem [4.2.1] are
conditions of external compatibility, that is respect of the restrictions imposed by
the constraints. Precisely the conditions (4.2.15) and (4.2.16) express the fact that
the rigidly built-in cross section z = 0 is not able to translate neither to rotate.
The condition (4.2.17) expresses the fact that in a hinge the bending moment is
zero. The condition (4.2.18) expresses the fact that a supported cross section cannot

Fig. 4.2.15
442 4 The One-Dimensional Problems

vertically move. We notice that the internal compatibility is banally insured from the
regularity of v. 

In the Mathematical theory of elasticity we prove the existence and uniqueness

theorem
[4.2.2] The problem [4.2.1] admits an unique solution. 
Remark 4.2.2 We notice that the unique solution v of the problem [4.2.1] furnishes
for the deflected beam of Fig. 4.2.15:

– through the (4.2.8), the rotations,

– through the (4.2.11), the bending moments,
– through the (4.2.11) and (4.2.2), the shears,
– through the (4.2.12), the σz ,
– through the shears and the technical theory of Zhurawskii (cfr. Sect. 2.6), the τz .


Clearly the problem [4.2.1] is mathematically linear. As a consequence also for

the deflected beams the superposition principle subsists:
[4.2.3] Let us consider any deflected beam (Fig. 4.2.15) and denote:

– with v1 the unique solution of the deflected beam submitted to a regular load p1
(Fig. 4.2.16),
– with v2 the unique solution of the deflected beam submitted to a regular load p2
(Fig. 4.2.17),
– with v3 the unique solution of the deflected beam submitted to a regular load p3
(Fig. 4.2.18),
– with c1 , c2 any two real numbers.

Well if p3 = c1 p1 + c2 p2 , then v3 = c1 v1 + c3 v3 .
Proof By hypothesis v1 is the unique solution of the problem

Fig. 4.2.16

Fig. 4.2.17
4.2 Beams 443

Fig. 4.2.18

[4.2.4] Supposing regular the data, find a real function v1 differentiable in ]0, 2 l[ at
least up to fourth order and such that:

d4 v1 p1 (z)
4
(z) = ∀z ∈ ]0, 2 l[ (4.2.19)
dz EI(z)
v1 (0) = 0 (4.2.20)
d v1
(0) = 0 (4.2.21)
dz
d2 v1
(0) = 0 (4.2.22)
dz2
v1 (2 l) = 0.  ; (4.2.23)

v2 is the unique solution of the problem.

[4.2.5] Supposing regular the data, find a real function v2 differentiable in ]0, 2 l[ at
least up to fourth order and such that:

d4 v2 p2 (z)
(z) = ∀z ∈ ]0, 2 l[ (4.2.24)
dz4 EI(z)
v2 (0) = 0 (4.2.25)
d v2
(0) = 0 (4.2.26)
dz
d2 v2
(0) = 0 (4.2.27)
dz2
v2 (2 l) = 0.  ; (4.2.28)

v3 is the unique solution of the problem

[4.2.6] Supposing regular the data, find a real function v3 differentiable in ]0, 2 l[ at
least up to fourth order and such that

d4 v3 c1 p1 (z) + c2 p2 (z)
4
(z) = ∀z ∈ ]0, 2 l[ (4.2.29)
dz EI(z)
v3 (0) = 0 (4.2.30)
d v3
(0) = 0 (4.2.31)
dz
444 4 The One-Dimensional Problems

d2 v3
(0) = 0 (4.2.32)
dz2
v3 (2 l) = 0.  (4.2.33)

Since obviously

d4 (c1 v1 + c2 v2 ) d4 v1 d4 v2 c1 p1 + c2 p2
4
= c1 4
+ c2 4
=
dz dz dz EI
(c1 v1 + c2 v2 )(0) = c1 v1 (0) + c2 v2 (0) = 0
d(c1 v1 + c2 v2 ) dv1 dv2
(0) = c1 (0) + c2 (0) = 0
dz dz dz
d2 (c1 v1 + c2 v2 ) d2 v1 d2 v2
2
(l) = c1 2 (l) + c2 2 (l) = 0
dz dz dz
(c1 v1 + c2 v2 )(2 l) = c1 v1 (2 l) + c2 v2 (2 l) = 0,

the function c1 v1 + c2 v2 is solution of the problem [4.2.6] and then, because of the
theorem [4.2.2], it results v3 = c1 v1 + c2 v2 . 

Remark 4.2.3 Reasoning as in Remark 4.2.2, we immediately get

v3 (z) = c1 v1 (z) + c2 v2 (z)

ϕ3 (z) = c1 ϕ1 (z) + c2 ϕ2 (z)
M3 (z) = c1 M1 (z) + c2 M2 (z)
T3 (z) = c1 T1 (z) + c2 T2 (z)
σz3 (P) = c1 σz1 (P) + c2 σz2 (P)
εz3 (P) = c1 εz1 (P) + c2 εz2 (P). 

Remark 4.2.4 The superposition principle [4.2.2] also holds if concentrated forces
and/or concentrated couples are present (Fig. 4.2.19). In fact these actions can be
easily approximated by a regular distributed load. 

Fig. 4.2.19
4.2 Beams 445

Fig. 4.2.20

Remark 4.2.5 Putting in the superposition principle c2 = 0, we obviously obtain

that if the load is amplified by the factor c1 , then every greatness of the beam results
exactly amplified by the same factor c1 (Fig. 4.2.20). 

Remark 4.2.6 The elastic line furnished by the adopted mathematical model is inde-
pendent from the way in which the loading intensity has been made to vary from the
initial instant to the final one. In fact the adopted mathematical model doesn’t take
into account such variation. 

Remark 4.2.7 Let us denote with v1 [resp. v2 ] the elastic line of the cantilever of
Fig. 4.2.21 [resp. Fig. 4.2.22]. From the superposition principle it immediately fol-
lows that if we apply on the deformed cantilever of Fig. 4.2.22 the load present in
Fig. 4.2.21, then the cantilever suffers an incremental deformation exactly equal to
the elastic line v2 present in Fig. 4.2.21 (Fig. 4.2.23). 

Let us denote with M the bending moment of the deflected beam of Fig. 4.2.15,
and consider the auxiliary problem
[4.2.7] Supposing regular the data, find a real function v differentiable in ]0, 2 l[ at
least up to second order and such that

d2 v M(z)
(z) = − ∀z ∈ ]0, 2 l[ (4.2.34)
dz2 EI(z)

Fig. 4.2.21

Fig. 4.2.22

Fig. 4.2.23
446 4 The One-Dimensional Problems

v(0) = 0 (4.2.35)
dv
(0) = 0 (4.2.36)
dz
v(l− ) = v(l+ ) (4.2.37)

v(2 l) = 0.  (4.2.38)

In the Mathematical theory of elasticity we prove the following equivalence

theorem between the problem [4.2.1] and the simpler problem [4.2.7]:
[4.2.8] Every solution of problem [4.2.1] is solution of problem [4.2.7] and vice
versa. 
Problem 4.2.1 You shall determine the elastic line of the cantilever of Fig. 4.2.24.
Solution Since the [4.2.8] we can determine the elastic line v solving the problem
[4.2.7], that in our case, in the frame of reference of Fig. 4.2.26, becames

d2 v M(z)
(z) = ∀z ∈ ]0, l[ (4.2.39)
dz2 EI
v(0) = 0 (4.2.40)
dv
(0) = 0. (4.2.41)
dz

To appraise M(z) we impose the equilibrium to rotation around B of the trunk of

beam of Fig. 4.2.25, getting M(z) = −F(l − z). So we must integrate the differential
equation

d2 v F
(t) = (l − t) ∀t ∈ ]0, l[. (4.2.42)
dz2 EI

Fig. 4.2.24

Fig. 4.2.25

Fig. 4.2.26
4.2 Beams 447

Let z ∈ ]0, l[. Integrating in [0,z] the (4.2.42) we have, taking into account the
(4.2.40)
 z   z
dv dv dv d dv M(t)
(z) = (z) − (0) = (t) dt = dt
dz dz dz 0 dz dz 0 EI
 z F F 2 Fl
= (l − t) dt = − z + z.
0 EI 2 EI EI

So

dv F 2 Fl
∀t ∈ ]0, l[ (t) = − t + t. (4.2.43)
dz 2 EI EI

Let z ∈ ]0, l[. Integrating in [0,z] the (4.2.43) and taking into account the (4.2.41)
we have
 z dv
v(z) = v(z) − v(0) = (t) dt
0 dz
 (4.2.44)
z F 2 Fl F 3 Fl 2
= − t + t dt = − z + z .
0 2 EI EI 6 EI 2 EI

So we have determined the elastic line (4.2.44) and the rotations (4.2.43) of the
cantilever of Fig. 4.2.24. The lowering and the rotation of the free extreme B have
great technical importance. They respective results are equal to (Fig. 4.2.26) 

Fl3
v(l) =
3EI

Fl2
ϕ(l) = .
2EI
As it was easy to conjecture on the physical plane, from the calculation we have
obtained that the free extreme lowers and rotates in clockwise sense (Fig. 4.2.26).
Problem 4.2.2 You shall determine the elastic line of the supported beam of
Fig. 4.2.27.

Fig. 4.2.27
448 4 The One-Dimensional Problems

Solution Since the [4.2.8] we can determine the elastic line v solving the problem
[4.2.7], that in our case, in the frame of reference of Fig. 4.2.27, becomes

d2 v M(z)
(z) = ∀z ∈ ]0, l[ (4.2.45)
dz2 EI
v(0) = 0 (4.2.46)

v(l) = 0. (4.2.47)

To appraise M(z) we impose the equilibrium to rotation at z = 0 of the trunk of

beam of Fig. 4.2.28, getting

p pl
M(z) = − z2 + z ∀z ∈ ]0, l[. (4.2.48)
2 2

Substituting the (4.2.48) in the (4.2.45) and integrating we get

 z   z
dv dv d dv M(t)
(z) − (0) = (t) dt = dt
dz dz 0 dz dz 0 EI
 z p pl p 3 pl 2
= − t2 + t dt = z − z
0 2 2 6 EI 4 EI

from which

dv dv p 3 pl 2
∀t ∈ ]0, l[ (t) = (0) + t − t
dz dz 6 EI 4 EI
that integrating in [0,z], with z ∈ ]0, l[, and taking into account the (4.2.46),
furnishes
 z dv
v(z) = v(z) − v(0) = (t) dt
0 dz
 z dv p 3 pl 2 p 4 p l 3 dv
= (0) + t − t dt = z − z + (0)z.
0 dz 6 EI 4 EI 24 EI 12 EI dz

Fig. 4.2.28
4.2 Beams 449

From this, taking into account the (4.2.47), we obtain

dv p l3
(0) = .
dz 24 EI

This way we obtain the elastic line and the rotations of the supported beam
subjected to uniformly distributed load of Fig. 4.2.27:

p 4 pl 3 p l3
v(z) = z − z + z (4.2.49)
24 EI 12 EI 24 EI
dv p 3 pl 2 p l3
= z − z + . (4.2.50)
dz 6 EI 4 EI 24 EI
Since the (4.2.50) the left supported cross section rotates of
p l3
ϕ(0) =
24 EI
radians in clockwise verse, as it is physically obvious, and the right supported cross
section rotates of
p l3
ϕ(l) = −
24 EI
radians in anticlockwise verse, as it still is physically obvious (Fig. 4.2.29).
The (4.2.49) furnishes for the lowering of the middle cross section of the beam
the value, technically very important

l 5 p l4
v = . 
2 384 EI

Fig. 4.2.29

4.2.2 The Analogy of Mohr

The analytical solution of the problem [4.2.1] or of the simpler problem [4.2.7]
can be practically gained only for elementary deflected beams and/or loads. A fast
method of solution of these problems was obtained by Mohr. He observed that the
problem to determine the elastic line of a deflected beam coincides with the problem
to determine the bending moment in another deflected beam subjected to a particular
450 4 The One-Dimensional Problems

Fig. 4.2.30

load. This observation is obviously useful for the practical calculation of the elastic
line or only of the lowering and/or of the rotation of a cross section.
Let’s us consider the deflected beam of Fig. 4.2.30. Since the [4.2.8], the problem
to determine its elastic line can be reduced to find a function v such that

d2 v M
=−
dz2 EI

dv dv
v(Bs ) = 0, v(Bd ) = 0, (Bs ) = (Bd )
dz dz
dv dv
v(Ds ) = 0, v(Dd ) = 0, (Ds ) = (Dd ) (4.2.51)
dz dz
dv
v(G) = 0, (G) = 0.
dz
where M denotes the function bending moment of the deflected beam (Fig. 4.2.31).
We consider the auxiliary problem constituted by the deflected beam of
Fig. 4.2.32. This new deflected beam has different constraints but situated at the
same abscissa and is loaded by the distributed load

M
p∗ = . (4.2.52)
EI

From the (4.2.3) and (4.2.2) we get that the problem to determine the bending
moment of this new deflected beam is to find a function M∗ such that

d2 M ∗
= −p∗
dz2

Fig. 4.2.31

Fig. 4.2.32
4.2 Beams 451

dM ∗ dM ∗
M ∗(Bs ) = 0, M ∗(Bd ) = 0, (Bs ) = (Bd )
dz dz
dM ∗ dM ∗
M ∗(Ds ) = 0, M ∗(Dd ) = 0, (Ds ) = (Dd ) (4.2.53)
dz dz
dM ∗
M ∗(G) = 0, (G) = 0,
dz

Since the (4.2.52) in the problems (4.2.51) and (4.2.53) it changes only the sym-
bol with which the unknown function is denoted. So such problems are identical
and then they have the same solution, that is

v = M∗ (4.2.54)

and as a consequence

dv dM ∗
ϕ= = = T ∗. (4.2.55)
dz dz

This way we conclude that the elastic line [resp. rotations] of the deflected
beam of Fig. 4.2.30 coincides with the bending moment M∗ [resp. shear T∗ ] of
the auxiliary deflected beam of Fig. 4.2.32.
We expressly notice that, since the (4.2.13), the load p∗ is a curvature, so that it
has dimensions

∗ M (kg cm)
[p ] = = = cm−1 .
EI kg
cm4
cm2

So the auxiliary problem is an abstract problem.

Since the (4.2.52), in the cross sections in which M extends the inferior fibers p∗
is positive and therefore is downward (Fig. 4.2.32).
From this reasoning it obviously follows that the auxiliary deflected beam must
have constraints such to impose to the bending moment M∗ [resp. to the shear
T∗ ] the same restrictions that the native constraints imposed in the beam in exam-
ination to the lowering v [resp. to the rotations ϕ]. Besides, the constraints that
don’t impose in the scheme of departure restrictions to v and ϕ must be replaced
with constraints that don’t impose in the scheme auxiliary restrictions neither to
M∗ neither to T∗ .
The substitutions of constraint to be effected are shown in Fig. 4.2.33.

Problem 4.2.3 You shall determine lowering and rotation of the free end B of the
cantilever of Fig. 4.2.24.

Solution We employ the analogy of Mohr. Since the Fig. 4.2.33, we must substitute
the free end to the fixed end and the fixed end to the free end. So the auxiliary beam
452 4 The One-Dimensional Problems

Fig. 4.2.33

is that of Fig. 4.2.34. Moreover in the beam of Fig. 4.2.33 the diagram M is that of
Fig. 4.2.35. So the auxiliary problem is that of Fig. 4.2.36. We have (Fig. 4.2.37)

F l3
v(B) = M ∗(B) =
3 EI

F l2
ϕ(B) = T ∗(B) = . 
2 EI

Fig. 4.2.34

Fig. 4.2.35

Fig. 4.2.36

Fig. 4.2.37
4.2 Beams 453

Fig. 4.2.38

Problem 4.2.4 You shall determine lowering and rotation of the free end B of the
cantilever of Fig. 4.2.38, submitted to an uniformly distributed load p.
Solution We employ the analogy of Mohr. Clearly the auxiliary problem is that of
Fig. 4.2.39, where the load p∗ has the parabolic law (Fig. 4.2.40)

p (1 − z)2
p∗ (z) = − .
2 EI

We have
 l  l
∗ ∗ p
ϕ(B) = T (B) = − p (z) dz = − − (l − z)2 dz
0 0 2 EI
 l
p p l3
= (l − z)2 dz =
2 EI 0 6 EI

so that in Fig. 4.2.38 the rotation of the free end B is clockwise, as physically
obvious. Furthermore
 l 
 l p
v(B) = M ∗(B) = − p∗ (z) dz (l − z) = − − (l − z)3 dz
0 0 2 EI
 l
p p l4
= (l − z)3 dz =
2 EI 0 8 EI

so that in Fig. 4.2.38 the free end B lowers, as it is physically evident. 

Fig. 4.2.39

Fig. 4.2.40
454 4 The One-Dimensional Problems

Fig. 4.2.41

Problem 4.2.5 You shall determine lowering and rotation of the free end B of the
cantilever of Fig. 4.2.41.
Solution We employ the analogy of Mohr. Clearly the diagram M is that of
Fig. 4.2.42, so that the auxiliary problem is that of Fig. 4.2.43. The reactions of
the fixed end in the auxiliary problem are shown in Fig. 4.2.44. So

M l2
v(B) = M ∗ (B) =
2 EI
∗ Ml
ϕ(B) = T (B) = . 
EI

Problem 4.2.6 You shall determine the rotations of the supported cross section A
and B of the beam of Fig. 4.2.45.
Solution We employ the analogy of Mohr. Clearly the diagram M is that of
Fig. 4.2.46, so that, taking into account the Fig. 4.2.33, the auxiliary problem is

Fig. 4.2.42

Fig. 4.2.43

Fig. 4.2.44

Fig. 4.2.45

Fig. 4.2.46
4.2 Beams 455

Fig. 4.2.47

Fig. 4.2.48

Fig. 4.2.49

that of Fig. 4.2.47. The reactions of the supports in the auxiliary problem are shown
in Fig. 4.2.48. So (Fig. 4.2.49)

Ml
ϕ(A) = T ∗ (A) = −
3EI
and as a consequence the supported cross section A rotates in verse anticlockwise,
as physically obvious. Moreover (Fig. 4.2.49)
Ml
ϕ(B) = T ∗ (B) =
6EI
and as a consequence the supported cross section B rotates in verse clockwise, as
physically obvious. 

Problem 4.2.7 You shall determine the rotations of the supported cross section A
and C and the lowering of the middle cross section B of the beam of Fig. 4.2.50.

Fig. 4.2.50
456 4 The One-Dimensional Problems

Fig. 4.2.51

Fig. 4.2.52

Solution We employ the analogy of Mohr. Clearly the diagram M is that of

Fig. 4.2.51, so that, taking into account the Fig. 4.2.33, the auxiliary problem is that
of Fig. 4.2.52. In the auxiliary problem, since the load is equivalent to the couple
clockwise

M l 1 1 l
· · · 2· · ,
2EI 2 2 3 2

the reactions of the supports are clearly the ones shown in Fig. 4.2.53. So
(Fig. 4.2.54)

Ml l M l 1 1 l
v(B) = M ∗ (B) = − · + · · · · = 0.
24 EI 2 2EI 2 2 3 2

Insofar, keeping in mind the fact that (Fig. 4.2.54) the extended fibers are the
superior in the trunk AB and the inferior in the trunk BC, the elastic line of the beam
of Fig. 4.2.50 is that shown in Fig. 4.2.54. 

Fig. 4.2.53

Fig. 4.2.54
4.2 Beams 457

Fig. 4.2.55

Problem 4.2.8 You shall determine the lowering of the cross section B of the
cantilever of Fig. 4.2.55.
Solution We calculate v(B) in two different ways. In the first way we employ the
analogy of Mohr. Clearly, taking into account the Fig. 4.2.33, the auxiliary problem
is that of Fig. 4.2.56. Imposing the equilibrium of the trunk AB in the auxiliary
problem, we get (Fig. 4.2.57)

Fl l 1 2 l Fl l 1 l 5 F l3
v(B) = M ∗ (B) = · · · · + · · · = 
2EI 2 2 3 2 2EI 2 2 2 48 EI

so that, as predictable physically (Fig. 4.2.55), the cross section B lowers.

Let us calculate now v(B) in a different way. Evidently the bending moment in
the cross section B of Fig. 4.2.55 is equal to Fl/2 (Fig. 4.2.58) and the superior fibers
are tense. Then, for the principle of dissection [1.2.1], the problem of Fig. 4.2.55 is
equivalent (for the common parts) to that of Fig. 4.2.59. Besides, since the schemes
of Figs. 4.2.55 and 4.2.59 have both zero degrees of freedom, their elastic lines

Fig. 4.2.56

Fig. 4.2.57

Fig. 4.2.58

Fig. 4.2.59
458 4 The One-Dimensional Problems

coincide. Insofar the lowering v(B) of the cross section B of Fig. 4.2.55 is equal to
that of the cross section B of Fig. 4.2.59. For the principle of superposition [4.2.2]
the lowering of the cross section B in Fig. 4.2.59 is sum of that of the cross section
3
B in Fig. 4.2.60 (that is equal is, as seen in the Problem 4.2.1, to 24FlEI ) and of that
of the cross section B in Fig. 4.2.61 (what is equal, as seen in the Problem 4.2.5, to
Fl3
16 EI ). So

3 2
F l Fl 1 l F l3 F l3 5 F l3
v(B) = + · = + = .
3EI 2 2 2EI 2 24 EI 16 EI 48 EI

Remark 4.2.8 If in a trunk AB of a deflected beam the bending moment M is identi-

cally zero, then after the deformation the trunk AB is still exactly rectilinear. In fact
from the (4.2.11) it follows that in the interval [a, b] relative to the trunk AB we get

d2 v
(z) = 0 ∀z ∈ ]a, b[ .
dz2
Then, from known theorems of the Mathematical analysis, it results

∃c, d ∈  : v(z) = c z + d ∀z ∈ ]a, b[ . 

Problem 4.2.9 You shall determine the lowering of the cross section C of the
cantilever of Fig. 4.2.62.
Solution We calculate v(C) in two different ways. In the first way we employ the
analogy of Mohr. Clearly, taking into account the Fig. 4.2.33, the auxiliary problem
is that of Fig. 4.2.63, in which they are shown also the fixed end reactions. Then the
required lowering is

Fl2 l 2 l 5 F l3
v(C) = + · = .
8EI 2 3 2 48 EI

Fig. 4.2.60

Fig. 4.2.61
4.2 Beams 459

Fig. 4.2.62

Fig. 4.2.63

Fig. 4.2.64

Let us calculate now calculate v(C) in a different way. Evidently the elastic line of
the beam of Fig. 4.2.62 is that of Fig. 4.2.64. Since the Remark 4.2.8, the part B C of
the elastic line is a straight line. So B C Q is a right triangle (Fig. 4.2.64). Moreover,
since the hypothesis of small deformations, ϕ(B) = tg (ϕ(B)). As a consequence
(Fig. 4.2.64)

l
v(C) = v(B) + ϕ(B) . (4.2.56)
2

Evidently in Fig. 4.2.62 in the cross section Bd immediately on the right of B it

results M = T = 0. Then, for the principle of dissection [1.2.1], v(B) [resp. ϕ(B)] is
perfectly equal in the problem of Fig. 4.2.62 and in the problem of Fig. 4.2.65. As a
consequence, as seen in Problem 4.2.1

3
F l F l3
v(B) = =
3EI 2 24 EI
2
F l F l2
ϕ(B) = = .
2EI 2 8 EI

Fig. 4.2.65
460 4 The One-Dimensional Problems

From this and from the (4.2.56) we get

F l3 F l2 l 5 F l3
v(C) = + · = . 
24 EI 8 EI 2 48 EI
Problem 4.2.10 You shall determine the lowering of the hinges B and C of the
deflected beam of Fig. 4.2.66.
Solution First of all, we observe that the diagram of the bending moment M is that
shown in Fig. 4.2.67. As a consequence, since in the trunk BD [AB, DF] the inferior
[superior] fibers are extended and taking into account the constraints, the elastic line
of the deflected beam is of the type shown in Fig. 4.2.68. Therefore from the calcu-
lation we must have that B lowers it and D raises it. We unthread now in Fig. 4.2.66
the pivot B and apply the principle of dissection to the trunk AB (Fig. 4.2.69). Since
(Fig. 4.2.67) the moment of the fixed end A is equal to M and the superior fibers
are tense, the action of the pivot B on the beam AB is that shown in Fig. 4.2.69.
Then, since the problems of Figs. 4.2.66 and 4.2.69 are equivalent, from the results
of Problem 4.2.1 we get

M l3 Ml2
v(B) = = .
l 3 EI 3EI

Around v(D), we consider the problem of Fig. 4.2.70, that obviously for the prin-
ciple of dissection is equivalent to that of Fig. 4.2.66. Insofar v(D) can be calculated

Fig. 4.2.66

Fig. 4.2.67

Fig. 4.2.68

Fig. 4.2.69
4.2 Beams 461

Fig. 4.2.70

in Fig. 4.2.70. For this equivalent problem the auxiliary Mohr’s problem is obvi-
ously that of Fig. 4.2.71. Clearly the constraint reactions in Fig. 4.2.71 are those of
Figs. 4.2.72 and 4.2.73. Insofar

M l 2 M l 2 Ml2
v(D) = M ∗ (D) = · 2· l − · l =− . 
EI 2 3 EI 2 3 3EI

Fig. 4.2.71

Fig. 4.2.72

Fig. 4.2.73

Fig. 4.2.74
462 4 The One-Dimensional Problems

4.2.3 Principle of Virtual Works

The analogy of Mohr is a fast technique of calculation but asks for the prelimi-
nary determination of the function bending moment arised in the assigned deflected
beam. So this technique can be applied only for statically determined structures. If
the deflected beam is not statically determined we can apply the principle of virtual
works, which is less fast but a very powerful tool of calculation, always applicable
in every case.
We denote with H the interval of  individualizing the deflected beam and with
W the set of the real functions differentiable in H at least up to second order and
satisfying the boundary conditions. So any v ∈ W individualizes a deformed con-
figuration of the deflected beam, compatible but not necessarily equilibrated. The
internal compatibility is guaranteed by the regularity of v. The external compatibility
is just the respect of the conditions of constraint.
We call problem of the virtual works the problem4.2.7
[4.2.9] Find v ∈ W:
 
d2 v d2 ṽ
p ṽ dz = −EI dz ∀ṽ ∈ W. 
H H dz2 dz2

In the Mathematical theory of elasticity we prove the following theorem of

existence and uniqueness, called principle of virtual works
[4.2.10] The problem [4.2.9] admits an unique solution. 
Furthermore in the Mathematical theory of elasticity we prove the following
equivalence theorem
[4.2.11] Every solution of problem [4.2.1] is solution of problem [4.2.9] and vice
versa. 
We underline that the principle of virtual works is a very powerful tool of the-
oretical investigation and allows further the practical calculation of any deflected
beam.
Remark 4.2.9 The principle of virtual works is also enunciated at times saying that
the virtual work of the external forces Lve done by the external forces present in
the deflected beam in examination (system of the forces) for the displacements that
their points of application have in the compatible configuration but not necessarily
balanced v (system of the displacements) is exactly equal to the virtual work of
the internal forces Lvi done on all elementary trunks dz of the deflected beam by
the internal forces present in the system of the forces (i.e. the bending moment M)

4.2.7 In the problem [4.2.9] we call virtual works of the external [resp. internal] forces, and denote
with the symbol Lve [resp. Lvi ] the left [resp. right] side of the equation. Clearly Lvi = ∫H M dϕ̃,
where M is the bending moment relative to elastic line v and ϕ̃ is the rotation relative to the
compatible deformed configuration ṽ (so that ϕ̃ = dṽdz ).
4.2 Beams 463

for the displacements (i.e. the relative rotations dϕ̃ of the faces of the trunk dz) that
their points of application (i.e. the cross sections) have in the system of the displace-
ments. 

Remark 4.2.10 We notice that in the [4.2.9] the virtual work of the internal forces
can be obviously read
 
d2 v d2 ṽ
Lvi = −EI dz = M dϕ̃. 
H dz2 dz2 H

Remark 4.2.11 We notice that the [4.2.9] is a particular case of the [1.3.1]. In fact,
taking into account only the stresses arising from the bending, we have

    
M σ̃z M M̃
Lvi = σz ε̃z dV = − y dV = − y − y dV
V V I E V EI I
       
M M̃ 2 M M̃ M M̃
= 2
y dV = 2
y2 dA dz = dz. 
V EI H EI A H EI

4.2.4 Strain Energy

As for as the energetic aspects, as in the three-dimensional problem of the elastic

equilibrium, in the theory of the deflected beams we suppose that

– the material remains in the elastic field and during the transformation the tem-
perature of all the points of the body is always equal, so that the transformation
from the state of thermodynamic equilibrium at the initial instant ti to the state of
thermodynamic equilibrium at the final instant tf is reversible,
– the deflected beam is unloaded up to the initial instant ti ,
– at the initial instant ti the load has intensity zero,
– at the initial instant ti we begin to apply on the deflected beam a load whose
intensity varies up to reach some value at the final instant tf ,
– in all the following instants to tf the value of the load applied on the deflected
beam indefinitely stays equal to the value assumed in tf ,
– the load is statically applied, so that in every instant any particle of material has
acceleration and velocity practically zero.

Insofar in the problem of the deflected beam the total energy E coincides in every
instant with the internal energy Ei , for which we assume

Ei (ti ) = E (ti ) = 0.
464 4 The One-Dimensional Problems

Then, denoting with the symbol L the strain work, i.e. the work that the load
performs on the deflected beam, since the first principle of the Thermodynamics
we have
   
L = Ei tf = E tf . (4.2.57)

As a consequence, the strain work performed for conducting the deflected beam
from the state of thermodynamic equilibrium at the initial instant ti to that at the final
instant tf is independent from the load path, i.e. from the law according to which the
statically applied loads vary in the time.
Proceeding as in the three-dimensional case, we can prove the theorem of
Clapeyron:
[4.2.12] In the problem of the elastic equilibrium of the deflected beam the strain
work L is exactly equal to half of the product between the external forces acting
at the final instant tf and the displacements that their points of application have
reached at the final instant tf . 
Problem 4.2.11 You shall determine the strain work for the problem of Fig. 4.2.74.
Solution As seen in Problem 4.2.9, at the final instant tf the lowering of the cross
section A is
Fl3
;
3EI
then, since the [4.2.12]

1 Fl3 F 2 l3
L= (F) = . 
2 3EI 6EI

Problem 4.2.12 You shall determine the strain work for the problem of Fig. 4.2.75.
Solution As seen in Problem 4.2.5, at the final instant tf the cross section B rotates
in verse clockwise of the amount
M(2 l)
;
EI
then, since the [4.2.12]

1 M(2 l) M2 l
L= (M) = . 
2 EI EI

Problem 4.2.13 You shall determine the strain work for the problem of Fig. 4.2.76.

Fig. 4.2.75
4.2 Beams 465

Fig. 4.2.76

Solution As we easily can verify, at the final instant tf the lovering of the cross
section A is

Fl3 Ml2
+
3EI 2EI
and the cross section B rotates in verse clockwise of the amount

Fl2 M(2 l)
+ .
2EI 2EI

Then, since the [4.2.12]

1 Fl3 Ml2 1 Fl2 M(2 l) F 2 l3 M2 l2 FMl3

L = (F) + + (M) + = + + . 
2 3EI 2EI 2 2EI 2EI 6EI EI 2EI

From the Problems 4.2.11, 4.2.12, and 4.2.13 it follows

[4.2.13] For the strain work (and then for the energy) the principle of superposition
is not valid. 
In analogy to the three-dimensional case, the (4.2.57) allows to express the strain
energy in terms of the characteristics of the solicitation. Let us consider in fact a
deflected beam submitted to a distributed load p, to n concentrated forces F1 , . . . , Fn
and to n concentrated couples M1 , . . . , Mn . We denote with v the elastic line and
with M the bending moment. We apply the principle of the virtual works assuming as
system of the forces [resp. as system of the displacements] the same deflected beam
in examination. We get, denoting with H the interval individualizing the deflected
beam

 
n 
n
pv dz + Fi v(zi ) + Mi ϕ(zi ) = Lve = Lvi
H i=1 i=1

M(z)M(z)
= dz.
H EI(z)
As a consequence
# " 
1 
n 
n
1 M2
p v dz + Fi v(zi ) + Mi ϕ(zi ) = dz
2 H 2 H EI
i=1 i=1
466 4 The One-Dimensional Problems

from which, taking into account the (4.2.57)


1 M2
E= dz = L. (4.2.58)
2 H EI
Remark 4.2.12 From the (4.2.58) it obviously follows

1 M2 1 M dz 1
dE = dz = M = M dϕ. 
2 EI 2 EI 2

Remark 4.2.13 From the (4.2.58) it obviously follows


1 M2
E=L= dz ≥ 0. 
2 H EI
We still denote with H the interval of  individualizing the deflected beam and
with W the set of the real functions differentiable in H at least up to second order and
satisfying the boundary conditions. So any ṽ ∈ W individualizes a deformed con-
figuration of the deflected beam, compatible but not necessarily equilibrated. The
internal compatibility is guaranteed by the regularity of ṽ. The external compatibility
is just the respect of the conditions of constraint.
We call potential energy of the deflected beam the functional
  2 
1 d2 ṽ
P: ṽ ∈ W → −EI 2 dz − p ṽ dz
H 2EI dz H
n  n
dṽ
− Fi ṽ(zi ) − Mi (zi )
dz
i=1 i=1

and consider the following minimum potential energy problem

[4.2.14] Find v ∈ W: P(v) ≤ P(ṽ) ∀ṽ ∈ W. 
In the Mathematical theory of elasticity we prove the following theorem of
existence and uniqueness, called principle of minimum of the potential energy
[4.2.15] The problem [4.2.14] admits an unique solution. 
Furthermore in the Mathematical theory of elasticity we prove the following
equivalence theorem
[4.2.16] Every solution of problem [4.2.14] is solution of problem [4.2.9] and vice
versa. 
We underline that the principle of minimum of the potential energy is a very
powerful tool of theoretical investigation.
An other energetic theorem that holds for the deflected beam problem is that
of Betti. We consider any deflected beam submitted to two loading conditions. In
4.2 Beams 467

the problem A [resp. Ã], the load related to the final instant tf is constituted by a
distributed load p [resp. p̃], by n concentrated forces F1 , . . . , Fn [resp. F̃1 , . . . , F̃n ]
and by n concentrated couples M1 , . . . , Mn [resp. M̃1 , . . . , M̃n ]. We denote with
v [resp. ṽ] the elastic line of problem A [resp. Ã] and with M [resp. M̃] the bending
moment in problem A [resp. Ã]. We call mutual work or dragging work among
the systems A and Ã, and we denote with the symbol LAÃ , the work done by the
external forces of the problem A when we engrave to their points of application the
displacements that such points have in the problem Ã. So
 
n 
n
LAÃ = (p dz) ṽ + Fi ṽ(zi ) + Mi ϕ̃(zi ) (4.2.59)
H i=1 i=1
 n n
LÃA = (p̃ dz) v + F̃i v(zi ) + M̃i ϕ(zi ). (4.2.60)
H i=1 i=1

It subsists the following result, obtained by Betti

[4.2.17] The mutual works are equal.
Proof We apply the virtual work principle to problems A and Ã. Assuming the
problem à as system of displacement and the problem A as system of forces, we
have
 
n 
n
p ṽ dz + Fi ṽ(zi ) + Mi ϕ̃(zi ) = Lve
H i=1 i=1
 (4.2.61)
M M̃
= Lvi = dz.
H EI

Assuming the problem à as system of forces and the problem A as system of

displacement, we have
 
n 
n
p̃ v dz + F̃i v(zi ) + M̃i ϕ(zi ) = Lve
H i=1 i=1
 (4.2.62)
M̃ M
= Lvi = dz.
H EI
From the (4.2.59) and (4.2.61) it follows

M̃ M
LAÃ = dz. (4.2.63)
H EI
and from the (4.2.60) and (4.2.62) it follows

M̃ M
LÃA = dz. (4.2.64)
H EI
468 4 The One-Dimensional Problems

From the (4.2.63) and (4.2.64) the thesis follows. 

Remark 4.2.14 Let us notice that the [4.2.17] can be proved by another reasoning.
We denote with  the problem constituted by the deflected beam of problem A sub-
mitted at the final instant tf to the distributed load p + p̃, to the n concentrated forces
F1 + F̃1 , . . . , Fn + F̃n and to the n concentrated couples M1 + M̃1 , . . . , Mn + M̃n .
Evidently the load of problem A is sum of the load of problem  and of the load of
problem Ã. Since the [4.1.12] the strain work L in problem A is

1 1
n n
1
L= p v dz + Fi v(zi ) + Mi ϕ(zi ),
2 H 2 2
i=1 i=1

the strain work L̃ in problem à is


1 1
n n
1
L̃ = p̃ ṽ dz + F̃i ṽ(zi ) + M̃i ϕ̃(zi ).
2 H 2 2
i=1 i=1

Since the (4.2.57) we can value the strain work L̂ in problem  by any load
path. In a first load path is applied on the deflected beam the load p, F1 , . . . , Fn ,
M1 , . . . , Mn of the problem A. Under this load the beam assumes the deformed
configuration v and the strain work L is done. Then, maintaining unchanged the
load p, F1 , . . . , Fn , M1 , . . . , Mn , we apply on the deformed beam the load p̃,
F̃1 , . . . , F̃n , M̃1 , . . . , M̃n of the problem Ã. As underlined in Remark 4.2.7, the
beam increases its deformed configuration just of ṽ, assuming, as required by the
principle of superposition, the elastic line v + ṽ. Since during this whole second
phase the forces p, F1 , . . . , Fn , M1 , . . . , Mn have constant intensity, they do the
work
 
n 
n
p ṽ dz + Fi ṽ(zi ) + Mi ϕ̃(zi )
H i=1 i=1

that is just the mutual work LAÃ . Besides in the second phase of application the load
p̃, F̃1 , . . . , F̃n , M̃1 , . . . , M̃n is statically applied. Then, since the [4.2.12], the work
L̃ is also done. This way, calculating the strain work of the problem  according to
this first load path, the strain work

L̂ = L + LAÃ + L̃ (4.2.65)

is done.
In a second load path is applied on the deflected beam the load p̃, F̃1 , . . . , F̃n ,
M̃1 , . . . , M̃n of the problem Ã. Under this load the beam assumes the deformed
configuration ṽ and the strain work L̃ is done. Then, maintaining unchanged the
load p̃, F̃1 , . . . , F̃n , M̃1 , . . . , M̃n , we apply on the deformed beam the load p,
F1 , . . . , Fn , M1 , . . . , Mn of the problem A. As underlined in Remark 4.2.7, the
4.2 Beams 469

beam increases its deformed configuration just of v, assuming, as required by

the principle of superposition, the elastic line v + ṽ. Since during this whole
second phase the forces p̃, F̃1 , . . . , F̃n , M̃1 , . . . , M̃n have constant intensity, they do
the work
 
n 
n
p̃ v dz + F̃i v(zi ) + M̃i ϕ(zi )
H i=1 i=1

which is just the mutual work LÃA Besides in the second phase of application the
load p, F1 , . . . , Fn , M1 , . . . , Mn is statically applied. Then, since the [4.2.12],
the work L is also done. This way, calculating the strain work of the problem Â
according to this second load path, the strain work

L̂ = L̃ + LÃA + L (4.2.66)

is done.
From the (4.2.65) and (4.2.66) the thesis follows. 

Problem 4.2.14 You shall determine the mutual works among the problem A of Fig.
4.2.77 and the problem à of Fig. 4.2.78.
Solution Evidently

LAÃ = F ṽ(H); LÃA = M ϕ(K).

Furthermore in problem à the cross section H raises of

M l2
ṽ(H) = ;
2 EI
in problem A the cross section K rotates in verse clockwise of

F l2
ϕ(K) = .
2 EI

Fig. 4.2.77

Fig. 4.2.78
470 4 The One-Dimensional Problems

So

M l2 F l2
LAÃ = −(F) ; LÃA = −(M) . 
2 EI 2 EI

Problem 4.2.15 You shall determine the lowering of the cross section K in the
problem of Fig. 4.2.77.

Solution We assume the problem of Fig. 4.2.77 [resp. 4.2.79] as problem A [resp.
Ã]. Clearly it results

F l3 (Fl) l2
ṽ(H) = + ,
3 EI 2 EI
and from [4.2.17] we get

F v(K) = LÃA = LAÃ = F ṽ(H).

So the cross section K of Fig. 4.2.77 and the cross section H of Fig. 4.2.79 lower
of the same value

F l3 F l3
v(K) = ṽ(H) = + . 
3 EI 2 EI

The Castigliano’s theorem, that we have proven in the three-dimensional case, is

also valid for the deflected beam. It results

[4.2.18] The partial derivative of the strain work L with respect to an external force
F concentrated in a cross section S of a deflected beam is equal to the lowering v(S):

∂L
= v(S).
∂F

Analogously:
∂L
= ϕ(S). 
∂M
Problem 4.2.16 You shall determine the lowering of the free end of the cantilever of
Fig. 4.2.80. The cross section is constant.

Fig. 4.2.79
4.2 Beams 471

Fig. 4.2.80

Fig. 4.2.81

Solution From the (4.2.58) it follows, in the system of coordinates of Fig. 4.2.81
 l  l  l
M2 1 F2 F 2 l3
L= dz = (− Fz)2 dz = z2 dz = .
0 2EI 2EI 0 2EI 0 6EI

So from the [4.2.18] we get

∂L ∂ F 2 l3 Fl3
v(A) = = = .
∂F ∂F 6EI 3EI

We notice that if F is downward it results F > 0 and then v(A) > 0 and then A
lowers. 

Problem 4.2.17 You shall determine the rotation of the free end of the cantilever of
Fig. 4.2.82. The cross section is constant.

Solution From the (4.2.58) it follows, in the system of coordinates of Fig. 4.2.81
 
l M2 1 l M2 l
L= dz = (− M)2 dz = .
0 2 EI 2 EI 0 2 EI

So from the [4.2.18] we get

∂L ∂ M2 l Ml
ϕ(A) = = = . 
∂M ∂M 2 EI EI

Fig. 4.2.82
472 4 The One-Dimensional Problems

Fig. 4.2.83

Problem 4.2.18 You shall determine the rotation of the free end of the cantilever of
Fig. 4.2.80. The cross section is constant.
Solution We preliminarily examine the problem of Fig. 4.2.83, for which we deter-
mine ϕ(A). Then, putting M = 0, we obtain the rotation of the free end of the
cantilever of Fig. 4.2.80. For the problem of Fig. 4.2.83 it follows from the (4.2.58),
in the system of coordinates of Fig. 4.2.81
 
l M2 1 l F 2 l3 M2 l FMl2
L= dz = (− Fz − M)2 dz = + + .
0 2 EI 2 EI 0 6 EI 2 EI 2 EI

So from the [4.2.18] we get

∂L ∂ F 2 l3 M2 l FMl2 Ml Fl2
ϕ(A) = = + + = + .
∂M ∂M 6 EI 2 EI 2 EI EI 2 EI

This result is obviously valid for any value of M. So, putting M = 0, we obtain
for the rotation of the free end of the cantilever of Fig. 4.2.80 the value

Fl2
ϕ(A) = . 
2 EI

Problem 4.2.19 You shall determine the lowering of the cross section B in the
cantilever of Fig. 4.2.84. The cross section is constant.
Solution Proceeding as in the Problem 4.2.18, we preliminarily examine the problem
of Fig. 4.2.85, for which we determine v(B). For such problem it follows from the
(4.2.58), in the system of coordinates of Fig. 4.2.86

  0  l
M2 (− M)2 (− M − Fz)2 M2 l F 2 l3 FMl2
L= dz = dz + dz = + + .
str 2 EI −l 2 EI 0 2 EI EI 6 EI 2 EI

Fig. 4.2.84

Fig. 4.2.85
4.2 Beams 473

Fig. 4.2.86

So for the problem of Fig. 4.2.85 from the [4.2.18] we get

∂L ∂ M2 l F 2 l3 FMl2 Fl3 Ml2

v(B) = = + + = + .
∂F ∂F EI 6 EI 2 EI 3EI 2 EI

So, putting F = 0 in the previous expression, we obtain that the cross section B
in the cantilever of Fig. 4.2.84 lowers of

Ml2
v(B) = . 
2 EI
Problem 4.2.20 You shall determine the rotation of the cross section A in the
cantilever of Fig. 4.2.87. The cross section is constant.
Solution Proceeding as in the Problem 4.2.18, we preliminarily examine the problem
of Fig. 4.2.88, for which we determine ϕ(A). For such problem it follows from the
(4.2.58), in the system of coordinates of Fig. 4.2.81, that
  2
l M2 1 l p z2 M2 l p2 l5 Mp l3
L= dz = −M − dz = + + .
0 2 EI 2 EI 0 2 2 EI 40 EI 6 EI

So

∂L ∂ M2 l p2 l5 Mp l3 Ml p l3
ϕ(A) = = + + = + .
∂M ∂M 2 EI 40 EI 6 EI EI 6 EI

As a consequence the rotation of the cross section A in the cantilever of

Fig. 4.2.87 is

p l3
ϕ(A) = . 
6 EI

Fig. 4.2.87

Fig. 4.2.88
474 4 The One-Dimensional Problems

An interesting consequence of the Betti’s theorem (or of the more general prin-
ciple of virtual works) they are the influence lines. We will say that a vertical
concentrated force is travelling when F can be applied in any section of the deflected
beam (Fig. 4.2.89). We denote in Fig. 4.2.89 with RC the action of the bogie C on
the deflected beam. We call line of influence of the reaction RC for vertical travel-
ling force F a function f whose value f(z) in the generic cross section S of abscissa
z coincides with the value that has the reaction RC when F is applied in the cross
section S. In perfect analogy we define the line of influence of the bending moment
in S for the travelling vertical force and the line of influence of the rotation of the
section S for the travelling vertical force.
Problem 4.2.21 You shall determine, for the deflected beam of Fig. 4.2.89, the line
of influence of the reaction RC of the bogie C for the travelling vertical force F.

Solution We consider the problem of Fig. 4.2.90, that since the dissection principle
is equivalent to that of Fig. 4.2.89. We also consider the problem of Fig. 4.2.91,
in which to the beam of Fig. 4.2.90 (what has one degree of freedom) a sys-
tem of rigid displacements w, gotten by lowering the hinge D of 2F, is imposed.
We apply the principle of the virtual works to the beam of Fig. 4.2.90, assuming
Fig. 4.2.91 as system of the displacements and Fig. 4.2.90 as system of the forces.
We have

F w(S) − RC w(C) = 0

so that

2F
F w(S) − RC =0
2

Fig. 4.2.89

Fig. 4.2.90

Fig. 4.2.91
4.2 Beams 475

so that
w(S) = RC .

This way the diagram w of Fig. 4.2.91 is the required line of influence. We get
that the greatest reaction that can expound the bogie is RC = 2F and that such event
is verified when the travelling force F is applied on the hinge D. 

Problem 4.2.22 You shall determine, for the deflected beam of Fig. 4.2.89, the line of
influence of the bending moment in the supported cross section C for the travelling
vertical force F.

Solution We consider the problem of Fig. 4.2.92, that since the dissection principle
is equivalent to that of Fig. 4.2.89. We also consider the problem of Fig. 4.2.93,
in which a system of rigid displacements w, gotten lowering the hinge D of Fl, is
imposed to the beam of Fig. 4.2.92 (what has one degree of freedom). We apply the
principle of the virtual works to the beam of Fig. 4.2.92, assuming Fig. 4.2.93 as
system of the displacements and Fig. 4.2.92 as system of the forces. We have

Fl
− (M(C)) + (F) (w(S)) = 0
l
from which

w(S) = M(C),

so that the diagram of Fig. 4.2.93 is the required line of influence. We notice that
the maximum bending moment that we can have in the supported cross section C is
Fl and that such value is reached when the travelling force F is applied on the hinge
D. Moreover if F is applied in a cross section of the trunk AC, then M(C) = 0. 

Problem 4.2.23 You shall determine, for the cantilever of Fig. 4.2.94, the line of
influence of the lowering of the cross section B for the travelling vertical force F.

Fig. 4.2.92

Fig. 4.2.93
476 4 The One-Dimensional Problems

Fig. 4.2.94

Fig. 4.2.95

Solution We consider the problem à [resp. A] of Fig. 4.2.95 [resp. 4.2.94] and denote
with w [resp. v] its elastic line. From the [4.2.17] we get

F v(B) = LÃA = LAÃ = F w(S).

Insofar w(S) coincides with the lowering that is verified in B when the force F is
applied in S (Fig. 4.2.94). So w is the required line of influence (Fig. 4.2.96). Insofar,
as physically obvious, the maximum lowering of B is verified when F is applied at
the free extreme. 

Problem 4.2.24 You shall determine, for the deflected beam of Fig. 4.2.97, the line
of influence of the rotation of the cross section B for the travelling vertical force F.
Solution We consider the problem à [resp. A] of Fig. 4.2.98 [resp. 4.2.95] and denote
with w [resp. ϕ] its elastic line [resp. rotation function]. From the [4.2.17] we get

F w(S) = LAÃ = LÃA = F ϕ(B).

Fig. 4.2.96

Fig. 4.2.97

Fig. 4.2.98
4.2 Beams 477

Insofar w(S) has the same value of the rotation suffered by the cross section B
when the force F is applied in S (Fig. 4.2.97). The required line of influence is the
elastic line w of the problem of Fig. 4.2.98. 

Problem 4.2.25 You shall determine, for the deflected beam of Fig. 4.2.99, the line
of influence of the fixed end moment M(A) for the travelling vertical force F.
Solution We consider the problem à [resp. A] of Fig. 4.2.100 [resp. 4.2.101] and
denote with w [resp. ϕ] its elastic line [resp. rotation function]. We notice that since
the dissection principle the problems 4.2.100 and 4.2.99 are equivalent and then
ϕ(A) = 0. So from the [4.2.17] we get

3FEI l
−M(A)F + F w(S) = −M(A) + F w(S)
l 3EI

dw 3FEI
= −M(A) (S) + F w(S) = LAÃ = LÃA = .ϕ(A)
dz l
3FEI
= ·0=0
l

from which

M(A) = w(S).

Insofar w(S) has the same value of the moment of the fixed end A when the force
F is applied in S (Fig. 4.2.101). The required line of influence is the elastic line w of
the problem of Fig. 4.2.100. 

Problem 4.2.26 You shall determine, for the deflected beam of Fig. 4.2.102, the line
of influence of the bending moment M(B) in the cross section B for the travelling
vertical force F.
Solution We consider the problem à of Fig. 4.2.103 and denote with w [resp. ψ] its
elastic line [resp. rotation function]. We consider the problem A of Fig. 4.2.104 and

Fig. 4.2.99

Fig. 4.2.100

Fig. 4.2.101
478 4 The One-Dimensional Problems

Fig. 4.2.102

Fig. 4.2.103

Fig. 4.2.104

denote with ϕ its rotation function. We notice that, since the dissection principle, the
problems 4.2.102 and 4.2.104 are equivalent. Then, denoting with ϕBs Bd the relative
rotation between Bs and Bd , we have ϕBs Bd = ϕ(Bd ) − ϕ(Bs ) = 0. So from the
[4.2.17] we get

M(B)ψBs Bd − F w(S) = LAÃ = LÃA = M · ϕBs Bd = 0

so that
F
M(B) = w(S).
ψBs Bd

Insofar, if the couples M in Fig. 4.2.103 has intensity such that

ψBs Bd = F,

then w(S) has the same value of the bending moment that arises in the cross section
B in Fig. 4.2.102 when the force F is applied in S. So the required line of influence
is the elastic line w of the problem of Fig. 4.2.103. 

Remark 4.2.15 Let us again consider the deflected beam of Fig. 4.2.99, for which
we have determined the line of influence of the fixed end moment M(A) for the
travelling vertical force F (Fig. 4.2.100). We notice that, since the superposition
principle, such line of influence allows the immediate evaluation of M(A) for any
condition of load. For instance, in the case of Fig. 4.2.105 it result
4.2 Beams 479
 K
F1 F2 pw
M (A) = w (S1 ) + w (S2 ) + dz.
F F H F

If the load is uniformly distributed, the integral precedent coincides, to less than
a constant factor, with the area outlined in Fig. 4.2.105. This observation allows, in
numerous cases of practical interest, to determine the more unfavorable conditions
of load. For instance, if the continuous beam of Fig. 4.2.106 has the line of influence
shown in the same Fig. 4.2.106, it is evident that Figs. 4.2.107 and 4.2.108 (where an
effect of opposite sign is induced) are the more unfavorable conditions of uniformly
distributed load. 

Remark 4.2.16 Let us consider the cantilever of Fig. 4.2.24 and assume the frame of
reference of Fig. 4.2.26. We denote with W the set
   
3 z 2 1  z 3
W = ṽ = − δ̃: δ̃ ∈ , δ̃  1 .
2 l 2 l

Clearly W is a subset of the set of all the compatible elastic line of the cantilever
in examination and δ̃ = ṽ(l). Well we consider the functional Lv of the work

1
Lv :ṽ ∈ W → F δ̃,
2

Fig. 4.2.105

Fig. 4.2.106

Fig. 4.2.107

Fig. 4.2.108
480 4 The One-Dimensional Problems

the functional Ev of the energy

 l  l 2
1 EI d2 ṽ 3 EI 2
Ev :ṽ ∈ W → M̃ dϕ̃ = dz = δ̃ ,
2 0 2 0 dz2 2 l3

the functional Epv of the potential energy

3 EI 2
Epv :ṽ ∈ W → Ev (ṽ) − F δ̃ = δ̃ − F δ̃.
2 l3
We notice that in solution the (4.2.57) is true and then it must be (Fig. 4.2.109)

F l3
δ̃ = .
3 EI

d Epv
We also notice that in solution, since the [4.2.15], it must be = 0 and then
dδ̃
(Fig. 4.2.109)
F l3
δ̃ = . 
3 EI

Remark 4.2.17 The mathematical model [4.2.1] correctly simulates the behavior of
a deflected beam if it is one-dimensional. If however the beam is stubby it needs
to also take into account the effect of the shear. You get then a new mathematical
model, more precise than the precedent, for which the theorem of existence and
uniqueness of the solution still subsist. The expression (4.2.58) of the strain energy,
obtained from the elastic potential only computing the bending term
 
1 1 M2
σz εz dV = dz,
V 2 2 H EI

must be corrected by adding the shear term

 
1  τz2
τzx γzx + τzy γzy dV = dV
V 2 V 2G

which we write, taking into account the (2.6.29)

Fig. 4.2.109
4.2 Beams 481


T2
χ dz. (4.2.67)
H 2GA

In the additional term (4.2.67) χ is called shear factor and, putting I = A!2 ,
obviously is equal to

1 Sn2
χ= dA. (4.2.68)
A!4 A c2
Clearly the shear factor is adimensional and only depends from the shape of the
cross section. Furthermore we can prove that always χ > 1.
From the (4.2.67) we obtain that the energy aliquot relative to shear is
 
1 T 1
T χ dz = T dv
H 2 GA H 2
so that the aliquot dv relative to shear is

T
dv = χ dz
GA
and then the aliquot of curvature due to shear is

d2 v χ dT χ
2
= =− p. (4.2.69)
dz GA dz GA
If the beam is stubby the term (4.2.69) is not negligible with respect to the
bending term

M
− .
EI
In such case the differential equation (4.2.11) becomes, taking into account the
(4.2.1)

d2 v M χ
2
=− − p. 
dz EI GA

4.2.5 Deflected Beams on Elastic Foundation

In the Civil engineering the deflected beam is often used as a structure of foundation,
that directly lies on the soil. The simplest but effective law that we can assume to
simulate the intensity of the reaction R (expressed in t/m) of the subgrade on the
deflected beam is due to Winkler. He proposed the expression

R=kv (4.2.70)
482 4 The One-Dimensional Problems

where the positive constant k is called modulus of the foundation and it depends
largely on the properties of the subgrade4.2.8 . In the (4.2.70) we suppose, in a
simplified analysis, that R can have any verse.
Clearly, denoting with A the interval of  interested by the deflected beam, its
behavior is simulated by a problem obtained modifying the differential equation of
the problem [4.2.1]. Precisely the modified equation is:

d4 v p − kv
4
= on A (4.2.71)
dz EI
and the mathematical model of the deflected foundation beam is the boundary
problem:
[4.2.19] In the hypothesis that the known term q is regular, find a real function
v defined in A, equipped in A with partial derivatives at least up to those of the
fourth order and satisfying in A the differential equation (4.2.71) and the boundary
conditions in the points constrained. 
For the problem [4.2.19] we can prove the following theorem of existence and
uniqueness:
[4.2.20] The problem [4.2.19] admits an unique solution. 
Evidently the problem [4.2.19] is mathematically linear. So
[4.2.21] For the problem [4.2.19] the principle of superposition holds. 
Remark 4.2.18 Another effective law that we can assume to simulate the intensity
of the reaction R of the subgrade on the deflected beam is due to Pasternak. He
proposed the expression

d2 v
R = −h +k v
dz2

where the positive constants h, k are called modulus of the soil. Also in this model
we suppose, in a simplified analysis, that R can have any verse. 

4.3 Arches

4.3.1 Arches with Small Curvature

We call arch a beam whose axis ρ is a regular curve of 2 contained in a plane α
(Fig. 4.3.1). Clearly in any point of ρ there is one and only one tangent line t and

4.2.8 For
a soil rich of sand k varies from 1 to 4 kg/cm2 ; for a soil rich of gravel k varies from 4 to
12 kg/cm2 ; for a soil rich of compact or cemented gravel k can assume great values.
4.3 Arches 483

Fig. 4.3.1

one and only one normal line n. We call cross section S of the arch (in a generic
point P of ρ) the intersection between the arch and the plane passing through P and
normal to t (Fig. 4.3.1). We suppose that all the cross sections of the arch have a
principal axis of inertia lying in the plane α, that contains the distributed load f too.
So the arch is in right bending.
Let us consider a generic cross section S of the arch and denote with N the axial
force in S according t, with T the shear in S according n, with M the bending moment
in S (Fig. 4.3.2). We assume that N is the resultant of σ distributed on S according
the (2.4.1), that T is the resultant of τ distributed on S according the (2.6.29), that M
is the resultant of σ distributed on S according the (2.3.1).
Let us denote with s the curvilinear abscissa on ρ and consider the element of
the arch individualized by two very near cross sections S, S1 (Fig. 4.3.2). By cor-
respondence an infinitesimal trunk of ρ is individualized, of length ds. Obviously
in the initial configuration, if we denote with dϑ the angle between S and S1 and
with r10 the curvature of ρ at the point P individualizing S, we have r0 dϑ = ds.
Furthermore we denote with p [resp. q] the orthogonal component of the distributed
load f according t [resp. n].
Let us impose now the equilibrium of the infinitesimal element S, S1 of the arch
(Fig. 4.3.2). Imposing the equilibrium to the translation according the tangent t, we
get

dϑ dϑ dϑ dϑ
(N + dN) cos − Ncos − Tsin − (T + dT) sin + pds = 0
2 2 2 2
from which, neglecting the second order infinitesimal ones, it follows

dN − T dϑ + pds = 0

Fig. 4.3.2
484 4 The One-Dimensional Problems

and finally

dN
T −r = r p. (4.3.1)
ds
Imposing the equilibrium to the translation according the normal n, we get

dϑ dϑ dϑ dϑ
(N + dN) sin + Nsin − Tcos + (T + dT) cos + qds = 0
2 2 2 2
from which, neglecting the second order infinitesimal ones, it follows

Ndϑ + dT + qds = 0

and finally

dT
r + N = r q. (4.3.2)
ds
Imposing the equilibrium to the rotation around the point O, we get

ds ds
M − (M + dM) + T + (T + dT) =0
2 2
from which, neglecting the second order infinitesimal ones, it follows

dM
= T. (4.3.3)
ds
Remark 4.3.1 Clearly the (4.3.1), (4.3.2), and (4.3.3) are valid for arches having
either small or great curvature. 

Remark 4.3.2 Evidently the (4.3.1), (4.3.2), and (4.3.3) are the generalization of the
(4.2.1) and (4.2.2). 

Remark 4.3.3 If the arch is circular or a ring, we have r0 = const. In such condition
from the (4.3.1), (4.3.2), and (4.3.3) we immediately get

d3 M dM dq
+ = r02 p + . (4.3.4)
dϑ 3 dϑ dϑ

So the bending moment M is given from the general integral M (ϑ) = c1 sinϑ +
c2 cosϑ + M0 , where M0 is a particular integral of the (4.3.4). 

The deformation of a curved beam is more complex than that of the rectilinear
beam because the generic cross section doesn’t move it orthogonally with respect to
the axis.
In the first place we will consider the case of arches with small curvature, that
is having the average dimension h of the generic cross section small enough in
4.3 Arches 485

comparison to the radius r of curvature. Clearly in such simpler case, more frequent
in the Civil engineering, every elementary trunk ds of the arch can be treated as
prismatic and then admits the application of the Saint Venant’s results.
So, since the (4.2.11), in any elementary trunk ds of the arch (Fig. 4.3.2) after the
application of the load we have an infinitesimal variation d (dϑ) of the infinitesimal
angle dϑ (between the cross sections S, S1 ) given by

M
d (dϑ) = ds. (4.3.5)
EI
Analogously, since the (2.4.6), after the application of the load we have an
infinitesimal variation d (ds), according t, of the infinitesimal length ds of the trunk
of Fig. 4.3.2, given by

N
d (ds) = ds. (4.3.6)
EA
Furthermore, since the (4.2.69), after the application of the load we have an
infinitesimal relative displacement dv, according n, of the cross sections S, S1 of
the infinitesimal trunk of length ds of Fig. 4.3.2, given by

T
dv = χ ds. (4.3.7)
GA

Let us consider now any cross section A [resp. S] [resp. B] of the arch and denote
(Fig. 4.3.3):

– with a [resp. s] [resp. b] its curvilinear abscissa,

– with ϑ(A) [resp. ϑ = ϑ(S)] [resp. ϑ(B)] the angle between the x axis and the
tangent to the curved axis ρ at A [resp. S] [resp. B], before the deformation
(Fig. 4.3.1),

Fig. 4.3.3

Fig. 4.3.4
486 4 The One-Dimensional Problems

– with A [resp. S ] [resp. B ] the deformed configuration of the cross section A [resp.
S] [resp. B],
    
– with ϑ A [resp. ϑ S ] [resp. ϑ B ] the angle between the x axis and the tangent
to ρ  at A [resp. S ] [resp. B ], after the deformation,
– with (xA , yA ) [resp. (x, y)] [resp. (xB , yB )] the coordinates of A [resp. S] [resp. B]
in the cartesian
 frame of reference
 O,
 x, y, 
– with xA , yA [resp. x , y ] [resp. xB , yB ] the coordinates of A [resp. S [resp.
B ] in the cartesian frame of reference O, x, y.

It is of technical interest the calculation of the relative displacement and rotation

of the cross sections A, B. Obviously such greatness are due to the deformation of
all the elementary trunks ds included among B and A (Fig. 4.3.4.).
The infinitesimal trunk ds has infinitesimal dimension according x

dx = ds cosϑ

and infinitesimal dimension according y

dy = ds sinϑ.

The infinitesimal dimension dx earns, during the deformation, an infinitesimal

variation d (dx) by a bending term, by an extensional term and by a shearing term.
The bending term is clearly originated by the (4.3.5). Exactly we have
   
d (dx) = ϑ S − ϑ(S) ds sinϑ
 s 
M   (4.3.8)
= dt + ϑ A − ϑ(A) sinϑ ds.
a EI

The extensional term is clearly originated by the (4.3.6). Easily we have

N Ncosϑ
d (dx) = ds cosϑ = ds. (4.3.9)
EA EA
The shearing term is clearly originated by the (4.3.7). Easily we have

T Tsinϑ
d (dx) = χ ds sinϑ = χ ds. (4.3.10)
GA GA
So, since the (4.3.8), (4.3.9), and (4.3.10) we get
 
xB − xA − [xB − xA ]
 b  s   b  b
M   Ncosϑ Tsinϑ
= dt + ϑ A − ϑ(A) sinϑ ds + ds + χ ds,
a a EI a EA a GA
(4.3.11)
4.3 Arches 487
   
yB − yA − yB − yA
 b  s   b  b
M   Nsinϑ Tcosϑ
= dt + ϑ A − ϑ(A) cosϑ ds + ds + χ ds,
a a EI a EA a GA
(4.3.12)

        b M
ϑ B − ϑ(B) − ϑ A − ϑ(A) = ds. (4.3.13)
a EI

Remark 4.3.4 If the funicular of the loads is near to the arch axis ρ, the contribution
of N is not negligible. Furthermore, if the arch is stubby, we must consider the
contribution of all terms. 

Problem 4.3.1 You shall determine the horizontal variation of the diameter 2 r of
the half ring of Fig. 4.3.5.
Solution Obviously we can execute the calculation employing the scheme of
Fig. 4.3.6. Since N = 0 and T = 0 identically, using the (4.3.11) we easily obtain
that the required variation is equal to
 πr   π π 
M s M r2 2 Mπ r2
dt cosϑ ds = − ϑ cosϑ dϑ = . 
0 EI 0 EI − π2 2 EI

In the analysis of an arch it is very advantageous to employ the funicular of the

loads discussed in Sect. 4.1.1. In fact the funicular curve allows to immediately
calculate in every cross section M, N, T. Furthermore the first side and the last side
of the funicular furnish the lines of action of the constraining reactions. As seen in
Sect. 4.1.1, their value and verse can be read on the polygon of the forces.
We expressly notice that if the loads are vertical, as it usually happens in the case
of an arch built in masonry, the constraining reactions are obligatorily oblique and
then the arch, whatever its form, always acts on the constraints with some thrusts
that have horizontal component H non zero (Fig. 4.3.7).

Fig. 4.3.5

Fig. 4.3.6
488 4 The One-Dimensional Problems

Fig. 4.3.7

Remark 4.3.5 If the arch is built in masonry, obviously every cross section must
be entirely pressed. As a consequence the funicular of the loads must be inside, in
every cross section, to the inertia centroidal kernel of the cross section. This can
be gotten shaping the arch according to the funicular of the loads or adding special
loads that exclusively have this function. We point out that if during an operation of
static reinforcing these particular weights are removed, part of the masonry enters
in traction with consequent collapse of the arch. 

4.3.2 Arches with Great Curvature

Many machine members have the shape of an arch with great curvature. In such
case in any cross section the arch has a radius of curvature R comparable with the
middle dimension of the cross section (Fig. 4.3.8). A very important consequence
is the different length ds of the fibers of any infinitesimal trunk (Fig. 4.3.8). On the
contrary, in the arch with small curvature the fibers of an infinitesimal trunk have
practically all the same length.
In the analysis of the static behavior of the arc with great curvature, both in the
case of centroidal axial load and in the case of the bending, we still suppose that the
cross sections are preserved plane.
This despite, as we will see in a short while, because of the aforesaid variation of
the length of the fibers of a same infinitesimal trunk, in the bending in every cross
section S, the σ z , where z is normal to S, doesn’t vary more linearly with the distance
from the neutral axis.

Fig. 4.3.8
4.3 Arches 489

Fig. 4.3.9

In fact let us consider an infinitesimal trunk SS1 of an arch submitted to bending

and denote with dϑ the angle between S and S1 (Fig. 4.3.9). We assume that the
deformed configuration of the cross section S [resp. S1 ] is a plane S’ [resp. S1 ]
and we denote with dϑ  the angle between S’ and S1 . Furthermore we put ddϑ =
dϑ  − dϑ.
We consider the cross section S and suppose that y is a principal direction of
inertia, so that the bending is right (Fig. 4.3.10). Finally we assume that S ∩ S is
a line n of the plane S, passing through a point Q = (0, c) of y and parallel to x
(Fig. 4.3.10), where c for the time being is unknown.
Clearly the fiber of SS1 individualized by the point P = (x, y) of S has, before
the deformation, length ds = (R − y) dϑ (Fig. 4.3.9) and suffers, because of the
deformation, the variation of length dds = (y − c) ddϑ (Fig. 4.3.10). 
Consequently

dds y−c
εz = =ω , (4.3.14)
ds R−y

where

ddϑ
ω= .
dϑ
Supposing σx = σy = 0, from (4.3.14) it follows

y−c
σz = ω E , (4.3.15)
R−y

so that n is the neutral axis of the bending.

We see from the (4.3.15) that σ z doesn’t linearly vary with the distance from the
neutral axis n (Fig. 4.3.11).

Fig. 4.3.10
490 4 The One-Dimensional Problems

Fig. 4.3.11

To operate with the (4.3.15), we impose that the σ z distributed on the area A of
the cross section S have resultant

σz dA = 0
A

getting
- y
dA
A R−y
c= - .
1
dA
A R−y

After that, imposing that the same distribution of σ z have, with respect to axis n,
the resultant moment M, we have

Eω σz (y − c) dA = M,
A

from which
M
ω= - .
(y − c)
E dA
A R−Y
Let us consider now the problem of the arch submitted to a centroidal axial load
N. We preliminarily observe that if the position of S is obtained rotating S around
O (Fig. 4.3.9), then (Fig. 4.3.12)

dds ddϑ(R − y)
εz = =
ds dϑ(R − y)

Fig. 4.3.12
4.3 Arches 491

Fig. 4.3.13

Fig. 4.3.14

so that σz = const, so that

N
σz = . (4.3.16)
A

Obviously the distribution (4.3.16) is equivalent to the centroidal axial load N.

Problem 4.3.2 You shall determine the σz in the cross section a of a lifting hook sub-
mitted to a load P = 5,000 kg (Fig. 4.3.13). The cross section a has the dimensions
of Fig. 4.3.14.
Solution We calculate from the Fig. 4.3.13 the area A and then determine G, from
which it follows the radius of curvature R. We transport P in G and then calculate
the σ z as sum of the (4.3.15) and (4.3.16).
Chapter 5
Thermoelasticity

5.1 Mechanics of Continuous Media

5.1.1 Introduction

Let us consider now the problem of simulating mathematically the behavior of a

solid body submitted simultaneously to external loads and thermal load. In this
analysis the Thermodynamics collaborate with the Mechanics holding a greater role
than the one played in the ancient Theory of elasticity. In fact a new type of problem
comes into play, that of the Transmission of the heat in the solid bodies.
Also in the presence of thermal effects, the analysis of the practical problems can
be conducted simulating the solid body with a continuous media. In rigor instead,
as it is well known in the Atomic physics, the volume occupied by the present atoms
in a body is alone a dwarfish part of the volume of the body. For instance, in a steel
piece in a solid state, the relationship is of the order of 10−14 .
In Thermoelasticity it can happen that, if the material remains in the linearly
elastic field, non negligible phenomena of dissipation of the mechanical energy are
verified. In such case to build a suitable mathematical model of the problem it is
necessary to resort to the Thermodynamics of the irreversible processes. In fact if a
mechanical external cause produces variations of the state of deformation in a body,
these are generally accompanied by variations of the temperature and consequently
from a flow of heat. The whole process results in an increase of entropy and therefore
in an increase of the energy stored in a way that is not mechanically recoverable.
This phenomenon is known as thermoelastic dissipation. It is brought in account in
the thermoelastic mathematical model by a term that appears in the equation of the
heat. As we will see, such term, said of joining, can be neglected for all the problems
except those in which the velocity of the deformation and that of the variation of the
temperature have the same order of greatness. Fortunately in the greatest part of the
usual practical applications the joining term is negligible. We will use to say, in such
cases, that the thermoelastic problem is uncoupled.
In the case, still simpler and frequent in the practice, in which it is possi-
ble to neglect the inertial terms in the equations of equilibrium, the problem is

A. Maceri, Theory of Elasticity, DOI 10.1007/978-3-642-11392-5_5, 493


C Springer-Verlag Berlin Heidelberg 2010
494 5 Thermoelasticity

divided into two separate matters, that we simply call Transmission of the heat and
Thermoelasticity. In this last case we can determine in a first phase the distribu-
tion of temperature that arises in the body under a thermal load assigned to the
contour. Then, in the second phase, we determine the state of stress induced in the
body by the applied external forces and by the distribution of temperature previously
determined.
To draw in a clear way the fundamental equation of the thermoelastic boundary
problem we use a method of analysis peculiar of the Fluid dynamics, which will be
explained in detail.

5.1.2 Classical Thermodynamics

We call system a portion of the three-dimensional space occupied by matter, envi-
ronment the rest of the universe. We denote with V(t) the portion of space occupied
by the system at the instant t∈ ]ti , tf [ and with Σ(t) its frontier, that is the surface
of the system. The symbol ti [resp. tf ] denotes the initial [resp. final] instant of the
interval of time in which it interests to analyze the phenomenon.
We formulate a theory of the classical Thermodynamics by postulates. We first
of all postulate
[5.1.1] A system can exchange mass and energy with its environment. 
In the [5.1.1] we must intend the mass in its different forms (i.e. in its different
states of aggregation) and the energy in its different forms (that is potential, kinetics,
electromagnetic, etc.). The postulate [5.1.1] affirms that everything that the system
exchanges with the environment is in partnership to an exchange of mass and/or
to an exchange of energy. For instance the exchange of mechanical momentum is
associated to an exchange of mass; the exchange of electromagnetic momentum is
associated to an exchange of energy.
Clearly this first postulate of the Thermodynamics is only valid out of the rela-
tivistic field. In fact in a relativistic field it is not possible to independently exchange
mass and energy.
We notice that this first postulation of the Thermodynamics implicates that the
system can exchange with the environment either mass and energy or energy but
not mass. In fact the mass exchanged carries with itself (also in non relativistic
field) its energy.
A greatness is said extensive [resp. intensive] when its numerical value depends
[resp. it doesn’t depend] from the- extension of the system. Evidently if G is an
extensive greatness it results G = v g dV, where g is the density of G.
Let us consider an extensive greatness G, a surface Λ(t), a point P(t) of Λ(t) and
an instant t ∈ ]ti , tf [. We call punctual (or local) flow of G through Λ(t) in P(t) at the
instant t, the quantity ϕG of G that crosses in the unity of time an unity of area of
Λ(t) to which P(t) belongs. The surface Λ(t) is supposed to be regular, so it admits
in every point a tangent plane and a normal line n. Evidently ϕG = gv × n, where v
is the velocity of P.
5.1 Mechanics of Continuous Media 495

We call flow (or global flow) of an extensive greatness G through a surface Λ at

an instant t ∈ ]ti , tf [ the real number
 
G (t) = ϕG dσ = gv × ndσ .
#(t) #(t)

We consider an extensive greatness G, a volume V(t), a point P(t) of V(t) and an

instant t ∈ ]ti , tf [. We call punctual (or local) production [resp. destruction] of G in
the point P(t) of V(t) at t, the quantity δG of G that is created (or destroyed) in the
unity of time in an unity of volume of V(t) to which P(t) belongs.
We call production (or global production) of an extensive greatness G in a
volume V(t) at an instant t ∈ ]ti , tf [ the real number

G (t) = δG dV .
V(t)

A system is said in thermodynamic equilibrium if in every point (inside or of

frontier) of the system all the local flows and all the local productions are null.
We now formulate the second postulate of the Thermodynamics, that is usually
called first principle of the Thermodynamics:
[5.1.2] There is an extensive greatness U (what we call inside energy) and an
extensive greatness S (what we call entropy), both function, under conditions of
thermodynamic equilibrium, of a finite number of extensive greatness of the system,
that we call variables of state. 
The postulate [5.1.2] allows us to assume S [resp. U] as variables of state. Insofar,
denoting with X1 , . . . , Xm (m∈N) the other variables of state, it results

U = U(S, X1 , . . . , Xm )[resp.S = S(U, X1 , . . . , Xm )] .

We also say that the system has the m + 1 degree of freedom

S, X1 , . . . , Xm [resp.U, X1 , . . . , Xm ] .

We now denote with T the absolute temperature5.1.1 of the system and postulate
[5.1.3] The functions U and S are continuous, first degree homogeneous and it
results

∂U ∂S
> 0, > 0.
∂S ∂U

5.1.1 T is also called thermodynamic temperature. The scale of the absolute temperatures was estab-
lished by Lord Kelvin. He assigned to the water and the ice, under conditions of thermodynamic
equilibrium and at pressure of 1 bar, the value T = 273.16 ◦ K.
496 5 Thermoelasticity

Furthermore in conditions of thermodynamic equilibrium it results

∂U ∂S 1
= T, = .  (5.1.1)
∂S ∂U T
We expressly notice that, because the postulation [1.2.3] we have ∀λ∈

U(λS, λX1 , . . . , λXm ) = λU(S, X1 , . . . , Xm )

S(λU, λX1 , . . . , λXm ) = λS(U, X1 , . . . , Xm ) .

We now formulate the fourth postulate, that usually is called second principle of
the Thermodynamics
[5.1.4] The inside energy U can be produced or destroyed. The entropy S can be
produced but not destroyed. 

Remark 5.1.1 We notice that it is not possible to attribute physical meaning to the
entropy. 

We now postulate the third principle of the Thermodynamics, also called

principle of Nernst:
[5.1.5] If T converges to zero then S converges to zero. 
Finally we denote with E the total energy of the system (in all its forms) and
postulate that following principle of conservation of the energy
[5.1.6] If a system is in thermodynamic equilibrium it results E = 0. 
We call transformation of the system an evolution of the system from an initial
state of thermodynamic equilibrium to a final state of thermodynamic equilibrium. If
the final state coincides with that initial the transformation is called closed or better
cycle. If the evolution of the system doesn’t arrive to a state of thermodynamic
equilibrium or it doesn’t depart from a state of thermodynamic equilibrium, it is
better to talk of process rather than of transformation.
A transformation is said reversible if from the final state of thermodynamic equi-
librium it is possible put the system and the environment back in the initial state of
thermodynamic equilibrium. We notice that a cycle generally turns the system into
the initial conditions but not the environment. Insofar it is reversible if is constituted
by two reversible transformations.
If during a transformation we have a production of entropy, since such production
cannot be annulled, we have to consider that the transformation is irreversible.
Let us denote with Q the energy flow in the form of heat. We say that a
transformation or a process is adiabatic if Q = 0.
We say that a transformation or a process is isentropic if ds dt = 0.
We consider now a system in thermodynamic equilibrium. Since the postulate
[5.1.2], if we know the value of the s ∈ N variables of state then it is possible to
5.1 Mechanics of Continuous Media 497

determine the value of any thermodynamic greatness of the system. Insofar the ther-
modynamic behavior of the system is completely characterized if we can formulate
s independent equations in the variables of state.
For a gas constituted by an only chemical phase and by an only physical phase
under conditions of thermodynamic equilibrium the parameters of state are the mass
M, the volume V and the entropy S, so that s = 3. As a consequence, from the [5.1.2]
we get

U = U(S, V, M)

from which it follows that the differential of the inside energy is

∂U ∂U ∂U
dU = dS + dV + dM . (5.1.2)
∂S ∂V ∂M

So ∂U ∂U
∂V dV is an energy and then ∂V is a pressure, which we denote with the sym-
∂U
bol p. Analogously ∂M is an energy by unity of mass, that we call electrochemical
potential μ. This way from the (5.1.2), taking into account the (5.1.1), for the gas in
examination we get the fundamental Euler relationship5.1.2

dU = TdS + pdV + μdM (5.1.3)

where in the right side of equation the third addend is zero if the system doesn’t
exchange mass.
Adding to the (5.1.3) two equations of state

p = p(S, V, M)
μ = μ(S, V, M)
we individualize completely the thermodynamic behavior of the system.
If a gas has r ∈ N components, in perfect analogy the fundamental Euler equation
is gotten

1 p  μi r
dS = dU − dV − dMi
T T T
i=1

to which, since s = r + 2, we must add s + 1 equations of state.

The equations of state express the extensive parameters in function of the inten-
sive parameters. An equation of state of a system composed by m grams of a gas of
molecular weight M can be approximate from the law5.1.3

5.1.2 As a rule, the fundamental relationship is gotten hypothesizing a model, that is conjecturing
an expression of it and then going to verify its field of validity.
5.1.3 In (5.1.4) R denotes an universal constant for all the gases, equal to R = 8,316.6 Nm/◦ K =
0 0
1.986 Cal/◦ K. We also sometimes put R = R0 /M in (5.1.4). The (5.1.4) summarizes the laws of
498 5 Thermoelasticity

pV = NR0 T (5.1.4)

where N = m/M.
We call perfect gas or ideal gas a substance that exactly5.1.4 obeys to the (5.1.4).
All the real gases have a behavior that is simulated only approximately by the
(5.1.4).
The second equation of state of a perfect gas is

U = cv T (5.1.5)

where the constant cv , what we call specific heat at constant volume, is equal to 3R/2
for a monatomic gas, to 5R/2 for a diatomic gas.
For a perfect gas Mayer easily got that

cp − cv
=R
cv
where cp denotes the specific heat at constant pressure.

Remark 5.1.2 We observe that an adiabatic transformation of a perfect gas has equa-
tion pV γ −1 = const, where γ = cp /cv . In fact we will see that every adiabatic
transformation is isentropic. So from (5.1.3) and (5.1.5) it follows

cv dT = p dV

from which, taking into account the (5.1.4), we get

dT cp − cv dV
+ =0
T cv V
from which

dT dV
+ (γ − 1) =0
T V
and then the thesis. 

Remark 5.1.3 We call specific heat c of a substance between the two absolute
temperatures T1 and T2 the quotient

q
c=
T2 − T1

   
Boyle (pV=const.), of Gay–Lussac (V = V0 1 + 272.24 T
, p = p0 1 + 272.24
T
) and of Avogadro
(equal volumes of any gas, in equal conditions of temperature and pressure, contain the same
number of molecules).
5.1.4 This result is drawn in a simple way employing the kinetic theory.
5.1 Mechanics of Continuous Media 499

where q denotes the necessary quantity of heat to bring the unity of mass of the
substance from T1 to T2 . This way the necessary quantity of heat to heat from T1 to
T2 a body that weighs P kg is Q = cP(T2 − T1 ).
You notice that the exact definition of c is
q dq
c = lim = .
T2 →T1 T2 − T1 dT

In the case of a gas [resp. solid] the specific heat to constant pressure [resp.
to constant stress] cp [resp. cσ ] and the specific heat to constant volume [resp. to
constant deformation] cv [resp. cε ] are significant. For the iron we have c = 0.115
cal/ ◦ C. Besides for the solids, and particularly for the metals, it results cσ/cε =
1.05. 

Problem 5.1.1 A fluid with γ = 2.4 evolves according the transformation A→ B→

C→ D of Fig. 5.1.1, where the transformation C → D is adiabatic. You shall
determine the done work and the exchanged heat.

Solution The transformation C → D has equation pV K = const, where K = γ −1 =

1.4. About the work, since in B → C the volume is constant, we have

 B  C  D
L = LAB + LBC + LCD = p dV + p dV + p dV
A B C
 B  D  D
=p dV + pD VDK V −K dV = p(VB − VA ) + pD VDK V −K dV
A C C
 D
V −K+1 pD VD − pC VC
= 32 · 7 + pD VDK = 224 + atm · lt
−K + 1 C 1−K

from which, by observing that from pC VC1.4 = pD VD1.4 it follows VD = 2.7lt, we get

32 · 2.71.4 − 10 · 81.4
L = 224 + = 48 atm · lt .
−0.4

Fig. 5.1.1
500 5 Thermoelasticity

As far as the flow of heat, we preliminarily observe that for a real gas 5.1.5

γ 1
cp = AR , cv = AR .
γ −1 γ −1
In the case of the air, since

1 cal kgm
A= , R = 29.27
427 kgm kg◦ C
we have

cal cal
cp = 0.239 , cv = 0.171 .
kg◦ C kg◦ C
Furthermore, by using equation (5.1.4), we get

TA = 11◦ C, TB = 87.5 ◦ C, TC = 27.5 ◦ C .

So

Q = QAB + QBC + QCD = QAB + QBC

cal 
= cp (TB − TA ) + cv (Tc − TB ) = 18.3 − 10.3 = 8 .
kg
Problem 5.1.2 Two systems are isolated by impermeable, unworking5.1.6 and adi-
abatic walls both from the external environment and from them. The system 1 is
constituted by two moles of He (what is monatomic), the system 2 from three moles
of O2 . The total energy of the system is of 6,000 kcal. You shall calculate U1 and U2
at the equilibrium.
Solution At the equilibrium T1 = T2 . Since
3 3
U1 = N1 R0 T1 = · 2 · 1.98T1 kcal
2 2
3 3
U2 = N2 R0 T2 = · 3 · 1.98T2 kcal ,
2 2
at the equilibrium we must have
U1 + U2 = 6, 000 kcal
U1 U2
=
3 · 1.98 7.5 · 1.98
so that U1 = 1, 720 kcal, U2 = 4, 280 kcal. 

5.1.5 For a perfect gas it results A = 1.

5.1.6 We call unworking a wall that doesn t allow exchanges of work.
5.1 Mechanics of Continuous Media 501

5.1.3 The Equations of Balance

We consider a greatness G of the system, as the mass or the energy, function of the
temporal variable5.1.7 t ∈ ]ti , tf [. We denote with g the quantity of G contained in
the unity of volume of the system. Insofar g is function of the spatial variables x, y,
z and of the temporal variable t and it results

∀t ∈ ]ti , tf [ G(t) = g(x, y, z, t) dx dy dz .
V(t)

Evidently G can vary in the infinitesimal interval of time ]t, t + dt[ only if we
introduce other from the outside or if other is created or destroyed inside the same
system. So the increase [resp. diminution] of G in the infinitesimal interval of time
]t, t + dt[ is equal to the sum of the entering [resp. going out] flow of G through Σ(t)
and of the production [resp. destruction] of G in V(t). Insofar

dG
= G + G in]ti , tf [ (5.1.6)
dt
The (5.1.6) is called equation of the balance of G.
The equation of the balance is also valid for thermodynamic greatness. However
when equation of balance is applied to a thermodynamic greatness, it is necessary
to suppose that the system is in thermodynamic equilibrium.
A greatness such that G = 0 is said conservative.
If in the (5.1.6) G denotes a conservative greatness, then the (5.1.6) is also called
equation of the conservation or principle of conservation.
Evidently in an isolated5.1.8 system every conservative greatness is constant.
Let us apply now the equation of the balance to the total energy E of the system,
which is a conservative greatness for postulation. We have

dE
= E . (5.1.7)
dt
We have postulated that the system can exchange with the environment or energy
but not mass or energy and mass. We separate therefore E in the sum of the flow
of energy EE which is not in partnership to exchange of mass and of the flow of
energy EM which is in partnership to exchange of mass.
It is worthwhile, and the reason will be clear in the succession, to consider the
flow of energy EE sum of an aliquot Q, that we call flow of heat for unity of time
or thermal power or simply heat and of an aliquot L, that we call flow of work for
unity of time or mechanical power or simply work for unity of time.

5.1.7 We denote with t ∈ [0, +∞[ the initial instant and with tf the final instant, that is a real number
tf > ti or the symbol +∞.
5.1.8 A system is said isolated if the flow of any greatness through the frontier (t) of the volume
V(t) occupied by the system at the instant t is zero.
502 5 Thermoelasticity

Insofar, if Σ doesn’t allow exchange of mass between system and environment,

it results

dE
= Q + L. (5.1.8)
dt

In the (5.1.8), that is one of the forms of the principle of conservation of the
energy, it is opportune to underline that Q and L are two forms of flows of energy,
i.e. they are both energies in transit.
The variation of E in an interval of time [t1 , t2 ] is gotten by integrating the two
members of the (5.1.8) with respect to the time, from t1 to t2 .
Let us perform now the balance of a greatness that doesn’t conserve it. A typical
greatness of the Thermodynamics that doesn’t conserve it is the entropy.
Also the inside energy U doesn’t conserve it. However of the entropy we know
something more: it does not only doesn’t conserve it but can never be destroyed.
We immediately observe that since the greatness of which we want to perform
the balance is a thermodynamic greatness, to be able to define the greatness the
system needs to be in thermodynamic equilibrium. From the (5.1.6) we have5.1.9

dS
− S = S in ]ti , tf [. (5.1.9)
dt

Since for the postulate [5.1.4] S ≥ 0, the (5.1.9) furnishes

dS
− S ≥ 0 . (5.1.10)
dt

We have postulated that the system can exchange with the environment or energy
but not mass or mass and energy. Insofar, as for the total energy, we separate S in
the sum of a flow of entropy SM associate to the exchange of mass and of a flow
of entropy SQ associate to the exchange of energy but not of mass.
Naturally if Σ doesn’t allow exchange of mass then it results SM = 0.
We now give a fundamental definition of the Thermodynamics that defines a
posteriori the heat and the work: SQ is associated to the flow of energy in the form
of heat. Insofar that particular form of flow of energy that we have called work do
not carry with itself a flow of entropy.
In conclusion the entropy can flow through Σ together with the mass or together
with the energy or with both, but it cannot flow alone through Σ. Then if the system
is isolated it results S = 0.

5.1.9 In this simplified formulation there is an incongruity, because under conditions of ther-
modynamic equilibrium the flows and the productions are zero. However it exists the exact
formulation, that preliminarily gives all the concepts of the Thermodynamics of equilibrium, where
this incongruity doesn t subsist.
5.1 Mechanics of Continuous Media 503

Let us denote now with ϕSQ the local flow of entropy through Σ not associated
to a flow of mass, so that

SQ = ϕSQ dσ .


We can prove in rigor that, said T the absolute temperature of the system, it results

TϕSQ = q .

Particularly, if T is constant on Σ, it results

Q
SQ = .
T

5.1.4 Thermodynamics of the Irreversible Processes

The Thermodynamics of the reversible processes, also called Classical thermo-

dynamics, correlates a state of thermodynamic equilibrium with another state of
thermodynamic equilibrium, however reached. It affirms that if energy is varied this
is due to the flows of heat and work. Furthermore it affirms that if the entropy is var-
ied this it is due to flows or productions of entropy. But it doesn’t give information
on what happens during the transformation if this doesn’t cross instant for instant
states of thermodynamic equilibrium. We remember that in such states the flows and
the productions are identically zero.
To give meaning to the flows and to the productions, and to draw operational
expressions of them, we build a new theory, that includes the precedent Classical
thermodynamics, still based on postulates. We formulate indeed the following pos-
tulate that allows us to also define the inside energy, the entropy and the other
thermodynamic greatness in states of the system that are not of thermodynamic
equilibrium.
[5.1.7] In any point P of a system, if I is a neighborhood of P, the subsystem
individualized by I is in state of thermodynamic equilibrium. 
The principal limitation of the postulate [5.1.7] implicates that the history of the
system, that is the processes suffered in precedence, doesn’t influence the state of
the system. In fact such postulate affirms that the inside energy in a point depends
only on the values that the thermodynamic parameters have in that point. In nature a
vast phenomenology exists to which this postulation cannot be applied. For instance
in the processes of hysteresis and in the thermal treatments the acquired properties
of the material depend on the past history.
After that we also formulate the following new postulate, that we call principle of
Curie, that allows us to appraise the flows and the productions of the thermodynamic
greatness.
504 5 Thermoelasticity

[5.1.8] If the system is isotropic every flow depends only on all the generalized forces
having the same tensorial5.1.10 order of the flow. 
In the following we will always deal with isotropic systems, and we will call gen-
eralized force5.1.11 any cause that provokes flows. The postulate [5.1.8] tells us that
a flow that is a scalar [resp. vector] [resp. tensor] depends only on all the generalized
forces that are scalars [resp. vectors] [resp. tensors]. For instance in the combustion
the chemical affinity is a generalized force of scalar type. It is the cause that pushes
the reaction in a sense or another. However, if the material is isotropic, it cannot
influence neither the flows of energy nor the flow of mass because they are vectors.
As a rule a generalized force directly provokes a flow and indirectly other flows.
For instance the generalized force
∂ 1
∂r T
is the direct cause of the flow of energy in the form of heat and is an indirect cause
of the flow of mass.
There is still to clarify the type of the functional dependence of effects, that is the
flows, from the causes, that is from the generalized forces. Well, we choose a linear
functional dependence, that is we limit the study to the linear Thermodynamics of
the irreversible processes. Moreover we choose the cause so that when the cause is
null the effect is null.
So we made a choice for which the effect is proportional to the cause and, impor-
tant thing, the coefficient of proportionality is independent from the cause. Insofar
such coefficient is a purely thermodynamic greatness, that is a function of state. For
instance the law of Fourier of the transmission of the heat is valid only if there is
the possibility of flows of mass:

∂T
Q=k (5.1.11)
∂r

where the constant k is a function of state of the material.

This way in an isotropic system the flows depend on all the direct and indi-
rect generalized forces of the same tensorial order. If the phenomenology is linear
then the flows are proportional to all the causes that can provoke them. Besides the
coefficients of proportionality are functions of state, that is depend only on the state
of the system.

5.1.10 A greatness associated to a point is said scalar or tensor of order 0 if is individualized by a

real number; vector or tensor of order 1 if is individualized by an ordered triplet of real numbers;
tensor or tensor of order 2 if is individualized by an ordered sextuple of real numbers or by two
vectors. It benefits to underline that to a vector a direction is associated; to a tensor two directions
are associated or, if we prefer, a plane and a direction.
5.1.11 This because the flow of work is provoked by a cause that we call force.
5.1 Mechanics of Continuous Media 505

Let us consider a system constituted by the mixture of two gases. There is flow
of energy in the form of heat Q and flow of mass of the first gas M1 . The relative
direct generalized forces are

∂ 1 ∂ μ
, .
∂r T ∂r T

By virtue of the principle of Curie, since these direct generalized forces are vec-
tors and both the flows are vectors, every flow must depend on both the generalized
forces. Then in the field of the Linear thermodynamics we will write

∂ 1 ∂ μ
Q = Lqq + Lqm
∂r T ∂r T
∂ 1 ∂ μ
M 1 = Lmq + Lmm .
∂r T ∂r T

The direct coefficients of proportionality Lqq , Lmm and those crossed Lqm , Lmq
are only functions of the state of the system.
It subsists the following theorem of Onsager, which is true also for systems with
n degree of freedom
[5.1.9] In absence of magnetic fields the crossed coefficients are equal:
Lqm = Lmq . 
Remark 5.1.4 These conclusions are of the maximum practical importance. It is
for instance thanks to them that it is possible to build refrigerators based on the
thermoelectric effect. In such case the exchange of energy in the form of heat is
gotten through a difference of electric potential. 

Let us denote now with δS the local production of entropy, with 1 , . . . , n all
the flows in game, with F1 , . . . , Fn all the generalized forces in game. We can prove
that


n
δS = i Fi . (5.1.12)
i=1

In the (5.1.12) the products are scalar products if the flows and the generalized
forces are vectors; double scalar products if the flows and the generalized forces
are tensors. We observe that in the (5.1.12) the energy’s flow in the form of work
doesn’t appear because the production of entropy is not accompanied.
We have postulated that δS ≥ 0. This is translated under conditions on the direct
and crossed coefficients Lij . For instance in the case already examined in which they
are Lqq , Lmm , Lqm , Lmq it results:

Lqq ≥ 0, Lmm ≥ 0, Lqq Lmm − Lmq

2
≥ 0.
506 5 Thermoelasticity

5.2 Fluid Dynamics

5.2.1 The Mathematical Model

In any problem of Fluid dynamics we meet unknowns of two types: the unknowns
of state (that characterize the thermodynamic state of the system) and the kinetic
unknown (that is the vector velocity). Obviously these unknowns are functions of
the point and of the time.
It goes from itself that it is necessary to formulate so many scalar equations
depending on how many thermodynamic unknowns of state there are. Furthermore
we need to formulate a vector equation in which the vector velocity intervenes. To
formulate the scalar equations we impose the fact that the evolution of the system
happens satisfying the equations of conservation or of balance.
Let us obtain at first the scalar equations expressing the balance of the mass. If
the system is constituted by an unique phase and the local production of mass δM is
zero, the balance becomes a conservation and we have, denoting with M the mass,
with ρ the density of mass and with v the vector velocity

G = M, g = ρ, ϕM = ρv × n, δM = 0 .

so that the (5.1.6) furnishes

 
∂
ρ dV + ρv × n dσ = 0 .
∂t V 

Applying the same reasoning to an elementary parallelepiped with faces parallel

to the coordinated planes and containing the generic point (x, y, z) ∈ V, keeping in
mind the theorem of the divergence, we obtain the scalar differential equation

∂ρ
+ div(ρv) = 0 in V × [ti , +∞[ . (5.2.1)
∂t
If instead the system is a mixture of j different masses, imposing the balance
of the single mass we obtain, for the mathematical model in building, j scalar
differential equation similar to the (5.2.1). If the possibility of chemical reactions
is excluded, such equations will be the equations of conservation of the single
masses.
We need now to obtain a vector equation in the unknown velocity v. Clearly
such vector equation will be able to decompose itself in so many scalar equations
depending on how many components of the vector there are. Well the suitable equa-
tion to write is the balance of the momentum. Since the production of the momentum
is due to the volumetric forces, as for instance the force of gravity, in absence of such
forces the balance of the momentum becomes a conservation.
The density of the momentum is g = ρv. Denoting with v0 the velocity of the
centroid of the system, from the theorem of the weighed average of the Theory of
5.2 Fluid Dynamics 507

the integration and from the definition of centroid it easily follows that the total
momentum is G = M v 0 .
Let us observe now that the flow of the momentum is the flow of a tensor. In fact
by definition the flow of a scalar greatness G is the flow of the vector gv as well as
the flow of a vector greatness u is the flow of the tensor u · v 5.2.1 .
We separate the tensor flow ρv of the vector momentum in an aliquot

(ρv · v) × n dσ


that we call convective flow of the momentum and in an aliquot


τ̆ × ndσ


that we call diffusive flow of the momentum.

The diffusive flow of quantity of motion is also it the flow of a tensor, that we
have denoted with the symbol τ̆ . The tensor τ̆ has the dimensions of a force for unity
of area and in effects is really the stress tensor. Precisely in a macroscopic analysis
of the molecular motions we see that the transport of the momentum manifests itself
with a system of superficial stresses.
We notice that the production of momentum is due to the volumetric forces f but
not to those superficial, i.e. 
ρv = fdV .
V

5.2.1 Letu = (ux , uy , uz ), v = (vx , vy , vz ) be two any vectors of 3 . We call tensor product of u and
v the tensor ⎡ ⎤
ux vx ux vy ux vz
u · v = uy vx uy vy uy vz ⎦ .
⎣
uz vx uz vy uz vz

Let ⎡ ⎤ ⎡ ⎤
τxx τxy τxz nx
τ̆ = ⎣ τyx τyy τyz ⎦ , n = ⎣ ny ⎦
τzx τzy τzz nz

be two any tensor and vector of 3 . We call scalar product of τ̆ and n the vector
⎡ ⎤ ⎡ ⎤
(τ̆ × n)x τxx nx + τxy ny + τxz nz
⎣
τ̆ × n = (τ̆ × n)y = τyx nx + τyy nz + τyz ny ⎦ .
⎦ ⎣
(τ̆ × n)z τzx nx + τzy ny + τzz nz

Finally we denote with x, y, z the unit vectors of the reference frame and consider any tensor τ̆ of
3 and any regular surface  of 3 . We call flow of the tensor τ̆ through Σ the vector
   
τ̆ ×n = (τ̆ × n)dσ = (τ̆ × n)x dσ x + (τ̆ × n)y dσ y + (τ̆ × n)z dσ z .
   
508 5 Thermoelasticity

So the equation of balance of the momentum is5.2.2

  
∂
ρvdV + (ρv · v + τ̆ ) × ndσ = fdV. (5.2.2)
∂t V  V

We gain now another scalar equation for the mathematical model in building
imposing the conservation of the total energy. The total energy of any system is
constituted by a lot of forms of energy. A first contribution to the total energy is the
inside energy U of the system. A second contribution is the kinetic energy

1 2
ρv dV .
V 2

Since the kinetic energy can be converted in inside energy and vice versa, it is
important to underline that every of these two contributions is able to change.
Let us consider a system that evolves in absence of gravitational or electromag-
netic fields. Then the total energy of the system is constituted by the sum of the
inside energy and of the kinetic energy, so that

v2G
G=U+M .
2
Clearly the density of G is
1
g = ρu + ρv2 .
2
where u denotes the mass unitary inside energy.
The convective flow of the total energy is the vector

v2
gv = ρ u + v.
2

Let us denote now with Jt the vector diffusive flow of the total energy. It is the
sum of three addenda. The first addendum is the diffusive flow of energy associated
with the possible flows of mass. The second addendum is the diffusive flow of
energy in the form of heat. The third addendum is the diffusive flow of energy in the
form of work.
The diffusive flow of energy associated with the flow of mass is given by the flow
of mass multiplied for the energy for unity of mass related to the chemical potential
of the substance.
The diffusive flow of energy in the form of heat, that we denote with the symbol
Q, is always irreversible since a production of entropy is always accompanied.

5.2.2 Applying the same reasoning to an elementary parallelepiped with faces parallel to the coor-
dinated planes and containing the generic point (x, y, z) ∈ V, we obtain three scalar differential
equations (found at first by Stokes-Navier) to satisfy in V.
5.2 Fluid Dynamics 509

The diffusive flow of energy in the form of work is the product τ̆ × v of the
superficial stress tensor and of the vector velocity with which the point of application
of the superficial stress moves. Such diffusive flow of energy in the form of work can
be separate in a reversible part, to which the production of entropy is not associate,
and in an irreversible part.
Clearly to separate the diffusive flow of energy in the form of work in a reversible
part and in an irreversible part we need to separate the stress tensor5.2.3
⎡ ⎤
σx τxy τxz
τ̆ = ⎣ τyx σy τyz ⎦
τzx τzy σz

in a reversible part and in an irreversible part.

The reversible [resp. irreversible] stress is that part of stress that results in
reversible [resp. irreversible] work. To individualize the reversible part of τ̆ we
recall there the Classical thermodynamics, that by definition is the thermodynamics
of the reversible phenomena. Well the stress considered in it is the pressure.
Insofar we decompose the stress tensor in the sum of an hydrostatic tensor p̆ and
of a remaining aliquot τ̆ d , that is

τ̆ × v = p̆ × v + τ̆ d × v = pv + τ̆ d × v .

We only need to add the constitutive equations, that simulate the mechanic
behavior of the matter. To characterize the state of a fluid is enough to assign its
volume. To characterize the state of a solid we need to assign not the volume but
its state of deformation. This way in the fluids the constitutive equation is the equa-
tion of state that express the dependence of p from the parameters of state. In the
solids the constitutive equations are the relationships that expresses the dependence
of the stress tensor from the strain tensor. In every case if we determine the constitu-
tive equations then the problem of the continuous, both fluid how solid, is correctly
formulated, since it has so many equations depending how many unknown ones
there are.
In conclusion, in absence of diffusive flows of mass and in absence of potential
energy consequential to forces of mass, the equation of conservation of the total
energy is written

∂ v2
ρ u+ dV
∂t V 2
   (5.2.3)
v2
+ u+ ρv + pv + τ̆ d × v + Q × n dσ = 0 .
 2

5.2.3 Usually we can denote τxx [resp. τyy ] [resp. τzz ] also with the symbol σx [resp. σy ] [resp. σz ].
510 5 Thermoelasticity

Remark 5.2.1 Every term of the equation of the energy has the dimensions of a
power, that is of an energy for unity of time. 

The difficulty of the formulation of the equation of conservation of the total

energy consists to recognize what the meaningful contributions that intervene in
a datum phenomenon are. For instance in a nuclear reaction it is necessary to bring
in account the variation of energy consequent to the variation of mass.
In every problem of Fluid dynamics or of Mechanics of the solid the engineer
must simulate the problem with a mathematical model obtained discarding all the
terms that are negligible in that particular problem. The choice of the terms to be
discarded can be effected with the theory of the characteristic numbers, which we
get in the following section. As we will see these numbers furnish a relative measure
of the greatness that intervene in the problem, so allow to individualize the terms of
the equations that can be neglected.

5.2.2 The Characteristic Numbers

Let us make dimensionless the equations of the continuous. First of all we have to
choose the unities of measure, a density of reference ρr and a velocity of reference
vr . As a consequence in the expressions ρ = ρ ∗ ρr , v = v∗ vr , the symbols ρ ∗ ,
v∗ denote dimensionless numbers. The job of the engineer lies in the opportune
choice of the reference’s greatness, meaning to gather the physical meaning of the
phenomenon that is considered.
The vr must be chosen in relationship to the phenomenon that we study. For
instance if we study some relativistic phenomena it will be the velocity of the light
because in these phenomena the relationship between the velocity of the system and
the velocity of the light interest us. On the contrary if we study the phenomenon of
the propagation of the sound it would be wrong to choose as velocity of reference
the light. In fact all the numbers would result extremely small and therefore neg-
ligible and the phenomenon could not be studied. Clearly in this second problem
the velocity of reference is that of the sound a, that is a thermodynamic greatness.
In the air and under standard conditions of temperature and of pressure, it results
a = 340 m/s.
Analogously for the time, if we study a phenomenon in which sinusoidal oscilla-
tions intervene, we will choose as time of reference the period tr , and we will write
t=t∗ tr , with t∗ dimensionless number.
Furthermore if we study the transmission of the heat in a cylinder, the surface of
reference will be the lateral surface, proportional to the diameter. Instead if we study
the flow of mass through the cylinder, the surface of reference will be the base, that
is proportional to the square of the diameter.
Let us make dimensionless the equation of the continuity (5.2.1). We have
 
ρr Vr ∂ ∗ ∗
ρ dV + ρr vr r ρ ∗ v∗ × n dσ ∗ = 0
tr ∂t∗ V ∗ ∗
5.2 Fluid Dynamics 511

from which
 
∂ vr r tr
ρ ∗ dV ∗ + ρ ∗ v∗ × n dσ ∗ = 0. (5.2.4)
∂t∗ V∗ Vr ∗

In the (5.2.4) the first addendum is a dimensionless number and the two factors
that constitute the second addendum are also dimensionless. The first one of such
factors is called number of Strouhal and is pointed out with the symbol

vr Σr tr
Str = .
Vr
It measures the relative importance of the contribution of the flow in comparison
to the contribution of the unsteadiness in the equation of the conservation of the
mass. If Str is very small in comparison to the unity, the contribution of the flow is
negligible in comparison to the contribution of the unsteadiness. Vice versa if Str
is very great then the unsteady term will be negligible in comparison to the term
related to the flow. If Str has order of greatness 1 then the unsteady term and that
related to the flow are of the same order of greatness for which they must be both
brought in account.
Let us observe that all the characteristic numbers, being dimensionless, are
always interpretable as ratio of two greatness having the same dimensions. This
way it is possible to give them various interpretations.
The most immediate interpretation of the number of Strouhal, since we are
studying the importance of the convection in comparison to the unsteadiness, is
the ratio of two times. We will call macroscopic time of reference or convective
time the term vrVr r , which has really the dimensions of a time. It is the character-
istic time of the convection and measures the middle permanence of the particle
in the volume of control. Instead the time tr is characteristic of the unsteadiness
and measures the rapidity of the unsteadiness. So if Str is very small the con-
vection is negligible in comparison to the unsteadiness and the motion is said
unsteady.

Remark 5.2.2 We notice that in the number of Strouhal the mass doesn’t appear
although to obtain it we have used the equation of conservation of the mass. In
reality, as it is immediate to verify, if we had written the equation of the conservation
of any other greatness G we would reached the same expression of Str . This means
that the relative importance of the convection in comparison to the unsteadiness is
always the same for any greatness that flows. 

Let us make dimensionless the equation (5.2.2) of balance of the momentum. We

have
  
∂
ρvdV + (ρv · v + τ̆ ) × n dσ = fdV
∂t V  V
512 5 Thermoelasticity

from which

 
ρr vr Vr ∂
ρ ∗ v∗ dV ∗ + ρr v2r Σr ρ ∗ v∗ · v∗ × n dσ ∗
tr ∂t∗ V∗ ∗
   (5.2.5)
+ pr Σr ∗ ∗
p̆ × n dσ + τdr Σr τ̆ ∗d ∗
× n dσ = fr Vr ∗ ∗
f dV .
∗ ∗ V∗

As it is easy to verify, in the (5.2.5) the coefficients of the dimensionless factors

have the dimensions of a momentum for unity of time, that is of a force. Dividing
left and right side of equation (5.2.5) for ρr v2r r it is gotten
 
1 ∂ ∗ ∗ ∗
ρ v dV + ρ ∗ v∗ · v∗ × n dσ ∗
Str ∂t∗ V∗ ∗
  
pr ∗ ∗ τdr ∗ ∗ fr Vr
+ p̆ × n dσ + τ̆ × n dσ = f∗ dV ∗ .
ρr v2r  ∗ ρr v2r  ∗ d ρr v2r r V ∗
(5.2.6)
In the (5.2.6) the three new characteristic numbers appear
pr
(5.2.7)
ρr v2r
τdr
(5.2.8)
ρr v2r
fr Vr
. (5.2.9)
ρr v2r r

We observe that in these three new characteristic numbers the surface of reference
appears only in the third one. In fact it measures the relative importance between the
contribution of volume and the superficial one.
With regard to the characteristic number (5.2.7) it is necessary to distinguish two
cases.
In the case of incompressible fluid ρpvr 2 is called number of Weber. As pres-
r r
sure of reference it is necessary to assume the pressure hydrostatics. If however we
are studying the phenomenon of the cavitations, as pressure of reference pr it is
necessary to assume the superficial tension of vapor.
Remark 5.2.3 It is important to notice that the number of Weber can be varied
through chemical additives that vary the value of the superficial tension of vapor.
The number of Weber is important for the study of the flows of mass to double phase,
in which the liquid phase and the phase of vapor coexist. In them the tension of vapor
is important to establish the conditions of equilibrium of a vapor bubble. 

In the case of compressible fluids the characteristic number ρpvr 2 is expressed

r r
through another characteristic number M. Precisely in such case the pressure of
5.2 Fluid Dynamics 513

reference of the compressible fluid is the thermodynamics one ∂U

∂V . It is the macro-
scopic result of the molecular motions. So it must be proportional to the mean
velocity of the molecules, that coincides with that of the sound velocity a. If the
compressible fluid is a perfect gas it results

p = ρRT, a2 = γ RT

where cp
γ = (5.2.10)
cv
R = cp − cv (5.2.11)

so that
a2 ρ
p=
γ
and the characteristic number (5.2.7) can be write

pr a2 1
= = .
ρr vr
2 γ vr
2 γ M2

Insofar in the compressible fluids the relative importance of the not dissipa-
tive flow of momentum in comparison to the convective flow depends on γ and
on the dimensionless characteristic number M, that we call number of Mach,
given by
vr
M= . (5.2.12)
a
Let us give now a kinetic interpretation of the number of Mach. M is the ratio
between the macroscopic velocity of the fluid and the velocity with which the sound
is propagated in the fluid. The sound velocity is the velocity with which the small
troubles are propagated in the matter, as the small variations of pressure. We can
interpret vr as the velocity with which the small troubles are created. If a body
moves in a continuous fluid with velocity vr , it creates with rapidity vr small troubles
what are propagated with velocity a in the matter. If the body moves at Mach < 1
(subsonic flow) the wave of pressure that it produces is perceived by the matter
before the arrival of the body. This way the matter knows the arrival of the body
before it come. Then the matter somehow prepares it to welcome it, creating a free
space for its passage. Instead if the body moves at Mach > 1 (supersonic flow), the
body will invest the fluid before the waves of pressure that it produces. This way the
environment, not having the opportunity to acknowledge the presence of the body
before its arrival, will be forced to brusquely open for allowing the body to pass. The
same considerations can be made if it is not the body that moves in the environment,
but it is the current of fluid that invests the body. Seen the different behavior of the
514 5 Thermoelasticity

subsonic flow and of the supersonic flow, an unusual behavior in correspondence of

M=1 (transonic flow) is probable.
Let us give now a dynamic interpretation of the number of Mach. We have see
that for compressible fluids

pr 1
= .
ρr vr
2 γ M2
Since γ always has order of greatness 1, M gives a measure of the importance
of the resultant one of the efforts of pressure in comparison to that of the forces of
inertia.
Let us give now an elastic interpretation5.2.4 of the number of Mach. The small
troubles are propagated through a succession of compressions and expansions of the
matter, for which the velocity of the sound is as great as the fluid is compressible.
Then if M ∼ = 0 or the body is practically motionless or the motion happens in a
highly incompressible fluid. For better saying, the fluid behaves as if were incom-
pressible. If the velocity with which the perturbations are produced is very small it is
M∼ = 0 and the air, which is a very compressible fluid, behaves as an incompressible
fluid. Vice versa the water, which is an a little compressible fluid, behaves as a very
compressible fluid when the velocity with which the perturbations are produced is
very elevated (hit of ram).
Let us give now an energy interpretation of the number of Mach. Since

ρr v2r vr
M2 = ,
γ pr vr
energetically M measures the relative importance between the convective flow of
kinetic energy and the reversible flow of work.
We now turn our attention to the characteristic number (5.2.8)

τdr
,
ρr v2r
that gives a measure of the importance of the dissipative diffusive flow of momen-
tum in comparison to the convective flow.
To have an idea of the criterion according to which τdr must be chosen we have
to understand before what it is τ̆ d , that is the dissipative part of the stress tensor
τ̆ = p̄ + τ̆ d that we have introduced when we have performed the balance of the
momentum. We have denoted with τ̆ the diffusive flow of the momentum and when
we have performed the conservation of the energy we have found again the super-
ficial stress τ̆ in the diffusive flow of energy in the form of work, distinguishing
in it a not dissipative isotropic part p̆ constituted by the thermodynamic pressure

5.2.4 However more than an elastic interpretation of the number of Mach, this is an interpretation of
the velocity of the sound.
5.2 Fluid Dynamics 515

and a dissipative part τ̆ d . Being a dissipative local flow of momentum, τ̆ d is of

the same nature of Q and of the flow of entropy, i.e. a greatness that needs the
Thermodynamics of the irreversible processes to be expressed clearly.
Proceeding according to such street, we must individualize the direct generalized
force that produces this dissipative flow. Then, invoking the principle of Curie, we
must say that in an isotropic system this flow, being a tensor greatness, must depend
on all and only the generalized forces of tensor type, linearly in the hypothesis of
a linear phenomenology. In this particular case, however, we prefer to look for the
expression of τ̆ d , i.e. for the constitutive relationships of the material, proceeding in
other way.
Preliminarily we assume that the tensor τ̆ d is symmetric, so that

τdxy = τdyx , τdxz = τdzx , τdyz = τdzy .

We now correlate the stress τ̆ to the strain of the fluid. We notice that in the case
of the solid τ̆ depends, linearly in the model of Navier, on the deformation. Instead
in the case of the fluid τ̆ depends not on the deformation, but on the contrary from
the velocity of deformation. In fact if we slowly dip a hand in a liquid we don’t
warn some stress on the hand. If instead the hand is quickly dipped we warn an
intense stress, increasing with the velocity of immersion, that is with the velocity
of deformation of the fluid. With this reasoning we have individualized the gener-
alized direct force of τ̆ d . In the hypothesis of linearity and isotropy we assume for
the fluids

∂εx ∂
σdx = 2μ1 + μ2 (εx + εy + εz )
∂t ∂t
∂εy ∂  
σdy = 2μ1 + μ2 εx + εy + εz
∂t ∂t
∂εz ∂  
σdz = 2μ1 + μ2 εx + εy + εz
∂t ∂t (5.2.13)
∂γxy
τdxy = μ1
∂t
∂γxz
τdxz = μ1
∂t
∂γyz
τdyz = μ1 .
∂t

We consider a fluid particle and denote with ux , uy , uz the components of its

displacement and with vx , vy , vz the components of its velocity v, so that

∂ux ∂uy ∂uz

vx = , vy = , vz = . (5.2.14)
∂t ∂t ∂t
516 5 Thermoelasticity

We know from the Strain analysis that in any continuous matter, in the hypothesis
of small deformations

∂ux
εx =
∂x
∂uy
εy =
∂y
∂uz
εz =
∂z
∂ux ∂uy
γxy = +
∂y ∂x
∂ux ∂uz
γxz = +
∂z ∂x
∂uy ∂uz
γyz = + .
∂z ∂y

As a consequence for the fluid, taking into account the (5.2.13) and (5.2.14)

∂vx
σdx = 2μ1 + μ2 div v
∂x
∂vy
σdy = 2μ1 + μ2 div v
∂y
∂vz
σdz = 2μ1 + μ2 div v
∂z
∂vx ∂vy (5.2.15)
τdxy = μ1 +
∂y ∂x
∂vx ∂vz
τdxz = μ1 +
∂z ∂x
∂vy ∂vz
τdyz = μ1 + .
∂z ∂y

In the constitutive relationships (5.2.15) the coefficients μ1 and μ2 are called

first and second coefficient of viscosity and are functions of state. Insofar they don’t
depend on the velocity of deformation, but only from the type of matter and from
the parameters that individualize the state of the system. The dependence of μ1 from
the temperature is different for the liquids and for the gases. Precisely for the liquids
μ1 decreases to the growth of the temperature while the opposite one occurs for the
gases. Besides the viscosity of the gases is always notably inferior to that of the
liquids.
Remark 5.2.4 The different dependence of μ1 from the temperature for the liquids
and for the gases can be explained on the molecular plan with the fact that viscosity
is due to two factors: the molecular attraction and the thermal molecular motions.
In the liquids the viscosity is essentially due to the first factor and as a consequence
5.2 Fluid Dynamics 517

while increasing the temperature, the molecular attraction decreases and then the
viscosity decreases. In the gases instead the viscosity is essentially due to the ther-
mal molecular motions and as a consequence, while increasing the temperature, the
molecular motions increase and then the viscosity increases. 

Remark 5.2.5 We observe that, resulting

∂  
div v = εx + εy + εz
∂t

div v is the variation of volume for unity of time, so that div v is a production of
volume. 

Remark 5.2.6 We also observe that

σdx + σdy + σdz 2

pd = = μ1 + μ2 div v
3 3
so that pd is an invariant with respect to the frame of reference and depends only
from div v. 

In the greatest part of the gases

2
μ1 + μ2 ∼
=0
3
so that pd ∼
= 0 and in consequence for the stress tensor τ̆ = p̆ + τ̆ d is gotten by

σx + σy + σz
p=
3
where p is the thermodynamic pressure.
In conclusion in the solids the direct generalized force of τ̆ d is the strain tensor
ε̆. In the fluids the direct generalized force of τ̆ d is the tensor
⎡ ∂v ∂v ∂vx ⎤
x x
⎢ ∂x ∂y ∂z ⎥
⎢ ⎥
∂v ⎢ ∂vy ∂vy ∂vy ⎥
=⎢ ⎥.
∂r ⎢
⎢ ∂x ∂y ∂z ⎥
⎥
⎣ ∂vz ∂vz ∂vz ⎦
∂x ∂y ∂z
The (5.2.15) are a generalization of the law of Newton

dv
F = μ1 A
dn
on the friction offered by a fluid situated among two parallel plane surfaces of area
A, in relative motion (Fig. 5.2.1). The fluids that obey to (5.2.15) are therefore called
518 5 Thermoelasticity

Fig. 5.2.1

newtonian. However we notice that there are many fluids that are not newtonian, as
for instance the heavy lubricating oils.
Remark 5.2.7 In the Solid mechanics in the elastic field the solids are all hookians,
in the obvious sense that they all obey the relationships of Navier. 

After that, we can choose τdr in the characteristic number (5.2.8)

τdr
.
ρr v2r
We suppose that the fluid is newtonian. Then the (5.2.15) hold and clearly we can
choose τdr as
vr
τdr = μ1r .
Lr
So it needs to choose a coefficient of viscosity of reference μ1r , a velocity of
reference vr and a length of reference Lr . The μ1r is a function of state and can be
selected in the initial state.
As a consequence for a newtonian fluid the characteristic number (5.2.8)
τdr
ρr v2r
can be written
μ1r
.
ρr vr Lr
Its reciprocal is a dimensionless characteristic number called number of Reynolds
and is denoted with the symbol Re

ρr vr Lr
Re = . (5.2.16)
μ1r
Introducing the kinematic viscosity
μ1ρ
νr =
ρr
we have
vr Lr
Re =
νr
5.2 Fluid Dynamics 519

Unlike the number of Mach, the number of Reynolds contains a length Lr , which
keeps the geometry of the system. For a sphere invested by a fluid tide, Lr will be the
diameter of the sphere. For a team of turbine blade, Lr will be the rope of the blade
or the pitch of the blades, indifferently since they have the same order of greatness.
For a nozzle Lr will be the diameter of the throat.

Remark 5.2.8 As a rule the choice of the length of reference, what must be consistent
with the phenomenon in study, is very delicate. 

Let us give a kinetic interpretation of the number of Reynolds. We write Re in the

form
vr
Re =   ,
νr
Lr

so that Re is the ratio among two velocity. The velocity Lνrr is that with which the
transversal waves are propagated. The transversal waves are the waves that are prop-
agated in the direction normal to the cause that produces it, for instance the bending
vibrations of a tree. The transversal waves are due to the tangential components of
the superficial stress. Contrarily to the velocity of the sound, the velocity Lνrr is a
function of the thermodynamic state and of the geometry of the system.
Let us give a dynamic interpretation of the number of Reynolds. We write Re in
the form

ρr v2r ρr v2r
Re = = (5.2.17)
μ1r vr /Lr τdr

Insofar Re measures the importance of the convective flow of momentum, that

is an ordered flow, in comparison to the diffusive flow of momentum, that is a
dissipative flow.
We notice that M measures the relative importance between the convective flow
of momentum and the diffusive but not dissipative flow of momentum, while Re
measures the relative importance between the convective flow of momentum and
the diffusive dissipative flow of momentum. We also notice that, in terms of forces,
M measures the importance of the resultant one of the inertia forces in compari-
son to the resultant one of the normal not dissipative stress, while Re measures the
importance of the resultant one of the inertia forces in comparison to the resultant
one of the superficial dissipative stresses.
When Re assumes very elevated values, then the importance of the resultant one
of the superficial dissipative stresses is negligible in comparison to the resultant
of the inertia forces. In such case the fluid can be considered as if the superficial
dissipative stresses didn’t exist, with the consequence that the stress tensor reduces
520 5 Thermoelasticity

it to the only isotropic part p̆. The inverse one happens for Re next to 0. In such case,
for instance the lubrication, the inertia forces are negligible in comparison to the
resultant of the superficial dissipative stresses.
Let us now give an energy interpretation of the number of Reynolds. From the
(5.2.17) it follows that Re is the ratio between the kinetic energy for unity of volume
and the dissipative work for unity of volume. This way at high numbers of Reynolds
the fluid is not able to express its ability to perform irreversible work.
Let us give finally an entropic interpretation of the number of Reynolds. From the
energetic interpretation of Re it follows that the number of Reynolds measures the
capacity of the viscosity to produce entropy. This makes the concept of reversibility
or irreversibility operational, because with Re we have found the measure of the
efficacy of the cause that produces the entropy.

Remark 5.2.9 We call more perfect gas a perfect gas deprived of the viscosity. 

Let us now consider the characteristic number (5.2.9) of the equation (5.2.6) of
balance of the momentum

fr Vr
 .
ρr v2r r

It represents the measure of the relative importance of the production of momen-

tum in comparison to the convective flow. In other words its value allows us to say
if the production of momentum is negligible or not in comparison to the convective
flow.
In the (5.2.9) it needs to choose fr , a force for unity of volume able to produce
momentum. The choice of fr depends on the type of force that acts on the system.
For fluid electrically neutral, since they don’t have free charges, obviously the elec-
tromagnetic forces don’t have importance. In such case as fr we choose the force
due to the gravitational field

fr = ρr g

where g is the acceleration of gravity and ρr is the density of reference.

With this the characteristic (5.2.9) number is written

fr Vr gVr
 = 2 .
ρr vr r
2 vr r

In the particular case in which we can set Vr = Lr3 and r = Lr2 , the characteristic
number (5.2.9) is called number of Froude and is given by

gVr
Fr = . (5.2.18)
v2r
5.2 Fluid Dynamics 521

Remark 5.2.10 The number of Froude is of fundamental importance in the naval

applications. 

Setting the number of Froude in the form

g
Fr =
vr /Lr

we see that it measures the relative importance among the forces of gravity and the
forces of inertia.
We have already underlined that the ability to create an effect is different from
to have the possibility and the time to produce it. We have in fact see that all the
fluids have the viscosity, however, dependently of the value of Re , they can behave
as if they were deprived of it. Likewise, for the number of Froude, it is obvious
that all the fluids are heavy. They suffer the effects of the acceleration of gravity.
However when Fr is very small it is as if the fluids were not heavy. The effects of
the acceleration of gravity are therefore negligible.
Finally, it is important to notice that in the number of Froude the greatness g,
independent from the material but not from the density, appears.

Remark 5.2.11 Let us consider the case of a hydrant. If the water flows with small
velocity, vr is small and therefore Fr is great so that the water suffers the effects
of the gravity. In fact, just when it escapes from the hydrant, the throw converges
downward immediately. Vice versa when the velocity of the throw of water is great
it preserves it horizontal a long line because doesn’t suffer any effects of the gravity.
Only in the most distant points, because of the diminution of the velocity of the
throw, the effects of the gravity return to predominate. 

Let us make now dimensionless the equation (5.2.3) of conservation of the

energy. We have

 
ρr ur Vr ∂ ρr νr2 Vr ρ ∗ ν ∗2 ∗
ρ ∗ u∗ dV ∗ + dV
tr ∂t∗ V∗ tr V∗ 2
 
∗ ∗ ∗ ∗ ρ ∗ v∗2 ∗
+ ρr ur vr r u ρ v × n dσ + ρr vr r 3
v × n dσ ∗
∗ ∗ 2
  (5.2.19)
+ pr vr r p∗ v∗ × n dσ ∗ + τdr vr r (τ̆ ∗d × v∗ ) × n dσ ∗
∗ ∗

+ Qr r Q∗ × n dσ ∗ .
∗

In the (5.2.19)all the dimensional groups have the dimensions of a power.

Dividing for ρr v3r r it is gotten
522 5 Thermoelasticity

 
ur Vr ∂ ∗ ∗ ∗ Vr ρ ∗ v∗2 ∗
ρ u dV + dV
tr v3r Σr ∂t∗ V ∗ tr vr Σr V ∗ 2
 
ur ρ ∗ v∗2 ∗
+ 2 u∗ ρ ∗ v∗ × n dσ ∗ + v × n dσ ∗
vr Σ ∗ Σ ∗ 2
  (5.2.20)
pr τdr
+ p ∗ ∗
v × n dσ ∗
+ (τ̆ ∗ × v∗ ) × n dσ ∗
ρr v2r Σ ∗ ρr v2r Σ ∗ d

Qr
+ Q∗ × n dσ ∗ .
ρr v3r Σ ∗

In the (5.2.20) they appear the characteristic numbers

pr τdr
, ,
ρr v2r ρr v2r

what we have already met, and the characteristic number

Vr
tr vr r
for which it obviously results
Vr 1
= .
tr vr r Str

This way the equation (5.2.20) furnishes the three new characteristic numbers

ur
(5.2.21)
v2r
ur Vr
(5.2.22)
tr v3r r
Qr
, (5.2.23)
ρr v3r

The characteristic number (5.2.21) measures the relative importance between

the convective flow of inside energy and the convective flow of kinetic energy. To
choose the specific inside energy of reference, we remember that for the (5.1.5) in a
more perfect gas the inside energy is only function of the temperature (cf. Remark
5.2.9) and then

u = cv T .

Insofar, since the velocity of the sound a, for a perfect gas, is

a2 = γ RT =γ (cp − cv )T = γ (γ − 1)cv T , (5.2.24)

5.2 Fluid Dynamics 523

we have

a2 = γ (γ − 1)u

so that
ur a2 1
= = .
2
vr γ (γ − 1)vr
2 γ (γ − 1)M 2

With this we have obtained that the relative importance among inside energy and
kinetic energy depends on the number of Mach.
Remark 5.2.12 It needs to expressly notice that, on the plan of the molecular struc-
ture of the matter, the inside energy is also a kinetic energy, due to the molecular
motions. The inside energy is also called disorderly energy. On the contrary the
macroscopic kinetic energy (due to the motion of the molecules in a same direction)
is also called ordered energy. 
About the characteristic number (5.2.22)

ur Vr
,
tr v3r r

if we can set Vr = Lr3 and r = Lr2 , we have the same conclusion

ur Vr ur 1
= 2 = .
tr v3r r vr γ (γ − 1)M 2

Let us now consider the last characteristic number (5.2.23). To choose Qr , we

remember that, in a linear phenomenology and in absence of diffusive flows of mass,
Q is given by the law of Fourier

∂T
Q = −λ . (5.2.25)
∂r
As a consequence we assume as heat flow of reference

λr Tr
Qr = .
Lr
So the characteristic number (5.2.23) can be written

Qr λr Tr
=
ρr v3r ρr v3r Lr
from which
Qr λr Tr
=
ρr vr
3 μ1r Re v2r
524 5 Thermoelasticity

from which, taking into account the (5.2.24)

Qr λr a2 λr
= · = .
ρr vr
3 μ1r Re vr γ (γ − 1)cv
2 μ1r Re M (γ − 1)cp
2

To sum up, the characteristic number (5.2.23) depends from the number of
Reynolds, from the number of Mach and from the dimensionless characteristic
number
cp μ1r
Pr = (5.2.26)
λr
that we call number of Prandtl.
Unlike Re and M, the number of Prandtl doesn’t contain the velocity of ref-
erence. As a consequence it doesn’t take into account the dynamics of the process.
Instead in Pr the specific heat to constant pressure cp , that is a thermodynamic great-
ness of equilibrium, appears. The kinetic coefficients λr and μ1r , introduced in the
Thermodynamics of the irreversible processes but function only of the state of the
system, appear too.
Insofar the number of Prandtl is a function of state. For small variations of pres-
sure, Pr varies only with the temperature, or rather such variation is so much weak
that as a rule we can assume Pr constant.
While Re and M have an ample field of variability, as for the order of greatness
of Pr it results

gas Pr ∼
=1
normal liquids Pr ∼= 10
∼
heavy oils Pr = 100
1
molten metals Pr ∼= .
100
Let us give an entropic interpretation of the number of Prandtl. Evidently since
the (5.2.26) Pr measures the relative importance of the effects associated to the vis-
cosity in comparison to the effects associated with the thermal conductibility. Then
on the entropic plan Pr measures the relative importance between the production of
entropy associated to the viscosity and the production of entropy associated to the
thermal conductibility.
Remark 5.2.13 Since the number of Prandtl is a function of state it is not possible
to give neither a kinetic interpretation nor a dynamic interpretation. 

5.2.3 Non Dissipative Flows

Let us study the motion of a gas in a divergent convergent plane nozzle (Fig. 5.2.2).
We suppose that the geometry of the nozzle is such that the motion can be considered
5.2 Fluid Dynamics 525

Fig. 5.2.2

steady and one-dimensional, so that the velocity has only one component, that we
denote with the symbol v. Furthermore we suppose that the walls of the nozzle
are impermeable, unworking and adiabatic. Last but not least, we suppose that the
dissipative causes are absent or ineffective.
The motion is isentropic. In fact from a side we have supposed ṡ = 0; from
the other one we have supposed that viscosity is ineffective to the objective of the
production of entropy.
In consequence, being Pr ∼ = 1, also the effects associate with the thermal
conductibility are negligible. So the fluid particle that crosses the duct stirs with
adiabatic and isentropic motion.
Let us impose the conservation of the mass. We observe that the flow of mass
is ρν. Since the motion is steady and one-dimensional, said A(x) the section of the
nozzle of abscissa x, the equation of the conservation of the mass furnishes

ρvA = const . (5.2.27)

Let us impose the conservation of the energy. Since the motion is steady only the
flows must be considered. The specific convective flow is

v2
ρv u + .
2

As seen previously, the diffusive flow in the form of heat is ineffective. The
diffusive flow in the form of irreversible work is ineffective since ṡ = 0. The spe-
cific diffusive flow in the form of reversible work is pv. Insofar the equation of
conservation of the energy is

p v2
ρv u + + A = const . (5.2.28)
ρ 2

We introduce the specific enthalpy

p
h=u+ (5.2.29)
ρ
526 5 Thermoelasticity

what evidently5.2.5 is a thermodynamic greatness, that is a function of state. Then

the (5.2.28) becomes

v2
pVA h + = const .
2
As a consequence, since the (5.2.27)

v2
h+ = const , (5.2.30)
2
namely the sum of the specific enthalpy and of the kinetic energy for unity of mass
is constant.
Let us notice that in the two equations (5.2.27) and (5.2.30), written for simulat-
ing the problem of the nozzle, the unknown ones are ρ, h, v In fact the geometry of
the nozzle is a datum of the problem, so the function A is known.
Remark 5.2.14 We observe that to derive the two equations (5.2.27) and (5.2.30) no
hypotheses on the thermodynamic nature of the matter were made. 

We denote with h0 and we call enthalpy of stagnation the greatness

v2
h0 = h + .
2
Since the (5.2.30) h0 is the specific enthalpy at v = 0, in conditions of stagnation.
In conditions of stagnation another important greatness is the pressure of stagna-
tion p0 . It is just the pressure that an one-dimensional tide reaches when is brought
to the velocity zero by a isentropic and reversible process.
Problem 5.2.1 We consider a container in which there is a gas of known ener-
getic content (Fig. 5.2.3). You shall analyze the behavior of the fluid when a hole
of communication with the external environment is opened in the container.
Solution If in the container the pressure is the exterior atmospheric pa , putting in
communication the container with the external environment through a hole or a
nozzle there is no outflow of the gas from the container. If instead the same gas has
the same energetic level but has a pressure greater than pa , an outflow of gas will be
produced if we put the container in communication with the outside. The energy in
the form of heat turns it into kinetic energy. 

Fig. 5.2.3

5.2.5 The enthalpy H = U + p V is a thermodynamic potential.

5.2 Fluid Dynamics 527

Remark 5.2.15 The energy in the form of heat is due to the irregular motion of the
molecules. It takes the name of disorderly energy. We have seen in Remark 5.2.12
that, in the opposite way, the kinetic energy takes the name of orderly energy. 

From the (5.2.27) it follows that for incompressible fluids, for which obviously
ρ = const, is

vA = const .

For them therefore the velocity ν is inversely proportional to the section A.

Let us suppose now that the fluid is compressible, i.e. that ρ is variable. The
elastic interpretation of M tells us that M plays in the phenomenon a non negligible
role. The kinetic interpretation of M tells us that if M < 1[resp. M > 1] the velocity
of the fluid is inferior [resp. superior] to that with which the small perturbations are
propagated in the fluid. It is therefore probable that at M = 1 the fluid has a peculiar
behavior. From the (5.2.27) for a more perfect gas we get, in the hypothesis that the
flow is steady, one-dimensional, isentropic

dA dv
= −(1 − M 2 ) . (5.2.31)
A v

A > 0. In con-
In the part of the nozzle in which the duct diverges it results dA
sequence if M 2 < 1 then it must be dv v < 0. Therefore in subsonic regime the
qualitative behavior of the gas is analogous to the one of the incompressible fluid.
An increase of section corresponds to a diminution of velocity. If M 2 > 1 it hap-
pens instead the opposite one, so in supersonic regime to an increase of section,
an increase of velocity happens. This is an entirely opposite behavior to that of
the incompressible fluid. The case M 2 = 1 cannot be verified because the second
member must be greater than 0.
Remark 5.2.16 If M 2 = 1 the case dv
v = +∞ is to discard for physical reasons and
has however never been found. 

A < 0. Reasoning
In the part of the nozzle in which the duct converges it results dA
as in the divergent case, we obtain that the velocity is or always increasing or always
decreasing and the case M = 1 cannot happen.
We now consider the case in which dA A = 0. In such case the duct has a point
of maximum or of minimum for the section. In the nozzle the point of minimum
only has practical interest. The corresponding section is called section of throat or
simply throat. In the throat, excluding the case dv v = +∞ (cf. Remark 5.2.16), it
must result or M = 1 or, if M = 1, dv v = 0. As a consequence the velocity in throat
must have a maximum or a minimum.
Remark 5.2.17 In the case M = 1 the (5.2.31) furnish dA A = 0 so in the throat the
fluid has the velocity of the sound. In other words, we can have M = 1 only in the
throat. Still, if M = 1 in throat then M = 1 in the whole nozzle, that is the nozzle
works all in subsonic regime or all in supersonic regime. In precedence we have
528 5 Thermoelasticity

observed that when M = 1 a peculiar behavior of the fluid is probable. We have now
seen that this peculiar behavior can happen only in a peculiar point of the nozzle,
that is in the throat. 

Having in precedence performed the qualitative analysis of the problem of the

nozzle, we now pass to the quantitative analysis. To do this we must use the
equations that simulate the phenomenon. In them we perform the simplifications
opportune in the case in examination, using the theory of the characteristic num-
bers. We have supposed that the walls of the nozzle are adiabatic, unworking,
impermeable and that in the motion all the causes of dissipation are ineffective.
Insofar the terms of the equations are negligible in which appear the numbers of
Reynolds, of Prandtl, of Strouhal, of Froude. The only cause able to provoke vari-
ations of the fluid dynamics state in the nozzle is the variation of the area of the
section. Accordingly if the duct’s section was constant then everything would stay
unchanged along the axis of the duct. Only when the geometry is varying the fluid
reacts to the new conditions. This expresses the physical fact that, without exchanges
with the outside and the inside, productions being absent, there is not possibility to
vary the thermodynamic state of the system.
The unknown greatnesses of the problem are the pressure p, the temperature T,
the velocity ν, the enthalpy h. We have already seen that they, provided that are
expressed in dimensionless way, can only depend on M, on γ ( that depend from the
considered gas) and on the distribution of the areas made dimensionless. So it needs
to made dimensionless5.2.6 p, T, ν, h.
As for the enthalpy

p
h=u+
ρ
2
we have already seen that in the problem of the nozzle h + v2 , a measure of the ener-
getic level of the fluid, is constant. So it appears opportune to assume as enthalpy of
reference that of stagnation, at v = 0. Equally we assume for Tr and pr the values
of T and p under conditions of stagnation. Similarly as density of reference ρr we
assume that of stagnation. For the velocity vr , as usual in Gas dynamics, we choose
that of the sound a. For the section of reference Ar we choose that of throat Ag .
Insofar for the (5.2.27) the number of Mach

v
M=
a
depends only from γ5.2.7 and A/Ag . Moreover

5.2.6 This is the most arduous assignment in the analysis of a fluid dynamics problem, because a
wrong choice leads as a rule to wrong results.
5.2.7 For the air γ =1.4.
5.2 Fluid Dynamics 529

p T h
, ,
pr Tr hr
depend only from M and from γ, i.e. from the type of gas. Such functions are
universal, meaning for a datum type of gas are the same for all the nozzles.
Let us determine at first the function TTr (γ , M). From the relationship h = cp T,
valid for a more perfect gas, being

v2
cp T + = const = hr = cp Tr ,
2
we obtain

v2
cp Tr = cp T 1 +
2cp T
so that

T v2
1= 1+
Tr 2cp T
so that

T 1
= 2
. (5.2.32)
Tr 1 + 2cvp T

It results

v2 v2 v2 v2
M2 = = = = ,
a2 γ RT γ (cp − cv )T γ (γ − 1)cp T

and then, since the (5.2.32), we obtain

T 1
= .
Tr M 2 γ (γ −1)
1+ 2

We observe that TTr is a monotonic decreasing function of M.

Let us now consider the dimensionless flow of mass. From the (5.2.27) we obtain

ρAv
= const (5.2.33)
ρr Ar vr

so the dimensionless flow of mass ρv/(ρr vr ) varies inversely to A/Ar . As it is easy

to verify, the diagram of the dimensionless flow of mass in function of M 2 is that of
Fig. 5.2.4.
We see in such diagram that at M = 0 it results v = 0 so that the flow of mass is
0. Moreover for M → +∞ it results ρ → 0 and ρv → 0 . Finally for M = 1 we
530 5 Thermoelasticity

Fig. 5.2.4

have a point of relative maximum, as we can obtain also by the fact that in throat
A/Ar has a point of relative minimum.
The curve of Fig. 5.2.4 is the fundamental curve of the one-dimensional motions.
It checks therefore also the working of the nozzle. Known A/Ar , with the (5.2.33)
we calculate the dimensionless flow of mass and with such value we draw M. From
the value of M we draw then all the other greatnesses.
Remark 5.2.18 The curve of Fig. 5.2.4 for M very great can not be used. In fact if
M → +∞ then ρ → 0. As a consequence the fluid cannot be more considered
continuous. 

Remark 5.2.19 The maximum flow of mass realizes it only if in the throat the value
M = 1 is reached. Values of M different from 1 in the throat give however an inferior
flow of mass. The maximum flow of mass multiplied for the section of the throat
furnishes the maximum flow that can be carried through the nozzle. It is absolutely
impossible to increase such flow. 

Remark 5.2.20 From the diagram 5.2.4 it is deduced that an assigned value of the
dimensionless flow of mass can be gotten both in subsonic regime and in supersonic
regime. 

Remark 5.2.21 From the diagram 5.2.4 it is deduced that in subsonic [resp.
supersonic] regime the flow is an increasing [resp. decreasing] function of M. 

Remark 5.2.22 The values that the greatness p, ρ, v, T assume in throat are called
critic. 

Let us suppose that in the nozzle the motion is subsonic. Then the distribution
of dimensionless pressure p/pr along the nozzle is given by the curves a and a of
Fig. 5.2.5. In the section of exit the pressure has a value pu that necessarily must be
equal to the pressure pa of the environment. In fact if it were pu = pa , some troubles
would originate that would be propagated toward the entry of the nozzle to vary the
fluid tide in such way that pu = pa . So in the section of exit p/pr has the value
A = pa /pr (Fig. 5.2.5). If pa decreases clearly it happens that A get down towards
B and the curves a and a get down towards the curves b and b that are when in
throat M = 1. Insofar b is a limit curve. In fact if it still goes down it would be in
throat M > 1.
5.2 Fluid Dynamics 531

Fig. 5.2.5

Analogously if in the nozzle the motion is supersonic. In such case the distri-
bution of dimensionless pressure p/pr along the nozzle is given by the curves that
have as limit curves the curves c and c (Fig. 5.2.5). The curves c and c are the same
when in throat M = 1.
When M = 1 in throat and the flow is subsonic in the convergent part and super-
sonic in the divergent part, then the working of the nozzle is shown by the curves b
and c . Denoting with pi the pressure at the entry section of the nozzle, in Fig. 5.2.5
the points B [resp. C] and B [resp. C ] give pi /pr and pu /pr if the motion is subsonic
[resp. supersonic] and M = 1 in throat.
On the basis of the preceding analysis, the nozzle cannot work if the pressure
of exit, i.e. of the environment, is such that pu /pr is included among B and C
(Fig. 5.2.5). However the experience in the laboratory reveals the contrary. This
means that under such conditions someone of the hypotheses is invalid.
The hypothesis of steady motion cannot be responsible of the gap that we have
found because the aforesaid experiment can be conducted in steady regime. The
same also happens for the hypothesis of one-dimensional motion. Evidently nothing
can be charged to the hypotheses of impermeability and unworking, since they are
properties of the walls of the nozzle. It stays only the hypothesis of ineffectiveness
of the dissipative phenomena, made when we have supposed that the production of
entropy is zero and that in consequence the motion is isentropic.
Therefore in the nozzle something happens which provokes an increase of
entropy. We had set ṡ = 0 in base to the fact that Re ∼ = +∞. Then we must
re-examine the discourse done for
ρvL
Re = .
μ1
To appraise Re we have chosen as ρr that of stagnation, vr = a and Lr = Ag ,
where Ag denotes the diameter of the throat. Evidently when pu /pr is included
among B and C the choice of Lr is incorrect. In fact under such conditions a
532 5 Thermoelasticity

phenomenon happens in which Lr is smaller, so Re → 0 and then a meaningful

dissipative phenomenon happens.
Such phenomenon is called shock wave and it is not necessary to prove its exis-
tence since we have met it by way of experiment. Said molecular free walk the
middle distance that intervenes among two collisions of the molecules, we call shock
wave a region of the space of the thickness of few molecular free walks through
which it had a conversion of orderly energy in disorderly energy. The orderly energy
is the kinetic energy ρv2 /2 while the disorderly energy is the thermal energy, i.e. the
energy of agitation. Through the shock wave some collisions among the molecules
are verified, that destroy the order of the fluid tide. As a consequence some entropy
is produced. It is clear that the number of Reynolds related to this phenomenon
must be appraised taking as length of reference the thickness of the shock wave, so
Re ∼= 0. Since the thickness of the shock wave is very small, we will treat it on the
macroscopic plan as a surface of discontinuity.
Remark 5.2.23 Obviously in the rarefied layers of the atmosphere the shock wave
can reach a thickness of some centimeters. Vice versa under normal conditions its
thickness is around 10–8 cm. 

Remark 5.2.24 We notice that the phenomenon of the shock wave doesn’t happen
only in the problem of the nozzle. An explosion or a water hammer are in fact other
examples of shock wave. 

To establish the properties of the shock wave, as usual we will give a qualitative
analysis speech to physically understand the phenomenon. After that we will pass
to the quantitative discourse keeping in mind that, anything happens in the shock
wave, neither the laws of conservation nor the equations of balance cannot be vio-
lated. However, since the phenomenon is macroscopically discontinuous, we cannot
formulate these laws or equations in differential terms, but on the contrary only in
finite terms.
Through the shock wave an abrupt conversion of orderly energy in disorderly
energy happens. So the kinetic energy behind the shock wave is smaller from the
kinetic energy before the shock wave5.2.8

v22 v2
< 1.
2 2
This way the velocity of the fluid vein decreases through the shock wave with
discontinuity from the value v1 to the value v2 .
Remark 5.2.25 If instead of using the macroscopic scale we use as unity of length the
molecular free walk, the shock wave occupies a finite region of the space, through
which v varies with continuity from the value v1 to the value v2 . Nevertheless inside
the shock wave there are very high gradients of velocity, such to give effectiveness
to the viscosity in the regards of the production of entropy. 

5.2.8 We denote with the index 1 [resp. 2] the greatness before [resp. after] the shock wave.
5.2 Fluid Dynamics 533

Let us now see in what conditions a shock wave can be verified. Since through the
shock wave it occurs the conversion of orderly energy in disorderly energy, we must
expect that such conversion can take place only if we have enough energy ordered to
convert. The number that gives the measure of the orderly energy in comparison to
that disorderly is M. Insofar we have enough energy ordered to convert in disorderly
energy only if M > 1. Then in the steady motions the shock wave can happen only
in supersonic regime.

Remark 5.2.26 The water hammer is a shock wave that is propagated. Since the
velocity of the sound in the water is elevated (1,500 m/s), in the water duct the super-
sonic regime is not attainable. In the case of the water the shock wave is established
in unsteady regime. In such conditions in fact the energy is the sum of that orderly
and of that disorderly associated to the unsteady motion. Insofar in the unsteady
regimes the shock wave can happen for every number of Mach. 

We also underline that at valley of the shock wave it is always M < 1. In fact at
valley of the shock wave the disorderly kinetics energy is greater than the orderly
kinetics energy. This way, if in a nozzle there is a current in steady supersonic
regime, a shock wave can happen and to valley of the shock wave the motion is
subsonic.

Remark 5.2.27 We expressly notice that in the phenomenon of the shock wave the
motion is macroscopically discontinuous and not the matter. 

Let us determine now the order of greatness of the time that the fluid current
employs to cross the shock wave. Since the velocity of the sound in the air, under
standard conditions, is of 340 m/s and the space to cover is of the order of 10–8 cm,
the fluid particle crosses the shock wave in a time that results to be of the order of
10–12 s. It is a very brief time and we ask us if it is possible to extract or to administer
energy to the particle. Or, in other terms, if the total energy of the particle

v2
H =h+
2
varies. Evidently in order that this happens it needs in action processes that have the
same rapidity of the shock wave. There are only the processes of radiation, in which
electromagnetic waves are manifested, that are propagated with the velocity of the
light. Then only in presence of intense electromagnetic fields the energetic level of
the mass that crosses the shock wave can be altered.
In absence of such fields it results

H1 = H2

and we will say that in the shock wave there is an abrupt passage from orderly
energy to disorderly energy, with production of entropy and with invariance of the
total energetic level.
534 5 Thermoelasticity

Remark 5.2.28 We point out that in Thermodynamics an order can be brusquely

breaken, but it is impossible to brusquely reestablish it. In other words we can-
not convert with an equally brusque passage the disorderly energy in orderly
energy. 
Let us now see the variations of the thermodynamic greatness through the shock
wave.
About the temperature, since v2 < v1 we have h2 < h1 from which, if the gas is
perfect, being the entropy an increasing function of the temperature, it follows

T2 > T1 .

About the density, the conservation of the mass involves

ρ1 v1 = ρ2 v2

so that
ρ2 < ρ1 .

About the pressure, we perform the balance of the momentum. There is not pro-
duction of momentum inside the shock wave. In fact the number of Froude is very
small and therefore the forces of mass don’t produce momentum. The convective
(orderly) flow of momentum is ρvv, where ρv is the density of momentum. The
pressure p is the diffusive not dissipative flow of momentum. The diffusive dissipa-
tive flow of momentum is zero. In fact we are considering flows, before and behind
the shock wave, in which gradients of velocity don’t exist. Besides the number of
Reynolds is small only inside the shock wave, where there are dissipative stresses,
while at the outside is sufficiently high. This way

(ρ1 v1 )v1 + p1 = (ρ2 v2 )v2 + p2

from which, being ρ1 v1 = ρ2 v2 and v2 < v1 , it follows

p2 < p1 .

About the entropy, it is produced inside the shock wave. These are not instead
exchanges with the outside because, to have them, the presence of incandescent
matter would be needed inside the shock wave, to radiate energy in the form of heat.
Insofar
S2 > S1 .

About the temperature of stagnation, for perfect and more perfect gas it stays
unchanged. In fact it is a measure of the energetic level of the current, that, as we
have seen, is constant.
About the pressure of stagnation, since it measures the possibility of kinetic
exploitation of a constant energetic level, from v2 < v1 it follows p2r < p1r .
5.2 Fluid Dynamics 535

Fig. 5.2.6

Problem 5.2.2 You shall determine the propelling force of the turbojet of Fig. 5.2.6.
The data are: altitude 7,000 m; M1 = 0.8; Ω2 = 1.35 Ω1 , T2 = 800 ◦ C, gνΩ =
15 kg/s.
Solution At the altitude of 7,000 m it results p1 = p2 = 0.40 bar, T1 = 242◦ K.
Denoting with S the propelling force of the turbojet, we have

S = (p2 + ρ2 v22 )Ω2 − (p1 + ρ1 v21 )Ω1 .

About v1 we have
3
2 1.4 · 242 · 8, 316 m
v1 = M1 γ RT1 = 0.8 = 249 .
28.96 s
About v2 we have

v1 Ω1 ρ1 v1 T2 249 · 1, 073 m
v2 = = = = 820 .
Ω2 ρ2 1, 35T1 1.35 · 242 s
About Ω1 we have

p1 p1 M12 γ 0.40 · 104 · 0.82 · 1.4 kg

ρ1 = = 2
= = 5.6 · 10−2 3
RT1 v1 2492 m
so that
ρvΩ 15
Ω1 = = = 1.07 m2 .
ρ1 v1 5.6 · 10−2 · 249
As a consequence

S = (p2 +ρ2 v22 )Ω2 −(p1 +ρ1 v21 )Ω1 = 10, 063 kg . 

Let us now pass to the quantitative analysis of the shock wave. We can proceed
by analytical way, with the equations

H1 = H2
ρ1 v1 = ρ2 v2
ρ1 v21 + p1 = ρ2 v22 + p2 ,

or making dimensionless the greatness, as done up to now. Following this second

way, it is necessary to choose the greatness of reference, as the temperature of
536 5 Thermoelasticity

stagnation and the pressure of stagnation. It deals with choosing among the values
before and behind the shock wave. It is convenient to adopt as values of reference
those before the shock wave.
The intensity of the shock wave depends on the level of the orderly energy in
comparison to the level of the disorderly energy. Therefore the intensity of the shock
wave grows to the growth of M.
If M < 1, as we have seen, the shock wave cannot verify it.
If M = 1 the shock wave is evanescent because the two energies, before and
behind the shock wave, are equal and there are not discontinuity through the shock
wave. In such case all the ratio between the values of the greatness before and behind
the shock wave have unitary value.
We suppose therefore M > 1. When in the nozzle the shock wave happens, the
distribution of the pressure before such wave will be given by the curve characteris-
tic of the supersonic motion. In correspondence of the section where the shock wave
is verified, there will be a jump of pressure. After this section the subsonic regime
is established and the pressure p will increase with the increase of the section until,
in the exit section, p will equalize pa (point Q of Fig. 5.2.7). After the shock wave
the nozzle continues on working but is degraded. In fact before the shock wave the
enthalpy is integrally converted in kinetic energy; behind the shock wave the nozzle
executes the conversion with a certain degradation. So the entropy of exit is greater
than the entropy of entry. If the section in which the shock wave is verified is that of
exit, then p2 is individualized by the point P4 of Fig. 5.2.7.
This way if the pressure of the environment pa is included among B and P4 ,
between the throat and the exit section a shock wave normal to the nozzle axis is
verified.
What happens if pa is smaller than the pressure corresponding to the point P4
of Fig. 5.2.7? We evidently must have through the shock wave a smaller jump of
pressure. As it is experimentally verified, this can be obtained by establishing an
oblique, rather than normal, shock wave (Fig. 5.2.8). In fact in such situation the
tangential component of the velocity doesn’t vary (v1t = v2t ), as we easily verify
by imposing the conservation of the tangential component of the momentum. In

Fig. 5.2.7
5.2 Fluid Dynamics 537

Fig. 5.2.8

Fig. 5.2.9

consequence the intensity of the shock wave is decreased, because of the orderly
energy only the part normal to the shock wave is turned into disorderly energy.
Obviously the shock wave is less intense as its inclination increases, because in
such way vn decreases more and more.
Remark 5.2.29 After an oblique shock wave the flow can be both subsonic and
supersonic. In any case v2n < v1n (Fig. 5.2.8). 

In conclusion, assigning the geometry and the fluid, the nozzle can work in end-
less ways. To individualize one of it, an additional boundary condition is needed,
that in an ordinary case is the pressure pa . With reference to the Fig. 5.2.9, if pa /pr
is included among P1 and P2 , the regime is subsonic and the shock wave is not
established. If pa /pr is included among P2 and P4 a normal shock wave is estab-
lished. If pa /pr is included among P4 and P3 an oblique shock wave is established
in the section of exit. If pa /pr is inferior to P3 a disorderly expansion is established
outside of the nozzle.

5.2.4 Dissipative Flows

Let us study now the one-dimensional motions in which the production of entropy
is efficacious. Preliminarily we observe that always, adherent to any solid surface,
there is a layer of fluid inside which viscosity is efficacious. We call it limit layer.
Inside the velocity of the fluid decreases up to annihilate on the wall, so that the
profile of velocity is of the type of Fig. 5.2.10. We denote with δ the thickness of
538 5 Thermoelasticity

Fig. 5.2.10

the limit layer. To make it dimensionless, we choose as greatness of reference D a

meaningful dimension of the problem in analysis. In the case of the nozzle, in which
the motion was isentropic, Re was so elevated that we can think Dδ = 0.
We now study the motion of a fluid in a duct with constant section. We want
analyze the influence on the motion of the viscosity, and therefore of the superficial
attrition on the walls. We will see that in this problem Re is not sufficiently high to
neglect δ/D.
Evidently in this problem it is worthwhile to choose as dimension of reference D
the diameter of the duct (Fig. 5.2.11).
To rigor we should also bring in account the effects of the thermal con-
ductibility. The number of Prandtl is a measure of the relative importance of the
effects associated to the viscosity in comparison to those associated to the thermal
conductibility.
However we will suppose Pr elevated, that is that the external surface of the duct
is protected in such way to prevent any thermal exchange. To the inside instead
every exchange of momentum will be associated with exchanges of energy in the
form of heat.
Insofar for hypothesis the motion is one-dimensional, Re is such that the effects
of the viscosity are not negligible and the wall of the duct is adiabatic, in addition
to impermeable and unworking. We can study this type of motion applying the law
of conservation of the mass. We get

ρvA = const

from which it follows ρv = const, being A = const. We impose then the conser-
vation of the energy and the balance of the momentum. About this last we should
say that the total flow of momentum, convective and diffusive, that enters is equal
to what goes out minus the contributions due to the walls, what are different from
zero because an efficacious tangential effort exists.
Instead of writing, in a purely analytical way, a system of three equations in three
unknowns, i.e. the velocity and two unknowns of state, we analyze the problem in
the way that follows.
As the laws of conservation of the mass and of conservation of the total energy
H = h + v2 /2 must be satisfied, the solution of our problem must be tried to

Fig. 5.2.11
5.2 Fluid Dynamics 539

find in the family of the fluid dynamics states characterized from ρv and H con-
stant. Obviously, since we have fixed two of the three available parameters, the fluid
dynamics states of this family will depend on one and on only one parameter.
This way in the plane s, h the set of the points individualized by this family must
be a figure to one degree of liberty, that is a curve. This curve, to ρv and H constant,
is called curve of Fanno.

Remark 5.2.30 With this method of analysis we don’t study only a particular one-
dimensional motion, but we get some general results that can also be used for the
examination of other problems. This also happens in thermodynamic field, when we
study the family of states characterized from the same entropy (isentropic state)
or from the same temperature (isotherm state) or from the same pressure (iso-
baric state) or from the same volume (isochoric state) or from the same enthalpy
(isenthalpic state). Gotten these general results, we will have some advantages. For
instance if an isentropic transformation is met then we will already know what its
properties are. In the case in examination the situation is analogous, with the only
difference that deals with a state fluid dynamics rather than thermodynamic. 

We study therefore the general properties of the families of states characterized

from ρv and H constant. These properties will be independent from the particular
problem in examination of the duct with constant section. We immediately observe
that these results can also be used for studying some questions related to the prob-
lem, already examined, of the nozzle. In that case, in fact, in the exit section ρv and
H had constant result, whatever was the position of the shock wave.
The curve of Fanno is gotten using the two equations that express the constancy
of ρv and H. So we can express two unknowns in function of the third one. Then
introducing the third equation of state, that specifies the mean in which the transfor-
mation is operating, in the plane s, h we have one equation in only one parameter.
Particularly as independent parameter it can be assumed M 2 . This way we trace
in the plane s, h the curve characterized by a particular value of ρv and from a
particular value of H = h + v2 /2. We get a curve of the type of Fig. 5.2.12.
The kinetic energies v2 /2 are measured beginning from the horizontal line
H=const. For M = 1 we have a maximum of entropy. In the superior branch we
have M < 1 and in the inferior M > 1. In fact it must be

γ −1 2
H =h 1+ M .
2

Fig. 5.2.12
540 5 Thermoelasticity

In the duct with constant section in examination the motion is not isentropic.
Obviously in this case we can go along the curve of Fanno or toward AB or toward
CB.
This way if the motion in the section of entry is subsonic, in the following sec-
tions the velocity will increase. In an isentropic regime the velocity would be stay
constant, therefore it is the viscosity that provokes the increase of velocity.
If instead to the entry the motion is supersonic the presence of the viscosity
creates a diminution of the velocity. Then M tends always to 1.
If we depart in a subsonic regime and the duct is sufficiently long, in a section
we will reach the value M = 1. In the following sections we can have or M = 1 or
M = 1. We experimentally verify that M = 1. Nevertheless the motion, if was simu-
lated by the curve of Fanno of Fig. 5.2.12, would be characterized by a smaller value
of the entropy, thing that cannot happen. Therefore a phenomenon arose which has
changed the state of the fluid dynamics current.
This phenomenon is said throttling. It deals with a trouble that is propagated
toward the mount provoking the diminution of the flow. After the throttling the
motion is represented by another curve of Fanno, related to the new value of the
flow. The new curve of Fanno stays on right of that preceding to be able to allow
greater production of entropy.
If instead the regime is initially supersonic, some shock waves are born that stay
in the duct or they travel toward mount to change the conditions of the system to
reduce the flow of mass.

Remark 5.2.31 We have seen that a connection exists among M and the entropy.
Precisely we have seen that at M = 1 the entropy has a maximum. 

Problem 5.2.3 You shall study the motion of a gas in a convergent nozzle to which a
duct with constant section is connected (Fig. 5.2.13).

Solution In the nozzle the motion can be considered isentropic, but we cannot affirm
the same thing for the duct. In the convergent nozzle the flow is maximum for
M = 1 in the exit section. If the fluid enter in the duct with M = 1 we have the
maximum entropy, as the curve of Fanno shows. Therefore the attrition met by the
fluid along the duct bring to a diminution of flow. Then the maximum flow allowed
by a convergent nozzle, that happens when M = 1 in the exit section, is greater than
that allowed by a convergent nozzle connected to valley to a duct with constant sec-
tion. Obviously the flow of mass allowed by this last will be much lower as the wall
of the duct becomes rougher. 

Fig. 5.2.13
5.2 Fluid Dynamics 541

Remark 5.2.32 In the case of duct with constant section the temperature decreases
if the regime is subsonic and increases if the regime is supersonic, while for the
velocity the inverse one happens. In fact h respectively decreases and increases. 

In the duct with constant section, in subsonic regime M varies because of the
production of entropy in the limit layer. The entropy is produced by the superficial
tangential stresses and increases with the growth of the distance from the section of
entry, up to a maximum value that is reached in the section in which the throttling
happens.
In the duct with constant section, in supersonic regime we are under more dis-
advantageous conditions because a shock wave can rise up that provokes an abrupt
variation of entropy. If it rises up, travels in the duct up to the section of entry and
changes the flow.
Another example of family of Fanno is the set of the fluid dynamics states in
the section of exit u of a convergent-divergent nozzle. We get, with reference to the
Fig. 5.2.9, that

– the point P3 represents the fluid dynamics state in u when the motion in the nozzle
is everything supersonic and isentropic,
– the point P2 represents the fluid dynamics state in u when the motion in the nozzle
is everything subsonic and isentropic,
– if in u there is a shock wave P3 represents the state before and P4 that behind,
– since P2 represents the fluid dynamics state in u when the shock wave is had in
the throat and P4 represents the fluid dynamics state in u when the shock wave is
in u, the points included among P4 and P2 represent the fluid dynamics state in u
when the shock wave is on the right of the throat and on the left of u,
– if in the curve of Fanno we trace the isobars, knowing pa /pr it is immediate to
graphically find the position of the shock wave.

Let us analyze now another fundamental type of motion, that will furnish a sec-
ond family of fluid dynamics states. We consider a duct with constant section in
which the motion is one-dimensional and steady. We suppose that through the walls
an exchange of energy happens, but only in the form of heat. We suppose that the
attrition along the walls is negligible5.2.9 . In this problem is interesting to know the
final conditions when the initial conditions and the administered power are assigned.
Or it interests to know the necessary power to pass from the initial state to that final.
As of custom, we write the equations of balance. The conservation of the mass
furnishes, since the duct has constant section

G = ρv = const . (5.2.34)

5.2.9 For instance a duct submitted to the solar energy. Another class of problems reenters in this
scheme, the rooms of combustion. They are ducts in which a chemical reaction happens that leaves,
in first approximation, the composition of the fluid unchanged and generates heat.
542 5 Thermoelasticity

Fig. 5.2.14

The balance of the momentum furnishes, since the tangential stresses are absent

I = p + ρv2 = const . (5.2.35)

Let us denote with P5.2.10 the thermal power changed through the surface of the
duct. Then, denoting with i [resp. u] the section of entry [resp. exit] of the duct
(Fig. 5.2.14), the conservation of the energy furnishes

P = ρu vu Au Hu − ρi vi Ai Hi = GA(Hu − Hi ) .

To these three equations the equations of state must be added.

This way the family of fluid dynamics states that appears in this problem is that
for which G = const, I = const. This family, what evidently has only a degree of
liberty, is said family of Reyleigh and is represented in the plane h, s from the curve
of Fig. 5.2.15, what exactly is called curve of Reyleigh. We notice in such curve
the characteristic points A and B. In A we have the maximum enthalpy, in B the
maximum entropy, so that M = 1.
Remark 5.2.33 When we go along the curve of Reyleigh, if the entropy increases
then we are administering heat, otherwise we are subtracting it. 

In supersonic regime the whole administered heat goes to increase the enthalpy.
Besides a part of the kinetic energy turns into enthalpy. Insofar in supersonic regime,
that in Fig. 5.2.15 is the part of curve below B, the administered heat provokes a
deceleration. The inverse one happens in subsonic regime, that in Fig. 5.2.15 is the
part of curve on the left of A. The trunk of curve among A and B is unstable. When
the representative point of the fluid dynamics state happens in this trunk, it jumps

Fig. 5.2.15

-
to the duct a flow of heat Jq is administered, it is Q =  Jq d. If in the duct a reaction
5.2.10 If
happens
- that generates a quantity q of energy for unity of time and for unity of mass, it is Q =
M qdM.
5.2 Fluid Dynamics 543

suddenly in B. This because it happens a chemical reaction due to the fact that
the system is not able to oppose it to the flow of heat caused by the difference of
temperature. Such chemical reaction accelerates the flow and a throttling is verified.
Remark 5.2.34 We denote with Pi the point of the curve of Reyleigh that represents
the fluid dynamics state in the section of entry. Since the transformation that happens
in the duct is reversible, the area below the part of curve BPi measures the maximum
energy for unity of mass that can be furnished to the fluid. In fact reached B and
still administering energy, the entropy should increase. However with that flow the
entropy cannot increase (Fig. 5.2.15). Then a throttling is verified, that varies the
flow. Insofar it is not possible to furnish energy in the form of heat to the fluid
current over a certain limit.
As a consequence in a room of combustion, as the velocity with which the fluid
comes in the duct is great as the energy that can be furnished to the fluid is small.
Vice versa, and this happens in the ram-jet engine, in supersonic regime it is
worthwhile that the fluid comes into the room of combustion with a much elevated
velocity. 

Remark 5.2.35 Since the (5.2.35), the family of Fanno is characterized from H =
const, I = const. Since the (5.2.34), the family of Reyleigh is characterized from
G = const, I = const. Then the points of intersection of the curves of Fanno and of
Reyleigh are characterized by H = const, I = const, G = const. These conditions are
just those that are verified simultaneously in the shock wave. Precisely the point of
intersection that is found on the subsonic branch furnishes the fluid dynamics state
behind the shock wave. That on the supersonic branch furnishes the fluid dynamics
state before the shock wave. 

Remark 5.2.36 The combustion in a duct doesn’t happen at constant pressure, as the
relationship p + ρν 2 = const shows. Only in subsonic motions with M << 1 it is
possible to assume, in first approximation, p = const. 

Let us give now a short notice of other dissipative motions of technical interest.
One is that of the motion in a duct with administration of heat without attrition and
in supersonic regime, in which energy is administered in discontinuous way. In such
case not only shock wave but also detonation wave5.2.11 can happen.
Another phenomenon of discontinuity can happen, analogous to the wave of
detonation, called non adiabatic shock wave. It happens with exchange of energy.
Another classical phenomenon, important for the vapor turbines, is that of the
shock wave of condensation. In it, besides the abrupt conversion of a form of energy
in another, the abrupt passage is by a phase to another. In this type of problem the
independent parameter is the energy generated from the condensation of the vapor.

5.2.11 The wave of detonation is a surface of discontinuity through which the conversion of a form
of energy in another happens. It, unlike the shock wave, doesn t leave the total energetic level
unchanged because some chemical reactions happen due to the collisions of the molecules in the
inside of the wave.
544 5 Thermoelasticity

Let us study now the laminar regime. In the relative motion between fluid and
solid the velocity of the fluid in correspondence of the wall is zero. This is always
true in the hypothesis of continuous fluid. Well a fluid can be assumed continuous
in analytical sense only if its density is sufficiently high5.2.12 . More precisely we
denote with l, and we call molecular free walk, the middle value of the distance
among two collisions of two molecules of fluid. Evidently in an assigned volume l
is as great as the number of molecules contained is small. Equally evident is that l
is a measure of the continuity of the fluid. However l must be compared to a length
Lr characteristic of the phenomenon that we study5.2.13 . We call number of Kundsen
the number

l
Kn = . (5.2.36)
Lr
If Kn is very small we can consider the fluid continuous. If Kn has order of
greatness 1 it doesn’t have sense to speak of adherence between fluid and wall.
Remark 5.2.37 To the height of 100 nautical miles l ∼ = 2 cm. To the altitude in which
they sail the satellites l is of the order of the meter. In the neon tubes l is much greater
than the diameter of the tube. In the whole technology of the extreme vacuum l is
very great. 

We suppose therefore continuous the fluid, so the velocity of creeping between

fluid and wall is zero. Let’s report us to a plane slab of length c invested by a fluid
current (Fig. 5.2.16). We denote with ν∞ the velocity of the current before it invests
the slab.
The slab cannot disturb the current to an endless height. In fact the slab, dis-
turbing the current, provokes a variation of the momentum, that corresponds to a
force. This way if the slab perturbed the whole mass of the fluid current we would
reach the paradox that a finite slab offers an infinite resistance. We must therefore
conclude that to a some distance δ from the slab the velocity of the current has the
same value v∞ of the unmolested current. Insofar the profile of the velocity at the
station x is that of Fig. 5.2.16. The velocity is zero on the slab because of the forces
of adhesion between the fluid molecules and the molecules of the slab. Afterwards it

Fig. 5.2.16

5.2.12 This happens if the number of elementary particles that constitute the fluid is very big.
5.2.13 In the case of motion in a duct we will choose as Lr the diameter of the duct.
5.2 Fluid Dynamics 545

goes increasing up to reach, to the height δ(x), the value v∞ . The region of the space
individualized by the function δ takes the name of zone of limit layer or simply of
limit layer.
We make δ dimensionless assuming as length of reference the length c of the
plate. To determine the order of greatness of δ/c various methods exist. Let us reason
in terms of characteristic numbers. Since δ/c is a dimensionless number of the phe-
nomenon in examination, it must depend on the other dimensionless numbers that
control him. These are5.2.14

Str , M, Re , Pr , Fr .

We suppose steady the motion and incompressible the fluid. Since the motion
is steady we must not take into consideration Str . Since the fluid is incompressible
we must not take into consideration M. Moreover, since in incompressible regime
the equation of the energy is uncoupled from that of the momentum, the field of
velocity is not influenced by the field of temperature. Then the parameters Pr , Fr
must not be taken in account because they intervene only through the equation of
the energy. This way in the problem in examination, for one assigned geometry, the
only parameter on which it depends δ/c is the number Reynolds.

Remark 5.2.38 In the general case the unknown ones of the problem are the velocity
v, the density ρ, the pressure p and the tangential stresses τ , that are functions of
the viscosity μ. For the incompressible fluids ρ doesn’t vary, so constitutes together
with μ a datum of the problem. To determine v and p it is enough to resolve the
system constituted by the equation of the continuity and by that of balance of the
momentum. Then the field of the velocity can be determined without resolving the
equation of the energy. Having determined the field of the velocity, the equation of
the energy allows to determine the field of the temperatures. Instead in compressible
regime to determine the field of the velocity it is necessary to also bring into account
the equation of the energy. 

As a consequence

δ
= f (Re ) . (5.2.37)
c
To render explicit the (5.2.37) it is necessary to have recourse to the balance
equations. We observe however that for the perfect fluid, that doesn’t exist in nature
and for which Re = +∞, the limit layer doesn’t exist5.2.15 .
Let us determine in (5.2.37) the order of greatness of δ/c. For definition the
limit layer is the region of the space inside which the viscous forces have the same
order of greatness of those of inertia. We consider the limit layer of a plane slab

5.2.14 The product Re Pr is called number of Peclet.

5.2.15 A gas perfect slips on the wall.
546 5 Thermoelasticity

Fig. 5.2.17

(Fig. 5.2.17) and an infinitesimal volume of fluid to its inside. The forces of inertia
acting on it are

∂v ∂v ∂x ∂v
ρ dx dy = ρ dx dy = ρv dx dy
∂t ∂x ∂t ∂x
whose order of greatness is

v∞ 2
ρv∞ ε .
c

The viscous forces acting on it are

∂τ ∂ ∂v
dy dx = μ dx dy
∂y ∂y ∂y
whose order of greatness is

v∞ 2
μ ε .
δ2
Equalizing the two orders of greatness we have

v2∞ v∞
ρ =μ 2
c δ
from which
3
δ μ 1
= =√ .
c ρc v∞ Re
Insofar
4
δ 1
=k (5.2.38)
c Re

where k is evidently function of the geometry. The (5.2.38) is very important because
it allows to immediately have some idea of the thickness of the limit layer.
5.2 Fluid Dynamics 547

Fig. 5.2.18

Problem 5.2.4 You should determine the thickness of the limit layer for the series of
turbine blades of Fig. 5.2.18.
Solution Around a blade the limit layer will be of the type of that outlined in
Fig. 5.2.18. In it the effects of the viscosity are suffered and the production of
entropy is concentrated. The profiles of velocity will have the course of Fig. 5.2.18,
so we must expect a production of entropy also after the fluid has abandoned the
blade. We assume as length of reference the pitch t of the blading (Fig. 5.2.18) and
we determine δ for the particular current in examination. If t = c/10 and Re has
order 106 , from the (5.2.38) it follows that δ/t has order 10–12 , that is δ is about
1/100 of t. 

Let us now see which are the effects of the presence of the limit layer.
A first effect is the local diminution of flow. In the general case the mass is
equally able to pass in the zones distant from the limit layer. In the case of a duct
the reduction of velocity in the limit layer irremediably involves a diminution of the
global flow. Then in the design of a duct it is necessary to take into account of the
presence of the limit layer.
A second effect are the tangential stresses, that are forces of resistance to the
relative motion. The efficiency of any form of conversion of energy is connected to
a large extent to the work that must be made for winning these resistances.
A third effect is the diminution of the kinetic exploiting of the fluid mass. When a
fluid current runs over a tapered body some gradient of pressure must be recovered.
To do this kinetic energy must be spent. In the limit layer the velocity is smaller, so
the fluid fillet doesn’t have enough kinetic energy to give and detaches it forming
the vortexes.
A fourth effect is that in the limit layer the regime is stable up to a value of Re .
After that becomes turbulent.
Remark 5.2.39 On the microscopic plan the viscosity is due to a molecular exchange
of momentum. When two fluid currents run in parallel5.2.16 , on molecular scale some
agitations happen such to make an exchange of particles among the two currents.
If the velocities of the two fluid fillets are equal, it is clear that this exchange of

5.2.16 It is the case of two fluid dynamics fillets of current.

548 5 Thermoelasticity

particles in motion (with consequent exchange of momentum) involves neither an

acceleration nor a deceleration of a fillet in comparison to the other. Instead this
happens if two velocities are different. Precisely the fluid fillet having a greater
velocity suffers a deceleration and that one having a smaller velocity suffers an
acceleration. This way the microscopic exchange of momentum has the tendency
to level the velocities of two contiguous fluid fillets. The result on the macroscopic
plan is a tangential stress given by (Fig. 5.2.17)

∂v
τ (y) = μ (y) .  (5.2.39)
∂y
In a fluid dynamics problem the limit layer is the zone within which the viscosity
shows its effects. The current external to the limit layer (Fig. 5.2.17) can be consid-
ered deprived of viscosity and is called external potential current or also external
potential field.
We say that the motion of a fluid current is a laminar regime when the exchange
of particles among two contiguous fluid fillets happens on microscopic scale, that is
molecular. In such case the fluid fillets are parallel, and it is for this that the motion
is called laminar. When instead the exchanges of particles happen on macroscopic
scale, we say that the motion is a turbulent regime. In such case violent exchanges
of particles are verified. As a consequence the contiguous fluid fillets don’t run in
parallel and the vortexes born.
Problem 5.2.5 You should calculate the resistance of attrition offered by the slab of
Fig. 5.2.17.
Solution From the (5.2.39) it follows that on the slab a tangential stress acts5.2.17
given by

∂v
τ0 = μ (0) .
∂y

So the slab is stressed by a force R having the direction of the current and of
intensity

Fig. 5.2.19

5.2.17 As the slab is parallel to the fluid current they are absent forces of pressure and for this the
unique resistance to the motion is due to the presence of the limit layer. If the slab is inclined with
respect to the current (Fig. 5.2.19) the resistance is also due to the forces of pressure. In such case
the current external to the limit layer is meaningfully different from that unmolested. This is not
verified if the slab is parallel to the current.
5.2 Fluid Dynamics 549

R= τ0 dA
A

where A denotes the surface of the slab.

Since we are in the limit layer, τ0 has the same order of greatness of the dynamic
pressure5.2.18 12 ρv2 , so that

1
τ0 = c0 ρv2 . (5.2.40)
2
In the (5.2.40) the coefficient c0 is dimensionless and takes the name of coef-
ficient of resistance or of attrition. It depends from Re , that measures the relative
importance between viscous forces and forces of inertia, from the geometry and
from the regime of motion. We find again the dependence of τ0 from Re by observing
that the order of greatness of the ratio

τ0
2 ρv
1 2

is
μ vδ∞ 1
=√ .
ρv∞2 Re

This way for the calculation of the resistance R it is enough to know the
coefficient c0 . 
Reynolds made some experiences to determine the velocity of transition from the
laminar motion to the turbulent motion. He made a liquid flow in a duct and he
introduced to the center of this current fluid a colored fillet through another small
duct. For some values of the velocity of the fluid tide the colored fluid fillet was
continuous, which meant that the motion was laminar. Increasing the velocity of
the current, a certain value was reached in correspondence of which the fluid fillet
broke it, coloring the whole current. Under such conditions the motion was evidently
characterized by violent and intense exchanges of molecules among the fluid fillets,
so it was turbulent. The value of this characteristic velocity resulted function of the
type of fluid and decreasing when the diameter of the duct increased.
This way a critical value Recr of the number of Reynolds exists below which the
motion is laminar and above which the motion is turbulent.
The experiment done for a circular duct resulted valid for any geometry of the
duct and particularly also for a plane slab. Also in this occasion the choice of the
length of reference to introduce in Recr must be done keeping in mind of the par-
ticular problem in examination and with prudence. Naturally for an assigned fluid
and for an assigned geometry it doesn’t exist a precise value of Recr . We only have

5.2.18 The dynamic pressure coincides with the kinetic energy for unity of volume.
550 5 Thermoelasticity

an interval of values outside of which the motion is certainly or laminar or tur-

bulent. For values included in this interval the motion can be both laminar and
turbulent5.2.19 .
Remark 5.2.40 For the tubes with constant circular section it results approximately
Recr = 2,300. 

Let us now see how the limit layer in laminar regime and the limit layer in tur-
bulent regime are differentiated. We perform the study reporting us to the problem
of an indefinite plane slab, of length c so great to be able to set c = +∞. The slab
is invested by a current parallel to it. The number of Reynolds that in this case we
must consider, contains as length of reference the distance x from the edge of attack
of the indefinite slab (Fig. 5.2.20), that is
ρvx
Rex = .
μ

Initially, being Rex small because x is small, the current is laminar.

The thickness δ(x) of √ the limit layer grows with the distance x. In fact, as we
know, δ/x varies√with 1/ Rex and since Rex is proportional to x we get that δ is
proportional to x. Well for the plane slab Recr = 500,000 approximately. Since
the slab is infinitely long, Rex overcomes Recr to a certain station xcr (Fig. 5.2.20).
In correspondence5.2.20 of xcr the motion from laminar becomes turbulent and the
thickness of the limit layer increases with rapidity greater than in the laminar motion
(Fig. 5.2.20). The thickening of the limit layer increases with rapidity greater than
in the laminar motion (Fig. 5.2.20). The thickening of the limit layer is due to the
fact that in the turbulent motion the agitations, i.e. the exchanges of mass, happen
on macroscopic scale, so they are more intense and therefore are strongly felt in an
amplest region of the space. The higher intensity of the exchanges of mass involves
a more rapid leveling of the velocity. So in the turbulent limit layer the profile of
velocity is flatter than that related to the laminar limit layer (Fig. 5.2.21). The level-
ing of the profile of velocity involves a greater resistance to the motion of the slab.

Fig. 5.2.20

5.2.19 The motion related to values of Re insides to the interval can be considered as an unstable
regime. If it is laminar, a small perturbation is enough to make it turbulent and vice versa. In nature
there are a lot of small perturbations so in the aforesaid interval the motion continually passes from
laminar to turbulent and vice versa.
5.2.20 As already specified on the interval of the values of R , in reality the transition from the
ecr
laminar regime to turbulent gradually happens in a neighborhood of xcr .
5.2 Fluid Dynamics 551

Fig. 5.2.21

In fact τ0 = μ ∂v∂y (0) and the value of

∂v
∂y (0) is greater in the turbulent motion than
that related to the laminar regime.
Remark 5.2.41 The results gotten for the limit layer of an indefinite slab are entirely
general. So we underline that, because of the greater resistance to the motion, as a
rule we must prefer the laminar regime to the turbulent one. 

Let us consider now a sphere immersed in a liquid current (Fig. 5.2.22), so that

1
p + ρv2 = const . (5.2.41)
2

From the (5.2.41) it follows that where the velocity increases the pressure
decreases and vice versa. So in P and P (Fig. 5.2.22) the velocity is maximum
and then the pressure is minimum; in Q and Q the velocity is zero and then the
pressure is maximum and clearly is just the pressure of stagnation.
From the symmetry of the current and of the body it results that the diagram of
the velocity and therefore of the pressures is symmetrical. It comes down from this
the paradox, called of D Alembert, that the resistance offered by the sphere to the
motion, in absence of viscosity, is zero. In fact the resultant one of the pressures p
on the sphere is zero.
On the contrary if we take into account the limit layer we see that the resultant
one of the forces of attrition τ0 , acting on the surface of the sphere and tangentially
to it, is different from zero. In the external potential field, out of the limit layer,
there is not degradation of energy and then p and v vary according the (5.2.41). In
the arc PQ of Fig. 5.2.22 the kinetic energy is entirely converted in energy of pres-
sure. However inside the limit layer, because of the viscosity, the energy degrades
it. In consequence the fluid particle reduces its velocity, cannot yet go up the gra-
dient of pressure, detaches it from the fluid fillet and forms a vortex. The vortex is
created because the fluid particle, under the pressure practiced by the fillets of the
potential field adjacent to it, stops and is pushed back. Since the gradient of pressure
among P and Q have not been entirely recovered, the diagram of the pressures is

Fig. 5.2.22
552 5 Thermoelasticity

not symmetrical, contrary to the case of non viscous motion. Therefore a resultant p
dS of the pressures is present that goes to add it to the resultant one of the forces of
attrition τo dS. The component according to v∞ of the sum of such resultants is the
force of resistance to the motion. The component according the normal to v∞ of the
sum of such resultants is the force of lift. The resistance due to the pressures p dS is
called resistance of wake or resistance of form. The resistance due to the gradient of
velocity in the limit layer is called resistance of attrition or viscous resistance. The
resistance of wake vanishes when the fluid particles don’t detach it from the fluid
fillet. Naturally the resistance of wake is predominant in comparison to the viscous
resistance.
Remark 5.2.42 A technique to decrease the resistance to the motion is to make
the transition from laminar regime into a turbulent regime. In fact the resistance
of wake happens for the fact that the fluid particle dissipates good part of its kinetic
energy. We can obviate to this fact furnishing energy to the particle. This can be
done furnishing to the particle an aliquot of the energy of the external potential
field. Precisely we can obtain such result making the limit layer become turbulent.
At that moment in fact the energy kinetic is just transferred by the external poten-
tial field inside the limit layer. Obviously in the turbulent regime the resistance of
attrition increases. However the resistance of wake, that is predominant on that of
attrition, decreases. So the global resistance to the motion decreases. 

5.3 Mechanics of Solids

5.3.1 The Dynamic Thermoelastic Problem

Let us consider a continuous three-dimensional solid C and denote with V the por-
tion of space that it occupies and with S the surface of V, that is the frontier of
V. We decompose S in the part Ss , whose points are constrained, and in the part
Sp = S − Ss . We assume a cartesian orthogonal frame of reference O, x, y, z and
suppose S regular, so that in any point of S the tangent plane exists.
We suppose that the material is homogeneous and isotropic. We suppose that
the temperature T can vary both with the point P = (x, y, z) and with the
time t. We remove the hypothesis of static application of the external load, so
the model that we are going to build will simulate the thermoelastic problem
both static and dynamic. The body is submitted to an external volumetric load
X(x, y, z, t), Y(x, y, z, t), Z(x, y, z, t) and, as regards the portion Sp of the surface S, to a
superficial external load px (x, y, z, t), py (x, y, z, t), pz (x, y, z, t) and to an external ther-
mal load, constituted by an heat flow through Sp . We suppose that in every instant
t (greater or equal than the initial instant ti ) the volumetric external load X, Y, Z,
the superficial external load px , py , pz and the reactions that the constraints act on C
through Ss , constitute a system of forces equivalent to zero. Finally we suppose that
the loads applied on the body, both mechanical and thermal, are such that we are in
every instant in the field of the small deformations.
5.3 Mechanics of Solids 553

The results obtained in Sect. 1.1 are not limited by hypotheses on the nature of the
mean and on the causes of the deformation. Insofar they are valid also in presence
of thermal effects and, in every instant t ≥ ti , also in dynamic field. We can express
this way the elongations and the shearing strains related to the coordinated axes in
function of the components of the displacement sx (x, y, z, t), sy (x, y, z, t), sz (x, y, z, t)
with the relationships, valid in V × [ti , +∞[

∂sx
εx =
∂x
∂sy
εy =
∂y
∂sz
εz =
∂z
(5.3.1)
∂sx ∂sy
γxy = +
∂y ∂x
∂sx ∂sz
γxz = +
∂z ∂x
∂sy ∂sz
γyz = + .
∂z ∂y

Similarly in presence of thermal field and in every instant t ∈ [ti , +∞[ the rela-
tionships (1.1.14) and (1.1.15) subsist and also the definitions of principal direction
of strain and of principal strain and all their properties.
Analogously, the compatibility equations still are, in V × [ti , +∞[

∂ 2 εx ∂ 2 εy ∂ 2 γxy
+ =
∂y 2 ∂x 2 ∂x∂y
∂ 2 εx ∂ 2 εz ∂ 2 γxz
+ =
∂z2 ∂x2 ∂x∂z
∂ 2 εy ∂ 2 εz ∂ 2 γyz
+ =
∂z 2 ∂y 2 ∂y∂z
(5.3.2)
∂ 2 εx ∂ ∂γxy ∂γyz ∂γzx
2 = − +
∂y∂z ∂x ∂z ∂x ∂y
∂ 2 εy ∂ ∂γyz ∂γzx ∂γxy
2 = − +
∂z∂x ∂y ∂x ∂y ∂z
∂ 2 εz ∂ ∂γzx ∂γxy ∂γyz
2 = − + .
∂x∂y ∂z ∂y ∂z ∂x

As already observed in Sect. 1.1, the displacements being small, the deformed
configuration is a little far from the initial one. From here it follows that we can
554 5 Thermoelasticity

apply the principle of dissection to any part of the body with reference to the
undeformed configuration.
Besides let us observe that also the results obtained in Sect. 1.2 can be applied
to every type of continuous and to every type of process, so they also subsist in
presence of thermal field. Besides if the forces of volume X̃, Ỹ, Z̃ applied on the
body are sum of the external volumetric load X, Y, Z and of the forces of inertia,
such results also subsist in every instant t ≥ ti for dynamic phenomena. Insofar in
V × [ti , +∞[ it holds the symmetry of τ , that is

τxy (x, y, z, t) = τyx (x, y, z, t)

τxz (x, y, z, t) = τzx (x, y, z, t)
τyz (x, y, z, t) = τzy (x, y, z, t)
or, briefly

τxy = τyx
τxz = τzx (5.3.3)
τyz = τzy .

Moreover, since in this linear treatment the mass density ρ of the material
can be assumed independent of the strain and then constant, and since X̃ =
∂2s
X − ρ ∂∂ts2x , Ỹ = Y −ρ ∂t2y , Z̃ = Z −ρ ∂∂ts2z , in V ×[ti , +∞[ the following differential
2 2

equations of equilibrium hold

∂σx ∂τxy ∂τxz ∂ 2 sx

+ + +X =ρ 2
∂x ∂y ∂z ∂t
∂τyx ∂σy ∂τyz ∂ 2 sy
+ + +Y =ρ 2 (5.3.4)
∂x ∂y ∂z ∂t
∂τzx ∂τzy ∂σz ∂ 2 sz
+ + +Z =ρ 2 ,
∂x ∂y ∂z ∂t

and still in V × [ti , +∞[, if P = (x, y, z) is a point of V, t ∈ [ti , +∞[, αn is a

plane passing through P, n = (nx , ny , nz ) is the normal to αn passing through P, the
following relationships hold

tnx (x, y, z, t) = σx (x, y, z, t) nx (x, y, z) + τxy (x, y, z, t) ny (x, y, z)

+ τxz (x, y, z, t) nz (x, y, z)
tny (x, y, z, t) = τyx (x, y, z, t) nx (x, y, z) + σy (x, y, z, t) ny (x, y, z)
+ τyz (x, y, z, t) nz (x, y, z)
tnz (x, y, z, t) = τzx (x, y, z, t) nx (x, y, z) + τzy (x, y, z, t) ny (x, y, z)
+ σz (x, y, z, t) nz (x, y, z)
or, briefly
5.3 Mechanics of Solids 555

tnx = σx nx + τxy ny + τxz nz

tny = τyx nx + σy ny + τyz nz (5.3.5)
tnz = τzx nx + τzy ny + σz nz .

Equally, if P = (x, y, z) is a point of V, t ∈ [ti , +∞[, αn is the plane tangent in

P to S, n = (nx , ny , nz is the normal to αn passing though P and oriented going out
from the material, in S × [ti , +∞[, the following Cauchy boundary conditions hold

px = σx nx + τxy ny + τxz nz
py = τyx nx + σy ny + τyz nz (5.3.6)
pz = τzx nx + τzy ny + σz nz .

Still equally, if P = (x, y, z) is a point of V, n = (nx , ny , nz ) is any line passing

through P, s = (sx , sy , sz ) is any line passing through P and orthogonal to n, t ∈
[ti , +∞[, it results in V × [ti , +∞[

σn = σx n2x + σy n2y + σz n2z + 2τxy nx ny

(5.3.7)
+2τxz nx nz + 2τyz ny nz
τns = σx nx sx + σy ny sy + σz nz sz + τxy (nx sy + ny sx )
(5.3.8)
+τxz (nx sz + nz sx ) + τyz (ny sz + nz sy )

where σn (x, y, z, t) is the normal stress acting on αn in P at the instant t ≥ ti and

τns (x, y, z, t) is the orthogonal component on s of the shearing stress acting on αn in
P at the instant t ≥ ti .
Similarly in presence of thermal field and in every instant t ∈ [ti , +∞[, the defi-
nitions of principal direction of stress and of principal stress and all their properties
subsist.
We must now add to the preceding ones a further differential equation, that
imposes the solution to satisfy the energy balance.
Let us consider a point (x, y, z) of the body and an instant t ∈ ]ti , tf [. We assume
as thermodynamic system an infinitesimal parallelepiped having edges dx, dy, dz
parallel to the coordinate axes and having the point (x, y, z) as a vertex. We denote
with qx [resp. qy ] [resp. qz ] the heat transferred by the environment to the system
in the unity of time through the unity of area of the face of the parallelepiped pass-
ing through (x, y, z) and normal to x [resp. y] [resp. z]. The function qx [resp. qy ]
[resp. qz ] is called local flow of heat according x [resp. y] [resp. z], is measured in
kcal/(cm2 s) and is given by the law of Fourier 5.3.1 (5.1.11). So
   
∂T ∂T ∂T
qx = −k resp. qy = −k resp. qz = −k .
∂x ∂y ∂z

5.3.1 In the (5.1.11), that Fourier get by experimental way, k is called constant of thermal
conductibility of the material and is measured in kcal/(cm ◦ C s).
556 5 Thermoelasticity

In virtue of the theorem of the divergence the flow of heat Q transferred by the
environment to the parallelepiped in the unity of time is, ∀t ∈ ]ti , tf [

∂ 2T ∂ 2T ∂ 2T
Q = div q dx dy dz = −k + + dx dy dz . (5.3.9)
∂x2 ∂y2 ∂z2

In the (5.3.9) q denotes the vector function with three components

q = qx x + qy y + qz z = (qx , qy , qz )

and usually is called thermal load5.3.2 .

Since during the phenomenon
 in examination the system remains in solid phase,
∀t ∈ ]ti , tf [ the surface (t) doesn’t allow exchange of mass between system and
environment. As a consequence in the energetic balance (5.1.8), that is in the

dE
= Q + L,
dt
the flow of work L5.3.3 done on the infinitesimal parallelepiped in the unity of time
has not convective aliquot associated to flow of mass through the faces of the par-
allelepiped. So L is only constituted by the diffusive flow of energy in the form of
work.
As in the case of the fluid, also for the solid the diffusive flow of energy in the
form of work is the scalar product of the stress tensor τ̆ acting on the faces of the
parallelepiped and of the vector velocity v with which the point of application of
such superficial stress moves it.
As in the case of the fluid, also for the solid the diffusive flow of energy in the
form of work L = τ̆ ×v can be separated in a reversible part Lr , to which it doesn’t
associate production of entropy, and in an irreversible part Ld , also said dissipative
part. So L = Lr + Ld .
As in the case of the fluid, we individualize Ld separating the stress tensor
⎡ ⎤
σx τxy τxz
τ̆ = ⎣ τyx σy τyz ⎦
τzx τzy σz

in a reversible part τ̆ (r) and in an irreversible part τ̆ (d) , still intending for reversible
[resp. irreversible] stress that part of stress that results in a reversible [resp.
irreversible] work.

5.3.2 Q is also called thermal power.

5.3.3 L is also called mechanical power.
5.3 Mechanics of Solids 557

To be able to practically perform this division it is necessary to determine the con-

stitutive equations of the mean in examination. In the solids these are relationships
that express the dependence of the stress tensor from the strain tensor.
Let us suppose the material isotropic and linearly elastic and denote:

– with α 5.3.4 the coefficient of free thermal linear expansion of the material, so that
α ∈ ]0, +∞[,
– with T(x, y, z, t) the absolute temperature of the material in the point P = (x, y, z)
at the instant t,
– with Ti (x, y, z, ti ) the absolute temperature of the material in the point P = (x, y, z)
at the initial instant ti .

Well it was experimentally seen that in the thermoelastic case the relation-
ships (1.4.1), proposed by Navier for the problem of the elastic equilibrium, are
generalized in

1  
εx = σx − ν σy + σz + α (T − Ti )
E
1 
εy = σy − ν (σx + σz ) + α (T − Ti )
E
1  
εz = σz − ν σx + σy + α (T − Ti )
E (5.3.10)
τxy
γxy =
G
τxz
γxz =
G
τyz
γyz =
G
where the modulus of Young E ∈ ]0, +∞[, the coefficient of Poisson ν ∈ ]0, 1/2[,
and the tangent modulus G = E/[2(1 + ν)] are independent from T.
We derive from the direct generalized relationships of Navier (5.3.10) those
inverse. We denote with λ the constant of Lamè (1.4.4), with ϑ[resp.Tσ ] the lin-
ear invariant of strain [stress], so that ϑ = εx + εy + εz [resp. Tσ = σx + σy + σz ].
From the (5.3.10) we have

1 − 2ν
ϑ= Tσ + 3α(T − Ti )
E
from which

E 3αE
Tσ = ϑ− (T − Ti ) . (5.3.11)
1 − 2ν 1 − 2ν

the iron α = 12 · 10−6 ◦ C−1 . For instance a track long 18 m lengthens from the winter to
5.3.4 For
the summer (T = 40 ◦ C) of l = alT = 12 · 10−6 · 40 · 18 = 0.0086 m = 8.6 mm.
558 5 Thermoelasticity

From the first of (5.3.9) and from the (5.3.10) it follows

Eεx = σx (1 + ν) − νTσ + aE(T − Ti )

from which
 
E ν E 3αE αE
σx = εx + ϑ− (T − Ti ) − (T − Ti )
1+ν 1+ν 1 − 2ν 1 − 2ν 1+ν

from which

E
σx = 2Gεx + λϑ − α(T − Ti ) .
1 − 2ν
This way we obtain the inverse generalized relationships of Navier:

  E
σx = 2Gεx + λ εx + εy + εz − α (T − Ti )
1 − 2ν
  E
σy = 2Gεy + λ εx + εy + εz − α (T − Ti )
1 − 2ν
  E
σz = 2Gεz + λ εx + εy + εz − α (T − Ti ) (5.3.12)
1 − 2ν
τxy = Gγxy
τxz = Gγxz
τyz = Gγyz .

Still proceeding by experimental way, we get that the reversible aliquot τ̆ (r) is
given from
 
σx(r) = 2Gεx + λ εx + εy + εz
 
σy(r) = 2Gεy + λ εx + εy + εz
 
σz(r) = 2Gεz + λ εx + εy + εz
(5.3.13)
τxy
(r)
= G γxy
τxz
(r)
= G γxz
τyz
(r)
= G γyz

and that the irreversible aliquot τ̆ (d) is given from

(d) E
σx =− α (T − Ti )
1 − 2ν
(d) E
σy =− α (T − Ti )
1 − 2ν
5.3 Mechanics of Solids 559

(d) E
σz =− α (T − Ti )
1 − 2ν
(d)
τxy = 0
(5.3.14)
(r)
τxz = 0
(r)
τyz = 0.

Thus
⎡ (d) (d) (d)
⎤ ⎡ (d)
⎤
σx vx + τxy vy + τxz vz σx vx
⎢ (d) ⎥ ⎢ (d) ⎥
τ̆ (d) × v = ⎢ (d) (d) ⎥ ⎢ ⎥
⎣ τyx vx + σy vy + τyz vz ⎦ = ⎣ σy vy ⎦ , (5.3.15)
(d) (d) (d) (d)
τzx vx + τzy vy + σz vz σz vz

where vx , vy , vz are the components of the vector velocity v, so that

∂sx ∂sy ∂sz

vx = , vy = , vz = .
∂t ∂t ∂t

This way from the (5.3.15) and (5.3.14) we obtain that the components δx , δy , δz
of the vector τ̆ (d) ×v have expression

Ea ∂sx
δx = − (T − Ti )
1 − 2ν ∂t
Ea ∂sy
δy = − (T − Ti )
1 − 2ν ∂t
Ea ∂sz
δz = − (T − Ti ) .
1 − 2ν ∂t
Insofar, keeping in mind of the hypotheses of small deformations
 and of small
temperature changes it was resulted that the flow through of τ̆ (d) ×v, the
power developed by the dissipative aliquot of the tensor τ̆ on the infinitesimal
parallelepiped, is
 
Ld = ∫ τ̆ (d) × v × n dσ

∂δx
= − δx dy dz + δx + dx dy dz − δy dx dz
∂x
∂δy ∂δz
+ δy + dy dx dz − δz dx dy + δz + dz dx dy
∂y ∂z
∂δx ∂δy ∂δz
= + + dx dy dz
∂x ∂y ∂z
 
Eα(T − Ti ) ∂ 2 sx ∂ 2 sy ∂ 2 sz
=− dx dy dz + +
1 − 2ν ∂x∂t ∂y∂t ∂z∂t
560 5 Thermoelasticity

so that the aliquot of mechanical power that turns in thermal power is

Eα(T − Ti ) ∂εx ∂εy ∂εz

Ld = − + + dx dy dz . (5.3.16)
1 − 2ν ∂t ∂t ∂t

Conclusively the (5.1.8) can be written

dE
= Lr + Ld + Q
dt
where Ld is given by the (5.3.16). Obviously

∂T
Q + Ld = c ρ
∂t
where c is the specific heat of the material (for unitary mass). So, because of (5.3.16)
and (5.3.9), we get the fundamental equation of the thermoelastic problem (obtained
first by Duhamel 5.3.5 )

∂ 2T ∂ 2T ∂ 2T
k + +
∂x2 ∂y2 ∂z2
(5.3.17)
∂T Eα(T − Ti ) ∂εx ∂εy ∂εz
=ρc + + + .
∂t 1 − 2ν ∂t ∂t ∂t

This way, assuming as unknown functions the components of the displacement

sx (x, y, z, t), sy (x, y, z, t), sz (x, y, z, t) and the absolute temperature T(x, y, z, t), and
denoting with ti [resp. tf ], with tf > ti or tf = +∞, the initial [resp. final] instant
of the process, the problem of the thermoelastic equilibrium admits the following
mathematical model:
[5.3.1] Find four real functions sx , sy , sz , T defined in V × ]ti , tf [, therein partially
differentiable at least twice and such that5.3.6 in V × ]ti, tf [
 
∂ ∂sx ∂sx ∂sy ∂sz Eα
2G +λ + + − (T − Ti )
∂x ∂x ∂x ∂y ∂z 1 − 2ν
   
∂ ∂sx ∂sy ∂ ∂sx ∂sz ∂ 2 sx (5.3.18)
+ G + + G + +X =ρ 2
∂y ∂y ∂x ∂z ∂z ∂x ∂t

5.3.5 Inhis fundamental paper back in 1837.

5.3.6 The (5.3.18), (5.3.19), and (5.3.20) are obtained from the (5.3.4) employing the (5.3.12) and
(5.3.1).
5.3 Mechanics of Solids 561

  
∂ ∂sy ∂sx ∂ ∂sy ∂sx ∂sy ∂sz
G + + 2G +λ + +
∂x ∂x ∂y ∂y ∂y ∂x ∂y ∂z
   (5.3.19)
Eα ∂ ∂sy ∂sz ∂ 2 sy
− (T − Ti ) + G + +Y =ρ 2
1 − 2ν ∂z ∂z ∂y ∂t
   
∂ ∂sz ∂sx ∂ ∂sz ∂sy
G + + G +
∂x ∂x ∂z ∂y ∂y ∂z
 
∂ ∂sz ∂sx ∂sy ∂sz Eα ∂ 2 sz
+ 2G +λ + + − (T − Ti) + Z = ρ 2
∂z ∂z ∂x ∂y ∂z 1 − 2ν ∂t
(5.3.20)
 
∂ 2T ∂ 2T ∂ 2T ∂T Eα ∂ 2 sx ∂ 2 sy ∂ 2 sz
k + 2 + 2 = ρc + (T − Ti ) + +
∂x2 ∂y ∂z ∂t 1 − 2ν ∂x∂t ∂y∂t ∂z∂t
(5.3.21)
and satisfying the boundary conditions

in Sp × ]ti, tf [, the mechanical conditions (5.3.6) expressed

(5.3.22)
in terms of sx , sy , sz , T
in Ss × ]ti, tf [, the constraint conditions; (5.3.23)

in Sp × ]ti, tf [, in case of solids in contact, the thermal

(5.3.24)
conditions5.3.7
∂T ∂T1
T = Ti, k = k1 ;
∂n ∂n

in Sp × ]ti, tf [, in case of solid immersed5.3.8 in a fluid having

(5.3.25)
temperature T0 and conductance h, the thermal condition

∂T
k = h(T0 − T) ;
∂n
in Sp × ]ti, tf [, in case of assigned thermal load, the thermal
(5.3.26)
condition
∂T
k = q;
∂n

5.3.7 If two particles have different temperatures, the heat moves from the greater temperature to
the smaller temperature by one of the three following ways: conduction, convection, radiation.
The first way is if the particles belong to solids in contact; the second if the particles are fluid; the
third one if the particles are separated from the void and at least one of the temperatures is elevated.
5.3.8 If the solid in examination is immersed in a fluid the phenomenon of the convection intervenes.
In such case in rigor the problem is rather complex, because the fluid dynamics problem and that
of the thermoelastic solid should be treated as an unique problem.
562 5 Thermoelasticity

in Sp × ]ti, tf [, in case of assigned superficial temperature,

(5.3.27)
the thermal condition
T=f;

in Sp × ]ti, tf [, in case of solid exposed to a source5.3.9 at high

(5.3.28)
temperature T0 , the thermal condition

∂T
k = c(T04 − T 4 ).
∂n 
From the solution of the problem of the thermoelastic equilibrium [5.3.1] we
obtain the state of strain through the (5.3.1) and from this the state of stress through
the (5.3.12).
In the mathematical theory of the Thermoelasticity the following theorem of
existence and uniqueness is proven
[5.3.2] If the frontier S of V and the mechanical and thermal loads are regular, the
problem of the thermoelastic equilibrium [5.3.1] admits an unique solution. 
Evidently the thermoelastic problem [5.3.1] is mathematically nonlinear because
of the term (5.3.29), and this is the unique cause of the nonlinearity. Insofar for the
problem [5.3.1] the principle of superposition [1.5.4] doesn’t hold.

5.3.2 The Thermoelastic Dissipation

When we apply a mechanical and/or thermal load on a solid body, its state of strain
varies. The term of mechanical coupling

Eα ∂εx ∂εy ∂εz

(T − Ti ) + + (5.3.29)
1 − 2ν ∂t ∂t ∂t

present in the thermoelastic problem [5.3.1] involves that to such variation of the
state of strain, a variation of the distribution of temperature T is accompanied. In
consequence of this, additional flows of heat are established in the body. Insofar the
energy stored in the body in form not mechanically recoverable is increased. This
phenomenon goes under the name of thermoelastic dissipation.
In the simple and ordinary case the phenomenon of the thermoelastic dissipation
is negligible. In fact, as it is intuitive, if the material remains in elastic field and the
deformations are small, the variations of temperature induced by the deformations
are negligible.

5.3.9 This is the phenomenon of radiation. In the law of Stefan-Boltzmann, q = c1 T14 −c2 T24 , c1 and
c1 depend from the relative orientation of the surfaces to which the two particles belong, from their
distance and from the properties of absorption and reflection of the surfaces. In this phenomenon
the transmission of the heat is meaningful only for high absolute temperatures (T ≥ 400◦ F).
5.3 Mechanics of Solids 563

A lot of numerical and in the laboratory experimentations allows to conclude

that in the greatest part of the problems of the Structural engineering it is possible
to neglect the term of mechanical coupling (5.3.29) present in the (5.3.21) of the
thermoelastic problem, without involve meaningful errors.
Proceeding with the theory of the characteristic numbers it is possible to make
dimensionless the (5.3.21) and to measure the importance of the phenomenon of
the thermoelastic dissipation with a pure number, that seems appropriate to call of
Duhamel. The obtained results show that the phenomenon of the thermoelastic dis-
sipation is not negligible when the velocity of deformation is meaningfully different
from the one with which the temperatures vary. This happens when we rapidly bend
a flexible bar up and down. This way in fact in the bar the temperature will start to
increase.
Remark 5.3.1 Because of the phenomenon of thermoelastic dissipation, when a body
in linearly elastic material is submitted only to mechanical loads, its deformation is
accompanied by dissipation, that is the entropy of the body increases. Insofar in
rigor the phenomenon is not conservative. 
Remark 5.3.2 We have seen in the problem [5.3.1] that the deformations produced
by mechanical and/or thermal loads are accompanied by variations of temperature
and consequently from flows of heat. The whole process results in an increase
of the entropy. The energy stored increased in way mechanically non recover-
able. Insofar if the phenomenon of the thermoelastic dissipation is not negligible,
the thermoelastic problem is not conservative and as a consequence doesn’t admit
potential. 
Remark 5.3.3 Usually the thermoelastic problem [5.3.1] is called coupled ther-
moelastic problem. This because the problem [5.3.1] couples the problems of the
Transmission of the heat and The theory of the elasticity. 

5.3.3 The Uncoupled Thermoelastic Problem

Let us now analyze the ordinary case of thermoelastic problem, in which all the
greatness are constant with respect to the temporal variable and the phenomenon of
the thermoelastic dissipation is negligible.
In such hypotheses the coupled thermoelastic problem [5.3.1] breaks it
into two independent boundary problems. The first one is called Transmission
of the heat and its unknown function is the absolute temperature T(x, y, z).
The second problem is called Uncoupled static thermoelasticity or simply
Thermoelasticity and its unknown functions are the components of the displace-
ment sx (x, y, z), sy (x, y, z), sz (x, y, z). In this second problem T(x, y, z) is a datum of
the problem, that we preliminary must determine solving the first problem.
We easily obtain, in the previous hypotheses, that the problem of the
Transmission of the heat has the mathematical model
[5.3.3] Find a real function T defined in V, therein differentiable at least up to the
second order, such that in V
564 5 Thermoelasticity

∂ 2T ∂ 2T ∂ 2T
+ + =0 (5.3.30)
∂x2 ∂y2 ∂z2

and satisfying in S the (5.3.24), (5.3.25), (5.3.26), (5.3.27), and (5.3.28). 

With equal facility we obtain, in the previous hypotheses, that the problem of the
Thermoelasticity has the mathematical model
[5.3.4] Denoting with T the solution of problem [5.3.3], find three real functions
sx , sy , sz defined in V, therein differentiable at least up to the second order, such that
in V
 
∂ ∂sx ∂sx ∂sy ∂sz Eα
2G +λ + + − (T − Ti )
∂x ∂x ∂x ∂y ∂z 1 − 2ν
    (5.3.31)
∂ ∂sx ∂sy ∂ ∂sx ∂sz
+ G + + G + +X =0
∂y ∂y ∂x ∂z ∂z ∂x

  
∂ ∂sy ∂sx ∂ ∂sy ∂sx ∂sy ∂sz
G + + 2G +λ + +
∂x ∂x ∂y ∂y ∂y ∂x ∂y ∂z
   (5.3.32)
Eα ∂ ∂sy ∂sz
− (T − Ti ) + G + +Y =0
1 − 2ν ∂z ∂z ∂y

   
∂ ∂sz ∂sx ∂ ∂sz ∂sy
G + + G +
∂x ∂x ∂z ∂y ∂y ∂z
 
∂ ∂sz ∂sx ∂sy ∂sz Eα
+ 2G +λ + + − (T − Ti ) + Z = 0
∂z ∂z ∂x ∂y ∂z 1 − 2ν
(5.3.33)
and satisfying in S the (5.3.22), and (5.3.23). 

Remark 5.3.4 To understand that a state of stress arises when a body not con-
strained and not submitted to mechanical load is heated, we can reason as it
follows. We divide the body in elementary right parallelepiped. Every of these
because of the thermal jump dilates or contracts of a quantity equal in all the direc-
tions and proportional to the thermal jump. This way after the thermal expansion
every parallelepiped is still right but is dilated or contracted in a different way.
So some thermal stress must rise up which will provoke in every parallelepiped
additional deformations, such that the deformed body still be a continuous one.
This way the total deformation in the generic point is composed of two parts.
The first part is an uniform expansion proportional to the thermal jump T − Ti .
Since this expansion is equal in all the directions, ε arise but not γ . The second
part is constituted by the deformations necessary to maintain the continuity of the
body. 
5.3 Mechanics of Solids 565

Fig. 5.3.1

Problem 5.3.1 You should determine the distribution of temperature in the disk of
turbine of Fig. 5.3.1. The flow of the heat q1 on the edge r = R of the disk is the heat
transmitted by the blades invested by the gas to high temperature. The flow of heat
q2 on the sides of the disk is the loss of heat due to the air of cooling, equal to −k1 T
on every face. Insofar the total loss of heat for unity of volume is −2k1 T/b.

Solution We suppose that the phenomenon is steady and the thermoelastic dissipa-
tion is negligible. Then equation

∂T
Q + Ld = c ρ
∂t
becomes Q = 0 and we must solve problem [5.3.3]. If b is small, we can substi-
tute problem [5.3.3] with another problem in which the effect of the cooling air is
computed not in the boundary conditions but in the differential equation. In this
new problem, approximate but analytically simpler, we admit that Q is given by two
aliquots. The first aliquot is (5.3.9), that in polar coordinates is written (taking into
account the axial-symmetry)

d2 T 1 dT
k + .
dr2 r dr
The second aliquot is

k1
−2 T.
b
So the differential equation of the approximating problem is

d2 T 1dT 2k1
2
+ − T = 0. (5.3.34)
dr rdr kb
566 5 Thermoelasticity

The (5.3.34) admits the solution

 3 
2k1
T = AI0 r
kb

where I0 is the function of Bessel of modified first type and of order zero. The
constant A is determined by the boundary condition

dT
k (R) = q1 .
dr
We get
5
2k1
I0 r kb
q1
T(r) = 5 5 . 
2 kkb1 dI0
R 2k 1
dr kb

Remark 5.3.5 In the Transmission of the heat a principle subsists analogous to that
of Saint Venant. It results in fact that if the thermal load is assigned only on a small
part of S and if the total flow of heat is zero, then the temperature of the body varies,
in comparison to the initial value Ti , only in a small portion of body next to the
thermal load. 

5.3.4 Thermoelasticity

Let us consider a solid body, that we suppose homogeneous, isotropic, linearly

elastic, in the field of the small deformations, submitted to a mechanical load
px , py , pz distributed on its surface S and to a mechanical load X, Y, Z distributed
on its volume V. A thermal load is also present. We suppose that the phe-
nomenon of the thermoelastic dissipation is negligible and that the problem is
static, i.e. that all the greatness in game are constant with respect to the temporal
variable t.
We have previously seen that in such hypotheses to determine the state of stress
in the body it is preliminarily possible to resolve the problem of Transmission of the
heat [5.3.3]. Its solution furnishes the distribution of temperature T that arises in the
body and that constitutes a datum for the subsequent problem of Thermoelasticity
[5.3.4].
In the Mathematical theory of thermoelasticity is proven that
[5.3.5] The problem [5.3.4] admits an unique solution. 
[5.3.6] The problem [5.3.4] is linear so it holds the principle of superposition, both
for the mechanical load px , py , pz , X, Y, Z and for the thermal load T. 
5.3 Mechanics of Solids 567

[5.3.7] In a unconstrained body submitted only to a linear thermal load, i.e. of the
type T(x, y, z) = ax + by + cz + d, the state of stress is zero, i.e. it identically results
σx , σy , σz , τxy , τxz , τyz = 0 in V. 
Besides, since in the thermoelastic problem dissipative phenomena are absent, it
results, as it is easy to verify
[5.3.8] The problem [5.3.4] admits a potential function, that we call thermoelastic
potential, given by

1 α(T − Ti )
ϕ= (σx εx + σy εy + σz εz + τxy γxy + τxz γxz + τyz γyz ) − (σx + σy + σz ). 
2 2
Finally, it is easy to verify that for the problem of Thermoelasticity [5.3.4] the
principle of the virtual works and the principle of minimum of the potential energy
subsist.

Problem 5.3.2 You should determine in a thin circular disk the stresses induced from
a distribution the temperature T steady and symmetrical with respect to the center
of the disk5.3.10 .

Solution We assume a polar frame of reference (r, t) with origin in the center of the
disk. Since the symmetry, the shearing stresses τrt are zero. So we must integrate
the equilibrium equation (cfr. Sect. 3.1)

dσr σr − σt
+ = 0. (5.3.35)
dr r
Putting T = T − Ti (where Ti denotes the initial temperature), the generalized
relationships of Navier become

1
εr = (σr − νσt ) + αT
E

1
εt = (σt − νσr ) + αT
E
so that

E
σr = [εr + νεt − (1 + ν)αT] (5.3.36)
1 − ν2
E
σt = [εt + νεr − (1 + ν)αT]
1 − ν2

5.3.10 This problem intervenes in the design of a gas turbine.

568 5 Thermoelasticity

and then the (5.3.35) becomes

d dT
r (εr + νεt ) + (1 − ν)(εr − εt ) = (1 + ν)αr . (5.3.37)
dr dr
Denoting with sr the radial displacement we easily have, taking into account the
symmetry of the problem

dsr sr
εr = , εt = (5.3.38)
dr r
so that the (5.3.35) becomes

d 2 sr 1 dsr sr dT
+ − 2 = (1 + ν)α
dr2 r dr r dr
or
 
d 1 d(rsr ) dT
= (1 + ν)α
dr r dr dr

that by integration furnishes

 r
1 c2
sr = (1 + ν)α Tzdz + c1 r + . (5.3.39)
r 0 r

It must result sr (0) = 0. From this, observing that

 r
1
lim Tzdz = 0
r→0 r 0

we get c2 = 0.
Denoting with b the radius of the disk, the second boundary condition to
impose is
σr (b) = 0 . (5.3.40)

From the (5.3.39), (5.3.38), and (5.3.36) we have

 r
1 E
σr = −αE Tzdz + c1 (1 + ν)
r2 0 1 − ν2

from which, taking into account the (5.3.40)


α b
c1 = (1 − ν) Tzdz
b2 0

so that
5.3 Mechanics of Solids 569
   
1 1 r b
σr = αE Tzdz − 2 Tzdz
0 b2 r 0
   
1 b 1 r
σt = αE −T + 2 Tzdz + 2 Tzdz .
b 0 r 0

We notice that, since

 r
1 T(0)
lim Tzdz = .
r→0 r 2 0 2

in the center of the disk the thermal stresses σr , σt have finite value. 

Problem 5.3.3 Let us consider a long cylinder with circular crown as section. You
should determine the stresses induced in the cylinder from a distribution of temper-
ature T symmetrical with respect to the axis of the cylinder and constant with the
axial coordinate z5.3.11 . The cylinder surface is unloaded. The initial temperature is
a real number Ti .

Solution We assume a cylindrical frame of reference (r, t, z) with origin in the infe-
rior base of the cylinder. We suppose, in a first step, that the axial displacement sz
satisfies anywhere the condition5.3.12

sz = 0 (5.3.41)

Denoting with T = T − Ti the increase of temperature, the generalized

relationships of Navier are written

1
εr = [σr − ν(σt + σz )] + αT
E
1
εt = [σt − ν(σr + σz )] + αT (5.3.42)
E
1
εz = [σz − ν(σr + σt )] + αT .
E
Since sz = 0 it results εz = 0. So the third of the (5.3.42) furnishes

σz = ν(σr + σt ) − αET . (5.3.43)

5.3.11 The determination of the thermal stresses in the wall of a cylinder is of great practical
importance in the design of the endothermic motors.
5.3.12 Such condition is exact in the indefinitely long cylinder, where every plane of normal z is of
symmetry.
570 5 Thermoelasticity

Substituting the (5.3.43) in the first two of the (5.3.42) we obtain

 
1 − ν2 ν
εr = σr − σt + (1 + ν)αT
E 1−ν
 
1−ν 2 ν
εt = σt − σr + (1 + ν)αT .
E 1−ν

Besides, because of the symmetry of the problem, it must be (denoting with sr

the radial displacement)
dσr σr − σt
+ =0
dr r
dsr
εr =
dr
sr
εt = .
r
Proceeding as in the problem 5.3.2, and denoting with a the inside radius of the
circular crown, we have

1+ν α r c2
sr = · Tzdz + c1 r +
1−ν r 0 r

αE 1 r E c1 c2
σr = − · 2 Tzdz + − 2 (5.3.44)
1−ν r a 1 + ν 1 − 2ν r

αE 1 r αET E c1 c2
σt = − · 2 Tzdz − + + 2
1−ν r a 1−ν 1 + ν 1 − 2ν r

from which, taking into account the (5.3.43), it follows

αET 2νEc1
σz = − + . (5.3.45)
1−ν (1 + ν)(1 − 2ν)
Such way we have obtained that if the cylinder is loaded by the distribution T of
temperature and, on the bases, by the axial uniform superficial load (5.3.43), then its
state of stress is (5.3.42). As a consequence, by the superposition principle, we can
obtain the solution of the problem in examination adding to the (5.3.42) the state of
stress of the cylinder submitted only to an axial uniform superficial load opposite
to (5.3.43). This is the Saint Venant problem of the axial centroidal load, whose
solution gives:
σr = 0
σt = 0.
So the second and third of (5.3.44) furnish the solution o the problem in
examination.
Denoting with b the external radius of the circular crown, the constants c1 and c2
are determined imposing the boundary conditions
5.3 Mechanics of Solids 571

σr (a) = 0, σr (b) = 0 .

We obtain

c1 c2
− 2 =0
1 − 2ν a

αE 1 b E c1 c2
− · 2 Tzdz + − 2 =0
1−ν b a 1+ν 1 − 2ν b

from which
 b
Ec2 αE a2
= · 2 Tzdz
1+ν 1 − ν b − a2 a
 b
Ec1 αE 1
= · 2 Tzdz .
(1 + ν)(1 − 2ν) 1 − ν b − a2 a

Of here, replacing c1 and c2 in the second and in the third one of the (5.3.44),
replacing c1 in the (5.3.43) and adding in the (5.3.43) the axial pressure c3 necessary
to annul the total axial pressure to apply on every base, it is gotten
   
αE 1 r2 − a2 b r
σr = · Tzdz − Tzdz
1 − ν r2 b2 − a2 a a
   
αE 1 r2 + a2 b r (5.3.46)
σt = · 2 Tzdz + Tzdz − Tr 2
1−ν r b2 − a2 a a

σz = 0 .

About Sr , we easily verify that it is the sum of the right side of the (5.3.44) and
of

νr αET 2νEc1
− . 
E 1−ν (1 − ν)(1 − 2ν)

Problem 5.3.4 Let us consider a cylinder with circular crown section. You should
determine the stresses induced in the cylinder from a thermal load constituted by a
constant flow of heat through the cylindrical walls and such to maintain the inside
surface of the cylinder to the initial uniform temperature Ti and that external to
uniform temperature Te .

Solution The thermoelastic problem in examination is uncoupled. So we can deter-

mine at first the cylinder temperature T by solving the heat transmission problem
[5.3.3]. Assuming the cylindrical frame of reference of problem 5.3.3, equation
(5.3.30) turns in
572 5 Thermoelasticity

d2 T 1 dT
2
+ =0 in ]a, b[
dr r dr
whose general integral is

T = c1 log r + c2 .

From the boundary conditions

T(a) = Ti , T(b) = Te

we get

Te − Ti r
T = Ti + b
log .
log a
a
After that we determine the state of stress in the cylinder by solving problem
[5.3.4]. Since the thermal distorsion T is axialsymmetric, the required state of stress
is given by (5.3.46) of problem 5.3.4, putting

Te − Ti r
T = b
log .
log a
a

This way we easily verify that if Ti < Te

• σr is of compression anywhere and is zero on the internal and external surface of

the cylinder (Fig. 5.3.2)
• σt assumes its maximum compression value on the internal surface of the cylin-
der and its maximum traction value on the external surface of the cylinder
(Fig. 5.3.2).

Remark 5.3.6 Often a body or a structure is subject to some initial stresses, that we
also call auto-stress, existing still before the application of the mechanical and/or
thermal loads. Naturally contextually to the auto-stress some initial deformations are
also present. The auto-stress can be unintentional, as those originated from defects
of assemblage or from processes of hot-working or cold-working of the metals. But

Fig. 5.3.2
5.3 Mechanics of Solids 573

they can also be provoked on purpose, for improving the static regime that will be
established at the end when the loads will be applied. For instance those originated
in a structure by inserting with a procedure of forcing a pendulum of smaller length
of the distance of the connected zippers.
Generally the auto-stress are harmful because can be of the same order of great-
ness of the stresses provoked by the loads. When this happens, after loading the real
state of stress results much more serious than in absence of initial tensions. This
way, except the case of auto-stress provoked on purpose, their effect can be a sud-
den breakup without apparent causes. When a piece of fusion is cooled, at first the
external parts solidify while those inside stay fused. When those inside also solid-
ify, these are contrasted in their shrinkage from the parts already solidified. This
way at the end in the material some stresses are established which also stay present
in absence of external mechanical or thermal loads. The auto-stress are also born in
the processes of lamination, in those of forging, in the temper of the steel pieces and
also in the moulding of concrete.
It is obvious that the auto-stress, any external force being absent, must be auto-
equilibrated. The presence of auto-stress is difficult to detect. As a rule if we cut
longitudinally a laminated beam the two separate parts bend him. 

5.3.5 The One-Dimensional Problem

The deflected beam is a static uncoupled thermoelastic problem of high technical

interest. This is an one-dimensional problem that admits a mathematical model
obtained by generalizing what was proposed in Sect. 4.2.
We denote with z the axis of the beam and suppose the thermal load consti-
tuted by a linear distribution of temperature T (Fig. 5.3.3). Furthermore we assume
the same hypotheses made in Sect. 4.2. Then, as it has been experimentally veri-
fied, we can generalize the differential equation (4.2.11) assuming, to control the
phenomenon, the law

d2 v M αT
=− − (5.3.47)
dz2 EI ho

Fig. 5.3.3
574 5 Thermoelasticity

where:
– the vertical displacement v is positive if downward,
– the bending moment M is positive if stretches the inferior fibers,
– z is positive if is oriented on right (Fig. 5.3.4),
– the beam has cross section constant or weakly varying,
– h0 (z) denotes the height of the cross section of the beam (Fig. 5.3.3),
– E is the Young modulus of the material,
– I(z) is the moment of inertia of the cross section with respect to the neutral axis of
the bending,
– α is the coefficient of linear elongation of the material,
– T = T − Ti is the temperature range, supposed variable with z and having in the
cross section the linear5.3.13 distribution of Fig. 5.3.3, positive if the inferior fibers
are lengthened.
About the signs of the (5.3.47) it is enough to observe that if M or T is positive
in the cross section z, then the infinitesimal trunk of beam deforms as in Fig. 5.3.4.
As a consequence it must result

d2 v
(z) < 0 .
dz2
About the expression assumed for the thermal aliquot of the curvature in the
(5.3.47), we observe that, applying only the thermal load on the infinitesimal
trunk of beam of Fig. 5.3.5, from the hypotheses (4.2.4) and (4.2.5) and from the
hypothesis of small displacements we obtain (Fig. 5.3.6)

dϕ dz
R =
2 2
h0 dϕ dz dz T
R+ = −α
2 2 2 2 2

Fig. 5.3.4

Fig. 5.3.5

5.3.13 This is the thermal load of greater technical interest for the deflected beams.
5.3 Mechanics of Solids 575

Fig. 5.3.6

and then

h0 dϕ α dz T
=
4 4
and then, taking into account the signs

d2 v d dv dϕ αT
= = =− .
dz2 dz dz dz ho
The (5.3.47) allows to simulate the behavior of a deflected beam, submitted to a
distributed mechanical load p and to the linear thermal load T of Fig. 5.3.3, with
a very good mathematical model. Without loss of generality, we formulate it for the
beam of Fig. 5.3.7.
Let us assume the frame of reference of Fig. 5.3.8 and suppose p and T regular
in ]0.2 l[. Keeping into account the (5.3.47) and the

dM
=T,
dz
the analytical problem admits the following formulation

Fig. 5.3.7

Fig. 5.3.8
576 5 Thermoelasticity

[5.3.9] Find a real function v defined in [0, 2 l], differentiable at least four times in
]0, l[ ∪ ]l, 2 l[ and such that

d4 v p α d2 T
= − in ]0, 2 l[
dz4 El h0 dz2
v(0) = 0
dv
(0) = 0
dz
v(l− ) = v(l+ )
d2 v − αT(l)
(l ) − =0
dz2 h0
d2 v + αT(l)
(l ) − =0
dz2 h0
d d2 v α d d d2 v α d
I 2 (l+ ) + (IT)(l+ ) − I 2 (l− ) − (IT)(l− ) = 0
dz dz h0 dz dz dz h0 dz
v(2 l) = 0
d2 v αT(2l)
2
(2 l) − = 0.
dz h0

In the Mathematical theory of thermoelasticity it is proven that:

[5.3.10] The problem [5.3.9] admits an unique solution. 
[5.3.11] The problem [5.3.9] is linear, so the principle of superposition, with regard
of the loads both mechanical and thermal, holds . 
[5.3.12] For the problem [5.3.9] the principle of the virtual works holds. 
[5.3.13] For the problem [5.3.9] the principle of the minimum potential energy
holds. 
Let us consider now the problem
[5.3.14] Find a real function ν defined in [0, 2 l], differentiable at least two times in
]0, 2 l[ and such that

d2 ν M αT
2
=− − in ]0, 2 l[
dz EI ho
ν(0) = 0
dν
(0) = 0
dz
ν(l− ) = ν(l+ )
ν(2 l) = 0 . 
5.3 Mechanics of Solids 577

In the Mathematical theory of thermoelasticity we prove that in Thermelasticity

the theorems of Clapeyron, Betti, Castigliano hold and, taking into account that

d2 M
= −p ,
dz2
that
[5.3.15] The problem [5.3.9] is equivalent to the problem [5.3.14]. 
[5.3.16] The elastic energy E stored in the beam is equal to the work of deformation
Ld and it results
 
1 2l M αT
Ld = E = Mdϕ = M + dz . 
2 0 EI h0

Problem 5.3.5 You should determine the elastic line of the cantilever of Fig. 5.3.9.
Solution We choose the reference of Fig. 5.3.10. We must determine a function ν
regular in ]0, 2 l[ and such that

d4 ν
= 0 in ]0, l[ ∪ ]l, 2 l[
dz4
ν(0) = 0
dν
(0) = 0
dz
d2 ν − d2 ν +
(l ) = (l )
dz2 dz2 (5.3.48)
d2 ν d2 ν
−EI 2 (l− ) + EI 2 (l+ ) = F
dz dz
d ν
2
EI 2 (2 l) = M
dz
d3 ν
= 0.
dz3

Fig. 5.3.9

Fig. 5.3.10
578 5 Thermoelasticity

Since the [5.3.15] the problem (5.3.48) is equivalent to the problem to find a ν
regular in ]0, 2 l[ and such that

d2 ν M
= in ]0, l[ ∪ ]l, 2 l[
dz2 EI
ν(0) = 0
(5.3.49)
dν
(0) = 0 .
dz

So it is enough to solve the simpler problem (5.3.49). To this aim we determine

the function bending moment M. Evidently it results M(z) = −M − F(l − z) [resp.
M(z) = −M] in ]0, l[ [resp. ]l, 2 l[,] (Fig. 5.3.11) and then in the two part of beam
AB and BC the moment M has two different analytical expressions. So to find the
solution of the problem (5.3.49) we determine, less than two constants of integra-
tion, a solution of the differential equation in the part AB. After that we determine,
less than other two constants of integration, a solution of the differential equation
in the part BC. Then we impose the boundary conditions of the problem (5.3.49) as
well as the condition that ν must be regular in the whole beam. This last is a condi-
tion of inside compatibility. Precisely to avoid the breaking of the beam in B, that is
to avoid laceration or overlap of material, it is necessary that the rotation [resp. the
lowering] of the cross section immediately on the left of B is equal to the rotation
[resp. the lowering] of the cross section immediately on the right of B:

dν − dν +
(l ) = (l )
dz dz
ν(l− ) = ν(l+ ) .

This way we get a linear algebraic system of four equations in four unknown
constants of integration. As it is easy to show, the determinant of its matrix of the
coefficients is different from zero. With this the solution of the problem (5.3.49) is
exactly determined.
Practically performing this procedure of calculation, we integrate in ]0, l[ the
differential equation

d2 ν1 −M − F(l − z)
=−
dz2 EI

Fig. 5.3.11
5.3 Mechanics of Solids 579

immediately obtaining

dν1 dν1 z M + F(l − t)
∀z ∈ ]0, l[ (z) − (0) = dt
dz dz 0 EI

so that
dν1 M F z2 dν1
∀z ∈ ]0, l[ (z) = z+ lz − + (0)
dz EI EI 2 dz
from which
 z 
M F t2 dν1
∀z ∈ ]0, l[ ν1 (z) − ν1 (0) = t+ lt − + (0) dt
0 EI EI 2 dz

so that
M 2 F z3 dν1
∀z ∈ ]0, l[ ν1 (z) = z + lz2 − +z (0) + ν1 (0)
2EI 2EI 3 dz

Analogously, integrating in ]l, 2 l[ the differential equation

d2 ν2 M
=
dz2 EI
we immediately obtain

dν2 dν2 z M
∀z ∈ ]l, 2 l[ (z) − (0) = dt
dz dz 0 EI

so that

dν2 M dν 2
∀z ∈ ]l, 2 l[ (z) = (z − l) + (l)
dz EI dz
from which

M 2 Ml dν2
∀z ∈ ]l, 2 l[ ν2 (z) = (z − l2 ) − (z − l) + (l)(z − l) + ν2 (l) .
2EI EI dz
Let us impose now the conditions of constraint and of regularity in l of ν, the
continuity in l of ν and of its first derivative. We have

ν1 (0) = 0
dν1
(0) = 0
dz
580 5 Thermoelasticity

M 2 F l3 dν1
l + l3 − +l (0) + ν1 (0) = ν1 (l− ) = ν2 (l+ ) = ν2 (l)
2EI 2EI 3 dz
M F l2 dν1 dν1 − dν2 + dν2
l+ l2 − + (0) = (l ) = (l ) = (l) .
EI EI 2 dz dz dz dz

It deals with a linear algebraic system of four equations in the four unknown
constants of integration

dν1 dν2
ν1 (0), (0), ν2 (l), (l) .
dz dz
Obviously in this elementary case its solutions are given by the solutions of the
linear algebraic system of two equations in two unknowns

M 2 F l3
l + l3 − = ν2 (l)
2EI 2EI 3

M F l2 dν2
l+ l2 − = (l)
EI EI 2 dz
so that

v1 (0) = 0
dν1
(0) = 0
dz
Fl3 Ml2
ν2 (l) = +
3EI 2EI
dν2 Fl2 Ml
(l) = + .
dz 2EI EI

Insofar the elastic line of the cantilever of Fig. 5.3.9 is

⎧
⎪
⎪ M 2 F z3
⎨ z + lz2 − in ]0, l[
ν(z) = 2EI 2EI 3
⎪
⎪ M 2 Fl2 Fl3 Ml2
⎩ (z − l2 ) + (z − l) + + in ]l, 2 l[ . 
2EI 2EI 3EI 2EI

Problem 5.3.6 You should determine the elastic line of the deflected beam of
Fig. 5.3.12.

Fig. 5.3.12
5.3 Mechanics of Solids 581

Fig. 5.3.13

Solution Let us introduce the frame of reference of Fig. 5.3.13. The problem is to
determine a function ν such that

d4 ν
= 0 in ]0, l[ ∪ ]l, 2 l[
dz4
ν(0) = 0
dν
(0) = 0
dz
d2 ν −
(l ) = 0
dz2
(5.3.50)
d2 ν + M
2
(l ) = −
dz EI
d3 ν − d3 ν +
(l ) = 3 (l )
dz3 dz
ν(2 l) = 0
d2 ν
(2 l) = 0 .
dz2
Since the [5.3.15] the problem (5.3.50) is equivalent to the problem to find a ν
regular in ]0, 2 l[ and such that

d2 ν M
2
=− in ]0, l[ ∪ ]l, 2 l[
dz EI
ν(0) = 0
(5.3.51)
dν
(0) = 0
dz
ν(2 l) = 0 .

Since in B there is a hinge rather than an inside fixed joint, the conditions of
regularity of ν reduce it to be alone

ν(l− ) = ν(l+ ) .

So it is enough to solve the simpler problem (5.3.51). To this aim we determine

the function bending moment M. Evidently it results (Fig. 5.3.14)
#
M
M(z) = l (l − z) in ]0, l[
M
l (2 l − z) in ]l, 2 l[ .
582 5 Thermoelasticity

Fig. 5.3.14

So by integration in ]0, l[ we obtain


dν1 z M dν1 M 2 M dν1
∀z ∈ ]0, l[ (z) = − (l − t)dt + (0) = z − z+ (0)
dz 0 lEI dz 2lEI EI dz
 z dν1 M 3 M 2 dν1
∀z ∈ ]0, l[ ν1 (z) = (t)dt + ν1 (0) = z − z +z (0) + ν1 (0)
0 dz 6lEI 2EI dz

and by integration in ]l, 2 l[ we obtain


dν2 z d2 ν2 dν2
∀z ∈ ]l, 2 l[ (z) = 2
(t)dt + (l)
dz 0 dz dz
M 2 2M dν2
=
(z − l2 ) − (z − l) + (l)
2lEI EI dz
 z
dv2
∀z ∈ ]l, 2 l[ ν2 (z) = (t)dt + ν2 (l)
0 dz

M 3 M 2 3Ml
= (z − l3 ) − (z − l2 ) + (z − l)
6lEI EI 2EI
dν2
+ (z − l) (l) + ν2 (l) .
dz
Imposing the boundary conditions in problem (5.3.51) we have

ν1 (0) = 0
dν1
(0) = 0
dz

Ml2 Ml2 dν1

− + +l (0) + ν1 (0) = ν1 (l− ) = ν2 (l+ ) = ν2 (l)
2EI 6EI dz

M M 2 3Ml2 dν2
(8 l3 − l3 ) − (4 l − l2 ) + +l (l) + ν2 (l) = ν2 (2 l) = 0 .
6lEI EI 2EI dz
5.3 Mechanics of Solids 583

This is a linear algebraic system of four equations in four unknowns, with non
singular matrix of the coefficients. Its unique solution is evidently

ν1 (0) = 0
dν1
(0) = 0
dz
Ml2
ν2 (l) = −
3EI
dν2 2Ml
(l) = ,
dz 3EI

so that it results

⎧
⎪ M 3 M 2
⎪
⎪ z − z in]0, l[
⎪
⎨ 6lEI 2EI
ν(z) =
⎪
⎪
⎪
⎩ M (z3 − l3 ) − M (z2 − l2 ) + 13Ml (z − l) − Ml
2 2
⎪
in ]l, 2 l[. 
6lEI EI 6EI 3EI

Problem 5.3.7 You should determine the elastic line of the deflected beam of
Fig. 5.3.15.

Solution Chosen the frame of reference of Fig. 5.3.16, we must determine a ν regular
in ]0, l[ ∪ ]l, 3 l[, continuous in l and such that

d4 ν
(z) = 0 ∀z ∈ ]0, l[ ∪ ]l, 3 l[
dz4

Fig. 5.3.15

Fig. 5.3.16
584 5 Thermoelasticity

ν(0) = 0
dν
(0) = 0
dz
d2 ν
(l) = 0
dz2
d3 ν − d3 ν
3
(l ) = 3 (l+ ) (5.3.52)
dz dz
ν(2 l) = 0
d2 ν
(3 l) = 0
dz2
d3 ν F
3
(3 l) = − .
dz EI

Since the [5.3.15] the problem (5.3.52) is equivalent to the problem to find a ν
regular in ]0, l[ ∪ ]l, 3 l[,, continuous in l and such that

d2 ν M
2
=− in]0, l[ ∪ ]l, 3 l[
dz EI
ν(0) = 0
(5.3.53)
dν
(0) = 0
dz
ν(2 l) = 0 .

Clearly the elastic line of the problem of Fig. 5.3.15 is of the type shown in
Fig. 5.3.17. So to determine the solution of the problem (5.3.52) we must integrate
in ]0, l[ and in ]l, 3 l[ and we must add the regularity condition

ν(l− ) = ν(l+ ) . (5.3.54)

To determine the solution of the problem (5.3.53), we observe that the bending
moment M in the beam of Fig. 5.3.15 is (Fig. 5.3.18)
#
−F(z − l) in [0, 2 l]
M(z) =
F(z − 3 l) in ]2 l, 3 l] .

Fig. 5.3.17
5.3 Mechanics of Solids 585

So we must integrate in ]0, l[, in ]l, 2 l[, in ]2 l, 3 l[ and we must add the (5.3.54)
and the new regularity condition

ν(2 l− ) = ν(2 l+ )
dν − dν + (5.3.55)
(2 l ) = (2 l ) .
dz dz

This way, proceeding to determinate the solution of problem (5.3.53), we

integrate in ]0, l[ obtaining

dν1 F 2 Fl dν1
∀z ∈ ]0, l[ (z) = z − z+ (0)
dz 2EI EI dz
F 3 F 2 dν1
∀z ∈ ]0, l[ ν1 (z) = z − z +z (0) + ν1 (0) ;
6EI 2EI dz

we integrate in ]l, 2 l[ obtaining

dv2 F 2 Fl dv2
∀z ∈ ]l, 2 l[ (z) = (z − l2 ) − (z − l) + (l)
dz 2EI EI dz
F 3 F 2
∀z ∈ ]l, 2 l[ ν2 (z) = (z − l3 ) − (z − l2 )
6EI 2EI
Fl2 dν2
+ (z − l) + (z − l) (l) + ν2 (l);
2EI dz

we integrate in ]2 l, 3 l[ obtaining

dv3 F 2 3Fl dv3

∀z ∈ ]2 l, 3 l[ (z) = − (z − 4 l2 ) + (z − 2 l) + (2 l)
dz 2EI EI dz
F 3 3Fl 2
∀z ∈ ]2 l, 3 l[ ν3 (z) = (z − 8 l3 ) + (z − 4 l2 )
6EI 2EI
8Fl2 dν3
− (z − 2 l) + (z − 2 l) (l) + ν3 (l).
EI dz

Furthermore the boundary conditions of problem (5.3.53), the (5.3.54) and the
(5.3.55) give

Fig. 5.3.18
586 5 Thermoelasticity

ν1 (0) = ν(0) = 0
dν1 dν
(0) = (0) = 0
dz dz
Fl3 Fl3 dν1
− +l (0) + ν1 (0) = ν1 (l) = ν(l− ) = ν(l+ ) = ν2 (l)
6EI 2EI dz
Fl Fl Fl2 dν2
(8 l3 − l3 ) − (4 l2 − l2 ) + l+l (l) + ν2 (l) = ν2 (2 l)
6EI 2EI 2EI dz
= ν(2 l− ) = ν(2 l) = 0
ν3 (2 l) = ν(2 l+ ) = ν(2 l) = 0
F Fl2 dv2 dν2 dν − dv dν3
(4 l2 − l2 ) − + (l) = (2 l) = (2 l ) = (2 l+ ) = (2 l) .
2EI EI dz dz dz dz dz
This is a linear algebraic system of six equations in six unknowns, with non
singular matrix of the coefficients. Its unique solution is evidently

ν1 (0) = 0
dν1
(0) = 0
dz
Fl3
ν2 (l) = −
2EI
dν2 Fl2
(l) =
dz 6EI
ν3 (2 l) = 0
dν3 2Fl2
(2 l) = ,
dz 3EI
so we get
⎧
⎪ F 3 Fl 2
⎪
⎪ z − z in [0, l]
⎪
⎪
⎨ 6EI 2EI
F 3 Fl 2 2Fl2 Fl
ν(z) = (z − l3 ) − (z − l2 ) + (z − l) − in ]l, 2 l]
⎪
⎪ 6EI 2EI 3EI 3EI
⎪
⎪ 2
⎪
⎩− F (z3 − 8 l3 ) − 3Fl (z2 − 4 l2 ) − 10 · Fl (z − 2 l) in ]2 l, 3 l[ . 
6EI 2EI 3 EI

Remark 5.3.7 It is possible in all the precedent problems to perform the integration
of the differential equation that simulates the problem on the whole interval of the
beam, saving the decomposition of it. To this aim however it is necessary to employ
the Theory of the distributions, a sophisticated tool of the modern Mathematical
analysis. 

Remark 5.3.8 As seen in the problem 5.3.5, in the problem of Fig. 5.3.19 the elastic
line ν1 of the trunk AB is (Fig. 5.3.20)
5.3 Mechanics of Solids 587

Fig. 5.3.19

Fig. 5.3.20

ν1 (z) = ν̃1 (z) + c1 z + c2 ∀z ∈ [0, l]

and that ν2 of the trunk BC is (Fig. 5.3.20)

ν2 (z) = ν̃2 (z) + c3 z + c4 ∀z ∈ [l, 2l] .

The constants of integration c2 and c4 evidently make the curves ν̃1 (Fig. 5.3.22)
and ν̃2 (Fig. 5.3.24) move rigidly in the plane. Besides, since we are in the field of
the small displacements, the constants of integration c1 and c3 make the curves ν̃1
(Fig. 5.3.23) and ν̃2 (Fig. 5.3.25) rotate rigidly in the plane, around the point A. This
way the elastic lines ν1 and ν2 are individualized by the constants c1 , c2 , c3 , c4 . To
determine their value we need resolve the system of Cramer obtained imposing the
compatibility in A and in B. 

Fig. 5.3.21

Fig. 5.3.22

Fig. 5.3.23
588 5 Thermoelasticity

Fig. 5.3.24

We have seen in the precedents problems that the presence of concentrated exter-
nal forces and/or couples and that the presence of external constraints in inside cross
sections of the beam involve the obligation to decompose the interval occupied by
the beam in a multiplicity of intervals. As a consequence in every interval of such
decomposition we must perform the integration of the differential equation of the
problem. Then the constants of integration are determined adding to the boundary
conditions of the problem, those of regularity of the elastic line to the ends of the
intervals of the decomposition.
This procedure, that obviously is very onerous, can be replaced by a simple
procedure that allows a great speed of calculation, called analogy of Mohr.
The analogy of Mohr available in Thermoelasticity is perfectly analogous to that
studied in Sect. 4.2. In fact everything obtained by Mohr in the case of mechanical
loads finds immediate extension applying in the auxiliary problem the load consti-
tuted by the curvature p∗ , sum of the mechanical curvature EI M
and of the thermal
αT
curvature h0
M αT
p∗ = + .
EI h0

Problem 5.3.8 You should determine with the analogy of Mohr the rotation of the
section A of the supported beam of Fig. 5.3.26.
Solution Obviously the reactions of the supports are zero (Fig. 5.3.27), so that M =
0 identically. Then in the trunk AC it results ν  = 0 identically. So the axis of the
trunk AC remains rectilinear. Furthermore in the trunk CB it results ν  = αT h0 =
const. Then the elastic line is an arc of parable, equal, in our hypothesis of small

Fig. 5.3.25

Fig. 5.3.26
5.3 Mechanics of Solids 589

Fig. 5.3.27

displacements, with an arc of circumference. Since the lengthened fibers, that are
not tense since σz = 0 anywhere, are those upper, the elastic line is necessarily that
of Fig. 5.3.28.
After these considerations, we observe that the auxiliary problem of Mohr is that
of Fig. 5.3.29. So

1 αT l αlT
ϕA = TA∗ = − · · =−
4 h0 2 8h0
and as a consequence the rotation of the cross section A is anticlockwise
(Fig.5.3.30), as anticipated (Fig. 5.3.28). 

Problem 5.3.9 You should determine the lowering of the cross section C in
Fig. 5.3.31 .
Solution Evidently the reactions of the fixed end are zero. Then ν  = 0 identically
in the interval ]0, 2l [. Then the axis of the trunk AB remains rectilinear. Then, being
rigidly inserted A, it stays horizontal. Reasoning as in the problem 5.3.8, we deduce
that the trunk BC deforms it according to an arc of circumference. This in B has
horizontal tangent, otherwise the beam breaks it in B. On the trunk BC it results

αT d2 ν d dν dϕ
− = ν  = 2 = = ,
h0 dz dz dz dz

Fig. 5.3.28

Fig. 5.3.29

Fig. 5.3.30
590 5 Thermoelasticity

Fig. 5.3.31

so
αT
dϕ = − dz ,
h0
so
αT l
ϕ = ϕ(C) − ϕ(B) = − · .
h0 2

Insofar the relative rotation among the cross sections B and C is equal to αlT
2h0
radians. This way the elastic line, since they lengthen the upper fibers, is that of
Fig. 5.3.32. In consequence C lowers of

αlT l αl2 T
νC = = .
2h0 4 8h0

Let us find again this result with the analogy of Mohr. The auxiliary problem of
Mohr is obviously that of Fig. 5.3.33, where p∗ is upward because T lengthens the
upper fibers. So C lowers of

αT l l αl2 T
νC = MC∗ = · = . 
h0 2 4 8h0

Remark 5.3.9 Obviously in any statically determinated structure, submitted only

to thermal loads, all the constraint reactions both external and internal are equal to
zero. So in such structures σ and τ are anywhere zero and ε and γ only have thermal
nature.
Instead in any statically indeterminated structure, submitted only to thermal
loads, a state of stress arises that can also provoke the structural collapse. 

Fig. 5.3.32
5.3 Mechanics of Solids 591

Fig. 5.3.33

Problem 5.3.10 You should determine with the analogy of Mohr the lowering of the
free extreme of the deflected beam of Fig. 5.3.34.
Solution Reasoning as seen in Sect. 4.2, the auxiliary beam is obtained replacing
(Fig. 5.3.35)

– in B an inside hinge to the inside bogie,

– in C a fixed joint to the free extreme,
– in A a sliding joint to the sliding joint, since it is necessary to replace to the sliding
joint a constraint such that M ∗ (A) can assume any value and T ∗ (A) = 0.

We immediately get the results of Fig. 5.3.36, so in the beam of Fig. 5.3.34 C
lowers of
αl2 T Fl3
ν(C) = M ∗ (C) = + . 
h0 3EI

Fig. 5.3.34

Fig. 5.3.35

Fig. 5.3.36
Chapter 6
Stability

6.1 Stability of the Elastic Equilibrium

6.1.1 The Buckling

The phenomenon that we meet in the Stability of the elastic equilibrium is so dif-
fused and dangerous that we have to talk about it, even though briefly. We consider
the problem of Fig. 6.1.1. It evidently reenters within the theory of the deflected
beams, whose one-dimensional mathematical model admits one and only one solu-
tion. Nevertheless this doesn’t happen anymore if the cross section of the beam has
moments of inertia, with respect to the principal axes of its inertia centroidal ellipse,
very different among them (Fig. 6.1.2). In such case in fact it is intuitive and experi-
mentally verified that the cantilever of Fig. 6.1.1 admits, for some values of F, other
deformed configurations (Fig. 6.1.3) over the one furnished by the one-dimensional
mathematical model. This phenomenon, that we call buckling by torsion and bend-
ing, happens even if the intensity of the load is such that in the solution furnished
by the one-dimensional mathematical model the deformations are small.
This way, the one-dimensional mathematical model of the deflected beam con-
stitutes, under certain conditions, an inadequate representation of the physical
phenomenon, in the sense that it furnishes partial indications on the behavior of the
structure. Such inadmissible inaccuracy is removed applying the principle of dissec-
tion to a generic configuration deformed of the structure rather than, as done till now
in the Theory of elasticity, to the undeformed initial configuration of the structure.
Simple and now classical examples show in fact that doing so we bring in account
terms that conduct to a formulation of the problem mathematically notably differ-
ent and that perfectly describes the behavior of the structure. If before, in fact, the
boundary problem of the one-dimensional mathematical model reentered within the
Theory of the elliptic equations, the boundary problem that is gotten imposing the
conditions of equilibrium on configurations deformed of the structure is a problem
of the Theory of the eigenvalues. The new boundary problem that is gotten, as we
will verify in some simple cases, is not linear, so that the principle of superposition
is not valid.

A. Maceri, Theory of Elasticity, DOI 10.1007/978-3-642-11392-5_6, 593


C Springer-Verlag Berlin Heidelberg 2010
594 6 Stability

Fig. 6.1.1

Fig. 6.1.2

Fig. 6.1.3

For some geometries and some load conditions a value Fcr exists of the load F,
that we call critical load, such that below Fcr the uniqueness theorem falls and the
structure can absorb the load in various deformed configurations. We also use to say
that in correspondence of a critical load a phenomenon of bifurcation of the equi-
librium happens. With this we intend to concisely express the fact that the structure
under smaller loads than the critical load can balance the load according to an only
configuration of equilibrium, while under loads greater than the critical load can
balance the load according to more configurations of equilibrium.
We notice right now that, as it is experimentally verified, the mathematical
models gotten operating on the configuration deformed of the structure have the
following properties. The values of the gotten critical load are in every case exact.
However if the hypothesis of small deformations is strictly preserved, wrong indica-
tions are gotten on the way according to which the structure can absorb the load, that
is on the number of the possible configurations and on the quality of their equilib-
rium. If instead the hypothesis of small deformations is at least partially removed,
the information that are gotten on the way according to which the structure can
absorb the load are exact. Obviously, partially removing the hypothesis of small
deformations we need more complex formulations of the mathematical model.
6.1 Stability of the Elastic Equilibrium 595

6.1.2 The Ultimate Strength

As it results from the experiences performed in the Laboratories material tests, the
critical load of a structure can be much superior or much inferior to the load Pmax
in correspondence of which the structure arrives at the ultimate strength, i.e. goes in
crisis.
We will see that in a supported beam, of length l and submitted to an axial com-
pressive load P, the lowering v0 of the middle cross section has the exact diagram
of Fig. 6.1.4. So if P > Pcr the beam can absorb the load according the three con-
figurations S1 , I, S2 , where the configurations S1 , S2 are clearly in the field of great
displacements. Obviously for this structure Pmax < Pcr . We also will see that for
this structure in presence of imperfections the diagram of Fig. 6.1.4 turns into the
diagram of Fig. 6.1.5. So in presence of imperfections the supported beam has Pmax
still inferior and in such case to assume Pcr as load of crisis means to dangerously
overestimate the carrying abilities of the structure.
The thin shells introduce instead another type of behavior. For them the ideal
curve load-displacements, that is the curve in absence of imperfections, is of
the type of the curve outlined in Fig. 6.1.6. Instead the real curve, that is in
presence of imperfections, is of the type of the continuous curve in Fig. 6.1.6.
Insofar in a real thin shell the load cannot go over the value of crisis Pmax
and the critical load Pcr is not attainable. Clearly if the imperfections are not
small, the continuous curve results much different from that ideal outlined and
in such case, as a consequence, the load of crisis is much inferior to that
critic. Then in the analysis of the stability of the thin shells a particular caution
is needed.

l
Fig. 6.1.4

Fig. 6.1.5
596 6 Stability

Fig. 6.1.6

Fig. 6.1.7

The rectangular panels submitted on their edge to loads of compression or to

shearing loads still introduce a different behavior. The ideal structure, that is absent
of imperfections, has a curve load-displacements of the type of that outlined in
Fig. 6.1.7. It is clear that over Pcr the panel enters in buckling. The real curve, that is
in presence of imperfections, is of the type of the continuous curve of Fig. 6.1.7. As
experimentally verified, the panels submitted on their edge to loads of compression
or to shearing loads enter in the crisis at values much greater than Pcr . For them,
therefore, the critical load doesn’t correspond to the load of crisis of the structure,
but only to a change of their way to absorb the load.

6.2 Energy Method

6.2.1 Introduction

Let us suppose that an elastic body absorbs an assigned external load disposing
it according to a deformed configuration of equilibrium Φ 0 . If we apply a pertur-
bation the body assumes a new configuration of equilibrium Φ. We will say6.2.1
that the configuration of equilibrium Φ 0 is stable [resp. unstable] if removing the

6.2.1 To
speak more precisely we must denote with C0 [resp. C] the coefficients of the bound-
ary problem whose solution is Φ 0 [resp. Φ]. Well we will say that Φ 0 is stable if, in opportune
topologies of opportune functional spaces, Φ has Φ 0 as limit when C tends to C0 .
6.2 Energy Method 597

perturbation the body abandons the perturbed configuration Φ and goes back [resp.
goes not back] to the configuration of equilibrium Φ 0 . Instead, if the body remains
in the perturbed configuration Φ we will say that the configuration of equilibrium
Φ 0 is indifferent.
In the Mathematical theory of elasticity the following theorem is proven:
[6.2.1] Let us suppose that an elastic body is submitted to an assigned external
load constant with respect to the time. Let us denote with Φ 0 the deformed con-
figuration of equilibrium assumed by the body. We consider the potential energy
functional Jp , defined in the subset W of the triplet (u, v, w) satisfying the internal
and external compatibility conditions (cfr. Sect. 1.5). Well the following statements
are equivalent:

1. Φ 0 is configuration of stable [resp. unstable] equilibrium,

2. Φ 0 is a point of relative minimum [resp. maximum] for Jp in W. 

Remark 6.2.1 The concept of configuration of stable equilibrium is present in many

sectors of the Physics. In the Mechanics of the solids the case of the equilibrium
of a rigid sphere supported on the surface of a rigid body is classical. If the sur-
face of support is concave (Fig. 6.2.1) the sphere is in stable equilibrium; if the
surface of support is convex (Fig. 6.2.2) the sphere is in unstable equilibrium; if the
surface of support is plane (Fig. 6.2.3) the sphere is in indifferent equilibrium. In
fact applying on the sphere a small horizontal force and afterwards removing the
force we see that

– in Fig. 6.2.1 the sphere has some initial movements and at the end assumes the
initial position;
– in Fig. 6.2.2 the sphere moves it estranging more and more itself from the initial
position;
– in Fig. 6.2.3 the sphere moves it in a new position and remains in such
position. 

Fig. 6.2.1

Fig. 6.2.2

Fig. 6.2.3
598 6 Stability

The [6.2.1] allows to determine the critical load of a structure and to appraise
the type of every possible configuration of equilibrium. When we proceed in such
way to analyze the behavior to the stability of a structure it is used to say that we
are employing the energetic method. In practice for an assigned structure, submitted
to an assigned external load, it is necessary to consider an arbitrary deformed con-
figuration of the structure, externally and internally compatible but not necessarily
balanced, and to write the expression of the relative potential energy. Then it needs
to determine the points of relative extreme of the potential energy when the config-
uration varies in all the possible way. In base to the [6.2.1], the points of relative
extreme are for the system configurations of stable or unstable equilibrium. During
the analysis we find that when the load overcomes a particular value, the possible
configurations of equilibrium change. In such way the critical loads of the structure
are individualized.

6.2.2 Some Elementary Application

In these brief signs on the Stability, we give only some elementary examples of
application of the energetic method. In particular we stop briefly to look at a problem
of concentrated elasticity, that with simple calculations offers a precise and complete
description of what happens in the general problem of the stability of the elastic
equilibrium. Instead the analysis with the energetic method of the behavior to the
stability of systems constituted by elastic structural elements is analytically complex
since it requires the use of the Functional analysis.
So let us consider the problem of concentrated elasticity constituted by the sup-
ported beam of Fig. 6.2.4, supposed rigid and of length l ∈ ] 0, +∞ [ . The elasticity
of the system is concentrated in the bogie, supposed elastically soft according the
Winkler law (4.2.70), with constant k > 0 (cm/kg). On the beam is applied a compres-
sive axial load P. Since the beam is rigid, an externally and internally compatible
configuration of the system is obviously individualized by the angle ϕ that the axis
of the beam forms with the horizontal one (Fig. 6.2.5).
The problem is to find in how many configurations the structure can absorb the
load P and the quality of any configuration of equilibrium. To perform the analysis,
we will resort to the energetic method, employing the [6.2.1].

Fig. 6.2.4

Fig. 6.2.5
6.2 Energy Method 599

Clearly the elastic energy of the system is concentrated in the bogie and is given,
since the theorem [1.6.5] and the (4.2.70), by
1 v(B)
v(B)
2 k
where v(B) denotes the vertical displacement of the cross section B constrained by
the bogie.
Then, because of the Remark 1.5.3, the potential energy functional Jp of the
system is given by

1 ϕl
Jp = (ϕl) − P (l − l cosϕ)
2 k
that is by

l2 2
Jp = ϕ − Pl (1 − cosϕ) . (6.2.1)
2k

We expressly notice that in the hypothesis of small deformation, that is supposing

ϕ small, in the configuration of Fig. 6.2.5 we have v(B) = ϕ l and u(B) = 0, where
u(B) denotes the horizontal displacement of B. However, as it is easy to verify, pro-
ceeding in such way we will not obtain results adherent to the real behavior of the
system. Instead, even only partially removing the hypothesis of small deformations,
as done in the (6.2.1) where we have assumed for u(B) the more precise expres-
sion (l − l cosϕ), we obtain, as we will see in the following, results that perfectly
describe the real behavior of the structure.
Since the [6.2.1], under the load P the possible configuration of stable or unstable
equilibrium of the structure are given by the values of ϕ such that

dJp
= 0,
dϕ
that is such that

dJp d l2 ϕ 2 l2 ϕ
0= = − Pl (1 − cosϕ) = − Pl sinϕ
dϕ dϕ 2k k
that is such that

l
ϕ = sinϕ. (6.2.2)
Pk

Let us solve the (6.2.2) graphically. First of all, in Fig. 6.2.6 we see that if a =
l
Pk < 0, that is if P < 0, the (6.2.2) has the unique solution ϕ = 0. In other words,
600 6 Stability

Fig. 6.2.6

if the beam is in traction the load can be absorbed only in the initial configuration.
Furthermore, since

d2 Jp l2 l
2
(0) = − Pl = Pl −1 >0
dϕ k Pk
and since the [6.2.1], such equilibrium is stable.
If a = Pk
l
> 1, in Fig. 6.2.6 we see that the (6.2.2) has the unique solution ϕ = 0.
So in such condition, in which the beam is compressed by a load 0 < P < kl , the
load can be absorbed only with the initial configuration. Furthermore, since

d2 Jp l
(0) = Pl −1 >0
dϕ 2 Pk

and since the [6.2.1], such equilibrium is stable.

If 0 < a = Pk l
< 1, in Fig. 6.2.6 we see that the (6.2.2) has three solutions −ϕ 0 ,
0, ϕ 0 , where ϕ0 , ∈ ]0, π [. So the structure in this case can absorb the load disposing
it in three different configurations. We notice that now P > 0, that is the beam is
compressed, and that
l
P> . (6.2.3)
k
Furthermore, since the initial hypothesis of small displacements gone putting
v(B) = lϕ, only the case ϕ0 ∈ ] 0, π2 [ is of technical interest. Then, observing that
because of the (6.2.2)

d2 Jp l2 Pl cosϕ0
(ϕ0 ) = − Pl cosϕ0 = (tgϕ0 − ϕ0 ) ,
dϕ 2 k ϕ0

since the Fig. 6.2.7 it results

d2 Jp
(ϕ0 ) > 0.
dϕ 2
6.2 Energy Method 601

Fig. 6.2.7

As a consequence, since the [6.2.1], the equilibrium is stable.

With analogous reasoning we obtain that the equilibrium configuration −ϕ 0 is
stable.
Instead for the equilibrium configuration ϕ = 0 we have

d2 Jp l
(0) = P l −1 <0
dϕ 2 Pk
and then the equilibrium is unstable.
From the (6.2.2) we easily have that in conditions of equilibrium it results

l ϕ
P= ,
k sinϕ
that admits the diagram of Fig. 6.2.8.
Summarizing, if P < 0 the structure can absorb the load only in the initial unde-
formed configuration, that results stable. If 0 < P < l/k the structure can absorb
the load only in the initial undeformed configuration, that results stable. If P > l/k
the structure can absorb the load only in three configuration (Fig. 6.2.8): the initial
undeformed one I, that results unstable, and the two deformed S (Fig. 6.2.9), that
both result stable. So when P becomes greater than l/k the equilibrium bifurcates.
Then the critical load of the structure is

l
Pcr = .
k

Fig. 6.2.8

Fig. 6.2.9
602 6 Stability

Fig. 6.2.10

The previous theoretical analysis results perfectly verified in the Laboratories

material tests.
Problem 6.2.1 You should determine the critical load of the structure of Fig. 6.2.10.
The beams are rigid and the bogies are elastically soft according the Winkler law
(4.2.70), with constant k>0 (cm/kg).

Solution Let us denote with P the intensity of the axial load applied on the beam
(Fig. 6.2.10). We consider an internally and externally compatible configuration
 of the system. Since the beams are rigid,  is individualized by the angles
ϕ 1 and ϕ 2 that the beams AB and CD respectively form with the horizontal axis
(Fig. 6.2.11). We notice that the beam BC forms with con the horizontal axis the
angle (Fig. 6.2.11)

lϕ1 + lϕ2
= ϕ1 + ϕ2 .
l

Obviously in correspondence of the configuration  the potential energy func-

tional Jp of the system has the value

1 v(B) 1 v (C)
Jp (ϕ1 , ϕ2 ) = v(B) + v (C)
2 k 2 k
− P [3 l − l cosϕ1 − l cosϕ2 − l cos (ϕ1 + ϕ2 )] .

Since, as well known,

x2 x4 x6 x2n−2
cosx = 1 − + − + · · · + (−1)n+1 + ··· ,
2! 4! 6! (2n − 2)!

Fig. 6.2.11
6.2 Energy Method 603

from the hypothesis of small displacements we get

ϕ12 ϕ22 (ϕ1 + ϕ2 )2

cosϕ1 = 1 − , cosϕ2 = 1 − , cos (ϕ1 + ϕ2 ) = 1 −
2 2 2
and (Fig. 6.2.11)

v(B) = ϕ1 l, v (C) = −ϕ2 l.

So

l2 2 l2 2
Jp (ϕ1 , ϕ2 ) = ϕ1 + ϕ
2k 2k 2
     !
ϕ2 ϕ2 (ϕ1 + ϕ2 )2
−P 3l − l 1 − 1 −l 1− 2 −l 1−
2 2 2

l2  2  Pl  2 
= ϕ1 + ϕ22 − ϕ1 + ϕ22 + (ϕ1 + ϕ2 )2 .
2k 2
(6.2.4)

Since the [6.2.1], the configurations with which the structure can balance the load
are given by the points (ϕ1 , ϕ2 ) of relative extreme for Jp . So we must find the points
(ϕ1 , ϕ2 ) such that
∂Jp ∂Jp
= 0, = 0.
∂ϕ1 ∂ϕ2
The technique to neglect, in the expression of the potential energy, the infinites-
imal terms of superior order to the second one, as proven by the theoretical and
experimental analysis of numerous and by now classical problems, furnishes the
exact values of the critical load.
Then we must find the points (ϕ1 , ϕ2 ) such that

l2
ϕ1 − Pl (2ϕ1 + ϕ2 ) = 0
k

l2
ϕ2 − Pl (2ϕ2 + ϕ1 ) = 0
k

that are the points (ϕ1 , ϕ2 ) such that

l2
− 2Pl ϕ1 − Plϕ2 = 0
k (6.2.5)
l2
−Plϕ1 + − 2Pl ϕ2 = 0.
k
604 6 Stability

This algebraic linear system of two equations in two unknowns (6.2.5) has the
following determinant of the matrix of the coefficients

l2
− 2Pl −Pl 4l l2
k = l2 3P2 − P+ 2 ,
l2 k k
−Pl − 2Pl
k

that vanish for

l
P=
k
and for
l
P= .
3k
So if P = l/k and P = l/(3 k) the system (6.2.5) admits only the trivial solution
(0, 0). If P = l/k or P = l/(3 k) the system (6.2.5) admits infinite solutions.
Furthermore we notice that from Eq. (6.2.5) it follows that if P = l/k it results
ϕ2 = −ϕ1 ; if P = l/(3k) it results ϕ2 = ϕ1 .
As a consequence the possible equilibrium configurations of the structure are

– (0, 0) if P = l/k and P = l/(3 k),

– (ϕ1 , −ϕ1 ), where ϕ 1 is a small real number, if P = l/k,
– (ϕ1 , ϕ1 ), where ϕ 1 is a small real number, if P = l/(3 k).

To see if these configurations are points of relative extreme for Jp , we must

appreciate the sign of the determinant

∂ 2 Jp ∂ 2 Jp
(ϕ1 , ϕ2 ) (ϕ1 , ϕ2 )
∂ϕ12 ∂ϕ1 ∂ϕ2
H (ϕ1 , ϕ2 ) = .
∂ 2 Jp   ∂ 2 Jp
ϕ1, ϕ2 (ϕ1 , ϕ2 )
∂ϕ2 ∂ϕ1 ∂ϕ22

Since

l2
− 2Pl −Pl l2  2 2 
H (ϕ1 , ϕ2 ) = k = 3 k P − 4klP + l 2
,
l2 k2
−Pl − 2Pl
k
6.2 Energy Method 605

  
we have that if p ∈ ]−∞, l (3 k)[ ∪ kl , +∞ it results H (ϕ1 , ϕ2 ) > 0 ∀ (ϕ1 , ϕ2 ) ∈
2 so that (0, 0) is a point of relative minimum for the functional Jp and then, since
the [6.2.1], the uniquepossible equilibrium configuration, (0, 0), is stable.
If P ∈ l/(3 k), l/k it results H (ϕ1 , ϕ2 ) < 0 ∀ (ϕ1 , ϕ2 ) ∈ 2 , so that (0, 0) is
a point of relative maximum for the functional Jp and then, since the [6.2.1], the
unique possible equilibrium configuration, (0, 0), is unstable.
If P = l/(3 k) from the (6.2.4) we get

l2
Jp (ϕ1 , ϕ2 ) = (ϕ1 − ϕ2 )2 ≥ 0
6k
and

Jp (ϕ1 , ϕ2 ) = 0.

So if P = l/(3 k) the point (ϕ1 , ϕ1 ) is ∀ϕ1 ∈  a point of absolute, and then

relative, minimum for Jp . As a consequence, since the [6.2.1], if P = l/(3 k) all the
configurations (ϕ1 , ϕ1 ) with ϕ1 ∈  are of stable equilibrium for the structure.
If P = l/k from the (6.2.4) we get

l2
Jp (ϕ1 , ϕ2 ) = − (ϕ1 + ϕ2 )2 ≤ 0
2k
and

Jp (ϕ1 , −ϕ1 ) = 0.

So if P = l/k the point (ϕ 1 , −ϕ 1 ) is ∀ϕ1 ∈  a point of absolute, and then

relative, maximum for Jp . As a consequence, since the [6.2.1], if P = l/k all
the configurations (ϕ 1 , −ϕ 1 ) with ϕ1 ∈  are of unstable equilibrium for the
structure.
In conclusion if the load P is smaller of l/(3k) the structure can balance the load
only in the undeformed initial configuration, of stable equilibrium. If the load P is
equal to l/(3k) the structure can balance the load only in the configurations (ϕ 1 , ϕ 1 )
with ϕ1 ∈  (Fig. 6.2.12), that are of stable equilibrium. If 3lk < P < kl the struc-
ture can balance the load only in the undeformed initial configuration, of unstable
equilibrium. If the load P is equal to l/k the structure can balance the load only in
the configurations (ϕ 1 , −ϕ 1 ) with ϕ1 ∈  (Fig. 6.2.13), that are of stable equilib-
rium. If the load P is greater than l/k the structure can balance the load only in the
undeformed initial configuration, that is of stable equilibrium.
606 6 Stability

Fig. 6.2.12

Fig. 6.2.13

Insofar in correspondence of the values l/(3k) and l/k the equilibrium bifurcates
and this means that the system admits the two critical loads

l l
Pcr1 = , Pcr2 = . 
3k k

Remark 6.2.2 The analysis performed in the problem 6.2.1 is only partially correct.
This because it conducts to the diagram of Fig. 6.2.14 while the real behavior, as
experimentally verified and as it would result from a theoretical analysis character-
ized by a greater degree of precision, is that of Fig. 6.2.15. These inaccuracies are
due to the fact that in the analysis of problem 6.2.1, to perform a slender analysis,
we are been too close to the hypothesis of small deformations. It is possible however
to verify, both theoretically and experimentally, that the values of the critical load
gotten with the simplified analysis performed in the problem 6.2.1 are exact, while
the results are wrong which gives us information on the way according to which the
structure balances the load.

Fig. 6.2.14

Fig. 6.2.15
6.3 Static Method 607

This same situation, as we will see in the next section, is verified when we con-
duct the analysis of the problem with the static method, if we are too close to the
hypothesis of small deformation. 

6.3 Static Method

6.3.1 The Beam Axially Loaded

The most important case of structural instability is the beam under compressive
axial load. We consider a supported beam, of length l and with constant section,
perfectly straight and submitted in the end supported by the bogie to an axial load of
compression P (Fig. 6.3.1). Let us increase the intensity of P. If the beam is slender
it is experimentally verified that until a certain value Pcr is not reached the beam has
an unique possibility to absorb the load, shortening itself and preserving rectilinear
the axis. Instead in correspondence of values of P greater than Pcr the beam can
absorb the load both preserving the rectilinear configuration and according to other
configurations, for instance that of Fig. 6.3.2.
So the one-dimensional mathematical model of the deflected beams, for which
the existence and uniqueness theorem subsists, doesn’t describe the behavior of the
structure of Fig. 6.3.1. It is necessary therefore for the structure of Fig. 6.3.1 to build
a new mathematical model that simulates such physical phenomenon.
In this section we will study such problem with a procedure different from the
energetic one, that is called static method. Substantially it consists in the formulation
of a boundary problem in which the differential equation, that controls the equilib-
rium of the structure, is gotten on a deformed configuration rather than, as done till
now, on that initial undeformed. We will see that, as it happened with the energetic
method, using the hypothesis of small deformations very tightly, the gotten results
are only partially exact.

Fig. 6.3.1

Fig. 6.3.2

6.3.2 Approximate Analysis

Let us consider a supported beam, of length l, with constant section and submitted
in the end supported by the bogie to an axial load of compression P (Fig. 6.3.1).
We begin the analysis supposing that during the phenomenon the deformations
608 6 Stability

stay small. To avoid to revert in the one-dimensional model of the deflected beams
we impose the equilibrium of the beam applying the principle of dissection to a
deformed configuration. On the basis of preliminary experimental results, we will
suppose that the deformed configurations v that the beam can assume under the axial
compressive load P are all of deflected type. So, particularly, the axis of the beam
turns it into a plane curve. We denote with n the neutral axis of the bending, with
In the moment of inertia of the cross section of the beam with respect to the axis n,
with E the modulus of Young of the material, supposed homogeneous, isotropic and
linearly elastic. We assume the frame of reference of Fig. 6.3.3, in which the origin
coincides with the section supported of left. Clearly at this point we can adopt all
the assumptions and approximations made in Sect. 4.2. So

dv
ϕ= (6.3.1)
dz

where ϕ [resp. v] is the rotation [resp. lowering] of the cross section of abscissa z;

d2 v M (z)
2
(z) = − ∀z ∈ ]0, l[ (6.3.2)
dz EIn

where M (z) denotes the bending moment at cross section of abscissa z, positive if
extends the inferior fibers.
Obviously the global equilibrium to the rotation around the support of left impli-
cates that in the beam of Fig. 6.3.1 the right constraint has zero reaction. We apply
therefore the principle of dissection to the deformed trunk of beam individualized by
the generic cross section S(z) of abscissa z and from the right end (Fig. 6.3.4). This
way we find that the system of forces constituted by P, N(z), T(z), M(z) is equivalent
to zero. As a consequence, denoting with Q the deformed position of S, the moment
of such system with respect to Q is zero, that is

P v (z) − M (z) = 0 ∀z ∈ [0, l] .

Fig. 6.3.3

Fig. 6.3.4
6.3 Static Method 609

Of here and from the (6.3.2) we get

d2 v P
(z) + v (z) = 0 ∀z ∈ ]0, l[ .
dz2 EIn
Insofar the mathematical model of the supported beam under compressive axial
load loaded of point is the boundary problem

[6.3.1] Find a real function v defined in [0, l], differentiable at least up to second
order and such that

d2 v P
2
(z) + v (z) = 0 ∀z ∈ ]0, l[ (6.3.3)
dz Eln

v (0) = 0 (6.3.4)

v (l) = 0.  (6.3.5)

For the problem [6.3.1] the theorem of existence is banal. In fact the function v
identically zero is, obviously, a solution.
About the uniqueness, from the Theory of the differential equations it easily fol-
lows that if P < 0 the problem [6.3.1] admits an unique solution and that if P > 0
the ordinary differential equation (6.3.3), linear of second order and with constant
coefficients, admits the general integral

1 ! 1 !
P 2 P 2
v (z) = c1 sin z + c2 cos z ∀z ∈ [0, l] (6.3.6)
EIn EIn

where c1 , c2 are two arbitrary real numbers. Imposing the first boundary condition
(6.3.4) we have

c2 = 0. (6.3.7)

From this, imposing the second boundary condition (6.3.5) we have

1 !
P 2
0 = v (l) = c1 sin l .
EIn

So if P is such that
1 !
P 2
sin l = 0, (6.3.8)
EIn
610 6 Stability

then the problem [6.3.1] admits the infinite solutions

1 !
P 2
v (z) = c1 sin z ∀z ∈ [0, l] (6.3.9)
EIn

where c1 is any small real number.

So, since the (6.3.8) the values Pcr of P that allows the beam to absorb the load
in configurations different from that initial straight, are

m2 π 2 EIn
Pcr = ∀m ∈ N .
l2
For m = 1 we obtain the first critical load of the supported beam axially
compressed, that obviously is the only value of real technical interest

π 2 EIn
Pcr = . (6.3.10)
l2

The (6.3.10) is called formula of Euler6.3.1 .

We immediately underline that the mathematical model [6.3.1] is only partially in
accord with the experiences performed in the Laboratories material tests. Precisely
all the experiences have shown that the formula of Euler (6.3.10) furnishes exact
values of the critical load, but the (6.3.9) is not experimentally verified. Precisely,
denoting with v0 the vertical displacement of the middle cross section of the beam,
the mathematical model [6.3.1] involves that for P = Pcr the beam can balance the
load in the infinite configurations (6.3.9) and for P > Pcr the beam can balance
the load only in the initial undeformed configuration. Such results are summarized
by the diagram of Fig. 6.3.5. On the contrary, the real behavior of the structure is
shown in Fig. 6.3.6. In fact in reality for P > Pcr the beam can absorb the load in
three configurations of equilibrium: in the two deformed stable configurations S1 ,
S2 and in the initial undeformed unstable configuration I. Such results are totally in
accord with the experiences performed in the Laboratories material tests.
So we must conclude that preceding analysis is approximate. Obviously its insuf-
ficiency can be attributed only to the much too narrow application of the hypothesis
of small displacements.

Fig. 6.3.5

6.3.1 Leonhard Euler, Basel 1707 – St. Petersburg 1783.

6.3 Static Method 611

Fig. 6.3.6

Fig. 6.3.7

Fig. 6.3.8

It is easy to extend the previous approximate analysis to beams constrained in

a different way. In the case of the cantilever of Fig. 6.3.7, proceeding as for the
supported beam of Fig. 6.3.1 we obtain the mathematical model (Fig. 6.3.8)

[6.3.2] Find a real function v defined in [0, l], differentiable at least up to second
order and such that

d2 v P P
2
(z) + v (z) = v (l) ∀z ∈ ] 0, l [
dz EIn EIn
v (0) = 0
dv
(0) = 0. 
dz

The problem [6.3.2] has the same analytic structure of the problem [6.3.1]. So
we have that if P > 0 the general integral of the differential equation of problem
[6.3.2] is

1 ! 1 !
P 2 P 2
v (z) = c1 sin z + c2 cos z + v (l) ∀z ∈ [0, l] .
EIn EIn
612 6 Stability

Imposing to such general integral to satisfy the boundary conditions we have

0 = v (0) = c2 + v (l)
1
dv P 2
0= (0) = c1
dz EIn

so that c1 = 0 and c2 = −v(l). Then the solutions of problem [6.3.2] are

# 1 !"
P 2
v (z) = v (l) 1 − cos z ∀z ∈ [0, l] (6.3.11)
EIn

where obviously P must be such that

1 !
P 2
cos l = 0.
EIn

So, according this analysis, the cantilever of Fig. 6.3.7 admits the infinite critical
loads

(2m − 1)2 π 2 EIn

Pcr = (m ∈ N ) ,
4 l2
every of which can be absorbed by the beam with one (Fig. 6.3.9) of the infinite
configurations (6.3.11). If P = Pcr the beam can absorb the load only in the initial
undeformed configuration.
For m = 1 we obtain the first critical load of the cantilever axially compressed,
that obviously is the only value of real technical interest

π 2 EIn
Pcr = . (6.3.12)
4 l2
Let us now consider the case of a built-in and supported beam submitted to an
axial compressive load P (Fig. 6.3.10). Proceeding as for the supported beam of
Fig. 6.3.1, we impose the equilibrium of the trunk of beam included among the

Fig. 6.3.9

Fig. 6.3.10
6.3 Static Method 613

Fig. 6.3.11

generic cross section of abscissa z and the supported cross section B (Fig. 6.3.11).
Denoting with R the reaction of the support B we obtain

P v (z) + R (l − z) − M (z) = 0 ∀z ∈ [0, l] .

So the behavior of the structure of Fig. 6.3.10 can be mathematically simulated

by model

[6.3.3] Find a real function v defined in [0, l], differentiable at least up to second
order and such that

d2 v P R
(z) + v (z) = − (l − z) ∀z ∈ ] 0, l [
dz2 EIn EIn
v (0) = 0
dv
(0) = 0
dz
v (l) = 0. 

The problem [6.3.3] has the same analytic structure of the problem [6.3.1]. So
we have that if P > 0 the general integral of the differential equation of problem
[6.3.3] is

1 ! 1 !
P 2 P 2 R
v (z) = c1 sin z + c2 cos z + (l − z) ∀z ∈ [0, l] .
EIn EIn P

Imposing to such general integral to satisfy the boundary conditions we have

Rl
0 = c2 −
P 1
dv P 2 R
0= (0) = c1 +
dz EIn P! !
1 1
P 2 P 2
0 = v (l) = c1 sin l + c2 cos l ,
EIn EIn
614 6 Stability

so that if P is such that

1 1 ! 1 !
R EIn 2 P 2 Rl P 2
− sin l + cos l =0
P P EIn P EIn

then ∀R ∈  the function

1 1 ! 1 !
R EIn 2 P 2 Rl P 2 R
v (z) = − sin z + cos z − (l − z) ,
P P EIn l EIn P

where z ∈ [0, l], is solution of problem [6.3.3].

So if P is such that
1! 1
P 2 P 2
tg l =l (6.3.13)
EIn EIn

then the problem [6.3.3] admits infinite solutions (Fig. 6.3.13).

If P doesn’t satisfy the (6.3.13) then the problem [6.3.3] admits only the solution
v = 0.
We solve the (6.3.13) graphically (Fig. 6.3.12), obtaining
1
P 2
l = 4.493
EIn
from which

4.4932 EIn 2π 2 EIn

Pcr = = . (6.3.14)
l2 l2

Remark 6.3.1 Let us notice that in Fig. 6.3.13 the support reaction must be down-
wards. In fact the elastic curve presents a point of inflection and then, since the
(6.3.2), in that point it must be M = 0. 

Fig. 6.3.12

Fig. 6.3.13
6.3 Static Method 615

Fig. 6.3.14

Finally we consider the case of a built-in and sliding jointed beam submitted to
an axial compressive load P (Fig. 6.3.14). Proceeding as for the supported beam
of Fig. 6.3.1, we impose the equilibrium of the trunk of beam included among
the generic cross section of abscissa z and the sliding jointed cross section B
(Fig. 6.3.15). Denoting with RB [resp. MB ] the reaction [resp. moment of reaction]
of the sliding joint B and imposing the equilibrium to the rotation around Q we
obtain

P v (z) + RB (l − z) + MB − M (z) = 0 ∀z ∈ [0, l] ,

from which, taking into account the (6.3.2)

d2 v P RB MB
2
(z) + v (z) = − (l − z) − ∀z ∈ ]0, l[ .
dz EIn EIn EIn
So the behavior of the structure of Fig. 6.3.14 can be mathematically simulated
by model
[6.3.4] Find a real function v defined in [0, l], differentiable at least up to second
order and such that

d2 v P RB MB
(z) + v(z) = − (l − z) − ∀z ∈ ] 0, l [
dz2 EIn EIn EIn
v(0) = 0
dv
(0) = 0
dz
v(l) = 0
dv
(l) = 0.
dz 
The problem [6.3.4] has the same analytic structure of the problem [6.3.1]. So
we have that if P > 0 the general integral of the differential equation of problem
[6.3.4] is

Fig. 6.3.15
616 6 Stability

1 ! 1 !
P 2 P 2 RB
v (z) = c1 sin z + c2 cos z − (l − z)
EIn EIn P
(6.3.15)
MB
− ∀z ∈ [0, l] .
P
Imposing to such general integral to satisfy the boundary conditions we have

RB l MB
0 = c2 − −
P P
1
dv P 2 RB
0= (0) = c1 +
dz EIn P
1! 1!
P 2 P 2 MB
0 = v (l) = c1 sin l + c2 cos l −
EIn EIn P
1 1 ! 1 1 !
dv P 2 P 2 P 2 P 2 RB
0= (l) = c1 cos l − c2 sin l +
dz EIn EIn EIn EIn P

from which it follows

1
RB EIn 2
c1 = − (6.3.16)
P P
RB l MB
c2 = + (6.3.17)
P P
# 1 ! 1 1 !"
P 2 EIn 2 P 2
l cos l − sin l RB
EIn P EIn
(6.3.18)
# 1 ! "
P 2
+ cos l − 1 MB = 0
EIn
# 1 1 ! 1 !"
P 2 P 2 P 2
−l sin l + 1 − cos l RB
EIn EIn EIn
(6.3.19)
1 1 !
P 2 P 2
− sin l MB = 0.
EIn EIn

Well the (6.3.18) and (6.3.19) constitute a homogeneous algebraic linear system
in the unknown RB , MB . Clearly if P is such that the matrix of the coefficients of
this system is not singular, since the (6.3.16) and (6.3.17) it results c1 = c2 = 0
6.3 Static Method 617

and then from the (6.3.15) it follows that the structure can assume only the initial
undeformed configuration.
On the contrary, if P is such that the matrix of the coefficients of this system is
singular, evidently the system (6.3.18) and (6.3.19) has infinite solution and then,
since the (6.3.16), (6.3.17), and (6.3.15), the structure can absorb the load in infinite
deformed configurations.
So to determine the critical loads of the structure we must impose that the deter-
minant of the matrix of the coefficients of the system (6.3.18) and (6.3.19) is equal
to zero. So proceeding we easily get
# 1 1 ! 1 !" 1 !
l P 2 l P 2 l P 2 l P 2
4 cos − sin sin = 0.
2 EIn 2 EIn 2 EIn 2 EIn

As a consequence the least critical load is the smaller value among the solutions
P > 0 of the equation
1 !
l P 2
sin =0 (6.3.20)
2 EIn

and the solutions P > 0 of the equation

1 ! 1
l P 2 l P 2
tg = . (6.3.21)
2 EIn 2 EIn

The smaller one of the solutions of the (6.3.20) is

4π 2 EIn
l2
and the smaller one of the solutions of the (6.3.21) is (Fig. 6.3.12)

4.4932 EIn 8π 2 EIn

= .
l2 l2
So the smallest critical load of a built-in and sliding jointed beam submitted to
an axial compressive load P (Fig. 6.3.14) is

4π 2 EIn
Pcr = . (6.3.22)
l2

Frequently rather than the (6.3.10), (6.3.12), (6.3.14), and (6.3.22) it is preferable
to employ the unique formula of Euler

π 2 EImin
Pcr = . (6.3.23)
l02
618 6 Stability

In the (6.3.23) l0 is called free bending length and denotes the distance among
two subsequent points of inflection of the elastic curve of the beam, that is, since
the (6.3.2), the distance among two subsequent points of the beam where M=0.
Therefore l0 = l √ in the case of the supported beam; l0 = 2l in the case of the
cantilever; l0 = l/ 2 in the case of the built-in and supported beam; l0 = l/2 in the
case of the built-in and sliding jointed beam. Furthermore Imin denotes the minimum
moment of inertia of the cross section allowed by the constraints.

6.3.3 Exact Analysis

The approximate analysis developed in Sect. 6.3.2 furnishes results only partially in
accord with the experiences performed in the Laboratories material tests. Precisely
all the experiences have shown that the formula of Euler (6.3.23) furnishes exact
values of the critical load but the real behavior of the structure is deeply different.
Precisely, denoting with v0 the vertical displacement of the middle cross section
of the beam, the mathematical model [6.3.1] involves that for P = Pcr the beam can
balance the load in the infinite configurations (6.3.9) and for P > Pcr the beam can
balance the load only in the initial undeformed configuration. Such results are sum-
marized by the diagram of Fig. 6.3.5. On the contrary, the experiences performed
in the Laboratories material tests show as real behavior of the structure that of
Fig. 6.3.6. In fact we verify that in reality for P > Pcr the beam can absorb the load
in three configurations of equilibrium: in the two deformed stable configurations S1 ,
S2 and in the initial undeformed unstable configuration I.
It is possible to obtain a more precise mathematical model for the problem of the
axially loaded beam by partially removing the hypothesis of small deformations.
Insofar with reference to the problem of Fig. 6.3.7 we still assume

1 M
=− (6.3.24)
r EIn
but simulate the curvature with the (Figs. 6.3.16 and 6.3.17)

1 dϕ
= (6.3.25)
r ds

Fig. 6.3.16

Fig. 6.3.17
6.3 Static Method 619

Fig. 6.3.18

rather than with the

1 d2 v
= 2
r dz
that strictly use the hypothesis of small deformations. In the (6.3.25) s denotes the
curvilinear abscissa, that we assume equal to zero at the built-in end. In Fig. 6.3.18
we see a deflected configuration of the beam in the field of the great deformations.
So we assume as differential equation of the problem of Fig. 6.3.7 the

dϕ M
=− . (6.3.26)
ds EIn
Let us denote with v(s) [resp. v0 = v(l)] the lowering of the cross section
S [resp. B] (Fig. 6.3.18) and with ϕ(s) [resp. ϕ0 = ϕ (l)] the rotation of the cross
section S [resp. B]. As a consequence

M (s) = −P [v0 − v (s)]

from which, since the (6.3.26)

dϕ P
=− (v0 − v) . (6.3.27)
ds EIn
Since (Fig. 6.3.19)

dv
= sinϕ,
ds
from the (6.3.27) we get

d2 ϕ P dv P
=− =− sinϕ.
ds2 EIn ds EIn

Insofar in the field of the great deformations we assume as mathematical model

of the cantilever axially loaded of Fig. 6.3.7 the boundary problem

Fig. 6.3.19
620 6 Stability

[6.3.5] Find a real function v defined in [0, l], differentiable at least up to second
order and such that

d2 ϕ P
2
(s) = − sinϕ (s) ∀s ∈ ]0, l[ (6.3.28)
ds EIn

ϕ (0) = 0 (6.3.29)

dϕ
(l) = 0.  (6.3.30)
ds

The problem [6.3.5] evidently admits the banal solution ϕ = 0, that evidently
corresponds to the initial undeformed configuration.
Let us see if the problem [6.3.5] admits solutions different from that banal.
From the (6.3.28), multiplying both the sides for dϕ
ds (s) and integrating on [t, l] ,
where t ∈ ]0, l[, we get

 
l d2 ϕ dϕ l P dϕ
ds = − sinϕ ds
t ds2 ds t EIn ds

from which

2 2
1 dϕ 1 dϕ P
(l) − (t) = (cosϕ(l) − cosϕ0 )
2 ds 2 ds EIn

from which

2
dϕ P
∀s ∈ ]0, l[ (s) =2 (cosϕ(s) − cosϕ0 )
ds EIn

from which, admitting that ϕ(s) is strictly increasing, we have

1
ds EIn 2 1
= 1
.
dϕ 2P (cosϕ − cosϕ0 ) 2

As a consequence, supposing that to deformation happened the beam still has

length l, we obtain

 l  ϕ0  ϕ0
1
ds EIn 2 1
l= ds = dϕ = 1
dϕ
0 0 dϕ 0 2P (cosϕ − cosϕ0 ) 2
6.3 Static Method 621

so that
1  ϕ0
2P 2 1
l = 1
dϕ
EIn 0 (cosϕ − cosϕ0 ) 2
 ϕ0
1
=    1 dϕ (6.3.31)
0 1 − 2sin2 ϕ2 − 1 − 2sin2 ϕ20 2
 ϕ0
1 1
=√ dϕ.
2 0 sin2 ϕ0 − sin2 ϕ  2
1

2 2

Let us put now

ϕ0 1 ϕ
p = sin , ϑ = arcsin sin
2 p 2
so that

ϕ 2p cosϑ dϑ
sin = p sinϑ, ϕ = 2arcsin (p sinϑ) , dϕ =  1 .
2 1 − p2 sin2 ϑ 2
Then from the (6.3.31) it follows

1  π/2
P 2 1 2p cosϑ
2l =  1   1 dϑ
EIn 0 p2 − p2 sin2 ϑ 2 1 − p2 sin2 ϑ 2

from which
1  π/2
P 2 1
l =     1 dϑ. (6.3.32)
EIn 0 1 − sin2 ϕ20 sin2 ϑ 2

The integral one that appears in the (6.3.32) is a first kind elliptic integral. Its
develop by series and the (6.3.32) furnish

1 +∞
 1 · 2 · 3 · . . . · (2n − 1) 2
P 2 π π 2n
l = = sin ϕ0 . (6.3.33)
EIn 2 2 · 4 · 6 · . . . · (2n) 2
n=1

So if ϕ0 = 0, that is if the problem [6.3.5] admits a solution not identically null,

it is necessary that

π 2 EIn
P> . (6.3.34)
(2 l)2
622 6 Stability

Fig. 6.3.20

As a consequence, if
π 2 EIn
P≤ ,
(2 l)2
the problem [6.3.5] admits only the banal solution.
It is possible to show that the (6.3.34) is also a sufficient condition for a non
banal solution of the problem [6.3.5] to exist. Furthermore and evidently if ϕ is a
non banal solution of the problem [6.3.5], then also −ϕ is a non banal solution of
the problem [6.3.5]. 
This way we easily can express P/Pcr in function of v0 l. The theoretical diagram
obtained is perfectly coincident with the experimental one of Fig. 6.1.4. We also ver-
ify that the configurations S1 , S2 are of stable equilibrium and that the configuration
I is of unstable equilibrium.
Remark 6.3.2 We easily verify that in the problem [6.3.5] when P increases then the
beam assumes the deformed configurations of Fig. 6.3.20. 

6.3.4 Effect of the Imperfections

As a rule in the reality the preceding schemes of perfectly rectilinear beams and
of perfectly axial loads can be seen only in the Laboratories material tests. In the
constructive practice in fact the axis of a beam is never perfectly rectilinear; an axial
load results always applied on a beam with a certain eccentricity; on a beam small
transversal loads are often present. Such imperfections have the obvious effect to
vary in a non negligible way the lowering of the axially loaded beam. Nevertheless,
as it is easy to theoretically and experimentally verify, their presence doesn’t vary
the value of the critical load.
We for instance consider the scheme of Fig. 6.3.21, in which the axis of the beam
in the initial configuration is not rectilinear. On the contrary, it has the equation
 
ṽ (z) = ṽ0 sin π z l .

Fig. 6.3.21
6.3 Static Method 623

Fig. 6.3.22

It is easy to verify that in the approximate static analysis when P tends toward the
critical load Pcr then the lowering v0 of the cross section z = l/2 tends toward +∞.
Precisely in the approximate static analysis, supposing non rectilinear the axis of
the beam in the initial configuration, we find that the theoretical curve of Fig. 6.3.5
turns into Fig. 6.3.22.
Partially removing the hypothesis of small deformations, analogously to the exact
static analysis, supposing not rectilinear the axis of the beam in the initial con-
figuration we find that the theoretical curve of Fig. 6.3.6 turns into Fig. 6.3.23.
Another imperfection which it is necessary to appraise is the eccentricity of the
axial load. We consider the problem of Fig. 6.3.24, in which the eccentricity e is
very small in comparison to l.
Similarly to the previous case, it is easy to verify that in the approximate static
analysis when we take into account the eccentricity of the axial load then the
theoretical curve of Fig. 6.3.5 turns into Fig. 6.3.25.
Partially removing the hypothesis of small deformations, analogously to the exact
static analysis, supposing eccentric the axial load, we find that the theoretical curve
of Fig. 6.3.6 turns into Fig. 6.3.26.

Fig. 6.3.23

Fig. 6.3.24

Fig. 6.3.25

Fig. 6.3.26
624 6 Stability

Fig. 6.3.27

Finally we consider the effect due to the presence on the beam of a small transver-
sal force q (Fig. 6.3.27). It is intuitive and easy to verify that in presence of q the
axially loaded beam never admits a rectilinear configuration of equilibrium.
Similarly to the previous cases, it is easy to verify that in the approximate static
analysis when we take into account the small transversal force then the theoretical
curve of Fig. 6.3.5 turns into Fig. 6.3.28.
Partially removing the hypothesis of small deformations, analogously to the exact
static analysis, taking into account the small transversal force, we find that the
theoretical curve of Fig. 6.3.6 turns into Fig. 6.3.29.
The theoretical curves of Fig. 6.3.23 and of Fig. 6.3.26 have always received the
complete experimental verification in the Laboratories material tests. Precisely it is
gotten that

– if P < Pcr the beam admits that initial undeformed as unique configuration of
equilibrium, that results stable,
– if P > Pcr the beam admits three configurations of equilibrium S1 , S2 , I
(Fig. 6.3.28). Of these S1 and S2 are configurations of stable equilibrium, while I
is configuration of unstable equilibrium.

6.3.5 Limit Slenderness

The formula of Euler (6.3.23) furnishes the critical load Pcr of the axially loaded
beam, that is the minimum value of the axial load of compression beginning from
which the equilibrium of the beam bifurcates. We have seen that when P < Pcr

Fig. 6.3.28

Fig. 6.3.29
6.3 Static Method 625

the beam can balance the load only in the initial undeformed configuration and
when P > Pcr the beam can balance the load also assuming deformed deflected
configurations different from the initial undeformed configuration. It is obvious but
very important to underline that in every deformed deflected configuration the state
of stress of the beam is much more severe than in an initial rectilinear configuration.
The formula of Euler (6.3.23) was obtained employing the (6.3.2), that is valid
in the hypothesis of linear elasticity. So we can use it only if, denoting with A the
area of the cross section, it results

Pcr
σcr = < σp , (6.3.35)
A
where σp denotes the proportionality stress (cfr. Sect. 1.4).
We call slenderness of the beam and denote with the symbol λ the ratio

l0
λ= (6.3.36)
ρmin

where l0 is the free bending length and ρmin denotes the least inertia radius, that is
the greatness
3
Imin
ρmin = . (6.3.37)
A
Since the (6.3.23) and (6.3.37) it results

Pcr π 2 EAρmin
2
π 2E
σcr = = = (6.3.38)
A Al02 λ2
 
and then the diagram of the function σcr = σcr λ2 is an equilateral hyperbola,
called of Euler. Its asymptotes are the coordinated axes (Fig. 6.3.30). We call limit
slenderness of the beam the slenderness
4
E
λl = π (6.3.39)
σp

and it is obvious (Fig. 6.3.30) that, since the (6.3.38), we can employ the formula of
Euler (6.3.23) only if

Fig. 6.3.30
626 6 Stability

π 2E
≤ σp
λ2
that is only if

π 2E
λ2 ≥ = λ2l .
σp
From the (6.3.39) it follows that λl depends from the material. As a rule it results

60 ≤ λl ≤ 100.

When the beam is stumpy to study its stability we must do it in the inelastic field.
On the subject some complex theories are available that, in conclusion, prolong the
diagram of Fig. 6.3.30 in the field of the stumpy beams (Fig. 6.3.31).
Otherwise it is possible to apply the method ω, that is an empirical procedure,
based on experimentally built schedules, that furnish the value of the critical load of
any axially compressed beam, both stumpy and slender. Precisely in correspondence
of λ the schedule furnishes a value ω and we empirically assume, denoting with σ 0
the yielding stress

σ0 A
Pcr = . (6.3.40)
sω
The Sched. 6.3.1 furnishes the values ω in correspondence of λ for structural
steel with σ0 /s = 1, 600 kg/cm2 .

Problem 6.3.1 You should determine by the method ω the maximum axial compres-
sive load Pcr for the cantilever of Fig. 6.3.32. The beam has length 1.5 m and a
cross section of area A = 33.4 cm2 and minimum moment of inertia Imin = 117 cm4 .
Moreover σ0 /s = 1,600 kg/cm2 .

Fig. 6.3.31

Fig. 6.3.32
6.3 Static Method 627

Solution It results ρmin = 1.87 cm, l0 = 300 cm, λ = 160.4. From the Sched. 6.3.1
we have ω = 4.33. So

σ0 A
Pcr = = 12.3t. 
sω

Problem 6.3.2 A rod hinged at the ends has a length of 3.2 m, is constituted by two
angle irons 100 mm × 100 mm × 10 mm. You should verify by the method ω the
structural safety of the beam knowing that it is submitted to an axial compressive
load P = 19 t and that σ 0 /s= 1,600 kg/cm2 .
Solution We have A = 38.4 cm2 , Imin = 354 cm4 , ρmin = 3.04 cm, l0 = 320 cm.
So λ = 105.3 and then, since the Sched. 6.3.1, ω = 2.12. As a consequence the
structure is in structural safety. In fact, since the (6.3.40)

P kg σ0 kg
σ = = 494.8 2
≤ σcr = = 754.7 . 
A cm sω cm2

6.3.6 Other Ways of Buckling

A beam submitted to an axial compressive load can present other types of instability
over the deflection. This happens especially if the beam is thin walled.
In fact it is experimentally verified that the compressed tube of Fig. 6.3.33 is able
to assume the buckled configuration of Fig. 6.3.34.
It is also experimentally verified that if the torsional rigidity of a compressed
beam is small, then it can assume an equilibrium configuration of torsion type.
Precisely the cross sections rigidly rotate and the axis of the beam stays rectilin-
ear. This phenomenon, said of torsional instability, essentially concerns the thin
walled beams with open cross section.

Fig. 6.3.33
628 6 Stability

Fig. 6.3.34

If instead the cross section of the beam is open but not very thin, it is experimen-
tally verified that the beam can assume a configuration contemporary of bending
and torsion type. We call such case bending-torsional instability.
Another phenomenon that we meet in the study of the stability of the compressed
beam is the local instability. It is experimentally verified that if a compressed thin
walled beam has a very thin cross section then a local buckling of the wall of the
beam can take place well below the critical Euler load (Fig. 6.3.35). According
to the cases this phenomenon is able both to decrease the carrying ability of the
beam and increase it. It’s because of this that the very thin walled beams, of forced
employment in the Aeronautical constructions, are stiffened by curves or swellings
to the ends of the cross section (Fig. 6.3.36).

Fig. 6.3.35

Fig. 6.3.36
6.4 Second Type Instability 629

Sched. 6.3.1

λ 60 61 62 63 64 65 66 67 68 69
ω 1.31 1.32 1.33 1.34 1.35 1.36 1.37 1.38 1.39 1.40
λ 70 71 72 73 74 75 76 77 78 79
ω 1.41 1.42 1.44 1.45 1.46 1.47 1.49 1.50 1.51 1.53
λ 80 81 82 83 84 85 86 87 88 89
ω 1.54 1.56 1.58 1.60 1.62 1.63 1.65 1.67 1.69 1.71
λ 90 91 92 93 94 95 96 97 98 99
ω 1.73 1.75 1.78 1.80 1.83 1.85 1.87 1.90 1.92 1.93
λ 100 101 102 103 104 105 106 107 108 109
ω 1.97 2.00 2.03 2.06 2.09 2.11 2.14 2.17 2.20 2.23

6.4 Second Type Instability

6.4.1 The Snapping

Till now in any problem of stability, a little variation of the applied loads has always
produced a little variation of the deformed configuration. However some structural
schemes exist in which such condition is not satisfied, in the sense that to a small
variation of the applied load an abrupt passage of the structure to a much different
configuration can correspond, with notable dynamic effects. This behavior, called
second type instability or snapping, is typical of the reduced arcs, that is of structures
of the type of that of Fig. 6.4.1, where h is much smaller than l.
In fact in such structure, denoting with v0 the lowering of the section A
(Fig. 6.4.1), if we increase the value of the load P then the value of v0 continu-
ously grows. This happens until the load P reaches the critical value given by the
ordinate of the point H (Fig. 6.4.2). If we try in such configuration H to increase
a little P, then the structure brusquely turns into the configuration K (Fig. 6.4.3),
obviously with dangerous dynamic effects. Clearly the critical load corresponding
to the configuration H of the structure (Fig. 6.4.2) is the load of crisis.

Fig. 6.4.1

Fig. 6.4.2

Fig. 6.4.3
630 6 Stability

Fig. 6.4.4

The phenomenon of the snapping previously described has been perfectly

experimentally confirmed in the Laboratories material tests.
The same phenomenon can theoretically be simulated analyzing the structure of
Fig. 6.4.4, where the rods BC, DC are axially deformable but can not deflect.6.4.1
Furthermore α = 20◦ .
We adopt the reference frame of Fig. 6.4.5. Since the geometric and load symme-
try, the hinge C lowers according the vertical in a position C (Fig. 6.4.6). Obviously
in the deformed configuration the rod BC [resp. DC ] is rotated of an angle ϕ with
respect to the initial configuration BC [resp. DC] (Fig. 6.4.6).
We use the energy method to find the deformed configurations with which the
structure can absorb a load P. So we must consider all the possible compatible
configurations of the structure. Clearly they are the deformed configurations of the
structure that respect the constraints, that is the configurations shown in Fig. 6.4.6
with an arbitrary ϕ.
In the compatible configuration individualized by ϕ the potential energy func-
tional Jp assumes the value

1      
Jp (ϕ) = 2 N dist B, C − dist (B, C) − P dist C, C (6.4.1)
2

where N is the axial force in the rod.

Since the (2.4.6) it results

  N dist (B, C)
dist B, C − dist (B, C) = . (6.4.2)
EA

Fig. 6.4.5

Fig. 6.4.6

6.4.1 The system is easily realizable with piston rods sliding in coupling sleeves and springs.
6.4 Second Type Instability 631

From the (6.4.1) and (6.4.2), it follows

2
EA cosα l l
Jp (ϕ) = − − P (l tgα − l tg (α − ϕ))
l cos (α − ϕ) cosα
so that

(cosα − cos (α − ϕ))2

Jp (ϕ) = EA l
cosα cos2 (α − ϕ)
(6.4.3)

P
− (tgα − tg (α − ϕ)) .
EA

Since the [6.2.1] the points of relative extremes of Jp individualize the deformed
configurations in which the structure can absorb the load P. The equilibrium will be
stable [resp. unstable] in the case of relative minimum [resp. maximum].
Proceeding analytically we determine the zeros of

dJp
dϕ
and secondly appraise the sign of

d2 Jp
dϕ 2
in such zeros. Precisely from the (6.4.3) we obtain

dJp Pl 2 EA l2 cosα − cos (α − ϕ)

(ϕ) = − 2 +
dϕ cos (α − ϕ) dist (B, C) cosα cos (α − ϕ)

−sin (α − ϕ) cosα cos (α − ϕ) − (cosα − cos (α − ϕ)) cosα sin (α − ϕ)

·
cos2 α cos2 (α − ϕ)
from which
dJp Pl
(ϕ) = −
dϕ cos2 (α − ϕ)
(6.4.4)
(cosα − cos (α − ϕ)) sin (α − ϕ)
− 2 EA l
cos (α − ϕ) cos2 (α − ϕ)
from which

d2 Jp Pl 2 EA l
(ϕ) = 2cos (α − ϕ) sin (α − ϕ) −

dϕ 2 cos (α − ϕ)
2 cos (α − ϕ) 
2

· cos2 (α − ϕ) − sin2 (α − ϕ) − (cosα − cos (α − ϕ)) cos (α − ϕ)


− (cosα − cos (α − ϕ)) sin2 (α − ϕ) 3cos2 (α − ϕ)
632 6 Stability

from which

d2 Jp 2EA l P
(ϕ) = · sin (α − ϕ) cos (α − ϕ)
dϕ 2 cos2 (α − ϕ) EA   
+ cosα 1 + 2sin2 (α − ϕ) − cos (α − ϕ) 1 + sin2 (α − ϕ) .
(6.4.5)

From the (6.4.4) we get that the zeros of

dJp
dϕ
are the ϕ such that

P
= 2 [cos (α − ϕ) − cosα] tg (α − ϕ). (6.4.6)
EA

To solve the (6.4.6) we proceed graphically tracing the diagram of Fig. 6.4.7,
where α = 20◦ . From the diagram of Fig. 6.4.7 we get that when P = 0 the struc-
ture can absorb the load in the three configurations: C1 individualized by ϕ1 = 0
(Fig. 6.4.5), C2 individualized by ϕ2 = 20◦ (Fig. 6.4.8), C3 individualized by
ϕ3 = 40◦ (Fig. 6.4.9). To study the type of such equilibriums, employing the (6.4.3)
we trace the diagram of Jp of Fig. 6.4.10, where α = 20◦ , P = 0. So, since the

Fig. 6.4.7

Fig. 6.4.8

Fig. 6.4.9

Fig. 6.4.10
6.4 Second Type Instability 633

Fig. 6.4.11

Fig. 6.4.12

[6.2.1], C1 and C3 are configurations of stable equilibrium but C2 is a configuration

of unstable equilibrium.
From the diagram of Fig. 6.4.7 we get that when P = 0.01 EA the structure
can absorb the load in the three configurations: C1 individualized by ϕ1 = 2, 9◦
(Fig. 6.4.11), C2 individualized by ϕ2 = 14, 9◦ (Fig. 6.4.12), C3 individualized
by ϕ3 = 42◦ (Fig. 6.4.13). To study the type of such equilibriums, employing the
(6.4.3) we trace the diagram of Jp of Fig. 6.4.14, where α = 20◦ , P = 0.01 EA.
So, since the [6.2.1], C1 and C3 are configurations of stable equilibrium but C2 is a
configuration of unstable equilibrium.
From the diagram of Fig. 6.4.7 we get that when P = 0.0164 EA the structure
can absorb the load in the two configurations: D1 = H individualized by ϕ1 = 8, 4◦
(Fig. 6.4.15), D2 = K individualized by ϕ2 = 43◦ (Fig. 6.4.16). To study the type of
such equilibriums, employing the (6.4.3) we trace the diagram of Jp of Fig. 6.4.17,

Fig. 6.4.13

Fig. 6.4.14

Fig. 6.4.15

Fig. 6.4.16
634 6 Stability

Fig. 6.4.17

where α = 20◦ , P = 0.0164 EA. So, since the [6.2.1], D2 is a configuration of stable
equilibrium but D1 is a configuration of equilibrium neither stable nor unstable. In
fact
dJp
(ϕ1 ) = 0
dϕ
but, since the (6.4.5)

d2 Jp d3 Jp
(ϕ1 ) = 0, (ϕ1 ) = 0.
dϕ 2 dϕ 3

So ϕ1 = 8, 4◦ is not a point of relative extreme for Jp .

In conclusion in making statically grow the value of P from 0 up to the value
0.0164 EA the structure assumes configurations all stable ones and finally assume
the configuration H of Fig. 6.4.7. If we try in such configuration H to increase a little
P, then the structure brusquely turns into the configuration K (Fig. 6.4.3). Obviously
dangerous dynamic effects arise that as a rule will engender the collapse of the
structure.
Remark 6.4.1 The phenomenon of the snapping can happen only for the reduced
arcs. In fact only in such hypothesis the critical load is smaller than the load of
crisis for the collapse of the material. For instance for the structure of Fig. 6.4.6
we compare (Figs. 6.4.18 and 6.4.19) the values of the critical loads relative
to α = 11.25◦ and to α = 22.5◦ . Such diagrams are obviously furnished by
the (6.4.6). 

Fig. 6.4.18

Fig. 6.4.19
Chapter 7
Anisotropy

7.1 The Three-Dimensional Anisotropic Problem

7.1.1 Introduction

Let us now consider the same problem analyzed in Chap. 1, only changing the
hypothesis that the material is isotropic. Precisely we consider the three-dimensional
solid body C of Fig. 1.1.1 and still denote with V the region of 3 occupied by C
and still assume that the frontier S = ∂V of C is a regular surface of 3 . We still
refer to the Cartesian reference frame O, x, y, z of Fig. 1.1.1. We still suppose that the
material is homogeneous and still admits the hypothesis of small displacements. We
still suppose that the superficial distributed load px , py , pz and the volumetric load
X, Y, Z are mathematically regular.
Since in Sect. 1.1 all the results were obtained without making hypothesis on the
mechanical characteristics of the material, all such results of strain analysis are also
valid for the anisotropic body in examination. In particular

– the relations (1.1.28) between the components of strain εx , εy , εz , γxy , γxz , γyz and
the components of displacement u, v, w are still valid;
– the quadratic expressions (1.1.14) and (1.1.15) of the elongation εa and of the
shearing strain γab according to two any directions a, b are still valid;
– the compatibility equations (1.1.29) are still valid.

As for as the stress, analogously, since in Sect. 1.2 all the results were obtained
without making hypothesis on the mechanical characteristics of the material, all
such results of stress analysis are also valid for the anisotropic body in examination.
In particular

– the shearing stress symmetry [1.2.1] is still valid;

– the differential equations of equilibrium [1.2.3] are still valid;
– the Cauchy’s relations [1.2.4] are still valid;
– the Cauchy’s boundary relations [1.2.5] are still valid;

A. Maceri, Theory of Elasticity, DOI 10.1007/978-3-642-11392-5_7, 635


C Springer-Verlag Berlin Heidelberg 2010
636 7 Anisotropy

– the quadratic expressions (1.2.8) and (1.2.9) of the normal stress σn and of the
shearing stress τns according to two any directions n, s are still valid.

Finally, since in Sect. 1.3 the result was obtained without making hypothesis on
the mechanical characteristics of the material, the principle of virtual works [1.3.1]
is still valid for the anisotropic body in examination.
So to build the mathematical model of the problem of the elastic equilibrium
of an homogeneous and anisotropic body, in the hypotheses of linear elasticity and
small displacements, we only need to detail the stress strain relations for the various
types of anisotropy.

7.1.2 Constituent Links

The experiences performed in the Laboratories material tests have suggested that
a linear link between stress and strain could describe, at least in the field of
small deformations, also the behavior of an anisotropic body. This conjecture was
favorable. In fact it conducts to a mathematical model of the problem of the
elastic equilibrium such that all the values furnished always have received bright
experimental confirmations.
So we assume that in any point (x, y, z) of the volume V occupied by the body it
results

εx = a11 σx + a12 σy + a13 σz + a14 τxy + a15 τxz + a16 τyz

εy = a21 σx + a22 σy + a23 σz + a24 τxy + a25 τxz + a26 τyz
εz = a31 σx + a32 σy + a33 σz + a34 τxy + a35 τxz + a36 τyz
(7.1.1)
γxy = a41 σx + a42 σy + a43 σz + a44 τxy + a45 τxz + a46 τyz
γxz = a51 σx + a52 σy + a53 σz + a54 τxy + a55 τxz + a56 τyz
γyz = a61 σx + a62 σy + a63 σz + a64 τxy + a65 τxz + a66 τyz

where the 36 constants a11 , . . ., a66 are called elastic constants of the material. The
(7.1.1) are called constituent link of the medium or generalized Hooke’s law.
The medium is called isotropic if the elastic constants a11 , . . ., a66 do not depend
from the choice of the reference frame O, x, y, z. Otherwise, the material is called
anisotropic. Clearly in an anisotropic body an infinitesimal parallelepiped, even if
equally loaded, differently deforms it depending on the orientations of the planes
of cut.
We can prove by energetic considerations that in any anisotropic medium the
distinct elastic constants are at most 21.
If in an anisotropic body there is symmetry of the elastic properties, some of
the elastic constants aij are zero and as a consequence the generalized Hooke’s law,
valid for the anisotropy of general type, is simplified. Obviously this fact involves
very notable analytical advantages.
An important particular case of these anisotropy is that of the material with a
plane of elastic symmetry. In such case a plane α exists such that in any point P of
7.1 The Three-Dimensional Anisotropic Problem 637

the body, denoting with α P the plane parallel to α and passing through P, every two
directions that are symmetrical with respect to α P are equivalent in reference to the
elastic properties. The normal to α is called principal direction of elasticity. It is
easy to show that if z is principal direction of elasticity then the elastic constants aij
are reduced to 13 and the (7.1.1) become

εx = a11 σx + a12 σy + a13 σz + a14 τxy

εy = a12 σx + a22 σy + a23 σz + a24 τxy
εz = a13 σx + a23 σy + a33 σz + a34 τxy
(7.1.2)
γxy = a14 σx + a24 σy + a34 σz + a44 τxy
γxz = a55 τxz + a56 τyz
γyz = a56 τxz + a66 τyz .

As a consequence an infinitesimal parallelepiped of material cut out according

the coordinate planes and only submitted to a compressive σ z

– shortens according to z,
– dilates it according to x and according to y,
– has its bases of normal z transformed from rectangles to parallelograms.

Another important particular case of anisotropy is the orthotropic material. We

call orthotropic a material with three planes of elastic symmetry two by two orthog-
onal. Assuming the coordinate axes normal to the planes of elastic symmetry we
find that there are twelve elastic constants equal to zero. Then the (7.1.1) become:

εx = a11 σx + a12 σy + a13 σz

εy = a12 σx + a22 σy + a23 σz
εz = a13 σx + a23 σy + a33 σz
(7.1.3)
γxy = a44 τxy
γxz = a55 τxz
γyz = a66 τyz .

Introducing the engineering elastic constants (modulus of Young E1 , E2 , E3 , coef-

ficients of Poisson ν12 , ν21 , ν13 , ν31 , ν23 , ν32 and tangent modulus G12 , G13 , G23 ), the
Eq. (7.1.3) can be written as

ν21 ν31
εx = E1 σx − E2 σy − E3 σz
1

εy = − νE121 σx + E12 σy − νE323 σz

εz = − νE131 σx − νE232 σy + E13 σz
(7.1.4)
γxy = G12 τxy
1

γxz = G13 τxz

1

γyz = G23 τyz .

1
638 7 Anisotropy

It is important to note that only nine of the twelve elastic constants E1 , E2 , E3 ,

ν12 , ν21 , ν13 , ν31 , ν23 , ν32 , G12 , G13 , G23 are independent. In fact since the symmetry
we always have

E1 ν21 = E2 ν12 , E2 ν32 = E3 ν32 , E3 ν13 = E1 ν31

Another important particular case of anisotropy is the transversely isotropic

material. We call transversely isotropic a material if a plane α exists such that in any
point P of the body, denoting with α P the plane parallel to α and passing through P,
all lines of α P are elastically equivalent. In such case the plane α is called plane of
transversal isotropy and any line normal to α is called axis of transversal isotropy.
Let the z axis be taken normal to the plane of isotropy α, with the axes x, y arbitrarily
directed in α. Then the (7.1.1) become:
εx = a11 σx + a12 σy + a13 σz
εy = a12 σx + a11 σy + a13 σz
εz = a13 σx + a13 σy + a33 σz
(7.1.5)
γxy = 2(a11 − a12 )τxy
γxz = a55 τxz
γyz = a66 τyz .
The number of independent elastic constants is five. Introducing the engineering
elastic constants Ez , Ex , Gxz , νxy , νxz the Eq. (7.1.5) can be written as

νxz
εx = Ex (σx − νxy σy ) − Ez σz
1

εy = E1x (σy − νxy σx ) − νExzz σz

εz = − νEzxx (σx + σy ) + E1z σz
(7.1.6)
γxy = G1xy τxy
γxz = G1xz τxz
γyz = G1xz τyz ,

where
νxz νzx Ex
= , Gxy = .
Ez Ex 2(1 + νxy )

A last important particular case of anisotropy is the material with cylindrical

anisotropy. In this case the anisotropy is curvilinear. We said that a material has a
cylindrical anisotropy if a line s exists, called axis of cylindrical anisotropy, such
that all lines normal to s are elastically equivalent. A particular case of cylindrical
anisotropy is when in any point three planes of elastic symmetry exist: a plane nor-
mal to s, another containing s and the third normal to both. A material of this kind
is called orthotropic with cylindrical anisotropy. For it the (7.1.1) become, in the
cylindrical coordinates O, r, ϑ, z, where z is the axis of cylindrical anisotropy:
7.1 The Three-Dimensional Anisotropic Problem 639

νϑr νzr
εr = Er σr − Eϑ σϑ − Ez σz
1

εϑ = − νErϑr σr + E1ϑ σϑ − νEzϑz σz

εz = − νErzr σr − νEϑzϑ σϑ + E1z σz
(7.1.7)
γrϑ = Grϑ τrϑ
1

γrz = Grz τrz

1

γϑz = Gϑz τϑz .

1

Remark 7.1.1 When we pass from the reference frame O, x, y, z to a new reference
frame O , x , y , z , the new elastic constants can be easily determined express-
ing εx , . . . , γy z in function of εx , . . . , γyz by means of the (1.1.14) and (1.1.15).
After that we express εx , . . . , γyz in function of σx , . . . , τyz by means of the (7.1.1).
Subsequently we express σx , . . . , τyz in function of σx , . . . , τy z by means of the
(1.2.8) and (1.2.9). This way we obtain the expressions of the elastic constants
a 11 , . . . , a 66 in the new reference frame in function of the old elastic constants
a11 , . . . , a66 . 
Remark 7.1.2 If there is symmetry of the elastic properties we can easily eliminate
from the generalized Hooke’s law some of the elastic constant a11 , . . . , a66 . In fact
it is sufficient to introduce a new reference frame O , x , y , z where for instance the
axes x, x are elastically equivalent. Then we determine the new elastic constants in
the way seen in Remark 7.1.1. Imposing the elastic equivalence between x and x ,
we obtain a link for the elastic constant a11 , . . . , a66 . 
As in the isotropic case, the inverse relationships of the (7.1.1) are very useful.
We easily obtain

σx = c11 εx + c12 εy + c13 εz + c14 γxy + c15 γxz + c16 γyz

σy = c21 εx + c22 εy + c23 εz + c24 γxy + c25 γxz + c26 γyz
σz = c31 εx + c32 εy + c33 εz + c34 γxy + c35 γxz + c36 γyz
(7.1.8)
τxy = c41 εx + c42 εy + c43 εz + c44 γxy + c45 γxz + c46 γyz
τxz = c51 εx + c52 εy + c53 εz + c54 γxy + c55 γxz + c56 γyz
τyz = c61 εx + c62 εy + c63 εz + c64 γxy + c65 γxz + c66 γyz

where obviously the coefficients cij are functions of the coefficients aij .

7.1.3 The Anisotropic Elastic Body

We consider a solid deformable body C and we denote with V its volume, i.e. the
portion of the three-dimensional space occupied by C. We suppose that the surface
(or boundary) S of V is regular, so in every point of S the tangent plane exists. Let us
refer to an orthogonal set of cartesian axes O, x, y, z and decompose S in two parts
Sp and Su . We suppose that the material is homogeneous, generally anisotropic and
linearly elastic, so that the (7.1.1) are true. We suppose that we are in the field of
640 7 Anisotropy

small deformations, so that the (1.1.28) are true. We suppose that the body is con-
strained in the points of Su . We suppose that, in any instant t, the body is submitted
to a volumetric load X, Y, Z and, in the points of Sp , to a distributed superficial load
px , py , pz . We suppose that the load X, Y, Z, px , py , pz and the constraint’s reactions
constitute a system of forces equivalent to zero.
We consider the elastic equilibrium problem of the anisotropic body, that is the
problem to find deformed configurations in which the constrained body supports
the applied load in equilibrium. Clearly the possible deformed configurations in
equilibrium must satisfy

– the inside compatibility, that is the (1.1.29),

– the outside compatibility, that is the conditions that the constraints impose to the
points of Su ,
– the inside equilibrium, that is the (1.2.3),
– the boundary equilibrium, that is the (1.2.5) in the points of Sp .

So mathematically the elastic equilibrium problem of an anisotropic body is

a boundary problem, i.e. the problem to find functions satisfying in V a system
of differential equations and satisfying in the boundary S of V some assigned
conditions.
We assume, according to Cauchy, as unknown quantities of the problem the
functions u, v, w. First of all, we observe that the inside compatibility (1.1.29) is
banally satisfied. As the system of differential equations, from the (1.2.3), taking
into account the (7.1.8) and (1.1.28), we have
  
∂
∂x c11 ∂u
∂x + · · · + c16
∂v
∂z + ∂w
∂y + ··· + X = 0
  
∂
∂x c41 ∂u
∂x + · · · + c46
∂v
∂z + ∂w
∂y + ··· + Y = 0 (7.1.9)
  
∂
∂x c51 ∂u
∂x + · · · + c56
∂v
∂z + ∂w
∂y + · · · + Z = 0.

As for the boundary conditions, they are assigned on Su directly in terms of u, v, w

and are assigned on Sp with the (1.2.5), which we explicit in terms of u, v, w by
means (7.1.8) and (1.1.28).
So for the generally anisotropic body the formulation of the elastic equilibrium
problem is the following boundary problem:
[7.1.1] In the hypothesis of regularity of S and of X, Y, Z, px , py , pz find the functions
u, v, w that are real, definite and regular in V and are such that

– the differential equations (7.1.9) are satisfied in V

– the boundary conditions are satisfied in S = Su ∪ Sp . 

We can prove the following existence and uniqueness theorem

[7.1.2] The problem [7.1.1] admits a solution and only one. 
7.1 The Three-Dimensional Anisotropic Problem 641

Furthermore, as a simple consequence of the mathematical linearity of the prob-

lem [7.1.1], we get that the principle of superposition [1.5.4] also subsists for the
anisotropic body.
Finally, we emphasize that in the thermoelastic field, putting T = T − Ti and
denoting with α x [resp. α y ] [resp. α z ] the coefficient of free thermal linear expan-
sion according the axis x [resp. y] [resp. z] and with α xy [resp. α xz ] [resp. α yz ] the
coefficient of free thermal shearing strain according the axes x, y [resp. x, z] [resp.
y, z], the (7.1.1) must be substituted by

εx = a11 σx + a12 σy + a13 σz + a14 τxy + a15 τxz + a16 τyz + αx T

εy = a21 σx + a22 σy + a23 σz + a24 τxy + a25 τxz + a26 τyz + αy T
εz = a31 σx + a32 σy + a33 σz + a34 τxy + a35 τxz + a36 τyz + αz T
γxy = a41 σx + a42 σy + a43 σz + a44 τxy + a45 τxz + a46 τyz + αxy T
γxz = a51 σx + a52 σy + a53 σz + a54 τxy + a55 τxz + a56 τyz + αxz T
γyz = a61 σx + a62 σy + a63 σz + a64 τxy + a65 τxz + a66 τyz + αyz T.

All the elastic constants aij and all the coefficients αx , αy , αz , αxy , αxz , αyz can be
valued directly from simple laboratory experiments. Obviously for the orthotropic
material it results αxy = αxz = αyz = 0; for the isotropic material it results αx =
αy = αz = α and αxy = αxz = αyz = 0.

7.1.4 Energetic Aspects

In the hypotheses of Sect. 1.6 we easily verify that also in the case of gener-
ally anisotropic body the Clapeyron’s theorem [1.6.5] subsists and that the elastic
potential ϕ still has the expression

1 
ϕ= σx εx + σy εy + σz εz + τxy γxy + τxz γxz + τyz γyz . (7.1.10)
2

From theorem [5.1.9] we obtain for the crossed coefficients aij the condition:

aij = aji ∀i, j ∈ {1, . . . , 6}. (7.1.11)

The (7.1.1) and (7.1.10) furnish

1    
ϕ= σx a11 σx + · · · + a16 τyz + · · · + τyz a61 σx + · · · + a66 τyz
2
642 7 Anisotropy

from which, taking into account the (7.1.11), obviously we have

∂ϕ ∂ϕ ∂ϕ ∂ϕ ∂ϕ ∂ϕ
= εx , = εy , = εz , = γxy , = γxz , = γyz . (7.1.12)
∂σx ∂σy ∂σz ∂τxy ∂τxz ∂τyz

From the (7.1.11) we deduce that even in the most general case of anisotropy of
the 36 elastic constant aij those independent are at most 21. In reality it is possible
to demonstrate that even in the most general case of anisotropy, of the 36 elastic
constant aij those independent are fewer than 18.

7.1.5 The Anisotropic Saint Venant’s Problem

Let us now consider the very important problem of Saint Venant, already analyzed
in Chap. 2. We assume all the hypotheses made in Chap. 2, except the hypothesis
of isotropy, that we substitute with the hypothesis of general anisotropy (7.1.1). We
assume now the centroidal of inertia frame of reference of Fig. 7.1.1 and assume
that the cylindrical prism of Saint Venant is built-in in the point (0,0,l). Moreover
we put

u0 = u(0, 0, 0), v0 = v (0, 0, 0) , w0 = w (0, 0, 0) ,

  
∂u ∂v ∂ ∂v ∂u
ψ0 = ∂z (0, 0, 0), ϕ0 = ∂z (0, 0, 0), ϑ0 = 2l ∂z ∂x − ∂y (0, 0, 0),

and denote with ϑ the unitary torsion angle, so that

 
1 ∂ ∂v ∂u
ϑ= − .
2 ∂z ∂x ∂y

Fig. 7.1.1
7.1 The Three-Dimensional Anisotropic Problem 643

As for as the problem of the centroidal axial load, we immediately verify that it
results
N
σx = σy = τxy = τxz = τyz = 0, σz = ,
A
and then
εx = N
A a13 , εy = N
A a23 , εz = N
A a33 ,
γxy = N
A a36 , γxz = NA a35 , γyz = N
A a34,
from which
 
u = NA a13 x + a236 y
 
v = NA a236 x + a23 y
w = NA (a35 x + a34 y − a33 (l − z)) .
As a consequence

– the axis of the beam remains rectilinear,

– the cross section of the beam remains plane,
– ϑ = 0,
– u0 = v0 = ψ0 = ϕ0 = ϑ0 = 0, w0 = − EA NI
a33 .

As for as the problem of the right bending of axis x, we immediately verify that
it results
Mx
σx = σy = τxy = τxz = τyz = 0, σz = y,
Ix
and then
Mx Mx Mx
εx = Ix a13 y, εy = Ix a23 y, εz = Ix a33 y,
Mx
γxy = Ix a36 y, yxz = MIxx a35 y, γyz = MIxx a34 y,
from which
Mx  
u = 2Ix 2a13 xy + a36 y2 − a35 (l − z)

Mx  
v = 2Ix −a13 x2 + a23 y2 − a33 (l − z)2 + a35 x(l − z)

Mx  
w= 2Ix a35 xy + a34 y2 − 2a33 y(l − z) .

As a consequence

– the axis of the beam turns into an arc of parabola contained in the plane yz,
– the cross section of the beam turns into a second order surface,
644 7 Anisotropy

and

ϑ = −M x
2Ix a35,
Mx l 2
u0 = 0, v0 = 2Ix a33 , w0 = 0,
Mx l
ψ0 = 0, ϕ0 = Ix a33 , ϑ0 = − M xl
2Ix a35.
As for as the problem of the right bending of axis y, we immediately verify that
it results
My
σx = σy = τxy = τxz = τyz = 0, σz = Iy x,
My My My
εx = Iy a13 x, εy = Iy a23 x, εz = Iy a33 x,
My M My
γxy = Iy a36 x, γxz = Iyy a35 x, γyz = Iy a34 x,

and then  
My
u = 2Iy a13 x2 − a23 y2 − a33 (l − z)2 + a34 y(l − z)
My  
v = 2Iy a36 x2 + 2a23 xy − a34 x(l − z)
My  
w = 2Iy a35 x2 + a34 xy − 2a33 x(l − z)
My
ϑ = 2Iy a34 ,

My l 2
u0 = − 2Iy a33 , v0 = 0, w0 = 0,
My l My l
ψ0 = Iy a33 , ϕ0 = 0, ϑ0 = 2Iy a34.

As for as the problem of the torsion, we immediately verify that for mono-
connected cross section it results
 
σx = ∂∂yF2 , σy = ∂∂xF2 , σz = 2aM33t aI34y x − aI35x y
2 2

 
− a133 a13 σx + a23 σy + a36 τxy + a35 τxz + a34 τyz
∂ψ
∂ F
τyz = − ∂ψ
2
τxy = − ∂x∂y , τxz = ∂y , ∂x ,

where F(x, y) and ψ (x, y) are solution of the system

L4 F + L3 ψ = 0
Mt a234 a235
L3 F + L2 ψ = −2ϑ̄ + 2a33 Iy + Ix

with the boundary conditions

∂F ∂F
= = ψ = 0.
∂x ∂y
7.1 The Three-Dimensional Anisotropic Problem 645

The operators L2 , L3 , L4 have expressions

∂ 2 ∂ 2 ∂ 2
L2 = β44 ∂x 2 − 2β45 ∂x∂y + β55 ∂y2

∂ 3 ∂ 3 ∂ 3 ∂ 3
L3 = −β24 ∂x 3 + (β25 + β46 ) ∂x2 ∂y − (β14 + β56 ) ∂x∂y2 + β15 ∂y3

∂ 4 ∂ 4 ∂ 4 ∂ 4 ∂ 4
L4 = β22 ∂x 4 − 2β26 ∂x3 ∂y + (2β12 + β66 ) ∂x2 ∂y2 − 2β16 ∂x∂y3 + β11 ∂y4

where
ai3 aj3
βij = aij − .
a33
The symbol ϑ̄ denotes a constant. His value is given by the condition that
the distribution of the τ z has resultant moment equal to Mt . Resolved the previ-
ous boundary problem, the displacement components for integration are gotten by
the strain components. Nevertheless their explicit expressions are worthwhile to
draw. Integrating the third one, the fourth one and the fifth one of the stress–strain
relationships and keeping into account the expression of σ z , it is gotten:
2  ∂D
u = − z2 + z a51 σx + a52 σy + a53 σz + a56 τxy + a55 τxz
∂y 
+a54 τyz − ∂W
∂x
0
+ U0 (x, y)
z2 ∂D

v = − 2 ∂y + z a41 σx + a42 σy + a43 σz + a46 τxy + a45 τxz

+a44 τyz − ∂W
∂y
0
+ V0 (x, y)
w = zD + W0 (x, y)
where
Mt a35 a34
D= − y+ x.
2 Ix Iy
Introducing such expressions of u, v, w in the remaining stress–strain relation-
ships we obtain the equations
 
∂
∂x a51 σx + a52 σy + a53 σz + a56 τxy + a55 τxz + a54 τyz − ∂W∂x
0
=0
 
∂ ∂W0
∂y a41 σx + a42 σy + a43 σz + a46 τxy + a45 τxz + a44 τyz − ∂y =0
 
∂ ∂W0
∂y a51 σx + a52 σy + a53 σz + a56 τxy + a55 τxz + a54 τyz − ∂x +
 
∂
+ ∂x a41 σx + a42 σy + a43 σz + a46 τxy + a45 τxz + a44 τyz − ∂W ∂y
0
=0

that by integration furnish

a51 σx + a52 σy + a53 σz + a56 τxy + a55 τxz + a54 τyz = −ϑ̄y + ω2
a41 σx + a42 σy + a43 σz + a46 τxy + a45 τxz + a44 τyz = ϑ̄x − ω1
646 7 Anisotropy

where ω1 and ω2 are constants. Introducing the constants ω3 , α, β, γ and putting

U = U0 + ω3 y − α
V = V0 + ω3 x − β
W = W0 − ω1 y + ω2 x − γ
we obtain
Mt
u = − 4Iy
a34 z2 − ϑ̄yz + U + ω2 z − ω3 y + α
Mt
v = − 4I a35 z2 + ϑ̄xz + V + ω3 x − ω1 z + β
x 
w = M2t aI34y x − aI35x y (z − l) + W + ω1 y − ω2 x + γ
from which, imposing the constraint condition of Fig. 7.1.1, we get

Mt 2 
u = − 4I y
a 34 z − 2lz + l 2 + ϑ̄y(l − z)

1  ∂U ∂V

− (0, 0) − (0, 0) y + U − U(0, 0)
2 ∂y  ∂x

Mt
v = − 4Ix
a35 z2 − 2lz + l2 − ϑ̄x(l − z)
1  
+ ∂U (0, 0) − ∂V
(0, 0) x + V − V(0, 0)
2 ∂y 
∂x

w = M2t a234 x − a235 y (z − l) + W − W(0, 0).

In conclusion, it is gotten ϑ = ϑ̄ and

Mt l 2 Mt l 2
u0 = 4Iy a34 , v0 = 4Ix a35 , w0 = 0,
Mt l Mt l
ψ0 = 2Iy a34 , ϕ0 = 2Ix a35 , ϑ0 = ϑ̄l.

As for as the problem of the shear of axis y, we immediately verify that for
mono-connected cross section it results

σx = ∂∂yF2 , σy = ∂∂xF2 , σz = − Ixy zy − 2Aa

2 2 T y 34 T a
33
T    
− 2Ix ay33 a35 xy + a34 y − a33 a13 σx + a23 σy + a36 τxy + a35 τxz + a34 τyz
2 1

∂ψ
∂ F
τyz = − ∂ψ
2
τxy = − ∂x∂y , τxz = ∂y + τ1 , ∂x + τ2 ,

where τ1 (x, y) and τ2 (x, y) are two arbitrary functions such that

∂τ1 ∂τ2 Ty
+ − y=0
∂x ∂y Ix

and F(x, y) and ψ(x, y) are solution of the system

7.1 The Three-Dimensional Anisotropic Problem 647

Ty ∂2
L4 F + L3 ψ = 2Ix a33 (2a13 a34 − a35 a36 ) − ∂y2
(β14 τ2 + β15 τ1 )
∂2 ∂2
− ∂x2 (β24 τ2 + β25 τ1 ) + ∂x∂y (β46 τ2 + β56 τ1 )
T   
L3 F + L2 ψ = 2Ix ay33 a235 − 4a13 a33 x + (a34 a35 − 2a33 a36 ) y
∂ ∂
− 2ϑ̄ + ∂x (β44 τ2 + β45 τ1 ) − ∂y (β45 τ2 + β55 τ1 )

with the boundary conditions


∂F ∂F s
= = 0, ψ= τ2 dx − τ1 dy.
∂x ∂y 0

The operators L2 , L3 , L4 and the constants βij have the same expression given for
the previous problem of the torsion. The symbol ϑ̄ denotes a constant. Its value is
given by the condition that the distribution of the τ z has a resultant moment with
respect to the barycentre G equal to zero. Having resolved the previous boundary
problem, the displacement components are gotten from the strain components by
integration. Of them nevertheless it is worthwhile to draw, with the procedure shown
in the previous problem of the torsion, the following expressions:

Ty  a36 2 
a35  2
u= −a13 xyz − 2 y z+ 4 y l
− z2
Ix  
+ϑ̄(l − z)y + U(x, y) − U(0, 0) + 12 ∂V∂x (0, 0) − ∂U
∂y (0, 0) y
  
Ty a33  3    
v = Ix 6 2 l − 3 l2 z + z3 + 12 a13 x2 − a23 y2 z − a435 l2 − z2 x
 
−ϑ̄(l − z)x + V(x, y) − V(0, 0) − 12 ∂V∂x (0, 0) − ∂U
∂y (0, 0) x
Ty  2   Ty
w = 2Ix a33 l − z y − a35 xyz − a34 y z + 2A a34 (l − z)
2 2

+W(x, y) − W(0, 0).

As a consequence
Ty a35
ϑ = ϑ̄ + 2Ix z
T l3 Ty l
u0 =0, v0 = 3Iy x a33 , w0 = 2A a34 ,
T l2 T l2
ψ0 = 0, ϕ0 = − 2Iy x a33 , ϑ0 = ϑ̄l + 4Iy x a35.

As for as the problem of the shear of axis x, we immediately verify that for
mono-connected cross section it results

σx = ∂∂yF2 , ∂2F
2 Tx a35
σy = ∂x2
, σz = − TIyx zx − 2Aa 33
   
− 2ITy ax33 a34 xy + a35 x 2 − a133 a13 σx + a23 σy + a36 τxy + a35 τxz + a34 τyz
∂ψ
∂ F
τyz = − ∂ψ
2
τxy = − ∂x∂y , τxz = ∂y + τ1 , ∂x + τ2 ,
648 7 Anisotropy

where τ1 (x, y) and τ2 (x, y) are two arbitrary functions such that

∂τ1 ∂τ2 Tx
+ − x=0
∂x ∂y Iy
and F(x, y) and ψ(x, y) are solution of the system

Tx ∂2
L4 F + L3 ψ = 2Iy a33 (2a23 a35 − a36 a34 ) − ∂y2
(β14 τ2 + β15 τ1 )
∂2 ∂2
− ∂x2 (β24 τ2 + β25 τ1 ) + ∂x∂y (β46 τ2 + β56 τ1 )
  
L3 F + L2 ψ = 2ITy ax33 a234 − 4a32 a33 y + (a34 a35 − 2a33 a36 ) x
∂ ∂
− 2ϑ̄ + ∂x (β44 τ2 + β45 τ1 ) − ∂y (β45 τ2 + β55 τ1 )

with the boundary conditions


∂F ∂F s
= = 0, ψ= τ2 dx − τ1 dy.
∂x ∂y 0

The operators L2 , L3 , L4 and the constants βij have the same expression given for
the previous problem of the torsion. The symbol ϑ̄ denotes a constant. Its value is
given by the condition that the distribution of the τ z has a resultant moment with
respect to the barycentre G equal to zero. Having resolved the previous boundary
problem, the displacement components are gotten by the strain components for inte-
gration. Of them nevertheless it is worthwhile to draw, with the procedure shown in
the previous problem of the torsion, the following expressions:
  3      
a33
u= Tx
Iy 6 a23 y2 − a13 x2 z − a434 l2 − z2 y
2 l − 3 l 3 z + z3 + 1
2
 
−ϑ̄(l − z)y + U(x, y) + U(0, 0) + 12 ∂V ∂x (0, 0) − ∂U
∂y (0, 0) y
Tx  a36 2 a34  2 
v = Iy −a23 xyz − 2 x z + 4 x l − z 2
 
+ϑ̄(l − z)x + V(x, y) − V(0, 0) − 12 ∂V ∂U
∂x (0, 0) − ∂y (0, 0) x
Tx     Tx
w = 2Iy
a33 l2 − z2 x − a34 xyz − a35 x2 z + 2A a35 (l − z)
+W(x, y) − W(0, 0).

As a consequence

Tx a34
ϑ = ϑ̄ − 2Iy z
Tx l 3 Tx l
u0 = 3Iy a33 , v0 = 0, w0 = 2A a35 ,
2 Tx l 2
ψ0 = − T2Ix ly a33 , ϕ0 = 0, ϑ0 = −ϑ̄l − 4Iy a34 .
7.2 The Macroscopic Anisotropy 649

7.2 The Macroscopic Anisotropy

7.2.1 Composite Materials

The composite materials have found by this time wide application in modern engi-
neering because of their valuable strength and lightness. The composite material is
a non traditional structural material since is a non homogeneous material. In fact it
is composed of two phases:

– the reinforcing elements, that are reinforcing fibers or reinforcing particles. They
have high strength and stiffness. The fibers can be continuous or not, unidirec-
tional or not, organic (polyamides) or mineral (glass, carbon) or metallic (boron,
aluminium). The particles can be mineral or metallic.
– a less strong and less stiff binder, usually called matrix, that has the assignment to
transmit the load to the reinforcing elements. The matrix can be organic (polymer
resin) or mineral (ceramic) or metallic.

The two phases are realized so that there is a perfect adherence to the interface
fiber-matrix. As a rule the phases are constituted by homogeneous and isotropic
materials. The reinforcing elements cover a volume fraction of the composite
material usually lying between 0.3 and 0.7.
A peculiar property of composite materials is that they can be tailor-made, in
order to fulfill the requirements necessary for any particular application. In other
words, the ensemble fiber-matrix, volume percent, type and orientation of the rein-
forcing elements can be chosen in the best way once the desired high temperature
behavior, density, strength and stiffness are assigned.
Extensive use is made of glass-fiber reinforced plastics. This is a composite
constituted by a plastic material reinforced with a very high number of very thin,
circular, continuous and unidirectional glass fibers.
Very often the glass-fiber reinforced plastic is employed to realize laminates.
Such laminate composite is constituted by several layers, everyone constituted by
glass-fiber reinforced plastic. Usually each layer has a different orientation of the
fibers. On average a glass-fiber has good mechanical properties. In fact it has density
ρ = 2.60 kg/dm3 , Young’s modulus E = 700,000 kg/cm2 , tensile strength σ r =
250 kg/cm2 .
Another very important reinforcing fiber is the carbon fiber. On average a
carbon-fiber has very good mechanical properties. In fact it has density ρ =
1.85 kg/dm3 , Young’s modulus E from 350,000 to 7,000,000 kg/cm2 , tensile strength
σ r from 15,000 to 50,000 kg/cm2 .
The aramid fibers with high mechanical properties, usually called kevlar, are also
frequently used.
The determination of the state of stress of a structure in composite mate-
rial can be made with good approximation by substituting the real material,
that is not homogeneous, with an ideal material anisotropic and homogeneous,
650 7 Anisotropy

macroscopically equivalent. The elastic constants of this equivalent material clearly

depend from the degree of elastic symmetry of the composite material.
The evaluation of the elastic constants of the ideal material macroscopically
equivalent to the composite can be done by calculation, knowing those of the con-
stituent materials and the geometrical parameters of their distribution, or better
experimentally.
After that we must solve the boundary problem [7.1.1] for the homogeneous
anisotropic body macroscopically equivalent to the composite.
Finally, as we will show, from its state of stress we can deduce the state of stress
of the fibers and of the matrix.
The basic configuration of fiber composite materials is the reinforcement of the
composite strengthened with continuous unidirectional fibers. Practically the fibers
dispose it according to hexagonal arrangement and then we have an hexagonal sym-
metry (Fig. 7.2.1). Thus such composite material is macroscopically transversely
isotropic with plane of isotropy normal to the fibers. So, taking the z axis along the
fibers, the macroscopically equivalent material has generalized Hooke’s equations
of the form (7.1.6).
The elastic constants Ez , Ex , Gxz , νzx , νxy can be estimate measuring the strains
in simple tests as tensile or shear tests, both longitudinal and transverse. It is also
possible to use for Ez the experimentally well confirmed law
 
Ez = Ef Vf + Em 1 − Vf (7.2.1)

where Ef [resp. Em ] denotes the Young’s modulus of the fibers [resp. matrix] and Vf
denotes the ratio (usually equal to 0.6) between the volume occupied by the fibers
and the volume occupied by the composite.
The law of the mixtures (7.2.1) is experimentally verified also for the stress. It
results
 
σz = σf Vf + σm 1 − Vf

where σ z is referred to the macroscopically equivalent material and σ f [resp. σ m ]

denotes the stress according z of the fiber [resp. matrix].
Remark 7.2.1 The problem of a multi-layered composite panel loaded in its plane
can be easily analyzed. We suppose that

– the structure works in the field of the linear elasticity and of small displacements,
– the composite panel consists of several layers, each constituted by a composite
reinforced with continuous unidirectional fibers,

Fig. 7.2.1
7.2 The Macroscopic Anisotropy 651

– the middle plane of the panel is plane of the geometrical symmetry of the structure,
so the n layers that constitute the panel are symmetrical two by two with respect
to the plane of symmetry, excluding the central layer if n is an odd number,
– external forces act only on the edges of the structure, parallel to the middle plane
and symmetrically ordered with regard to it,
– the phenomena of the buckling is absent.

To determine the distribution of stress in the outlined structure we assimilate

the composite panel to a homogeneous anisotropic panel equivalent to it. So we
must determine the basic equations of this equivalent structure. More precisely,
we must determine the stress–strain, equilibrium and compatibility relations of the
homogeneous anisotropic panel equivalent.
We assume a cartesian system of reference O, x, y, z with the plane x, y coincident
with the middle plane of the panel. Clearly the equations of the equilibrium of the
homogeneous anisotropic panel are

∂σx ∂τxy
+ +X =0
∂x ∂y
∂τyx ∂σy
+ +Y =0
∂x ∂y

the compatibility equation is

∂ 2 εx ∂ 2 εy ∂ 2 γxy
+ =
∂y2 ∂x2 ∂x∂y

and the generalized Hooke’s law is

σx = c11 εx + c12 εy + c13 γxy

σy = c21 εx + c22 εy + c23 γxy
τxy = c31 εx + c32 εy + c33 γxy

where cij = cji ∀i, j ∈ {1, 2, 3}.

The problem consists in determining the elastic constants cij . We observe that for
(k)
the layer k of the panel, denoting ∀i, j ∈ {1, 2, 3} with cij its elastic constants, the
generalized Hooke’s law can be written
(k) (k) (k) (k)
σx = c11 εx + c12 εy + c13 γxy
(k) (k) (k) (k)
σy = c21 εx + c22 εy + c23 γxy (7.2.2)
(k) (k) (k) (k)
τxy = c31 εx + c32 εy + c33 γxy .

Then, denoting with h the width of the composite panel and supposing that the n
constituent layers are of equal width, it results
652 7 Anisotropy

h  (k)
n
cij = cij ∀i, j ∈ {1, 2, 3} .
n
k=1

Obviously the layer k is orthotropic and denoting with

– ϑ k the angle that the fibers of the layer k form with the axis x (Fig. 7.2.2),
– Ell [resp. Ett ] the longitudinal [resp. transverse] Young’s modulus,
– νlt , νtl the orthotropic Poisson’s modulus,
– Glt the orthotropic tangent modulus

it results

(k) Ell cos4 ϑk 2Ett νlt cos2 ϑk sin2 ϑk Ett sin4 ϑk

c11 = 1−νlt νtl + 1−νlt νtl + 1−νlt νtl + 4Glt cos2 ϑk sin2 ϑk
(k) Eu cos2 ϑk sin2 ϑk Ett νlt (cos4 ϑk +sin4 ϑk )
c12 = 1−νlt νtl + 1−νlt νtl
ϑk sin
2 2ϑ
+ Ett cos
1−νlt νtl
k
− 4Glt cos2 ϑk sin2 ϑk
 
(k) ϑk sinϑk
3 Ett νlt cos3 ϑk sinϑk −cosϑk sin3 ϑk
c13 = − Ell cos
1−νlt νtl + 1−νlt νlt
3ϑ  
+ Ett cosϑk sin
1−νlt νtl
k
+ 2Glt cos3 ϑk sinϑk − cosϑk sin3 ϑk
(k) Ell sin4 ϑk 2Ett νlt cos2 ϑk sin2 ϑk tt cos ϑk
4
c22 = 1−νlt νtl + 1−νlt νtl + E1−ν lt νtl
+ 4Glt cos2 ϑk sin2 ϑk
 
(k) k sin ϑk
3 Ett νlt cosϑk sin3 ϑk −cos3 ϑk sinϑk
c23 = − Ell cosϑ
1−νlt νtl + 1−νlt νtl
3 ϑ sinϑ  
+ Ett cos k
1−νlt νtl
k
+ 2Glt cosϑk sin3 ϑk − cos3 ϑk sinϑk
(k) Ell cos2 ϑk sin2 ϑk 2Ett νlt cos2 ϑk sin2 ϑk
c33 = 1−νlt νtl − 1−νlt νtl
ϑk sin
2 2ϑ  2
+ Ett cos
1−νlt νtl
k
+ 2Glt cos2 ϑk − sin2 ϑk . 

Remark 7.2.2 Also the problem of a multi-layered composite plate can be easily
analyzed. We suppose that

Fig. 7.2.2
7.2 The Macroscopic Anisotropy 653

Fig. 7.2.3

– the structure works in the field of the linear elasticity and of small displacements,
– the composite laminate consists of n layers, each constituted by a composite
reinforced with continuous unidirectional fibers,
– the cartesian reference system O, x, y, z has the plane z = 0 coincident with the
upper surface of the plate (Fig. 7.2.3),
– the thickness of the k-th layer is sk = hk − hk−1 (Fig. 7.2.3),
– the surface of the structure is loaded by distributed loads px , py , pz both orthogonal
and parallel to its plane,
– the phenomena of the buckling is absent.

To determine the distribution of stress in the outlined structure we assimilate the

composite plate to a homogeneous anisotropic plate equivalent to it. So we must
determine the basic equations of this equivalent structure.
Clearly the equations of the equilibrium of the homogeneous anisotropic equiva-
lent plate are (Figs. 7.2.4 and 7.2.5)

∂Nx ∂Nxy
∂x + ∂y + px = 0
∂Nxy ∂Ny
∂x + ∂y + py = 0
∂Mx ∂Mxy
∂x + ∂y − Qx = 0
∂Mxy ∂My
∂x + ∂y − Qy = 0
∂Qx ∂Qy
∂x + ∂y + pz = 0

where
 hn  hn  hn
Nx = σx dz, Ny = σy dz, Nxy = τxy dz, (7.2.3)
0 0 0
 hn  hn  hn
Mx = σx z dz, My = σy z dz, Mxy = τxy z dz. (7.2.4)
0 0 0
654 7 Anisotropy

Fig. 7.2.4

Fig. 7.2.5

Clearly in the usual hypotheses of Euler-Bernouilli the compatibility equations

of the homogeneous anisotropic equivalent plate are

εx = εx0 + z kx
εy = εy0 + z ky
γxy = γxy
0 +z k
xy

where εx , εy , γxy [resp. kx , ky , kxy ] denote the strain components [resp. the bending
curvatures and twist] and
∂u ∂v ∂u ∂v
εx0 = ∂x (0), εy0 = ∂y (0), γxy
0 =
∂y (0) + ∂x (0)

− ∂∂xw2 , − ∂∂yw2 , ∂2w

2 2
kx = ky = kxy = − ∂x∂y .

To have the mathematical model of the homogeneous anisotropic equivalent plate

we must add to the previous equilibrium and compatibility equations the stress–
strain relations
7.2 The Macroscopic Anisotropy 655
⎡ ⎤ ⎡ ⎤⎡ 0 ⎤
Nx A11 A12 A13 B11 B12 B13 εx
⎢ Ny ⎥ ⎢ A21 A22 A23 B21 B22 B23 ⎥ ⎢ εy0 ⎥
⎥ ⎢
⎢ ⎥ ⎢ ⎥
⎢ Nxy ⎥ ⎢ A31 B33 ⎥ ⎢ 0⎥
⎢ ⎥ ⎢ A32 A33 B31 B32 ⎥ ⎢ γxy ⎥ .
⎢ Mx ⎥ = ⎢ C11 C12 C13 D11 D12 D13 ⎥ ⎢ ⎥
⎢ ⎥ ⎢ ⎥ ⎢ kx ⎥
⎣ My ⎦ ⎣ C21 C22 C23 D21 D22 D23 ⎦ ⎣ ky ⎦
Mxy C31 C32 C33 D31 D32 D33 kxy

To obtain the elastic constants Aij , Bij , Cij , Dij we consider the stress–strain rela-
tions for the k-th layer of the plate. Since each layer of the composite laminate is
reinforced by continuous unidirectional fibers parallel to the l axis which makes
the angle ϑ k with the x axis (Fig. 7.2.2), the stress–strain relations for the k-
(k)
th layer of the plate are the (7.2.2), whose elastic constants cij are given in
Remark 7.2.1.
Therefore, taking into account the (7.2.3) and (7.2.4), we get ∀i, j ∈ {1, 2, 3}


n
(k)
Aij = cij sk
k=1
n
(k) hk −hk−1
2 2
Bij = cij 2
k=1
Cij = Bij

n
(k) hk −hk−1
3 3
Dij = cij 3 .
k=1

Employing the previous relationships we can formulate the mathematical model

of the homogeneous anisotropic equivalent plate only in terms of u, v, w. We easily
obtain

L1 u + L2 v + L3 w = px
L4 u + L5 v + L6 w = py
L7 u + L8 v + L9 w = pz

where the operators L1 , . . . , L9 are given by

∂ 2 ∂ 2∂ 2
L1 = −A11 ∂x 2 − (A13 + A31 ) ∂x∂y − A33 ∂y2

∂ 2 ∂ 2∂ 2
L2 = −A13 ∂x 2 − (A12 + A33 ) ∂x∂y − A32 ∂y2

∂ 3 ∂ 3 ∂ ∂ 3 3
L3 = B11 ∂x 3 + (2B13 + B31 ) ∂x2 ∂y + (B12 + 2B33 ) ∂x∂y2 + B33 ∂y3

∂ 2 ∂ 2∂ 2
L4 = −A31 ∂x 2 − (A33 + A21 ) ∂x∂y − A23 ∂y2
656 7 Anisotropy

Fig. 7.2.6

∂ 2 ∂ ∂
2 2
L5 = −A33 ∂x 2 − (A32 + A23 ) ∂x∂y − A22 ∂y2

∂ 3 ∂ 3 ∂ ∂ 3 3
L6 = B31 ∂x 3 + 2 (B33 + B21 ) ∂x2 ∂y + (B32 + 2B23 ) ∂x∂y2 + B22 ∂y3

∂ 3 ∂ 3 ∂ ∂ 3 3
L7 = −C11 ∂x 3 − (C13 + 2C31 ) ∂x2 ∂y − (2C33 + C21 ) ∂x∂y2 − C23 ∂y3

∂ 3 ∂ 3 ∂ ∂ 3 3
L8 = −C13 ∂x 3 − (C12 + 2C33 ) ∂x2 ∂y − (2C32 + C23 ) ∂x∂y2 − C22 ∂y3

∂ 4 ∂ 4 ∂ 4
L9 = D11 ∂x 4 + 2 (D13 + D31 ) ∂x3 ∂y + (D12 + D21 + 4D33 ) ∂x2 ∂y2

∂ 4 ∂ 4
+2 (D32 + D23 ) ∂x∂y 3 + D22 ∂y4 .

The boundary conditions at the edges of the plate are (Fig. 7.2.6)

∂w ∂ w̄
un = ūn , ut = ūt , w = w̄, ∂n = ∂n
∂Mnt ∂ M̄nt
Nn = N̄n , Nnt = N̄nt , Qn + ∂t = Q̄n + ∂t , Mn = M̄n . 

7.2.2 Structural Anisotropy

We have seen in Sect. 7.1.1 that a structure constituted by a non homogeneous
isotropic material can be analyzed in approximate way by assimilating it to a macro-
scopically equivalent structure constituted by a homogeneous anisotropic material.
This same technique can be employed with success also in the case of particular
structures constituted from homogeneous isotropic material. A classical example is
the oblique beam bridge. Such structure is constituted by a grid of beams with an
overlooking co-working slab (Fig. 7.2.7). It is absolutely proven that in such struc-
ture the contribution of the slab to the resistance is not negligible. So to calculate
the structure we cannot use the scheme of a grid of beams even in approximate way.
On the contrary Guyon, Massonnet and Bares proposed a satisfactory technique of
analysis. They replaced the real structure with a homogeneous anisotropic plate,
macroscopically equivalent to the oblique beam bridge.
Let us consider the plane grid constituted by m parallel beams, spaced of b0 , and
by n parallel traverses, spaced of a0 (Fig. 7.2.7). Beams and traverses are rigidly
built-in at the junctions. The grid has only two opposite sides supported (Fig. 7.2.7).
The applied loads are normal to the plane of the grid. We assume the reference
7.2 The Macroscopic Anisotropy 657

Fig. 7.2.7

frame O, x, y, z with the axis z normal to the plane of the grid. The axis x [resp. y]
is parallel to the beams [resp. traverses]. The angle between x and y is π2 + α
(Fig. 7.2.7). We denote with (EI)P [resp. CP ] the flexural [resp. torsion] rigidity
of the beams and with (EI)E [resp. CE ] the flexural [resp. torsion] rigidity of the
traverses.
We simulate the real structure with the macroscopically equivalent homogeneous
anisotropic plate having

– length 2a according x (Fig. 7.2.7),

– length 2b = 2b + b0 according y (Fig. 7.2.7),
– unitary flexural rigidity ρP = (EI) (EI)E
b0 , ρE = a0 ,
P

CP CE
– unitary torsion rigidity γP = b 0 , γE = a0 .

We notice that for the orthogonal reference frames x , y , z and x , y , z of
Fig. 7.2.8 it results

y
x = x + y tgα, y= , z = z
cosα
x
x= , y = y + x tgα, z = z .
cosα

Fig. 7.2.8
658 7 Anisotropy

From this we get

∂ ∂ ∂ ∂ 1 ∂
= , = tgα + ,
∂x ∂x ∂y ∂x cosα ∂y
∂ 1 ∂ ∂ ∂ ∂
 = + tgα ,  = .
∂x cosα ∂x ∂y ∂y ∂y

Let us notice now that because of the geometrical curvatures

∂ 2w ∂ 2w
,
∂x2 ∂y2

of the deflected surface w(x, y), the unitary bending moments arise

∂ 2w ∂ 2w
MfP = −ρP , MfE = ρE
∂x2 ∂y2
and because of the geometrical torsions

∂ 2w ∂ 2w 1 ∂ 2w
  = tgα 2 +
∂x ∂y ∂x cosα ∂x∂y
∂ 2w ∂ 2w 1 ∂ 2w
= tgα +
∂x ∂y ∂y2 cosα ∂x∂y

of the deflected surface w(x, y), the unitary torsion moments arise

∂ 2w 1 ∂ 2w
MtP = γP tgα +
∂x2 cosα ∂x∂y
∂ 2w 1 ∂ 2w
MtE = −γE tgα + .
∂y2 cosα ∂x∂y

So on an infinitesimal element of the equivalent plate the moments

MfP , MfE , MtP , MtE , the shears Qx , Qy and the unitary external load p(x, y) act
together (Fig. 7.2.9).
The equilibrium to the rotation around the axis y furnishes

∂MfP ∂MtP 1 ∂MtE

Qx = − tgα + ,
∂x ∂x cosα ∂y
the equilibrium to the rotation around the axis x furnishes

∂MfE ∂MtE 1 ∂MtP

Qy = − + tgα − ,
∂y ∂y cosα ∂x
7.2 The Macroscopic Anisotropy 659

Fig. 7.2.9

the equilibrium to the vertical translation furnishes

∂Qx ∂Qy
+ = −p cosα.
∂x ∂y

From these equilibrium equations, taking into account the previous expressions
of the bending and torsion moments, we obtain the differential equation of the
macroscopically homogeneous anisotropic plate equivalent to the real homogeneous
isotropic structure (Fig. 7.2.10)

  ∂ 4w ∂ 4w
ρP cos2 α + γP sin2 α + 2γP sinα 3
∂x 4 ∂x ∂y
∂ 4w ∂ 4w
+ (γP + γE ) + 2γ E sinα (7.2.5)
∂x2 ∂y2 ∂x∂y3
  ∂ 4w
+ ρE cos2 α + γE sin2 α = cos2 α p(x, y).
∂y4
Clearly the boundary conditions of this problem are, on the supported sides x = a
and x = −a

w=0
MfP cosα − MtP sinα = 0

Fig. 7.2.10
660 7 Anisotropy

that is

w=0
(7.2.6)
(ρP cosα + γP sinα tgα) ∂∂xw2 + γP tgα ∂x∂y
∂2w
2
=0

and on the free sides x = b and x = −b

MfE cosα − MtE sinα = 0

∂MfE
Q̄y = Qy + sinα ∂x + cosα ∂M
∂x = 0
tE

that is

(ρE cosα + γE sinα tgα) ∂∂yw2 + γE tgα ∂x∂y ∂ w

2 2
=0
 
tgα ∂ 3 w
+ γE + γP cos12 α ∂x∂ 2w∂y
3
γP cosα ∂x3
(7.2.7)
 tgα  ∂3w   ∂3w
+ γE cosα + γE sinα − ρE sinα ∂x∂y 2 + ρE + γE tg α ∂y3 = 0.
2

In the boundary condition on the free side we have taken into account the con-
dition that the torsion moment Mc must be zero (Fig. 7.2.11). To this aim we have
added, according Kirchhoff, to the real shears the additional shears

∂Mc
,
∂x
getting

Mc = MtE cosα + MfE sinα

∂Mc ∂MtE ∂MfE
Q̄y = Qy + = Qy + cosα + sinα .
∂x ∂x ∂x

The analytical solution w(x, y) of the problem of the anisotropic plate (7.2.5),
(7.2.6), and (7.2.7) can be found, since the principle of superposition, as the sum of
two aliquots, i.e.

w(x, y) = ws (x, y) + was (x, y). (7.2.8)

Fig. 7.2.11
7.2 The Macroscopic Anisotropy 661

In the problem in examination to proceed with the (7.2.8) we can decompose the
external load p(x, y) in an aliquot with polar symmetry

p(x, y) + p( − x, −y)
ps (x, y) =
2
and in an aliquot with polar anti-symmetry

p(x, y) − p( − x, −y)
pas (x, y) = ,
2
where obviously p(x, y) = ps (x, y) + pas (x, y). Then we denote with ws (x, y) [resp.
was (x, y)] the solution of the problem (7.2.5), (7.2.6), and (7.2.7) when p(x, y) =
ps (x, y) [resp. p(x, y) = pas (x, y)].
After that, as banal consequence of the (7.2.8), we get

MfP = MfP
s + M as , M = M s + M as , M = M s + M as ,
fP fE fE fE tP tP tP

MtE = MtE
s + M as , Q = Qs + Qas ,
tE x x x Qy = Qsy + Qas
y .

For the problem (7.2.5), (7.2.6), and (7.2.7), developing the known terms ps (x, y)
and pas (x, y) and the unknown functions ws (x, y)and was (x, y) in double Fourier
series it is possible to analytically obtain the solution.
Chapter 8
Nonlinear Elasticity

8.1 Nonlinear Problems

8.1.1 The Nonlinearity Causes

As shown in Theory of structures and as confirmed by all the experiences done in the
Laboratories material tests, the greatest part of the problems of the Structural engi-
neering can be analyzed with excellent approximation by the linear mathematical
models furnished in the previous chapters. However many other important struc-
tural problems exist which, in the elastic field, are correctly simulated only by a
nonlinear mathematical model. In such cases obviously the mathematical tools to
be used are much more complex.
In the Theory of elasticity a first important cause of loss of linearity is the
discharge from the field of the small deformations, as already previously found
in the stability problems. Such case of finite deformations is technically very
important. Cleary in such case the relationships (1.1.28) cannot be employed
anymore.
In the Theory of elasticity a second important cause of loss of linearity is the
presence of inequalities in the differential relationships and/or in the boundary con-
ditions of the mathematical model of the problem. This happens in all the unilateral
problems. Also such problems are technically very important. They can be analyzed
only by employing the Modern mathematical analysis.
In the Theory of elasticity a third cause of loss of linearity is the discharge of the
state of stress from the field of the linear elasticity. In such case the relationships
(1.4.1), (1.4.5), (7.1.1), and (7.1.8) clearly cannot be employed anymore. However,
as underlined in Sect. 1.4, as a rule any resistant material has an elastic load limit
very close to its limit proportionality load.
In the Theory of elasticity a fourth cause of loss of linearity is, in presence of
thermal effects, the term of mechanical coupling (5.3.29). However, as underlined
in Sect. 5.3, as a rule such term is negligible.

A. Maceri, Theory of Elasticity, DOI 10.1007/978-3-642-11392-5_8, 663


C Springer-Verlag Berlin Heidelberg 2010
664 8 Nonlinear Elasticity

8.2 Finite Deformations

8.2.1 Three-Dimensional Problem

We consider here the elastic problems in which the deformations are not infinites-
imal. The technical problems that can arise are many and important. For instance
this happens, as already previously seen, in the stability problems.
We assume a cartesian reference frame O, x, y, z and still denote with V the
open of 3 occupied before the deformation from the three-dimensional body in
examination. We retain the definition (1.1.2) of the elongation εr in the direction r
at the point P = (x, y, z) of V and the definition (1.1.3) of the shearing strain γrt
between the directions r and t at the point P = (x, y, z) of V.
Let P = (x, y, z) and Q = (x + dx, y + dy, z + dz) two any points of V, where
dx, dy, dz are any three infinitesimal real numbers. Let P and Q be the positions of
the points P and Q after the finite deformation of the body. So
P = (x + u (x, y, z) , y + v (x, y, z) , z + w (x, y, z)) , (8.2.1)

Q = (x + dx + u (x + dx, y + dy, z + dz) , y + dy

(8.2.2)
+v (x + dx, y + dy, z + dz) , z + dz + w (x + dx, y + dy, z + dz)) ,

and, in approximate way

∂u
u (x + dx, y + dy, z + dz) = u (x, y, z) + (x, y, z) dx
∂x
∂u ∂u
+ (x, y, z) dy + (x, y, z) dz
∂y ∂z
∂v
v (x + dx, y + dy, z + dz) = v (x, y, z) + (x, y, z) dx
∂x
(8.2.3)
∂v ∂v
+ (x, y, z) dy + (x, y, z) dz
∂y ∂z
∂w
w (x + dx, y + dy, z + dz) = w (x, y, z) + (x, y, z) dx
∂x
∂w ∂w
+ (x, y, z) dy + (x, y, z) dz.
∂y ∂z
From the (8.2.2) and (8.2.3) we easily get

∂u
Q = x + u (x, y, z) + 1+ (x, y, z) dx
∂x
 
∂u ∂u ∂v
+ (x, y, z) dy + (x, y, z) dz , y + v (x, y, z) + (x, y, z) dx
∂y ∂z ∂x
 (8.2.4)
∂v ∂v
+ 1+ (x, y, z) dy + (x, y, z) dz , z + w (x, y, z)
∂y ∂z
 
∂w ∂w ∂w
+ (x, y, z) dx + (x, y, z) dy + 1 + (x, y, z) dz .
∂x ∂y ∂z
8.2 Finite Deformations 665

As a consequence

dist2 (P, Q) = dx2 + dy2 + dz2 , (8.2.5)

 2
  ∂u ∂u ∂u
dist P , Q =
2
1+ (x, y, z) dx + (x, y, z) dy + (x, y, z) dz
∂x ∂y ∂z
 2
∂v ∂v ∂v
+ (x, y, z) dx + 1 + (x, y, z) dy + (x, y, z) dz
∂x ∂y ∂z
 2
∂w ∂w ∂w
+ (x, y, z) dx + (x, y, z) dy + 1 + (x, y, z) dz .
∂x ∂y ∂z
(8.2.6)

After that, considering the point A = (x + dx, y, z) where dx is an infinitesimal

positive real number, from the (8.2.5) and (8.2.6) we have

dist (P, A) = dx
!1
  ∂u 2
∂v 2
∂w 2 2
dist P , A = 1+ (x, y, z) + (x, y, z) + (x, y, z) dx
∂x ∂x ∂x
(8.2.7)

from which, since the (1.1.2)

  !1
dist P , A − dist (P, A) ∂u 2
∂v 2
∂w 2 2
εx = = 1+ + + −1
dist (P, A) ∂x ∂x ∂x

so that
 # "! 1
2 2 2
∂u 1 ∂u ∂v ∂w 2
εx = 1 + 2 + + + − 1. (8.2.8)
∂x 2 ∂x ∂x ∂x

With a perfectly analogous procedure we get

 # "! 1
2 2 2
∂v 1 ∂u ∂v ∂w 2
εy = 1 + 2 + + + −1 (8.2.9)
∂y 2 ∂y ∂y ∂y
 # "! 1
2 2 2
∂w 1 ∂u ∂v ∂w 2
εz = 1 + 2 + + + − 1. (8.2.10)
∂z 2 ∂z ∂z ∂z

As for as γxy , considering the point B = (x, y, +dy, z) where dy is an infinitesimal

positive real number, from the (8.2.6) we have
666 8 Nonlinear Elasticity
 
dist P , B
!1
∂u 2
∂v 2
∂w 2 2 (8.2.11)
= (x, y, z) + 1+ (x, y, z) + (x, y, z) dy.
∂x ∂x ∂x

−−→ −−→
Denoting with ϑ  the angle between the segments P A and P B , since the (1.1.3)
we have
π
γxy = − ϑ (8.2.12)
2

and then

sin γxy = cos ϑ  . (8.2.13)

Observing that

−−→ −−→    
P A × P B = dist P , A dist P , B cos ϑ  ,

from the (8.2.13), (8.2.7), and (8.2.11) we get

−−→ −−→
sin γxy = P A × P B
: #  # ""
2 2 2
∂u 1 ∂u ∂v ∂w
1+2 + + +
∂x 2 ∂x ∂x ∂x
#  # ""! 1 ⎞
2 2 2
∂v 1 ∂u ∂v ∂w 2
· 1+2 + + + dx dy⎠
∂y 2 ∂y ∂y ∂y

and then, because of the (8.2.4)

∂u ∂v ∂u ∂u ∂v ∂v ∂w ∂w
sin γxy = + + + +
∂y ∂x ∂x ∂y ∂x ∂y ∂x ∂y
: #  # ""
2 2 2
∂u 1 ∂u ∂v ∂w
1+2 + + +
∂x 2 ∂x ∂x ∂x (8.2.14)
#  # ""! 1 ⎞
2 2 2
∂v 1 ∂u ∂v ∂w 2
⎠·
· 1+2 + + +
∂y 2 ∂y ∂y ∂y
8.2 Finite Deformations 667

With a perfectly analogous procedure we get

∂u ∂w ∂u ∂u ∂v ∂v ∂w ∂w
sin γxz = + + + +
∂z ∂x ∂x ∂z ∂x ∂z ∂x ∂z
: #  # ""
2 2 2
∂v 1 ∂u ∂v ∂w
1+2 + + +
∂y 2 ∂x ∂x ∂x (8.2.15)
#  # ""! 1 ⎞
2 2 2
∂w 1 ∂u ∂v ∂w 2
⎠
· 1+2 + + +
∂z 2 ∂y ∂y ∂y

∂v ∂w ∂u ∂u ∂v ∂v ∂w ∂w
sin γyz = + + + +
∂z ∂y ∂y ∂z ∂y ∂z ∂y ∂z
: #  # ""
∂v 1 ∂u 2 ∂v 2
∂w 2
1+2 + + +
∂y 2 ∂x ∂x ∂x
#  # ""! 1 ⎞
2 2 2
∂w 1 ∂u ∂v ∂w 2
⎠·
· 1+2 + + +
∂z 2 ∂y ∂y ∂y

(8.2.16)
This way we can formulate the mathematical model of the elastic equilibrium
problem for the three-dimensional body in the field of finite deformations. In the
hypotheses that the medium is homogeneous, isotropic and linearly elastic, so that
the (1.4.1) are true, we get the following nonlinear boundary problem:

[8.2.1] In the hypothesis of regularity of S and of X, Y, Z, px , py , pz , find the func-

tions u, v, w real, definite and regular in V and such that the differential equations
obtained by substituting the (8.2.8), (8.2.9), (8.2.10), (8.2.14), (8.2.15), and (8.2.16)
in the
 
∂ 2Gν   ∂   ∂
2Gεx + εx + εy + εz + Gγxy + (Gγxz ) + X = 0
∂x 1 − 2ν ∂y ∂z
 
∂   ∂ 2Gν   ∂  
Gγyx + 2Gεy + εx + εy + εz + Gγyz + Y = 0
∂x ∂y 1 − 2ν ∂z
 
∂ ∂   ∂ 2Gν  
(Gγzx ) + Gγzy + 2Gεz + εx + εy + εz + Z = 0
∂x ∂y ∂z 1 − 2ν

are satisfied in V and the boundary conditions are satisfied in S = Su ∪ Sp . 

This very complex problem admits a simplified formulation in the case in which
 # "
2 2 2
∂u 1 ∂u ∂v ∂w
2 + + +  1, (8.2.17)
∂x 2 ∂x ∂x ∂x
668 8 Nonlinear Elasticity
 # "
2 2 2
∂v 1 ∂u ∂v ∂w
2 + + +  1, (8.2.18)
∂y 2 ∂y ∂y ∂y
 # "
2 2 2
∂w 1 ∂u ∂v ∂w
2 + + +  1. (8.2.19)
∂z 2 ∂z ∂z ∂z

In fact in such case, since the Taylor’s formula, the (8.2.8) becomes
# "
2 2 2
∂u 1 ∂u ∂v ∂w
εx = + + + , (8.2.20)
∂x 2 ∂x ∂x ∂x

the (8.2.9) becomes

# "
2 2 2
∂v 1 ∂u ∂v ∂w
εy = + + + , (8.2.21)
∂y 2 ∂y ∂y ∂y

the (8.2.10) becomes

# "
2 2 2
∂w 1 ∂u ∂v ∂w
εz = + + + . (8.2.22)
∂z 2 ∂z ∂z ∂z

Furthermore, in the supplementary hypotheses

∂u ∂v ∂u ∂u ∂v ∂v ∂w ∂w ∼
+ + + + = 0, (8.2.23)
∂y ∂x ∂x ∂y ∂x ∂y ∂x ∂y
∂u ∂w ∂u ∂u ∂v ∂v ∂w ∂w ∼
+ + + + = 0, (8.2.24)
∂z ∂x ∂x ∂z ∂x ∂z ∂x ∂z
∂v ∂w ∂u ∂u ∂v ∂v ∂w ∂w ∼
+ + + + = 0, (8.2.25)
∂z ∂y ∂y ∂z ∂y ∂z ∂y ∂z
the (8.2.14) becomes

∂u ∂v ∂u ∂u ∂v ∂v ∂w ∂w
γxy = + + + + , (8.2.26)
∂y ∂x ∂x ∂y ∂x ∂y ∂x ∂y
the (8.2.15) becomes

∂u ∂w ∂u ∂u ∂v ∂v ∂w ∂w
γxz = + + + + , (8.2.27)
∂z ∂x ∂x ∂z ∂x ∂z ∂x ∂z
the (8.2.16) becomes

∂v ∂w ∂u ∂u ∂v ∂v ∂w ∂w
γyz = + + + + . (8.2.28)
∂z ∂y ∂y ∂z ∂y ∂z ∂y ∂z
8.2 Finite Deformations 669

As a consequence in the previous hypotheses (8.2.17), (8.2.18), (8.2.19), (8.2.23),

(8.2.24), and (8.2.25), the problem [8.2.1] assumes the simpler formulation
[8.2.2] In the hypothesis of regularity of S and of X, Y, Z, px , py , pz , find the
functions u, v, w real, definite and regular in V and such that the differential equa-
tions obtained by substituting the (8.2.20), (8.2.21), (8.2.22), (8.2.26), (8.2.27), and
(8.2.28) in the
 
∂ 2Gν   ∂   ∂
2Gεx + εx + εy + εz + Gγxy + (Gγxz ) + X = 0
∂x 1 − 2ν ∂y ∂z
 
∂   ∂ 2Gν   ∂  
Gγyx + 2Gεy + εx + εy + εz + Gγyz + Y = 0
∂x ∂y 1 − 2ν ∂z
 
∂ ∂   ∂ 2Gν  
(Gγzx ) + Gγzy + 2Gεz + εx + εy + εz + Z = 0
∂x ∂y ∂z 1 − 2ν
are satisfied in V and the boundary conditions are satisfied in S = Su ∪ Sp . 
Obviously, since the problems [8.2.1] and [8.2.2] are mathematically nonlinear,
for them the principle of superposition is not true.

8.2.2 Large Deflections of Thin Plates

To build the model [3.2.1] we have supposed that in the plate the stresses have
resultants parallel to the middle plane null or negligible. If the deflections of the plate
are not small in comparison with its thickness, these supplementary internal forces
must be taken into consideration in deriving the differential equation of plates. In
this way we obtain nonlinear equations and the solution of the problem becomes
much more complicated.
We will say that the plate is in the field of the large deformations if the deflections
w are no longer small in comparison with the thickness s of the plate but are still
small as compared with the other dimensions, and the first order partial derivatives
of u, v, w are anywhere in modulus near to zero.
Since this problem has too some technical interest, we will propose an approx-
imated but mathematically correct model to make calculations. The mathematical
model will be built in the hypothesis that on the plate a superficial distributed load
of component qx [resp. qy ] [resp. q] according the axis x [resp. y] [resp. z] is applied.
In fact, if in addition to lateral load q there are forces acting in the middle plane of
the plate, these latter forces may have a considerable effect on the bending of the
plate and must be considered in deriving the differential equations of the problem.
Remark 8.2.1 In the case of large deflections we have to distinguish between immov-
able edges and edges free to move in the plane of the plate. Calculation show that
the magnitude of deflections and stresses in a plate with immovable edges is con-
siderably different from the one of the plate with edges free to move in the middle
plane. 
670 8 Nonlinear Elasticity

Remark 8.2.2 The theories of thin plates become unreliable in the case of plates of
considerable thickness. In such a case we must consider the problem of the plate as
a three-dimensional problem of elasticity. 
First of all, let us suppose true the (3.2.6), (3.2.7), (3.2.9), (3.2.10), (3.2.20), and
(3.2.22). After this approximation, we consider the equilibrium of the infinitesimal
element cut from the plate of Fig. 3.2.1. Since the hypothesis of large deflections,
in addition to the forces discussed in Sect. 3.2.1 we now have forces acting in the
middle plane of the plate. We denote the magnitude of these forces per unit length
by Nx , Ny , Nxy = Nyx , as shown in Fig. 8.2.1 and in Fig. 8.2.2.
Clearly the internal forces Nx , Ny , Nxv depend not only on the external forces qx ,
qy applied in the x, y plane, but also on the strain of the middle plane of the plate
due to bending. Evidently the equations of equilibrium of the infinitesimal element
to the translation according x and y are respectively

∂Nx ∂Nxy
+ + qx = 0 (8.2.29)
∂x ∂y
∂Nxy ∂Ny
+ + qy = 0. (8.2.30)
∂x ∂y
A third equation in the unknown Nx , Ny , Nxy is obtained considering the strain
in the middle surface of the plate during the bending. We denote with u [resp. v ]
the component according x [resp. y] of the displacement of a point P of the middle
plane of the plate. Furthermore we denote with εx , εy , γxy
 the strain components

in P.
Since the hypothesis of large deflections, we must employ an expression more
precise than the (1.1.28). So we consider an infinitesimal fibre of the middle plane
parallel to x and of length dx. We denote its end points with P1 = (x, y, 0) and

Fig. 8.2.1

Fig. 8.2.2
8.2 Finite Deformations 671

P2 = (x + dx, y, 0). During the deformation P1 moves it in P1 = (x + u, y + v, w)

and P2 moves it in

∂u ∂v ∂w
P2 = x + dx + u + dx, y + v + dx, w + dx .
∂x ∂x ∂x
Then
 
dist P1 , P2 − dx
εx =
dx !1
2 2 2 2
∂u ∂v ∂w
dx + dx + dx + dx − dx
∂x ∂x ∂x
=
# dx !" 1
∂u 2 ∂v 2 ∂w 2 2
= 1+ + − 1.
∂x ∂x ∂x

From this, approximating the variation of the function square root with its first
order differential and taking into account the hypothesis of large deflections, it
follows
2
∂u 1 ∂w
εx = + . (8.2.31)
∂x 2 ∂x

In a similar way we obtain

2
∂v 1 ∂w
εy = + , (8.2.32)
∂y 2 ∂y
 ∂u ∂v ∂w ∂w
γxy = + + . (8.2.33)
∂y ∂x ∂x ∂y
By taking the second order derivatives of the (3.2.35), (3.2.36), and (3.2.37) and
combining the resulting expressions, we get

∂ 2 εx ∂ 2 εy 
∂ 2 γxy ∂ 2w
2
∂w 2
∂w 2
+ − = − . (8.2.34)
∂y2 ∂x2 ∂x∂y ∂x∂y ∂x ∂y

By replacing the strain components by the obvious expressions

1  
εx = Nx − νNy (8.2.35)
sE
1  
εy = Ny − νNx (8.2.36)
sE
 1
γxy = Nxy , (8.2.37)
sG
672 8 Nonlinear Elasticity

we get

∂ 2 Nx ∂ 2 Nx ∂ 2 Ny ∂ 2 Ny ∂ 2 Nxy
−ν + + − ν − 2 (1 + ν)
∂x2 ∂y2 ∂x2 ∂y2 ∂x∂y
2
! (8.2.38)
∂ 2w ∂w 2 ∂w 2
= sE − .
∂x∂y ∂x ∂y

The fourth equation necessary to determine the unknown functions is now

obtained modifying the (8.2.29) taking into account Nx , Ny , Nxy (Fig. 8.2.2). In the
equilibrium of the infinitesimal element to the translation according z, to evaluate
the projections of the internal forces Nx , Ny , Nxy and of the external forces qx , qy on
the z axis we must take into account the bending of the plate and the resulting small
angles that these forces form with the middle plane (Fig. 8.2.1). As a result of the
bending, the projection of the forces Nx on the z axis gives

∂w ∂Nx ∂w ∂ 2 w
− Nx dy + Nx + dx + 2 dx dy,
∂x ∂x ∂x ∂x

from which, if the small quantities of order higher than the second are neglected,
we get

∂ 2w ∂Nx ∂w
Nx dx dy + dx dy. (8.2.39)
∂x2 ∂x ∂x

The same happens for qx . In the same way the projection of the forces Ny on the
z axis furnishes

∂ 2w ∂Ny ∂w
Ny dx dy + dx dy. (8.2.40)
∂y2 ∂y ∂y

The same happens for qy . In the same way the projection of the shearing forces
Nxy on the z axis is equal to

∂ 2w ∂Nxy ∂w
Nxy dx dy + dx dy (8.2.41)
∂x∂y ∂x ∂y

and the projection of Nyx = Nxy on the z axis is equal to

∂ 2w ∂Nxy ∂w
Nxy dx dy + dx dy. (8.2.42)
∂x∂y ∂y ∂x

This way, equalizing to zero the sum of the projections of all the acting forces on
the z axis and taking into account the (8.2.29) and (8.2.30), we get the final equation
8.3 Unilateral Problems 673

∂ 4w ∂ 4w ∂ 4w
+ 2 +
∂x4 ∂x2 ∂y2 ∂y4
(8.2.43)
1 ∂ w
2 ∂ 2w ∂ 2w ∂w ∂w
= q + Nx 2 + Ny 2 + 2Nxy − qx − qy .
D ∂x ∂y ∂x∂y ∂x ∂y

About the boundary conditions, in an immovable part of the edge ∂A we have

(since the displacements according x and y are null)
u = 0, v = 0. (8.2.44)

So the mathematical model of the thin plate loaded laterally and submitted to the
action of forces in the middle plane of the plate, in the field of large deflections is
the problem:
[8.2.3] In the hypothesis in which the known terms qx , qy , q and the open one A are
regular, find fourth real functions w, Nx , Ny , Nxy defined in A, equipped in A with
partial derivatives (w at least up to those of the fourth order and Nx , Ny , Nxy at least
up to those of the second order) and such that they satisfy in A the differential equa-
tions (8.2.29), (8.2.30), (8.2.38), and (8.2.43), and in ∂A the boundary conditions
(3.2.27), (3.2.28), and (3.2.29) [resp. (3.2.27), (3.2.28), (3.2.29), and (8.2.44)] for
edges free to move in the plane of the plate [resp. for immovable edges]. 
For the problem [3.2.9] we can prove an existence and uniqueness theorem.
Furthermore, since the problem [3.2.9] is mathematically nonlinear, for the prob-
lem [3.2.9] the principle of superposition doesn’t hold. As a rule, in any case the
unique solution w of problem [3.2.9] is obtained by using the Finite element method
of the Numerical analysis. In turn, from w, employing the (3.2.6), (3.2.7), (3.2.9),
(3.2.20), and (3.2.22), we determine in all the points of the plate the bending and
twisting moments and the shears. Finally, from w, Nx , Ny , Nxy , employing the (3.2.2),
(3.2.3), and (3.2.4), we determine in all the points of the plate the state of stress.
Remark 8.2.3 For very thin plates the problem [8.2.3] can be reduced to that of find-
ing the deflection of a flexible membrane. In fact in this case, where the deflections
can be many times larger than their thickness, the resistance of the plate to bending
can be neglected. 

8.3 Unilateral Problems

8.3.1 Introduction

We call unilateral problem a mathematical problem in which inequalities appear

in the differential relationships and/or in the boundary conditions. This happens in
many problems of the Theory of elasticity. The most important, both in the field of
small deformations and for finite elasticity, are:
– the contact between an elastic body and an elastic or a rigid body,
– the problems with unilateral constraints,
674 8 Nonlinear Elasticity

– the friction problems,

– the collision problems.

We immediately underline that any unilateral problem necessarily asks for the
massive employment of the Functional analysis, both for its formulation and for its
theoretical analysis and for the computation of its solution. As a rule in any unilat-
eral problem P a convenient Sobolev space with integer or fractional exponent is
assumed as environment. The formulation can be weak, if the weak derivatives are
employed, or distributional, if the Theory of distributions is employed. The solu-
tions of problem P are pursued in the Sobolev space, employing the methods of the
Functional analysis.
The analysis starts with the study of the existence and uniqueness of the solu-
tion u. Then its qualitative properties are determined. As for as the computation of
the solution u of problem P, usually a succession {Pn } of approximating problems is
built. In each problem P n a convenient subspace of finite dimension (of the Sobolev
space where problem P is defined) is assumed as environment. Then the problem
P n is formulated as a discrete model of the problem P, usually employing the Finite
element method. The finite element discretization must be such that

– every problem P n has one and only one solution un ,

– every solution un can be numerically obtained by few iterations of a convergent
algorithm,
– the sequence {un } is strongly convergent to u in V.

We will analyze in the following some simple example of unilateral problems.

Remark 8.3.1 As a rule any unilateral problem is a free boundary problem. In fact
the open where the differential relationships of the problem must be satisfied is an
unknown of the problem. 
Remark 8.3.2 Let us notice that the frictional contact is a very complex problem.
This since it is governed by multivalued tribological laws. 
Remark 8.3.3 The particular problem studied in Sect. 2.4.3 is clearly an unilateral
problem. It is the unique unilateral problem for which a simple technical theory is
allowable. 

8.3.2 Contact Problems

The problem to mathematically simulate the contact among two elastic bodies is par-
ticularly complex. In fact the accurate description of the phenomenon certainly asks
for the use of the Microscopic physics. However it is sometimes possible to apply a
simpler approach. This happens for instance in the following contact problem.

Let us formulate the problem of the contact between two beams rectilinear,
parallel, with axes at distance δ > 0. We denote with
8.3 Unilateral Problems 675

– x the axis of the upper beam,

– a1 ∈  [resp. b1 ∈ ]a1 , +∞[] the abscissa of the left [resp. right] end of the upper
beam,
– y the axis normal to x and downwards,
– a2 ∈  [resp. b2 ∈ ]a2 , +∞[] the abscissa of the left [resp. right] end of the lower
beam,
– B1 [resp. B2 ] the flexural rigidity of the upper [resp. lower] beam,
– 1 [resp.2 ] the open ]al , b1 [ [resp. ]a2 , b2 [],
– β1 [resp. β2 an element of ]0, +∞[ ,
– V1 [resp. V2 ] a closed subspace of the Sobolev space H 2 (1 ) [resp. H 2 (2 )],
– u1 [resp. u2 ] the elastic curve of the upper [resp. lower] beam, where the displace-
ments according y are positive and are measured from the axis of the upper [resp.
lower] beam.
Furthermore we suppose that
– the ends of the beams can be supported or built-in,
– B1 [resp. B2 ] is measurable and essentially bounded in 1 [resp. 2 ], so that
B1 ∈ L∞ (1 ) [resp. B2 ∈ L∞ (2 )],
– B1 ≥ β1 [resp. B2 ≥ β2 ] a.e. on 1 [resp. 2 ],
– 1 ∩ 2 = ∅
and we put

K = {(v1 , v2 ) ∈ V1 × V2 : v2 (x) + δ ≥ v1 (x) ∀x ∈ 1 × 2 }

and, for every (u1 , u2 ) ∈ V1 × V2

u1 on 1 u2 on 2
ũ1 = , ũ2 =
0 on  − 1 0 on  − 2

 
u1 = 1 − 2 ∪ {x ∈ 1 : ũ2 (x) + δ > u1 (x)}
   
u2 = 2 − 1 ∪ x ∈ 2 : u2 (x) + δ > ũ1 (x) .

Finally, we simulate the loads applied on the upper [resp. lower] beam with the
functional

   −
q1 : v ∈ H 2 (1 ) → q10 v dx + q11 b+ 1 v (b1 ) − q11 a1 v (a1 )
1
 
    − 
− q11 v dx − q12 b+

1 v (b1 ) + q 12 1a v (a1 ) + q12 v dx
1 1
 
 +  
resp. q2 : v ∈ H (2 ) →
2
q20 v dx + q12 b2 v (b2 ) − q21 a− 2 v (a2 )
2
  
    − 
− q21 v dx − q2 b+2 v (b2 ) + q22 2 a v (a2 ) + q22 v dx
2 2
676 8 Nonlinear Elasticity

where

– q10 [resp. q20 ] is an element of L2 (1 ) [resp. L2 (2 )] that simulates the dis-
tributed load applied on the upper [resp. lower] beam (parallel to y and positive if
downwards),
– q11 [resp. q21 ] is an element of L2 (1 ) [resp. L2 (2 )] that simulates the load
constituted by a finite number of concentrated forces applied on the upper [resp.
lower] beam (parallel to y and positive if downwards),
– q21 [resp. q22 ] is an element of L2 (1 ) [resp. L2 (2 )] that simulates the load
constituted by a finite number of concentrated couples applied on the upper [resp.
lower] beam (positive if anticlockwise).

With such positions we assume as mathematical model, in the case of contact

between two beams built-in at all their ends, the problem
[8.3.1] Supposing, in addition to the previous hypotheses, that V1 ∈ H02 (1 ) and
V2 ∈ H02 (2 ), find (u1 , u2 ) ∈ K such that, in the distribution sense
   
B1 u1 + B2 u2 = q1 + q2 on 1 ∩ 2 (8.3.1)
 
B1 u1 − q1 ≤ 0 on 1 (8.3.2)
 
B2 u2 − q2 ≥ 0 on 2 (8.3.3)
 
B1 u1 − q1 = 0 on u1 (8.3.4)
 
B2 u2 − q2 = 0 on u2 ·  (8.3.5)

and in the case of contact between an upper beam supported at both its ends and a
lower beam built-in at both its ends, as mathematical model the problem
[8.3.2] Supposing, in addition to the previous hypotheses, that V1 ∈ H01 (1 ) ∩
 (1 ), V2 ∈ H0 (2 ) and q12 is left and right convergent in a1 and in b1 , find
H 2 2

u1, u2 ∈ K such that, in the distribution sense

   
B1 u1 + B2 u2 = q1 + q2 on 1 ∩ 2 (8.3.6)
 
B1 u1 − q1 ≤ 0 on 1 (8.3.7)
 
B2 u2 − q2 ≥ 0 on 2 (8.3.8)
 
B1 u1 − q1 = 0 on u1 (8.3.9)
 
B2 u2 − q2 = 0 on u2 (8.3.10)
    +  −
B1 u1 a+
1 = q12 a1 − q12 a1 (8.3.11)
8.3 Unilateral Problems 677
    −  +
B1 u1 b−
1 = q12 b1 − q12 b1 ·  (8.3.12)

and in the case of contact between two beams supported at all their ends, as
mathematical model the problem
[8.3.3] Supposing, in addition to the previous hypotheses, that V1 ∈ H10 (1 ) ∩
H 2 (1 ), V2 ∈ H01 (2 ) ∩ H 2 (2 ), q12 [resp.
 q22 ] is left and right convergent in a1
[resp. a2 ] and in b1 [resp. b2 ], find u1, u2 ∈ K such that, in the distribution sense
   
B1 u1 + B2 u2 = q1 + q2 on 1 ∩ 2 (8.3.13)
 
B1 u1 − q1 ≤ 0 on 1 (8.3.14)
 
B2 u2 − q2 ≤ 0 on 2 (8.3.15)
 
B1 u1 − q1 = 0 on u1 (8.3.16)
 
B2 u2 − q2 ≤ 0 on u2 (8.3.17)
    +  −
B1 u1 a+
1 = q12 a1 − q12 a1 (8.3.18)
    −  +
B1 u1 b−
1 = q12 b1 − q12 b1 (8.3.19)
    +  −
B2 u2 a+
2 = q22 a2 − q22 a2 (8.3.20)
    −  +
B2 u2 b−
2 = q22 b2 − q22 b1 ·  (8.3.21)

Let us first of all obtain some general properties of the possible solutions of
problems [8.3.1], [8.3.2], and [8.3.3].
Let (u1 , u2 ) ∈ V1 × V2 be solution of problem  [8.3.1].
 Since  the (8.3.2) [resp.

(8.3.3)] the distribution on 1 [resp. 2 ] q1 − B1 u1 [resp. B2 u2 − q2 ] is a
positive measure μ1 [resp. μ2 ]. So, denoting with c1 [resp. c2 ] a point of 1 [resp.
2 ], the function
⎧
⎨ μ1 ([c1 , x[) if x > c1
g1 : x ∈ 1 → 0 if x = c1 , (8.3.22)
⎩
−μ1 ([x, c1 [) if x < c1

and respectively
⎧
⎨ μ2 ([c2 , x[) if x > c2
g2 : x ∈ 2 → 0 if x = c2 , (8.3.23)
⎩
−μ2 ([x, c2 [) if x < c2

is real, increasing and left continuous on 1 [resp. 2 ]. Furthermore it results

(x, y) ∈ 21 and x<y ⇒ μ1 ([x, y[) = g1 (y) − g1 (x) , (8.3.24)

678 8 Nonlinear Elasticity

and respectively

(x, y) ∈ 22 and x<y ⇒ μ2 ([x, y[) = g2 (y) − g2 (x) , (8.3.25)

and, in the distribution sense

 
q1 − B1 u1 = g1 on 1 (8.3.26)
 
B2 u2 − q2 = g2 on 2 . (8.3.27)

The (8.3.26) [resp. (8.3.27)] and the (8.3.4) [resp. (8.3.5)] implicate that the
restriction of g1 [resp. g2 ] to an open interval contained in u1 [resp. u2 ] is
constant. Moreover, because the (8.3.1), we have

g1 − g2 = const. a.e. on 1 ∩ 2 . (8.3.28)

This obviously involves that g1 and g2 are bounded. As a consequence from the
(8.3.26) and (8.3.27), four real constants d1 , d2 , d3 , d4 exist such that
 
B1 u1 (x) = d2 + q12 (x)
 x  t 
+ d1 + q11 (t) − g1 (t) + q10 (z) dz dt a.e. on 1
a1 a1
(8.3.29)
 
B2 u2 (x) = d4 + q22 (x)
 x  t 
+ d3 + q21 (t) − g2 (t) + q20 (z) dz dt a.e. on 2 .
a2 a2
(8.3.30)
From the (8.3.29) [resp. (8.3.30)] it obviously follows

– if q12 [resp. q22 ] is right convergent in a1 [resp. a2 ] and left convergent in b1 [resp.
b2 ], the same it happens for B1 u1 [resp. B2 u2 ],
– if B1 ∈ H 1,p (1 ) [resp. B2 ∈ H 1,p (2 )], q11 ∈ Lp (1 ) [resp. q21 ∈ Lp (2 )] and
q12 ∈ H 1,p (1 ) [resp. q22 ∈ H 1,p (2 )], where ≥ 2, then u1 ∈ V1 ∩ H 3,p (1 )
[resp. u2 ∈ V2 ∩ H 3,p (2 )].

To analyze the problems [8.3.1], [8.3.2], and [8.3.3] it is convenient to introduce

some new problems, equivalent to them. We put
  
∀v ∈ H02 (1 ) < L1 , v >= q10 v dx− q11 v dx+ q12 v dx
1 1 1
  
∀v ∈ H02 (2 ) < L2 , v >= q20 v dx− q21 v dx + q22 v dx
2 2 2
8.3 Unilateral Problems 679

and, if q12 [resp. q22 ] is left convergent in a1 [resp. a2 ] and right convergent in b1
[resp. b2 ]
 
∀v ∈ H01 (1 ) ∩ H 2 (1 ) < L1 , v >= q10 v dx− q11 v dx
1 1

  +   −   
− q12 b1 v (b1 ) − q12 a1 v (a1 ) + q12 v dx
 1 
∀v ∈ H0 (2 ) ∩ H (2 )
1 2
< L2 , v >= q20 v dx− q21 v dx
2 2

  +   −  
− q22 b2 v (b2 ) − q22 a2 v (a2 ) + q22 v dx.
2

Furthermore we introduce the potential energy functional


1 b1   2
J: (v1 , v2 ) ∈ V1 × V2 → B1 v1 dx
2 a1

1 b2   2
+ B2 v2 dx− < L1 , v1 > − < L2 , v2 >∈ .
2 a2

We consider the new problems constituted by the variational inequality

[8.3.4] Find
 b1  
(u1 , u2 ) ∈ K: B1 u1 v1 − u1 dx
a1
 b2  
+ B2 u2 v2 − u2 dx ≥< L1 , v1 − u1 >
a2
+ < L2 , v2 − u2 > ∀ (v1 , v2 ) ∈ K. 

and by the potential energy minimum problem

[8.3.5] Find

(u1 , u2 ) ∈ K: J (u1 , u2 ) ≤ J (v1 , v2 ) ∀ (v1 , v2 ) ∈ K. 

It holds the equivalence theorem

[8.3.6] For every (u1 , u2 ) ∈ V1 × V2 the following statements are equivalent

(a) (u1 , u2 ) is solution of the problem [8.3.1] or of the problem [8.3.2] or of the
problem [8.3.3]
(b) (u1 , u2 ) is solution of the problem [8.3.4]
(c) (u1 , u2 ) is solution of the problem [8.3.5].

Proof We demonstrate the thesis in the case in which (u1 , u2 ) is solution of the
problem [8.3.2]. It is easy to verify that if (u1 , u2 ) is solution of the problem [8.3.1]
680 8 Nonlinear Elasticity

or of the problem [8.3.3], the thesis can be demonstrated with a perfectly analogous
procedure.
(a) ⇒ (b). Let (u1 , u2 ) ∈ K be solution of problem [8.3.2], (v1 , v2 ) an element of
K. Let us first of all observe that

μ1 = μ2 on 1 ∩ 2 (8.3.31)

v2 − u2 + v1 − u1 ≥ 0 μ1 − a.e. on 1 ∩ 2 . (8.3.32)

The (8.3.31) follows from the (8.3.28) because of the (8.3.24) and (8.3.25). The
(8.3.32) follows from the hypothesis (v1 , v2 ) ∈ K and from the relationship

μ1 ({x ∈ 1 ∩ 2 : u2 (x) + δ > u1 (x)}) = 0,

due to (8.3.24) and (8.3.25). Let us observe now that, by using (8.3.11), (8.3.12),
(8.3.29), and (8.3.30), we get
 
b1   b2  
B1 u1 v1 − u1 dx + B2 u2 v2 − u2 dx− < L1 , v1 − u1 >
a1 a2
 b1  b2
− < L2 , v2 − u2 >= g1 (v1 − u1 ) dx − g2 (v2 − u2 ) dx.
a1 a2

Then to gain the thesis it is enough to prove that

 b1  b2
g1 (v1 − u1 ) dx − g2 (v2 − u2 ) dx ≥ 0. (8.3.33)
a1 a2

We confine the reasoning to the case a1 < a2 and b1 < b2 , since in the other cases
it is possible to apply the same reasoning. Assuming in the (8.3.22) and (8.3.23)
c1 = a2 and c2 = b1 , we have
 b1  b2  b1
g1 (v1 − u1 ) dx − g2 (v2 − u2 ) dx = g1 (v1 − u1 ) dx
a1 a2 a2
 b1  
−g2 (v2 − u2 ) dx + g2 b+ 1 [v2 (b1 ) − u2 (b1 )]
a2

 
= − g1 a+ 2 [v (a
1 2 ) − u (a
1 2 )] − (v1 − u1 ) dμ1
1 ∩2

 
+ (v2 − u2 ) dμ1 +g2 b+ 1 [v2 (b1 ) − u2 (b1 )]
1 ∩2

 
≥ g1 a+ 2 [u (a
1 2 ) − δ] + (v2 − u2 + u1 − v1 ) dμ1
1 ∩2
 
− g2 b+ 1 [u2 (b1 ) + δ] .
8.3 Unilateral Problems 681

From this, from the (8.3.32) and from the obvious implications
 
u1 (a2 ) − δ < 0 ⇒ g1 a+
2 =0
 
u2 (b1 ) + δ > 0 ⇒ g2 b+
1 = 0,

the (8.3.33) follows.

(b) ⇔ (c). Obvious.
(b) ⇒ (a). Let (u1 , u2 ) ∈ H01 (1 ) ∩ H 2 (1 ) × H02 (2 )be solution of problem
[8.3.4]. Evidently (u1 , u2 ) satisfies the (8.3.6), (8.3.7), and (8.3.8). As for as the
(8.3.9), let ϕ be an element of C0∞ u1 − {0}. If suppϕ ⊂ 1 −  ¯ 2 , since

(u1 − ϕ, u2 ) ∈ K , (u1 + ϕ, u2 ) ∈ K ,

we have
 b1
B1 u1 ϕ  dx =< q1 , ϕ > .
a1

¯ 2 = Ø, since for
If suppϕ ⊂ 
min
¯2
x∈suppϕ∩
[u2 (x) + δ − u1 (x)]
0<ε<
max |ϕ|
it results
(u1 − εϕ, u2 ) ∈ K , (u1 + εϕ, u2 ) ∈ K ,

we still have
 b1
B1 u1 ϕ  dx =< q1 , ϕ > .
a1

Proceeding in perfect analogy, we verify that also the (8.3.10) is satisfied.

As for as the (8.3.11) let us consider before the case a1 = a2 . Let v ∈ H01 (1 ) ∩
H 2 (1 ) ∩ H02 (2 ) such that v (a1 ) = 0 and v (b1 ) = 0. Since

(u1 + v, u2 + v) ∈ K , (u1 − v, u2 − v) ∈ K ,

we have
 
B1 u1 v dx + B2 u2 v dx =< L1 , v > + < L2 , v > . (8.3.34)
1 2

Furthermore, having already acquired the (8.3.6), (8.3.7), (8.3.8), (8.3.9), and
(8.3.10), they hold the (8.3.29) and (8.3.30). As a consequence, taking into account
the (8.3.34), we get
682 8 Nonlinear Elasticity

 b1  b1
− d2 v (a1 ) + q12 v dx − (q11 − g1 ) v dx
a1 a1
 b2  b2  b2  b2
+ q22 v dx − q21 v dx − g2 v dx + q20 v dx
a2 a2 a2 a2
   b1
b1 b1   
= q10 v dx − q11 v dx + q12 a−
1 v (a1 ) + q12 v dx
a1 a1 a1
 b2  b2  b2
+ q20 v dx − q21 v dx+ q22 v dx,
a2 a2 a2

that is
 
b1 b2   
−d2 v (a1 ) + g1 v dx− g2 v dx = q12 a−
1 v (a1 ) ,
a1 a2

from which, since

 b1  b2
g1 v dx = g2 v dx,
a1 a2

we obtain
 
− d2 = q12 a−
1 .

As a consequence, since the (8.3.29), we have

    +  −
B1 u1 a+
1 = q12 a1 − q12 a1 .

If a1 = a2 preliminarily we observe that a point ξ ∈ ]a1 , min {b1 , b2 }[ exists such

that

u2 (x) + δ > u1 (x) ∀x ∈ [a1 , ξ ] . (8.3.35)

We select a v ∈ H01 (]a1 , ξ [) ∩ H 2 (]a1 , ξ [) such that v (a1 ) = 0 and v (ξ ) = 0 and

consider an
! min
x∈[a1 ,ξ ] [u2 (x) + δ − u1 (x)]
ε ∈ 0, .
x∈[a1 ,ξ ] |v (x)|
max

Then we put for every x ∈ [a1 , b1 ]

u1 (x) + εv1 (x) if x ∈ [a1 , ξ ]

w̄ (x) =
u1 (x) if x ∈ ]ξ , b1 ]
u1 (x) − εv1 (x) if x ∈ [a1 , ξ ]
w̄¯ (x) =
u1 (x) if x ∈ ]ξ , b1 ]
8.3 Unilateral Problems 683

and observe that, since

 
(w̄, u2 ) ∈ K , ¯ u2 ∈ K ,
w̄,

it results
 ξ  ξ  ξ
B1 u1 v dx = q10 v dx− q11 v dx
a1 a1 a1
 ξ
(8.3.36)
  
+ q12 a−
1 v (a1 ) + q12 v dx.
a1

On the other hand, taking into account of the (8.3.29) and (8.3.35), we have a.e.
on [a1 , ξ ]
 x  t 
B1 u1 (x) = d2 + q12 (x) + d 1 + q11 (t) + q10 (z) dz (8.3.37)
a1 a1

where d1 ∈ . Consequently

 ξ  ξ  ξ 
B1 u1 v dx = q12 v dx − d2 v (a1 ) − v d 1 + q11 (x)
a1 a1 a1
 x   ξ
+ q10 (z) dz dx = q12 v dx (8.3.38)
a1 a1
 ξ  ξ
− d2 v (a1 ) − q11 v dx + q10 v dx.
a1 a1

From the (8.3.36) and (8.3.38) it follows

 
− d2 = q12 a−
1 .

from which, because of the (8.3.37), it follows

    +  −
B1 u1 a+
1 = q12 a1 − q12 a1 .

Proceeding in perfect analogy, we verify that also the (8.3.12) is satisfied. 

Finally, we observe that the following existence and uniqueness theorem subsists

[8.3.7] The problem [8.3.1] [resp. [8.3.2]] [resp. [8.3.3]] admits an unique solution.

Proof Since
; ;
c1 v V1 ≤ ;v ;L2 ( ) ≤ c1 v V1 ∀V ∈ V1
1
684 8 Nonlinear Elasticity
; ;
c2 v V2 ≤ ;v ;L2 ( ) ≤ c2 v V2 ∀v ∈ V2
2

where c1 , c1 , c2 , c2 are real numbers independent from v, the thesis follows from a
known theorem of Lions-Stampacchia. 

8.3.3 Unilateral Constraints

When an elastic body is constrained by rigid or elastic unilateral constraints, some
inequalities appear in the boundary conditions of the mathematical model and again
we are in presence of a problem of nonlinear elasticity. In fact, considering a linearly
elastic beam with an unilateral rigid support at z = a, so that the condition v (a) ≥ 0
holds, the linearity implication

c1 , c2 ∈ , v1 (a) ≥ 0, v2 (a) ≥ 0 ⇒ c1 v1 (a) + c2 v2 (a) ≥ 0

is obviously false. As done in the preceding section where we have analyzed the
problem of the contact among one-dimensional bodies, we will now perform the
theoretical analysis of a bi-dimensional problem of this type.

We consider a linearly elastic plate under transverse loads, resting on elastic uni-
lateral boundary support. Given the bounded and connected domain  occupied by
the plate in its middle plane x1 , x2 , let us assume external forces q and displace-
ments v to be positive in x3 direction (the orthogonal reference frame O, x1 , x2 , x3
is anticlockwise). The reaction r of the unilateral elastic support has the Winkler
expression

r = −Ev+

where E is a nonnegative function.

It is convenient to formulate the elastic equilibrium problem like an energetic
one, assuming as environment the Sobolev space W 2 () and considering a general
fourth order operator and taking into account distributed and/or concentrated forces.
Hence we denote with

–  a bounded and connected open of 2 of class R(0),1 ,

– $ the boundary of ,
– s the curvilinear measure on $,  
– E ∈ L∞ (Γ ) − {0}, E≥0 s − a.e. on $, q ∈ W 2 () ,
A = ∞
|r|=2 D (ars D ), with ars ∈ L () and ars = asr , a fourth order
– r s
|s|=2 - -
 
differential operator such that |r|=2  ars Ds vDr v dx ≥ a0 |r|=2  |Dr v|2 dx
|s|=2

∀v ∈ W 2 (), where a0 ∈ ]0, +∞ .
8.3 Unilateral Problems 685

Furthermore we consider the potential energy functional


1 
J: v ∈ W () → J (v) =
2
ars Ds vDr v dx
2  |r|=2
|s|=2

1  2
+ E v+ ds− < q, v >∈ 
2 $

and assume as mathematical model of the plate under transverse loads constrained
by an elastic unilateral boundary support the following potential energy minimum
problem
[8.3.8] Find u ∈ W 2 () such that J (u) ≤ J (v) ∀v ∈ W 2 () . 
The theoretical analysis of the problem [8.3.8] will be executed giving some
equivalent formulations and then studying the solution’s existence, uniqueness and
regularity.
We preliminarily establish the lemmas
[8.3.9] The functional J is convex, Gateaux-differentiable in W 2 () and such that
 2
∀ (u, v) ∈ W 2 ()

 
J  (u, v) = ars D uD v dx +
s r
Eu+ vds− < q, v > .
|r|=2  $
|s|=2

Consequently J is weakly lower semicontinuous on W 2 () .

Proof Convexity is obvious. To prove that J is differentiable it is sufficient to prove
that the functional

1  2
v ∈ W () →
2
E v+ ds
2 $
is differentiable. The Lebesgue theorem on dominated convergence applies. 
[8.3.10] For any u ∈ W 2 () the functional J  (u, ·) is linear and continuous on
W 2 (). Moreover the operator

B : u ∈ W 2 () → J  (u, ·)

is monotone and hemicontinuous.

Proof Linearity of J  (u, ·) is obvious. As for continuity, it is obviously sufficient to
prove it for the functional

F : u ∈ W 2 () → Eu+ vds.
$
686 8 Nonlinear Elasticity

Since  ∈ R(0),1 , we have

v L2 (Γ ) ≤ const. v W 1 () ∀v ∈ W 2 () . (8.3.39)

Continuity of F is then acquired by observing that


 2 2 ; +;
∀ (u, v) ∈ W () E u+ v ds ≤ const. E ;u ; v
L∞ (Γ ) L2 (Γ ) L2 (Γ ) .
$
(8.3.40)
Monotonicity and hemicontinuity are obvious. 
Now, by using lemma [8.3.9], we easily prove that
[8.3.11] For any u ∈ W 2 () the following statements are equivalent

(a) u is solution of problem [8.3.8],

(b) u is solution of the variational equation of the virtual works

u ∈ W 2 () : ars Ds u Dr v dx
|r|=2 
|s|=2 (8.3.41)

+ Eu+ v ds =< q, v > ∀v ∈ W 2 () ,
$

(c) u is solution of the mixed type variational inequality


u ∈ W 2 () : ars Ds u Dr (v − u) dx− < q, v − u >
|r|=2 
|s|=2
 
1  2 1  2
+ E v+ ds− E u+ ds ≥ 0 ∀v ∈ W 2 () . 
2 $ 2 $
(8.3.42)

After that we study the existence and uniqueness of the problem [8.3.8] solution.
Let us denote with P1 the subspace of W 2 () of the polynomials having degree not
greater than the first one. Let us recall that, because  ∈ R(0),1 , it results
⎛ ⎞1

2

c1 ⎝ Dr v dx⎠ ≤ ṽ
2
W 2 ()
|r|=2 
P1

⎛ ⎞1 (8.3.43)

2
W 2 ()
≤ c1 ⎝ Dr v dx⎠
2
∀ṽ = [v] ∈
P1
|r|=2 

where the positive constants c1 and c1 are independent of v.

8.3 Unilateral Problems 687

We let ΓE = {x ∈ Γ : E (x) > 0} and ∀x = (x1 , x2 ) ∈ 2

1 (x) = 1, p1 (x) = 1, p2 (x) = 1.

Moreover if < q, 1 >> 0 we let

< q, p1 > < q, p2 >

ξ= ,
< q, 1 > < q, 1 >
and we remark that < q, 1 > is the component in the x3 direction of the external
forces resultant, applied, if nonzero, at ξ .
We preliminary establish some equilibrium necessary conditions with the follow-
ing theorem
[8.3.12] If problem [8.3.8] has solution, then < q, 1 >≥ 0. If < q, 1 >= 0 and
problem [8.3.8] has solution, then < q, p1 >=< q, p2 >= 0. If < q, 1 >> 0 and
problem [8.3.8] has solution, then

p ∈ P1 − {0} and p (ξ ) = 0 ⇒ s ({x ∈ ΓE : p (x) ≥ 0}) > 0. (8.3.44)

If < q, 1 >> 0 and

∃p0 ∈ P1 − {0} such that p0 (ξ ) = 0 and s ({x ∈ ΓE : p (x) ≥ 0}) > 0 (8.3.45)

and if problem [8.3.8] has solution, then (8.3.44) is true and

p ∈ P1 and p (ξ ) = 0 and p = λp0 ∀λ ∈  ⇒

(8.3.46)
⇒ s ({x ∈ ΓE : p0 (x) = 0, p (x) > 0}) > 0.

Proof Let problem (8.3.8) admits a solution u. Because u satisfies (8.3.41), we must
have

E u+ ds =< q, 1 > (8.3.47)
$

so that < q, 1 >≥ 0.

If < q, 1 >= 0, from (8.3.47) and from the equality

Eu+ pi ds =< q, pi >
$

it follows < q, pi >= 0.

If < q, 1 >> 0, because (8.3.47) is true, we have

s ({x ∈ ΓE : u (x) > 0}) > 0.

688 8 Nonlinear Elasticity

Hence, for any p ∈ P1 − {0} with p (ξ ) = 0, because


Eu+ p ds =< q, p >= p (ξ ) < q, 1 >= 0,
$

it is obvious that

s ({x ∈ ΓE : p (x) ≥ 0}) > 0.

Let us assume now < q, 1 >> 0 and let (8.3.45) be true. At first, because

Eu+ p0 ds =< q, p0 >= 0
$

and, by (8.3.45),

Eu+ p0 ≤ 0 s − a.e. on $,

we have

Eu+ p0 = 0 s − a.e. on $. (8.3.48)

After that, by absurd, let us suppose that p̃ ∈ P1 , with p̃ (ξ ) = 0 and p̃ =

λp0 ∀λ ∈ , exist such that

s ({x ∈ ΓE :p0 (x) = 0, p̃ (x) > 0}) = 0. (8.3.49)

From (8.3.48) and (8.3.49) we have

Eu+ p̃ ≤ 0 s − a.e. on $. (8.3.50)

Then, because u is solution of the (8.3.41), we must have


Eu+ p̃ ds =< q, p̃ >= 0. (8.3.51)
$

From (8.3.50) and (8.3.51) it follows

u+ p̃ = 0 s − a.e. on ΓE .

Hence, taking into account of (8.3.48) and observing that

 
x ∈ 2 : p̃ (x) = p0 (x) = 0 = {ξ } ,
8.3 Unilateral Problems 689

we obtain

u+ = 0 s − a.e. on ΓE

and, consequently

Eu+ ds = 0.
Γ

But this is absurd, because


Eu+ ds =< q, 1 >> 0.
$

Hence the (8.3.46) is true. 

From theorem [8.3.12] it follows that problem [8.3.8] can allow a solution only
in the following cases

(α) < q, 1 >= 0, < q, p1 >=< q, p2 >= 0,

(β) < q, 1 >> 0 and it results

p ∈ P1 and p (ξ ) = 0 ⇒ s ({x ∈ ΓE : p (x) > 0}) > 0, (8.3.52)

(γ) < q, 1 >> 0 and (8.3.44), (8.3.45), and (8.3.46) are true,

i.e., with different terminology, in following cases

(α) the external forces system is self-equilibrated,

(β) the external forces resultant has the direction of the x3 axis and is applied at a
point ξ such that any straight line passing through leaves on the left and on the
right a set of constrained points whose measure is positive,
(γ) the external forces resultant has the direction of the x3 axis and is applied at
a point ξ such that any straight line passing through leaves on the right or on
the same straight line, and on the left or on the same straight line, a set of con-
strained points whose measure is positive. Moreover a straight line r of equation
p0 (x) = 0 exists such that on its right, or its left, the set of the constrained points
has measure zero and such that all straight line passing through ξ but different
from r leaves on the right and on the left a set of points of r whose measure is
positive.

Remark 8.3.4 Let us notice that from the proof of (8.3.46) it follows that any possible
solution u of problem [8.3.8] in the (γ) case is such that

Eu+ p0 = 0 s − a.e. on $. (8.3.53)


Let us prove now the existence theorem
690 8 Nonlinear Elasticity

[8.3.13] In the (α) case problem [8.3.8] allows infinite solutions, whose set coincides
with the set of solutions of the variational equation


u ∈ W () :
2
ars Ds uDr v dx
|r|=2  (8.3.54)
|s|=2

=< q, v > ∀v ∈ W 2 ()

(relative to a free plate problem) non-positive on ΓE . In the (β) case problem [8.3.8]
allows at least a solution.

Proof About the (α) case, by using (8.3.43), we immediately verify that (8.3.54)
allows infinite solutions, whose set is an element of W P()
2

1
. Thus the thesis is easily
proven.

About the (β) case, using again the problem [8.3.8] equivalence with (8.3.41) and
taking account of lemma [8.3.10], by a known Browder’s theorem it is sufficient to
prove that, as v W 2 () → +∞, it results

 
 2
ars D vD v dx − < q, v > +
s r
E v+ ds → + ∞. (8.3.55)
|r|=2  $
|s|=2

By absurd, let us suppose that k > 0 and a sequence {vn } of elements of W 2 ()
exist such that

vn W 2 () >n
∀n ∈ N (8.3.56)
 
 2
ars Ds vn Dr vn dx + E v+ n ds
|r|=2  $
(8.3.57)
|s|=2
≤< q, vn > + k ∀n ∈ N.
vn
By putting wn = vn we have, from (8.3.57)
W2 ()

 2 1 k
a0 Dr wn dx ≤ q (W 2 ())
+ 2
∀n ∈ N
v v
|r|=2  W 2 () W 2 ()

from which

for |r| = 2 limn→+∞ ||Dr wn ||L2 () . (8.3.58)

Because wn W 2 () = 1 ∀n ∈ N, a subsequence of {wn } exists, which we denote

with the same symbol, weakly-convergent in W 2 () (and hence strongly in W 1 ())
towards an element w. From this, from (8.3.57) and because the functional
8.3 Unilateral Problems 691

 
 2
v ∈ W () →
2
ars D vD v dx +
s r
E v+ ds
|r|=2  $
|r|=2

is weakly lower-semicontinuous, we have

 
 2
a0 D w dx + E w+ ds = 0
r 2

|r|=2  $

and hence

Dr w = 0 for |r| = 2, Ew+ = 0 s − a.e. on Γ .

Hence
w ∈ P1 (8.3.59)

s ({x ∈ ΓE : w (x) > 0}) = 0. (8.3.60)

From (8.3.59) and (8.3.60), and because

lim wn − w W 1 () = 0,
n→+∞

we have
lim wn − w W 2 () = 0,
n→+∞

and this, because wn W 2 () = 1 ∀n ∈ N, implies

w = 0. (8.3.61)

Let us observe now that, from the (8.3.57)

< q, w > ≥ 0,

from which, because < q, w >= w(ξ ) < q, 1 >

w(ξ ) ≥ 0.

Moreover, if w(ξ ) = 0, from (8.3.59), (8.3.61), and (8.3.52) we obtain

s({x ∈ ΓE : w(x) > 0}) > 0

and this contrasts with (8.3.60). Hence

w(ξ ) > 0. (8.3.62)

Let us prove that (8.3.62) is false. To see this we let, ∀x ∈ 2 , Q(x) = w(x)−w(ξ ).
692 8 Nonlinear Elasticity

If Q = 0, because ∀x ∈ 2 w(x) = w(ξ ) > 0, we have

s({x ∈ ΓE : w(x) > 0}) = s(ΓE ) > 0,

which contrasts with (8.3.60).

If Q = 0, because Q(ξ ) = 0, from (8.3.52) we have

s({x ∈ ΓE : Q(x) > 0}) > 0

which implies

s({x ∈ ΓE : w(x) > 0}) > 0

and this is impossible by (8.3.60). Hence (8.3.62) is false. This absurd proves
(8.3.55). 
About the solution’s existence in the γ) case, it is convenient to study an auxiliary
problem. We fix in  a point x1 such that p0 (x1 ) = 0, and we put

V1 = {v ∈ W 2 () : v(x1 ) = 0}.

¯ with continuous imbedding, V1 , equipped by the norm

Because W 2 () ⊆ C0 ()
of W (), is a closed subspace of W 2 (). Now we let s-a.e. on Γ
2

E(x) if p0 (x) ≥ 0
E1 (x) =
0 if p0 (x) < 0
and we consider the variational equation

 
u1 ∈ V1 : ars D u1 D v dx +
s r
E1 u+
1 v ds
|r|=2  $
(8.3.63)
|s|=2
=< q, v > ∀v ∈ V1

describing the elastic equilibrium of a plate supported only along r in the same way
as the given plate, and moreover with imposed displacement equal to zero at x1 .
Well in the γ) case the following existence theorems subsist

[8.3.14] In the hypotheses of the γ) case the variational equation (8.3.63) allows an
unique solution.

Proof About the existence of a solution of the (8.3.63), it is sufficient, as already

done for theorem [8.3.13], to prove that, as v W 2 () → +∞ on V1
8.3 Unilateral Problems 693

 
ars D v D v dx +
s r
E(v+ )2 ds
|r|=2  $
(8.3.64)
|s|=2
− < q, v >→ +∞.

Proceeding in a similar way as for theorem [8.3.13], denying the (8.3.64) we

prove the existence of a w ∈ V1 such that

w ∈ P1 − {0} (8.3.65)

s({x ∈ Γ : E1 (x) > 0, w(x) > 0}) = 0 (8.3.66)

< q, w >≥ 0. (8.3.67)

Let us prove that

w(ξ ) = 0. (8.3.68)

Since by (8.3.67)

w(ξ ) − < q, 1 >≥ 0,

we must have w(ξ ) ≥ 0. By absurd, let us suppose w(ξ ) > 0. We put, ∀x ∈ 2 ,
Q(x) = w(x) − w(ξ ) and at first we remark that

Q ∈ P1 − {0}, Q{ξ } = 0.

Consequently from (8.3.44), (8.3.45), and (8.3.46)

s ({x ∈ ΓE : p0 (x) = 0, Q(x) ≥ 0}) > 0

and consequently

s ({x ∈ Γ : E1 (x) > 0, Q(x) ≥ 0}) > 0. (8.6.69)

The (8.3.69) implies

s ({x ∈ Γ : E1 (x) > 0, w(x) ≥ 0}) > 0

which contrasts with (8.3.66). Hence (8.3.68) is true.

Now we observe that, because w = 0, w(x1 ) = 0, p0 (x1 ) = 0 it results

w = λp0 ∀λ ∈ 
694 8 Nonlinear Elasticity

and hence, taking into account (8.3.65), (8.3.68), and (8.3.46),

s ({x ∈ ΓE : p0 (x) = 0, w(x) > 0}) > 0,

i.e.
s ({x ∈ Γ : E1 (x) > 0, p0 (x) = 0, w(x) > 0}) > 0,

which is absurd by (8.3.66). Thus (8.3.64) is proven; consequently at least one

solution u1 of (8.3.63) exists.
Now we prove that u1 is the unique solution of (8.3.63). By absurd, let u2 a
solution of (8.3.63) different from u1 . Putting p̃ = u2 − u1 , by the obvious relations
 
ars Ds (u2 − u1 ) Dr (u2 − u1 ) dx+ E1 (u2 − u1 )(u+ +
2 − u1 ) ds = 0
|r|=2  $
|s|=2

E1 (u2 − u1 )(u+ +
2 − u1 ) ≥ 0,

and because p̃(x1 ) = 0, p0 (x1 ) = 0, we have

p̃ ∈ P1 , p̃ = λp0 ∀λ ∈  (8.3.70)

u+ +
1 = u2 s − a.e. on {x ∈ Γ : E1 (x) > 0}. (8.3.71)

Let us notice that from (8.3.44) and (8.3.45) we get

s ({x ∈ Γ : E1 (x) > 0, p0 (x) > 0}) = 0

(8.3.72)
s ({x ∈ Γ : E1 (x) > 0, p0 (x) > 0}) > 0,

and that from (8.3.71) we get

u+ +
1 = u2 s − a.e. on {x ∈ Γ : E1 (x) > 0, p0 (x) = 0} .

Hence, taking into account (8.3.70), we must have

s ({x ∈ Γ : E1 (x) > 0, p0 (x) = 0, u1 (x) > 0}) = 0,

and this, together with (8.3.72), implies

 
|r|=2 ars Ds u1 Dr v dx =< q, v > ∀v ∈ V1 .
|s|=2 

Putting then λ̄ = − p0 (x
1
1)
, because 1 + λ̄p0 ∈ V1 and < q, p0 >= 0, from the
previous relation it follows
< q, 1 >= 0

against the hypothesis. 

8.3 Unilateral Problems 695

[8.3.15] In the (γ) case, problem [8.3.8] allows solution if, and only if, called u1 the
solution of (8.3.63), a real number λ1 exists such that

(u1 − λ1 p0 )+ = 0 s − a.e. on {x ∈ ΓE : p0 (x) < 0} . (8.3.73)

When this condition occurs, u1 − λ1 p0 is solution of problem [8.3.8].

Proof About the necessity, given a solution u of problem [8.3.8], we let

u (x1 )
λ1 = − , u1 = u − λ1 p0 ,
p0 (x1 )
so that u1 ∈ V1 . Observing that E1 = E,

u1 = u on {x ∈ Γ : p0 (x) = 0} ,

and that, because (8.3.53), it results

Eu+ = 0 = E1 u1 on {x ∈ Γ : p0 (x) < 0} ,

we have, taking into account (8.3.45)

E1 u+
1 = Eu
+
s − a.e. on Γ

and consequently
 
E1 u+
1 v ds = Eu+ v ds ∀v ∈ W 2 ().
$ $

Hence, because u is solution of (8.3.41), u1 is the solution of (8.3.63). Moreover

from (8.3.53) it results

(u1 − λ1 p0 )+ = 0 s − a.e. on {x ∈ ΓE : p0 (x) < 0} .

Let us prove now that the condition is sufficient. Let v be an element of W 2 ().
1)
Putting η = − pv(x
0 (x1 )
, because u1 is solution of (8.3.63) and v + ηp0 ∈ V1 , it results


ars Ds (u1 − λ1 p0 ) Dr v dx
|r|=2 
|s|=2 (8.3.74)

+ E1 u+
1 (v + ηp0 ) ds =< q, v > .
$

On the other hand, from (8.3.73) it results

E1 u+ +
1 (v + ηp0 ) = E(u1 − λ1 p0 ) v s − a.e on {x ∈ Γ : p0 (x) ≤ 0}
696 8 Nonlinear Elasticity

and therefore, taking into account (8.3.45)

 
E1 u+
1 (v + ηp0 ) ds = E(u1 − λ1 p0 )+ v ds. (8.3.75)
$ $

From (8.3.74) and (8.3.75) it follows that u1 − λ1 p0 is solution of (8.3.41). 

To study the uniqueness of problem [8.3.8] in β) and γ) cases we put, for any
v ∈ W 2 ()

Γv+ = {x ∈ ΓE : v(x) > 0} , Γv− = {x ∈ ΓE : v(x) < 0} ,

and, for any p ∈ P1 − {0}

1 if p(x) = 0
∀x ∈ Γ χp (x) =
0 if p(x) = 0.

Let us notice that, if u is solution of problem [8.3.8], because


Eu+ ds =< q, 1 >> 0,
$
 
we have s Γu+ > 0.
It holds the lemma

[8.3.16] In (β) and (γ) cases, if u and ũ are solutions of problem [8.3.8], then

u+ = ũ+ s − a.e. on ΓE
u − ũ = p̃ with p̃ ∈ P1 , p̃(ξ ) = 0.

Proof From the relations

 
ars Ds (ũ − u) Dr (ũ − u) dx + E(ũ − u)(ũ+ − u+ ) ds = 0
|r|=2  $
|s|=2
E(ũ − u)(ũ+ − u+ ) ≥ 0,

we have

u − ũ = p̃ with p̃ ∈ P1
E(ũ − u)(ũ+ − u+ ) = 0 s − a.e. on Γ .
Hence

ũ+ = u+ s − a.e. on ΓE
8.3 Unilateral Problems 697

and, consequently

Eu+ p̃ ds = 0.
$

Therefore p̃(ξ ) = 0, since


Eu+ p̃ ds = p̃(ξ ) < q, 1 > . 
$

Let us now consider the problem

[8.3.17] Find u ∈ W 2 () such that

χp Eu+ ds =< q, 1 > ∀p ∈ P1 − {0} with p(ξ ) = 0. 
$

We will say that problems [8.3.8] and [8.3.17] are compatible if they allow a
common solution.
It holds the theorem
[8.3.18] If problems [8.3.8] and [8.3.17] are compatible then problem [8.3.8] allows
an unique solution. If problems [8.3.8] and [8.3.17] are not compatible and if u is
a solution of problem [8.3.8], then, denoted as p̄ an element of P1 − {0} such that

p̄(ξ ) = 0, χp̄ Eu+ ds =< q, 1 >, (8.3.76)
$

all and only solutions of problem [8.3.8] are of the type u + λ p̄, where
⎡ ⎤
< u u
λ∈ ⎣ sup − , inf − ⎦
  p̄   p̄
Γ0 ⊆Γ + ∪Γ − −Γ + Γ − − Γu+ ∪Γ0
p̄ p̄ u p̄ Γ + − Γu+ ∪Γ0 p̄
s(Γ0 )=0
   
if s $p̄+ > 0 and s $p̄− > 0,
⎡ ⎡
< ⎢ u ⎢    
λ∈ ⎢ sup − , +∞ ⎢if s $ + = 0 and s $ − > 0,
⎣   p̄ ⎣ p̄ p̄
Γ0 ⊆Γp̄− −Γu + Γp̄− − Γu+ ∪Γ0
s(Γ0 )=0 ⎤ ⎤
< ⎥ u⎥    
λ∈ ⎥ −∞, inf − ⎥ if s $p̄+ > 0 and s $p̄− = 0,
⎦   p̄ ⎦
Γ0 ⊆Γp̄+ −Γu + Γp̄+ − Γu+ ∪Γ0
s(Γ0 )=0
   
λ is any real number if s $p̄+ = s $p̄− = 0.
698 8 Nonlinear Elasticity

Proof Let us suppose that problems [8.3.8] and [8.3.17] are compatible and let u be
a common solution. By absurd, let us suppose that problem [8.3.8] has a solution
ũ = u. Putting p̃ = ũ − u from lemma [8.3.16] we have

p̃ ∈ P1 − {0}, p̃(ξ ) − 0, p̃ = 0 s − a.e. on Γu+ . (8.3.77)

From first and second of (8.3.77) we have


χp̃ Eu+ ds =< q, 1 > ; (8.3.78)
$

moreover because u is solution of (8.3.41), we have


Eu+ ds =< q, 1 > . (8.3.79)
$

The (8.3.78) and (8.3.79) imply


 
1 − χp̃ Eu+ ds > 0. (8.3.80)
$

On the other hand, taking account of third of (8.3.77), we must have

 
1 − χp̃ Eu+ = 0 s − a.e. on Γu+ ,

and this contrasts with (8.3.80).

Let us suppose now that problems [8.3.8] and [8.3.17] are not compatible. If u is
solution of problem [8.3.8], p̄ an element of P1 − {0} satisfying (8.3.76), let us at
first notice that

p̄ = 0 s − a.e. on Γu+ . (8.3.81)

In fact from (8.3.76) and because u is solution of (8.3.41) it results


 
1 − χp̄ Eu+ ds = 0,
$

which implies
 
1 − χp̄ Eu+ = 0 s − a.e. on Γu+

from which the (8.3.81) follows.

After that, let us analyze the case
   
s Γp̄+ > 0 and s Γp̄− > 0.
8.3 Unilateral Problems 699

Let ũ be a solution
 of problem [8.3.8]. From (8.3.81), taking into account lemma
[8.3.16] and that s Γu+ > 0, we have the existence of a real number λ such that
ũ = u + λp̄. Hence, by lemma [8.3.16]

u+ = (u + λp̄)+ s − a.e. on ΓE

which implies

u + λp̄ ≤ 0 s − a.e. on Γp̄+ ∪ Γp̄− − Γu+ .

Hence a subset Γ 0 of Γp̄+ ∪ Γp̄− − Γu+ exists such that s (Γ0 ) = 0 and

u u
sup
 
− ≤λ≤ inf − . (8.3.82)
p̄ +  +  p̄
Γp̄− − Γu+ ∪Γ0 Γ − Γu ∪Γ0
p̄

Inversely if Γ 0 is a subset of Γp̄+ ∪ Γp̄− − Γu+ with zero s-measure, and if λ is a

real number satisfying (8.3.82), taking into account (8.3.81), we have

Eu+ = E (u + λp̄)+ s − a.e. on Γ .

Then u + λp̄ is obviously a solution of (8.3.41), i.e. of problem [8.3.8].

Similarly we proceed in the other cases. 

Remark 8.3.5 Obviously in β) case, if ∀p ∈ P1 − {0}, with p{ξ } = 0, s({x ∈

ΓE : px = 0}) = 0, problems [8.3.8] and [8.3.17] are compatible. On the contrary,
by (8.3.53), they are incompatible if γ ) case is true. 

To study the regularity of the problem [8.3.8] solutions, we put for any δ ∈
]0, +∞[

Sδ = {y ∈ n : |y| < δ} , δ = {y ∈  : |y| < δ, yn > 0} (n ≥ 2)
n
  '   (
δ = v∈L δ : ∃δv ∈ ]0, δ[: v(y) = 0 as |y| > δv .
W 2

We also put, ∀i ∈ {1, . . . n − 1} and ∀h ∈ 

i,h '   (
hi = (0, . . . , h, . . . , 0), δ = y ∈ δ : y + hi ∈ δ

and, if v ∈ L2 (δ )
 
v y + hi − v(y) i,h
!i,h v(y) = a.e. on δ (h = 0).
h
700 8 Nonlinear Elasticity

We recall by known results that ∀v ∈ W 2 (δ ) ∩ W(δ )

⎛ ⎞1

2

v W 2 (δ ) ≤ const. ⎝ r 2
D v dy ⎠ (8.3.83)
|r|=2 δ
' (
and moreover, if δ̄ ∈ ]δv , δ[ and h0 = min δ − δ̄, δ̄ − δv
; ;
; ; ; ∂v ;
0 < |h| < h0 ⇒ ;!i,h v;L2 ( ) ≤ ; ;
; ∂y ; 2 . (8.3.84)
δ̄ i L (δ )

Let

b(u, v) = brs Ds u Dr v dy ∀(u, v) ∈ W 2 (δ )2
|r|=2 δ
|s|=2

be a bilinear integro-differential form with brs ∈ L∞ (δ ) and such that

 2
b(v, v) = b0 Dr v dy ∀v ∈ W 2 (δ ) ∩ W (δ ) (8.3.85)
|r|=2 δ

where b0 ∈ ]0, +∞[ .

Moreover let L be a real-valued linear functional, continuous in W 2 (δ )∩W(δ ),
equipped with the W 2 (δ ) norm and such that

L(χ ) = 0 ∀χ ∈ C0∞ (δ ) (8.3.86)

∂v
|L(v)| + L ≤ b v w1 (δ )
∂yi (8.3.87)
∀v ∈ C0∞ (Sδ ) and ∀j ∈ {1, . . . , n − 1}

where b ∈ ]0, +∞[ .

It holds the theorem
[8.3.19] If

 ∈ R(2),1 , ¯
ars ∈ C0,1 (), q ∈ (W 1 ()) ,

then any solution u of problem [8.3.8] belongs to W 3 () and it results

 ; ; 
u W 3 () ≤γ q (w1 ()) + E L∞ ($) × ;u+ ;L2 ($) + u W 2 ()

where γ = γ (ars , ).

Proof Taking into account (8.3.39) and (8.3.40), the thesis can be easily gained. 
8.3 Unilateral Problems 701

Remark 8.3.6 In the hypotheses of theorem [8.3.19], because W 3 () ⊆ C1,λ () ¯
∀λ ∈ ]0, 1[, any problem [8.3.8] solution has first order partial derivatives satisfying
¯ a Holder condition with any exponent λ ∈ ]0, 1[.
in  

The following lemma allows us to state a further regularity result

 
[8.3.20] Let brs ∈ C1,1 δ . If u ∈ W 2 (δ ) is such that

|b(u, v) + L(v)| ≤ b v L2 (δ ) ∀v ∈ W 2 (δ ) ∩ W(δ ) (8.3.88)

where b ∈ ]0, +∞[, then for any δ  ∈ ]0, δ[

 
u ∈ W 4 (δ  ), u W 4 (  ) ≤ γ b + b + u W 2 (δ )
δ

where γ = γ (δ  , brs ).
Proof At first from (8.3.87) and (8.3.88) we have

|b(u, v)| ≤ (b + b ) v W 1 (δ ) ∀v ∈ W 2 (δ ) ∩ W(δ ),

and from this we easily can prove that for any δ  ∈ ]0, δ[

 
u ∈ W 3 (δ  ), u W 3 (  ) ≤ γ b + b + u W 2 (δ ) (8.3.89)
δ

where γ = γ (δ  , brs ) .  
After that, let δ  < δ2 < 'δ3 < δ4 < δ, η ∈ C0∞ Sδ3 with (η = 1 on Sδ2 , ũ = u η,
δη < δ  < δ3 , h0 = min δ − δ4 , δ4 − δ3 , δ3 − δ  , δ  − δη . Let us prove that if
|r| = 2 and if i, j ∈ {1, . . . n − 1}
; ;
∂ ∂ r ; ∂ ∂ r ;
D ũ ∈ L2 (δ ), ; ;
∂yi ∂yj ; ∂y ∂y D ũ ; 2
i j L (δ ) (8.3.90)
  

≤ const. b + b + u W 2 (δ ) .

 
Given a sequence {vn } of elements of C0∞ Sδ3 such that

; ;
; ∂ ũ ;
lim ; vn − !i,h ; =0 (8.3.91)
n→+∞ ; ∂yj ;W 2 (δ )

we have, ∀n ∈ N and as 0 < h < h0

702 8 Nonlinear Elasticity

∂ ũ ∂vn
b !i,h , vn = b u, η !i,−h
∂yj ∂yj
  
r ∂  
+ (brs (y)Ds u(y)Dα η(y))Dr−α !i,h vn (y) dy
α δ3 ∂yj
|r|≤2 α≤r
|s|≤2 α =(0,...,0)
 ∂ brs (y + hi ) − brs (y)
− Dr vn (y) η(y + hi ) DS u(y + hi ) dy
∂yy h
|r|≤2 δ3
|s|≤2
  
s ∂   
+ Dr vn brs !i,h Dα ηDs−α u dy
α δ3 ∂yj
|r|≤2 α≤s
|s|≤2 α =(0,...,0)
 ∂brs s
− D !i,h (η u) Dr vn dy.
∂yj
|r|≤2 δ3
|s|≤2

Hence, because

∂ ũ ∂ ũ ∂vn
b !i,h , vn = b !i,h , vn + b u, η !i,−h
∂yj ∂yj ∂yj
∂vn ∂vn ∂vn
+L η !i,−h − b u, η !i,−h + L η !i,−h
∂yj ∂yj ∂yj

we have, taking into account (8.3.87), (8.3.88), (8.3.89), and (8.3.84)

∂ ũ  
b !i,h , vn ≤ const. b + b + u W 2 (δ ) vn W 2 (δ )
∂yj

from which, taking into account (8.3.91), (8.3.83), and (8.3.85), we have the
(8.3.90).
Because the (8.3.86) is true, we can easily complete the proof.
Finally, we prove the regularity theorem 

[8.3.21] If

 ∈ R(3),1 , ¯
ars ∈ C1,1 (), q ∈ L2 (), E ∈ W 1 ($),

then any solution u of problem [8.3.8] belongs to W 4 () and it results

 ; ; 
u W 3 () ≤γ q (W 1 ()) + E L∞ ($) · ;u+ ;L2 ($) + u W 2 ()

where γ = γ (ars , ).

8.3 Unilateral Problems 703

Proof At first we recall the known property that for any open  with closure
contained in  it results
 
u ∈ W 4 ( ), u W 4 ( )
≤γ q L2 () + u W 2 () (8.3.92)
 
where γ = γ ars ,  .
Now we fix on Γ a point x̄. Because  ∈ R(3),1 , an open neighborhood U of x̄,
a δ ∈ ]0, +∞[ and an one-to-one transformation  = (1 , 2 ) mapping Sδ onto U
exists such that

 ∈ C3,1 (Sδ ) , −1 ∈ C3,1 (U),

∂(1 , 2 )
(y) = 1 ∀y ∈ Sδ , (δ ) = U +
∂(y1 , y2 )

where U + =  ∩ U.
 2  
We put, for any (ṽ, w̃) ∈ W 2 (δ ) resp.ṽ ∈ W 2 (δ ) ∩ W(δ ) , v = ṽ ◦ −1
 
ṽ ◦ −1 on U +
and w = w̃ ◦ −1 resp. v = , and we consider the sum [resp.
0 on  − U +
the function]
    
+
ars (x)D v(x) D w(x) dx resp. L(ṽ) =
s r
Eu v ds .
$ (8.3.93)
|r|=2 U +
|s|=2

Replacing x by (y) and writing Ds v (((y)) and Dr w ((y)) in terms of

derivatives of ṽ and w̃ respectively, the (8.3.93) becomes

bαβ (y)Dβ ṽ(y) Dα w̃(y) dy,
|α|≤2 δ
|β|≤2
 
where bαβ ∈ C1,1 δ and depends on ars and  only.
Obviously the bilinear integro-differential form
  2
b(ṽ, w̃) = bαβ (y)Dβ ṽ(y) Dα w̃(y) dy ∀(ṽ, w̃) ∈ W 2 (δ )
|α|≤2 δ
|β|≤2

satisfies (8.3.85). It is also obvious that the functional L verifies (8.3.86) and
(8.3.87). Then, we put ũ = u ◦  and note as ṽ any element of W 2 (δ ) ∩ W(δ ).
Because
  
ars Ds u Dr v dx + E u+ v ds = q v dx,
|r|=2  $ 
|s|=2
704 8 Nonlinear Elasticity

we have

|b(ũ, ṽ) + L(ṽ)| ≤ q L2 () ṽ L2 (δ ) .

Consequently, by lemma [8.3.20]

 
ũ ∈ W 4 (δ  ), ũ 4   ≤ γ , ars , δ 
W 
 δ
; ; 
· q L2 () + E W 1 ($) · ;u+ ;C0,1 ($) + u W 2 ()

 
for any δ  ∈ ]0, δ[. Hence, putting U  =  Sδ  and U  + = U  ∩ , we have

u ∈ W 4 (U  + ), u  + ≤ γ (, ars , δ  )
W 4 U
 ; +; 
· q L2 () + E ; ; 0,1 + u
W 1 ($) · u C ($) W 2 () .

The regularization at the boundary is acquired and therefore, taking account of

the (8.3.92), the thesis is proven. 
Bibliography

1. Boley B.A. and Weiner J.H., Theory of thermal stresses, Krieger, Florida, 1960
2. Duhamel J.M.C., Seconde mémoire sur les phénomènes thermo-mécaniques, Journal de
l’Ecole Polytechnique (vol. 15, cahier 25, pp. 1–57), 1837
3. Glowinski R., Lions J.L. and Tremolieres R., Analyse numérique des inéquations variation-
nelles, Dunod, Paris, 1976
4. Landau L. and Lifshitz E., Theory of elasticity, Pergamon Press, Oxford, 1970
5. Lekhnitskii S.G., Theory of elasticity of an anisotropic elastic body, MIR Publishers, Moscow,
1981
6. Lions J.L. and Stampacchia G., Variational inequalities, Communications on Pure and
Applied Mathematics, (vol. 20, pp. 493–519), 1967
7. Napolitano L.G., Equazioni macroscopiche della magnetofluidodinamica, “Fisica del plasma”
CNR serie “Congressi, convegni e simposi scientifici”, 1965
8. Necas J., Les méthodes directes en théorie des équations elliptiques, Masson, Paris, 1976
9. Necas J. and Hlavacek I., Mathematical theory of elastic and elastoplastic bodies an
introduction, Elsevier, Berlin, 1981
10. Sokolnikoff I.S., Mathematical theory of elasticity, Krieger, Florida, 1956
11. Timoshenko S. and Gere J., Theory of elastic stability, Mc Graw-Hill, New York, 1961
12. Timoshenko S. and Goodier J.N., Theory of elasticity, Mc Graw-Hill, New York, 1951
13. Timoshenko S. and Woinowsky-Krieger S., Theory of plates and shells, Mc Graw-Hill, New
York, 1959
14. Todhunter I. and Pearson K., A history of the theory of elasticity and of the strength of
materials, Cambridge University Press, London, 1893
15. Zemansky M., Heat and thermodynamics, Mc Graw-Hill, New York, 1968

A. Maceri, Theory of Elasticity, DOI 10.1007/978-3-642-11392-5, 705


C Springer-Verlag Berlin Heidelberg 2010
Author Index

A L
Airy, G. B., 364 Lamè, G., 77
Laplace, P. S., 220
B
Beltrami, E., 88 M
Betti, E., 111 Mises, R. von, 118
Boussinesq, J. V., 382 Mohr, C. O., 43
Bredt, R., 243
N
Navier, L. M. H., 76
C
Neumann, K. G., 220
Castigliano, A. C., 112
Cauchy, A. L., 31
P
Clapeyron, B. P. E., 104
Poisson, S. D., 77
Clausius, R. J. E., 99
Prandtl, L., 410
Pythagoras, 147
D
Dirichlet, P. G. L., 220 S
Duhamel, J. M. C., 560 Saint Venant, A. J. C. B., 165
Sobolev, V. J., 90
E Stokes, G. G., 248
Euclid, 154
Euler, L., 610 T
Tresca, H. E., 116
F
Fubini, G., 244 V
Von Mises, 118
G
W
Gauss, K. F., 69
Wheatstone, C., 26
Winkler, E., 402
H
Hooke, R., 72 Y
Young, T., 77
K
Kelvin, Lord, 495 Z
Kirchhoff, G. R., 89 Zhuravskii, D. I., 294

707
Subject Index

A Beam
Adiabatic, 96, 101, 496, 498–500, 525, 528, axially loaded, 607
538, 543 deflected, 433–449, 458, 460–470,
Analogy 474–478, 480–482, 573–574, 580, 583,
hydrodynamic, 232–234, 237, 241, 591, 593, 607–608
245–246, 251, 255, 259 differential equation of deflected, 441
of the membrane, 409 on elastic foundation, 481–482
of Mohr, 449–461 mathematical model of deflected, 241, 434,
Angle 445, 482, 573, 607, 611, 677
unitary of twist, 256, 263–264, 289–290 rotation of a cross section of, 450
unitary of twist, average, 289–292 thermoelastic deflected, 573
unitary of twist, local, 285 Bending
Anisotropic of anisotropic beam, 643–644
material, 656 deviated, 190–195, 205, 211
plate, 653, 656, 659, 660 differential, 277
problem of Saint Venant, 642–648 optimal form of the cross section, 403
Anisotropy, 635–661 right of isotropic beam, 173–195
structural, 656–661 Bifurcation of the equilibrium, 594
Antipolar, 141–143, 153, 155, 207–209, 212, Bogie, 434, 474–475, 591, 598–599, 602, 607
218 Buckling by torsion and bending, 593
Antipolarity, 141–159
Antipole, 142–144, 146, 151, 153–154, 156,
158, 160, 162–163, 207, 218, 301, 309 C
Arches Castigliano’s theorem, 470
with great curvature, 488–492 Catenary, 432–433
in masonry, 404 Cauchy
with small curvature, 482–488 boundary conditions of, 31–38, 67, 555
Area convention of, 33, 36, 38, 47, 51, 56
concentrated, 127, 129, 131, 134, 138, 159, relations of, 635
162 Center of pressure, 201, 207–208, 215–216,
distributed, 127, 130, 138, 155 218
sectoral, 268 Centroid, 127–131, 133–135, 137–146,
Auto-stress, 572–573 148–151, 153–165, 175–177, 184, 198,
217, 265, 284, 289–290, 300, 315, 327,
B 333, 437, 506
Balance Circular
equations, 545 disk loaded by two diametral forces, 381,
of the momentum, 506–508, 511, 514, 520, 384
534, 538, 542 hole in a panel, 387

709
710 Subject Index

Coefficient Destruction
crossed, 505 global, 495
of cubic elongation, 23 local, 495
of thermal linear expansion, 557, 641 punctual, 495
of viscosity, 516, 518 Diameter conjugate, 144, 146, 153, 342, 345,
Comparator, 26 351, 439
Compatibility equations, 25, 68, 256, 363–364,
426, 553, 635, 651, 654
components E
of displacement, 1–2, 635 Elastic
of strain, 9 constants, 76, 110, 636–639, 641–642,
of stress, 351 650–651, 655
Constant curve, 72, 614, 618, 675
of Lamé, 77, 557 foundation, 402–403, 481–482
of thermal conductibility, 555 modulus, 77–80
Constituent link, 76, 636–639 potential, 107–110, 119–121, 480, 641
Constraint’s reaction, 81, 640 surface, 393, 396
Contact problems, 674–684 Electric-resistance strain gauge, 26
Convective time, 511 Elliptical hole in panel, 392
Crisis Elongation
parameter of, 116, 118, 120–122, 125 breaking, 72, 75
threshold of, 115–116, 118, 120–121, 125 principal, 15–16, 18–19, 23–24
Criterion Energy
energetic, 118–125 disorderly, 523, 527, 532–533, 536–537
of the intrinsic curve, 124–125 internal, 96–103, 107–108, 244, 253,
of the maximum elongation, 125 463
of the maximum normal stress, 125 kinetic, 96–97, 99, 102, 508, 514, 520,
of the maximum shearing stress, 116–117, 522–523, 526–527, 532, 536, 542, 547,
122, 239, 262, 264 549, 551–552
of the octahedral shearing stress, 117–118 method, 596–607, 630
of Tresca, 116–117 minimum of the complementary, 91–92,
of von Mises, 118, 120–122 110
Critical load, 594–596, 598, 601–603, 606, minimum of the potential, 466, 567
610, 612, 617–618, 622–624, 626, 629, orderly, 523, 527, 532–534, 536
634 Enthalpy, 525–526, 528, 536, 539, 542
Curvature of a regular curve, 396 of stagnation, 526
Curve Entropy, 100, 493, 495–497, 502–503, 505,
of Fanno, 539–541, 543 508–509, 515, 520, 524–525, 531–534,
of Reyleigh, 542–543 536–537, 539–543, 547, 556, 563
Equation
D of balance, 17, 508
Deflectometer, 26 of balance of the mass, 506
Deformation of balance of the momentum, 506, 511,
finite, 664–673 520, 538, 545
infinitesimal, 2–3 of compatibility, 24–25, 233
permanent, 70–71 of conservation, 509–510, 525
plastic, 71–72, 75, 115 of conservation of the total energy, 509–510
pure, 6–12, 15, 17, 23 of Duhamel, 563
small, 5–9, 12, 14, 31, 68, 75, 81, 104, of equilibrium, 405–406, 409–410,
106–107, 111, 182, 225, 266, 400, 402, 414–415
437–438, 440, 552, 594, 618, 663, 673 of Euler, 617–618, 654
state of plane, 422 of Laplace, 220
Subject Index 711

Equilibrium of Zhuravskii, 298, 300, 309, 337, 342,

indifferent, 597 345, 349, 351, 353
stable, 597, 605, 622, 624, 633–634 of Zhuravskii for tapered beam, 547
unstable, 597–598, 605, 622, 624, 633 Function
Extensometer, 26 of Airy, 365, 376, 382–384, 391
of Airy in polar coordinates, 376
harmonic, 220, 227, 230
F Functional
Fiber of the complementary energy, 92, 110
of aramid, 649 of the potential energy, 91, 93–94, 401,
of carbon, 649 466, 480, 597, 599, 602, 630, 679, 685
of glass, 649 Funicular
of kevlar, 649 curve, 428–429, 487
Finite polygon, 427–428, 430
deformations, 664–673
element method, 401, 403, 673–674 G
Gas
Fixed end, 434, 451, 454, 458, 460, 477–478,
ideal, 498
589
more perfect, 520, 522, 527, 529, 534
Flexural
perfect, 498, 500, 513, 520, 522, 527, 534
axis, 184, 189, 195, 212, 294–295
Geometry
plane, 184, 189, 195, 212, 294 of areas, 127–164, 325
rigidity, 182, 189, 395, 421, 657, 675 of mass, 526
Flow Greatness
constant, 234, 241, 246, 571 conservative, 501
convective of the momentum, 507–508 extensive, 494–495
diffusive of the momentum, 507, 514 intensive, 494
dissipative, 513, 515, 519, 524, 534,
537–552 H
global, 495, 547 Harmonic
of heat, 99, 493, 500–501, 542–543, function, 220, 227, 230, 365
555–556, 565–566, 571 operator of, 220
local, 494–495, 503, 515, 555 Heat
non dissipative, 524–537 specific at constant pressure, 498–499, 524,
punctual, 241, 494–495 543
subsonic, 513–514 specific at constant volume, 498–499
supersonic, 513–514 Hinge, 200, 434, 441, 460, 474–475, 581, 591,
of work, 501, 504, 514, 556 627, 630
Fluid Hydrodynamic analogy, 232–234, 237, 241,
ideal, 649–650 245–246, 251, 255, 259–260
Hydrostatic pressure, 29, 78–79
more perfect, 520
newtonian, 518
I
perfect, 545 Inertia
Forces centroidal ellipse, 131–141, 143–146,
curve of the, 428–429 148–151, 153–158, 162, 165, 176, 184,
generalized, 504–505, 515, 517 189, 194–195, 207–208, 212–213, 218,
polygon of, 427–428, 430, 487 282, 293, 300–301, 309, 317, 324, 329,
system of parallel, 31 331–333, 338, 342, 345, 351, 355, 439,
Formula 593
of Bredt, 243–245, 247 centroidal kernel, 159–164, 208, 215–216,
of Euler, 610, 617, 624–625, 654 218, 488
of Navier, 177, 188, 194, 206, 284, 368, radius, 140, 141, 144, 152, 625
370 Influence lines, 474
712 Subject Index

Instability M
local, 628 Macroscopic anisotropy, 649–661
of second type, 629–634 Material
of type bending-torsional, 628 anisotropic, 656
of type torsional, 627–628 brittle, 75
composite, 649–656
with cylindrical anisotropy, 638
K
ductil, 75
Kinematic viscosity, 518
homogeneous, 76
Kronecker delta, 21
isotropic, 76
linearly elastic, 76
L non resistant to traction, 215–219
Laminate, 73, 573, 649, 653, 655 orthotropic, 637, 641
Law orthotropic with cylindrical anisotropy, 638
of Fourier, 504, 523, 555 with a plane of elastic symmetry, 636–638,
of Hooke, 72, 75, 636, 650–651 650
of the mixtures, 650 transversely isotropic, 638, 650
of Newton, 517 Matrix, 21, 246, 250, 256, 259, 272, 354, 578,
Limit 583, 586, 604, 616–617, 649–650
layer, 537–538, 541, 545–552 Mechanics
slenderness, 624–627 of continuous media, 493–505
Load of fluids, 494, 509, 518, 556
axis, 184, 189–190, 193–195, 202, of solids, 518
208–209, 211, 213, 292–293, 302, 309, Membrane
317, 324, 329, 331–333, 342, 345 analogy, 409
breaking, 71 axialsymmetric, 407
cell of, 66 circular cylindrical, 408
centroidal axial, 165, 195–201, 203–209, curvature of a, 404, 411
211–213, 215, 218, 488, 490–491, 643 meridian of a, 404–406, 412, 416
components of the, 2, 28, 31, 37, 70, 113, parallel circle of a, 404–408, 412, 415–416
301, 405, 413, 553, 563 spherical, 408
concentrated, 28, 166 Method
critical, 594–596, 598, 601–603, 606, 610, of the admissible stress, 115–116
612, 617–618, 622–624, 626, 629, 634 energetic, 598, 607
distributed, 28, 31, 37–38, 165, 381, 393, static, 607–629
434, 436, 441, 444, 449–450, 453, 465, of the ultimate limit states, 115
467–468, 479, 483, 635, 653, 669, 676 Modulus
linear thermal, 567, 575 bounds for the elastic, 78–80
of mass, 28, 375–376 bulk, 79
non centroidal axial, 201–215, 218 of the cross section, 189
path of, 89 of the foundation, 402
plane of, 656, 184, 189–190, 195 longitudinal elastic, 77
of proportionality, 70, 663 of transversal contraction, 77
statically applied, 463 of poisson, 77, 278, 557
superficial, 2, 28, 31, 66, 81, 102–103, of the soil, 482
165, 167–168, 174–176, 188, 190, 193, tangent, 77, 557, 637, 652
197–198, 201, 212–213, 360, 640 of Young, 77, 200, 278, 439, 557
thermal, 68, 493–494, 552, 556, 561–563, Mohr
566–567, 571–574, 590 analogy of, 449–461, 588, 590–591
volumetric, 28, 31–32, 38, 66, 81, 165, circle of, 43–63, 79, 116–117, 122–125,
174, 197, 222, 233, 280, 296, 298, 359, 238, 308
552, 554, 635, 640 principal circle of, 57–63, 116–117,
yelding, 71 124–125
Subject Index 713

Moment Plate
centrifugal, 132–133, 135, 142, 157, 176, anisotropic, 653, 656–657, 659–660
194 curvature of the elastic surface of a, 393,
of inertia, 131–134, 173, 184, 188, 204, 396
212, 217, 278, 286, 301, 310, 368, 439, on elastic foundation, 402–403, 481–482
574, 608, 618, 626 flexural rigidity of a, 395, 657
polar of inertia, 227 multilayered composite, 652
static, 128–129, 142, 175, 205, 217, 297, small deflections of thin, 392–402, 411
305, 312, 323, 327, 333 with large deflections, 669–673
twisting, 224, 242, 247, 252, 255–256, 259, with unilateral constraint, 684–704
261–262, 264, 267, 269, 274–275, 283, Pneumostructure, 403
288–291, 294, 301, 354, 394–401, 403, Polygon of the forces, 427–428, 430, 487
411, 419 Potential
Motion elastic, 107–110, 119–121, 641
steady, 531, 533 elastic of disfiguration, 121–122
unsteady, 533 electrochemical, 497
minimum of the, 90, 401–402, 466, 567
N thermoelastic, 567
Necking, 71 Power
Neutral axis, 177, 184, 189, 191, 193–195, mechanical, 501, 556
204–205, 207–209, 211–213, 215–218,
thermal, 501, 542, 556
292–302, 304–306, 309–310, 312–314,
Principal
316–317, 324, 328–329, 331–333, 336,
axes of inertia, 140, 278, 483, 593
338, 340, 342, 345, 351, 355–356,
direction of strain, 14–21, 23, 553
439–440, 488–489, 574, 608
direction of stress, 40–42, 58–60, 64, 79,
Nonlinear elasticity, 663–704
122, 238, 555
Normal stress, 29–30, 57–58, 61–63, 404, 413,
elongation, 15–16, 18–19, 23–24
555, 636
inertia radius, 140–141
Nozzle, 519, 524–533, 536–541
strain, 553
throat of the, 527, 531
Number stress, 40, 42, 58–60, 63, 65, 79, 116–118,
characteristic, 510–524, 528, 545, 563 120, 122, 124, 238, 306–308, 555
of Duhamel, 560, 563 Principle
of Froude, 520–521, 528, 534 of action and reaction, 29, 435
of Mach, 513–514, 519, 523–524, 528, 533 of Curie, 503, 505, 515
of Peclet, 545 of dissection, 31, 33, 36, 96, 245, 255–256,
of Prandtl, 410, 524, 528, 538 259, 296, 385, 388, 457, 459–460,
of Reynolds, 518–520, 524, 528, 532, 534, 474–475, 477–478, 554, 593, 608
545, 549–550 of energy conservation, 98, 496, 502
of Strouhal, 511, 528 of maintenance of the energy, 496
of Weber, 512 of minimum potential energy, 576
of superposition, 89, 104, 106, 168,
O 190–191, 202–203, 211, 290–291, 328,
Orthotropic material, 637, 641 348, 384, 386, 388, 401, 403, 442, 444,
458, 465, 468–469, 478, 482, 562, 566,
P 576, 593, 641, 660, 669, 673
Panel of symmetry, 292
circular, 370–387 of virtual works, 567
hole in a, 387, 392 Problem
multilayered composite, 650–652 of the anisotropic body, 640
rectangular, 364–370, 387, 596 of the anisotropic thermoelastic body,
Photoelasticity, 66 101–103
Plane strain, 422–426 of Beltrami, 82, 88
Plane stress, 359, 361, 364 of Boussinesq, 381–383
714 Subject Index

of Cauchy, 31, 81, 640 Rigid

of contact, 674–684 rotation, 6–7, 10–11, 31, 68, 70, 177, 198
coupled thermoelastic, 563 translation, 6–7, 11, 31, 70, 177, 198, 283,
of Dirichlet, 220, 410 289
of the elastic equilibrium, 2, 75, 80–89, Rigidity
101–104, 106–107, 110–111, 359, flexural of a beam, 182, 189
361–362, 463–464, 557, 636, 640, 667 flexural of a plate, 395
free boundary, 674 torsional, 224–225, 228, 259, 267, 275, 627
of the minimum complementary energy, 91 Rope, 430–433, 519
of the minimum potential energy, 90,
401–402, 466, 576 S
of Neumann, 220, 227, 278–279, 286, 294, Safety factor, 116, 118, 120, 185, 192, 200,
410 205
non linear, 70 Saint Venant
of plane strain, 422–426 anisotropic problem of, 656–661
with polar symmetry, 377, 388 beam of, 164–165, 185, 222, 238–240, 280,
of the rope, 430–433, 519 294
of the thermoelastic equilibrium, 560, 562 postulate of, 167–168, 190–191, 202–203,
of the transmission of the heat, 493, 563, 211, 290–291, 348–366, 368–369, 388
566 problem of, 127–357, 642
uncoupled thermoelastic, 563–566, 573 Secular equation, 16, 19, 41–42, 65
unilateral, 663, 673–704
Shear
variational, 90–95, 401, 686
approximate solution, 294–304
of the virtual works, 90, 111, 462, 467,
of anisotropic beam, 651, 653, 657, 664
566, 576
center of, 289, 291–292, 338–339, 355
weak, 402
of circular cross section, 304–308
Production
of closed thin walled cross section, 275,
global, 495
351–357
local, 495, 505–506
determination of the shear center, 337–338
punctual, 495
exact solution of the, 351–352
Q of open thin walled cross section, 262–267,
Quadratic expressions 308
of ε and γ , 14 of rectangular cross section, 261
of σ and τ , 651 Shearing strain, 4–5, 14, 101, 553, 635, 641,
664
R Shearing stress
Regime in circular cross section, 264
laminar, 544, 550, 552 in I cross section, 184
turbulent, 548, 550, 552 in rectangular cross section, 262
Reinforcing in T cross section, 317
fibers, 649 symmetry, 32
particles, 649 Shell
Relations axialsymmetric, 419
direct of Navier, 76 circular cylindrical, 419, 421
inverse of Navier, 77 curvature of, 416, 417
between strain and displacements, 14 meridian of, 405, 412–413, 416–417
between stress and strain, 70–80 parallel circle of, 415
Resistance spherical, 419
of attrition, 548–549, 552 theory of thin, 411
of form, 552 Shock wave
of lift, 552 of condensation, 543
viscous, 552 non adiabatic, 543
of wake, 552 oblique, 537
Subject Index 715

Snapping, 629–630, 634 linear invariant of, 42, 77

Sobolev space, 401, 674–675, 684 measurement of, 66
Source plane of, 63–65, 170–172, 359
cold, 99 plane state of, 63–65
warm, 99 principal, 40, 42, 58–61, 63–65, 79,
Stability 116–118, 120, 122, 238, 306–308, 555
approximate analysis, 607 principal triplet of, 42, 59, 77–78, 117–118
effect of the imperfections, 622 of proportionality, 71
exact analysis, 618 quadratic invariant of, 43
State reversible, 509, 556
function, 97–98, 100 tensor, 42, 507–509, 514, 517, 556–557
hydrostatics of stress, 42 yelding, 72, 75, 116
of plane deformation, 422 Stress vector, 27–32, 34–40, 44, 63–65, 118,
of plane stress, 359, 361 170
plane of stress, 63–65, 359 components of the, 29
uniaxial of stress, 65 Surface
Static-method curvature of a, 396
moment, 128–129, 142, 157, 175, 205, 217, elastic, 393, 396
296, 297–298, 301, 305, 310–312, 314, torsion of a, 397
326–327, 333 Symmetry
safety, 115–116, 185–187, 195, 200, 205, axis, 292, 342
208, 212, 218–219, 239, 306–308, of shearing stress, 32, 297, 635
348–351, 627 of stress vector, 31–32, 38–39
Strain System
analysis of, 1–27 of displacements, 31, 68, 111, 462, 465,
components of, 9–14, 108, 635 474
cubic invariant of, 22 of forces, 31, 66, 81, 111, 290, 348, 435,
energy, 95–114, 463–481 608, 640
linear invariant of, 22, 77, 557 isolated, 500
measurement, 25–27
plane state of, 23–24 T
principal direction of, 14–21, 23, 553 Tensor
principal triplet of, 18, 22, 77–79 of order one, 21
quadratic invariant of, 22 of order two, 21
tensor, 22, 509, 517, 557 of order zero, 21
work, 103–107, 112–113, 243, 253, product, 507
464–465, 468–470 Test of tensile breaking, 70–75, 116, 118, 120,
Strain gauge, 26, 66 215
rosette, 26 Theorem
Streamline, 234, 236–238, 251–252, 254–255, of Betti, 466
261 of Castigliano, 470
Strength of Clapeyron, 641
criterions, 115–125 of Kirchhoff, 89
ultimate, 595–596 of Onsager, 505
stress of Stokes, 249–250, 270, 273, 354
admissible, 115–116 of transport, 134–135, 137, 147, 153
analysis of, 27–66 Thermal machine, 99
components of, 30–31, 33, 39–40, Thermodynamics
107–108, 372, 375 cycle, 96, 99–100, 496
concentration of, 232–239, 324, 387 equilibrium, 96–98, 100–103, 463,
cubic invariant of, 43 495–497, 501–503
initial, 572 first principle of, 97–99, 101, 103, 464, 495
irreversible, 509, 556 of the irreversible processes, 503–505
716 Subject Index

linear, 504–505 of open thin walled cross section, 250–264

process, 96, 493, 496, 503, 515, 524 of rectangular cross section, 261
second principle of, 99–100, 496 optimal form of the cross section, 250
third principle of, 496 unitary angle of, 226, 243, 246, 256,
transformation, 101 263–266, 275, 285, 289–292, 354
Thermoelastic dissipation, 493, 562–563, 565 Transformation
Thermoelasticity, 101–102, 493–591 adiabatic, 499
uncoupled static, 563 irreversible, 99
Throttling, 540–541 isentropic, 498, 539
Torsion reversible, 100, 496
of anisotropic beam, 657, 659 Transverse isotropy, 638
of circular cross section, 227–228, 264
of closed thin walled cross section,
251–253, 261–263, 354–357 U
of elliptic cross section, 229 Unilateral, 663, 673–704
non uniform, 264–277 contact problem, 674