Department of Civil and Structural Engineering

Conrad Arnan Ribas

FINITE ELEMENT ANALYSIS OF STRESSES IN BEAM

STRUCTURES

Final project

Espoo, 7.6.2011

Supervisor: Professor Jukka Aalto

Instructor: Professor Jukka Aalto

ABSTRACT
AALTO UNIVERSITY ABSTRACT OF THE FINAL PROJECT
SCHOOLS OF TECHNOLOGY
PO Box 11000, FI-00076 AALTO

http://www.aalto.fi

Author: Conrad Arnan Ribas

Title: Finite element analysis of stresses in beam structures

School: Engineering

Department: Civil and Structural Engineering

Professorship: Structural mechanics Code: Rak-54

Supervisor: Jukka Aalto

Instructor(s): Jukka Aalto

Abstract:

This Final project deals with determining of stresses in beam structures. First, warping
function based on finite element formulations for determining shear stresses at beam
cross-sections under biaxial bending and twisting are developed and discretized. Second,
finite element programs for analysis of beams and three-dimensional frames are
developed. These programs, which have been implemented in MATLAB® environment, are
then combined to a program system, which determines efficiently the combined state of
stresses at any cross-section of a beam or 3D frame.

In order to compare the results of the presented approximate technique to practically exact
ones, a program, based on the finite prism method, for solving stresses in a simply
supported prismatic beam has further been developed. The results of the presented
technique and finite prism method are compared using example problems. In general,
closed thin-walled and solid cross-sections give really accurate results, whereas open thin-
walled cross-sections do not. Some discrepancies in the shear stresses values can be
found usually in the points of the beam where any load starts or finishes. Composite cross-
sections can be analysed applying this methodology.

Finally, applications of the developed program are demonstrated with some more practical
examples.

Date: 7.6.2011 Language: English Number of pages: 90

Keywords: Finite element analysis, 3D frame, Bernoulli-Euler kinematics, warping

function, combined stress state, irregular cross-section geometry, composite cross-section
Finite element analysis of stresses in beam structures 2

TABLE OF CONTENTS

Abstract.................................................................................................................................... 1

Table of contents ..................................................................................................................... 2

1 Preface ............................................................................................................................. 4

2 Introduction ....................................................................................................................... 5

2.1 Aims and objectives of the research ......................................................................... 5

2.2 Research methods .................................................................................................... 5

3 Finite element method ...................................................................................................... 7

3.1 Quadratic isoparametric elements 1-D ...................................................................... 7

3.2 Biquadratic isoparametric Lagrange elements 2-D ................................................... 9

3.3 Numerical integration .............................................................................................. 13

4 Method development ...................................................................................................... 15

4.1 Assumptions ............................................................................................................ 15

4.2 Frame analysis ........................................................................................................ 15

4.2.1 Finite element equations of a bar element ....................................................... 15

4.2.2 Coordinate change ........................................................................................... 24

4.2.3 Global system .................................................................................................. 28

4.2.4 Stress resultants .............................................................................................. 28

4.3 Stress distribution .................................................................................................... 28

4.3.1 Cross-sectional properties ............................................................................... 29

4.3.2 Tension / Compression .................................................................................... 30

Finite element analysis of stresses in beam structures 3

4.3.3 Biaxial bending ................................................................................................. 31

4.3.3.1 Shear centre ................................................................................................. 40

4.3.4 Torsion ............................................................................................................. 41

4.3.5 General case .................................................................................................... 47

5 Biquadratic finite prism method ...................................................................................... 48

5.1 Assumptions ............................................................................................................ 48

5.2 Method description .................................................................................................. 48

6 Validation of the method ................................................................................................. 62

6.1 Comparison between the two methods ................................................................... 62

6.2 Discussion of the results ......................................................................................... 75

7 Demonstrative examples ................................................................................................ 78

Conclusions ........................................................................................................................... 84

Acknowledgements................................................................................................................ 86

References ............................................................................................................................ 87

Appendix ................................................................................................................................ 89
Finite element analysis of stresses in beam structures 4

1 PREFACE

Determining of stresses in beam structures is standard teaching material in basic courses on

mechanics of materials and structural mechanics [1], [2]. However, there are two topics
which are not dealt with enough depth at this level.

The first thing is torsion. Torsion of circular and thin-walled hollow shafts and free torsion of
thin-walled open cross-sections are usually considered in these texts carefully. Introduction
to torsion of beams with general cross-section is, however, done using so-called membrane
analogy, which is based on stress function formulation of free torsion of a cross-section with
general shape [3]. This membrane analogy has been chosen in basic teaching texts,
because it is an excellent tool in explaining the behaviour of shear stresses in a cross-
section. The membrane analogy as such cannot, however, give definite numerical results to
the shear stresses in a twisted cross-section. In order to do this, one should be able to solve
the original problem of free torsion. Different formulations of this problem are presented in
the literature (for example [3], [5] and [6]). Solution of this problem based on stress function
formulation [3] and the finite element method is well known and demonstrated in reference
[4]. This formulation is not, however, straightforward and easy to implement, if the cross-
section contains holes. In this context, the warping function formulation [6] is simpler and the
corresponding finite element solution using linear triangular elements has first been
presented in reference [7]. In this project, the finite element equations for solving the
torsional problem are based on this warping function formulation and biquadratic
isoparametric elements [8].

The second thing, which is not tackled through in standard texts on structural mechanics, is
shear stresses caused by bending. Typically shear flux on any longitudinal section between
two parts of the beam can be determined using simple equilibrium consideration. With the
help of such shear fluxes average, but not exact, shear stresses on the cross-section can
further be determined. However, determination of the shear stresses distribution caused by
bending on a cross-section of any shape is possible. Corresponding formulations have been
presented for example in references [3] and [6]. Finite element solution based on warping
function formulation and using linear triangular elements has been first presented in
reference [9]. In this thesis the finite element equations for solving the bending shear
stresses are based on this warping function formulation and biquadratic isoparametric
elements [10], [11].
Finite element analysis of stresses in beam structures 5

2 INTRODUCTION

2.1 AIMS AND OBJECTIVES OF THE RESEARCH

The goal in this final project is to develop a finite element based program in MATLAB®
environment, which solves the stress resultants of a beam structure and using these, the
combined state of stresses at any point in this structure. The program thus consists of two
main parts. The first part solves the stress resultants of the structure using cubic C 1 −
continuous space frame elements under given loading. The second part performs stress
analysis of the cross-section by combining the effects of axial loading, torsion and biaxial
bending using the obtained stress resultants as data.

2.2 RESEARCH METHODS

The space frame program uses typical straight, cubic C 1 − continuous space frame elements
[12] of constant stiffness. The loading consists of linearly distributed axial and transverse
load and twisting moment within elements and concentrated loads and moments at the
nodes. Although the elements are rather simple, they produce exact solution for the stress
resultants at the element nodes if the elements are straight and uniform and the loading is
just as described. They can also be used in connection with moderately curved beam
structures and structures with variable thickness if the finite element grid is chosen to be
sufficiently dense.

The program for calculating the cross-sectional stresses uses C 0 − continuous isoparametric
biquadratic quadrilateral elements with nine nodes and numerical integration. It calculates
the effects of normal force, shear forces, twisting moment and bending moments to the
cross-sectional stresses and thus obtains the combined state of stresses.

In order to study the efficiency of the developed technique, a program, based on the finite
prism method [4], [13], [14] and [15], has also been developed in this thesis. This program is
based on three dimensional elasticity equations and does not have the restrictions of
classical beam theory. Thus the results obtained using this program give good comparison
data for studying the accuracy of the enhanced beam theory of this project. The finite prism
program also uses C 0 − continuous isoparametric biquadratic elements and numerical
integration. It can handle simply supported straight beams with arbitrary cross-section. The
loading can be surface load, which is piecewise constant in axial direction and quadratically
Finite element analysis of stresses in beam structures 6

distributed in transverse direction, volume force, which is piecewise constant in axial

direction, or distributed nodal line load, which is piecewise constant in axial direction.

Three simply supported example beams, with solid rectangular, open U-shaped and hollow
rectangular cross-section are considered in the comparison study. Each beam is loaded by
narrow uniform rectangular surface load acting on the second half of the beam. Both centric
loading causing pure bending and eccentric loading causing mixed torsion and bending have
been studied. The stress results of the two methods are compared at quarter points and
midpoint of the beam.

The final example problems of the thesis have been selected to demonstrate the applications
of the developed program. The first problem considers a circular beam in the horizontal plane
loaded by uniform line load along the axis of the beam. The cross-section is trapezoidal and
non-symmetric. Both torsion and bending are present. The second example problem is a
slightly curved symmetric girder in the vertical plane loaded by uniform transverse load in the
symmetry plane. The cross-section is solid and T-formed, but to avoid stress singularities at
the corners, they have been smoothed using circles. Both axial loading and bending is
present in this problem. The third example problem considers the application of the
developed method to a composite beam. The cross-section consists of concrete deck and
steel U-shaped thin-walled plate. Both pure bending and twisting cases are studied
separately.
Finite element analysis of stresses in beam structures 7

3 FINITE ELEMENT METHOD

In order to solve the elastic problem, the finite element method will be used with modelling
and discretization of the object under study.

One- and two-dimensional elements are needed, so the basics of both are going to be
described [16].

In all cases, the sort of element is going to be the same, which is quadratic in one dimension
and biquadratic Lagrange in two dimensions, to obtain more accurate results. Moreover, the
use of isoparametric elements is important to be able to model cross-sections of any shape.

Furthermore, numerical integration [16] for solving one- and two-dimensional integrals is
explained in connection with the application of the finite element method.

3.1 QUADRATIC ISOPARAMETRIC ELEMENTS 1-D

A quadratic line element with three equally spaces nodes in natural coordinates, −1 ≤ ξ ≤ 1 ,
is considered (Fig. 3.1). The shape functions of this element are obtained using quadratic
Lagrange polynomials as
(ξ − ξ 2 )(ξ − ξ3 ) (ξ − 0)(ξ − 1) 1
N1e (ξ ) = = = − ξ (1 − ξ ),
(ξ1 − ξ 2 )(ξ1 − ξ3 ) (−1 − 0)(−1 − 1) 2
(ξ − ξ1 )(ξ − ξ3 ) [ξ − (−1)](ξ − 1)
N 2e (ξ ) = = = (1 − ξ )(1 + ξ ), (3.1)
(ξ 2 − ξ1 )(ξ 2 − ξ3 ) [0 − (−1)](0 − 1)
(ξ − ξ1 )(ξ − ξ 2 ) [ξ − (−1)](ξ − 0) 1
N 3e (ξ ) = = = ξ (1 + ξ ).
(ξ3 − ξ1 )(ξ3 − ξ 2 ) [1 − (−1)](1 − 0) 2

Fig. 3.1: Shape functions of a three node quadratic line element in natural coordinate ξ

Derivatives Ni ,ξ ≡ dNi / dξ of the 1-dimensional shape functions (3.1) with respect to the
e e

natural coordinates ξ are

Finite element analysis of stresses in beam structures 8

1
N1,eξ (ξ ) = − (1 − 2ξ ),
2
N 2,ξ (ξ ) = −2ξ ,
e
(3.2)
1
N3,e ξ (ξ ) = (1 + 2ξ ).
2

In Fig. 3.2 (a), a three node quadratic isoparametric line element is presented by a line
segment of length 2 in the natural ξ − line and in Fig. 3.2 (b) by a curved line segment in the
physical x,y-plane.

Fig. 3.2: A three node quadratic isoparametric line element (a) in natural coordinate ξ and (b) in
physical coordinates x, y

The shape of the element in the physical plane is expressed by position vector
r (ξ ) = x (ξ )i + y (ξ ) j , (3.3)

where the functions x (ξ ) and y (ξ ) are

3
x(ξ ) = ∑ N ie (ξ ) xie ,
i =1
3
(3.4)
y (ξ ) = ∑ N (ξ ) y .
i
e e
i
i =1

Differential vector between points ξ and ξ + d ξ in the physical plane is

dr ⎛ dx dy ⎞
dr = dξ = ⎜ i+ j ⎟ dξ = ( x,ξ i + y,ξ j)d ξ . (3.5)
dξ ⎝ dξ dξ ⎠

The length of this differential vector is

ds = x,ξ2 + y,ξ2 dξ (3.6)

Finite element analysis of stresses in beam structures 9

and it is the length of a differential line element corresponding to differential change d ξ of

the natural coordinate. The derivatives of the coordinates functions x (ξ ) and y (ξ ) in

equation (3.6) are obtained using formulas
3
x,ξ (ξ ) = ∑ N ie,ξ (ξ ) xie ,
i =1
3
(3.7)
y,ξ (ξ ) = ∑ N (ξ ) y .
e
i ,ξ
e
i
i =1

3.2 BIQUADRATIC ISOPARAMETRIC LAGRANGE ELEMENTS 2-D

Shape functions of a biquadratic quadrilateral Lagrangian element in natural co-ordinates

−1 ≤ ξ ≤ 1 and −1 ≤ η ≤ 1 (Fig. 3.3) can be obtained with the help of shape functions (3.1)
using the formula
N3(e i −1)+ j (ξ ,η ) = Nie (ξ ) N ej (η ), i, j = 1, 2,3. (3.8)

Fig. 3.3: Three shape functions of a nine node biquadratic quadrilateral Lagrange element in natural
co-ordinates ξ and η

Therefore, one obtains

Finite element analysis of stresses in beam structures 10

1
N1e (ξ ,η ) = N1e (ξ ) N1e (η ) = ξ (1 − ξ )η (1 − η ),
4
1
N 2e (ξ ,η ) = N 2e (ξ ) N1e (η ) = − (1 − ξ )(1 + ξ )η (1 − η ),
2
1
N 3e (ξ ,η ) = N 3e (ξ ) N1e (η ) = − ξ (1 + ξ )η (1 − η ), (3.9)
4
1
N 4e (ξ ,η ) = N1e (ξ ) N 2e (η ) = − ξ (1 − ξ )(1 − η )(1 + η ),
2
etc.

Partial derivatives Ni ,ξ ≡ ∂Ni / ∂ξ and Ni ,η ≡ ∂Ni / ∂η of the 2-dimensional shape functions

e e e e

(3.9) are

1
1 N1,eη (ξ ,η ) = ξ (1 − ξ )(1 − 2η ),
N1,eξ (ξ ,η ) = (1 − 2ξ )η (1 − η ), 4
4 1
N 2,ξ (ξ ,η ) = ξη (1 − η ),
e N 2,e η (ξ ,η ) = − (1 − ξ )(1 + ξ )(1 − 2η ),
2
1 1
N 3,e ξ (ξ ,η ) = − (1 + 2ξ )η (1 − η ), N 3,e η (ξ ,η ) = − ξ (1 + ξ )(1 − 2η ), (3.10)
4 4
1 N 4,η (ξ ,η ) = ξ (1 − ξ )η ,
e
N 4,e ξ (ξ ,η ) = − (1 − 2ξ )(1 − η )(1 + η ),
2 etc.
etc.

Fig. 3.4: A nine node quadratic isoparametric quadrilateral element (a) in natural coordinate ξ ,η and
(b) in physical coordinates x, y
Finite element analysis of stresses in beam structures 11

In Fig. 3.4 (a), a nine node quadrilateral Lagrange isoparametric element is presented by a
square of side lengths 2 in the natural ξ ,η − plane and in Fig. 3.4 (b) by a curved
quadrilateral in the physical x,y-plane.

The shape of the element in the physical plane is expressed by position vector
r (ξ ,η ) = x(ξ ,η )i + y (ξ ,η ) j , (3.11)

where the functions x (ξ ,η ) and y (ξ ,η ) are

9
x(ξ ,η ) = ∑ N ie (ξ ,η ) xie ,
i =1
9
(3.12)
y (ξ ,η ) = ∑ N (ξ ,η ) y .
i
e e
i
i =1

Differential vector between points ξ ,η and ξ + d ξ ,η in the physical plane is

∂r ⎛ ∂x ∂y ⎞
dr1 = dξ = ⎜ i + j ⎟ dξ = ( x,ξ i + y,ξ j)d ξ . (3.13a)
∂ξ ⎝ ∂ξ ∂ξ ⎠

and a differential vector between points ξ ,η and ξ ,η + dη in the physical plane is

∂r ⎛ ∂x ∂y ⎞
dr2 = dη = ⎜ i+ j ⎟ dη = ( x,η i + y,η j)dξ . (3.13b)
∂η ⎝ ∂η ∂η ⎠

The differential area of the element in the physical plane corresponding to a differential
square d ξ dη in the natural plane is

x,ξ y,ξ 0
x,ξ y,ξ
dA = dr1 × dr2 ik = x,η y,η 0 d ξ dη = dξ dη . (3.14)
x,η y,η
0 0 1

This result is usually expressed as

dA = det Jd ξ dη , (3.15)

where det J is determinant of so-called Jacobian matrix

⎡ ∂x ∂y ⎤
⎢ ∂ξ ∂ξ ⎥ ⎡ x,ξ y,ξ ⎤
J=⎢ ⎥≡⎢
y,η ⎥⎦
. (3.16)
⎢ ∂x ∂y ⎥ ⎣ x,η
⎢ ∂η ∂η ⎥⎦
⎣

The partial derivatives of the functions x (ξ ,η ) and y (ξ ,η ) in formula (3.12) are obtained
using formulas
Finite element analysis of stresses in beam structures 12

9 9
x,ξ (ξ ,η ) = ∑ N ie,ξ (ξ ,η ) xie , y,ξ (ξ ,η ) = ∑ N ie,ξ (ξ ,η ) yie ,
i =1 i =1
9 9
(3.17)
x,η (ξ ,η ) = ∑ N (ξ ,η ) x , y,η (ξ ,η ) = ∑ N (ξ ,η ) y .
e
i ,η
e
i
e
i ,η
e
i
i =1 i =1

Later, partial derivatives of the shape functions N ie with respect to the physical coordinates

x and y in terms of the corresponding derivatives with respect to the natural coordinates ξ
and η are needed. Using the chain rule of differentiations one can write

∂N ie ∂x ∂N ie ∂y ∂N ie ⎫
= + ⎪
∂ξ ∂ξ ∂x ∂ξ ∂y ⎪ ⎧⎪ Nie,ξ = x,ξ Nie, x + y,ξ Nie, y
⎬ or with our notation ⎨ e (3.18)
∂N ie ∂x ∂N ie ∂y ∂N ie ⎪ ⎪⎩ Ni ,η = x,η Ni , x + y,η Ni , y
e e
= +
∂η ∂η ∂x ∂η ∂y ⎪⎭

or

⎪⎧ N i ,ξ ⎫⎪
e
⎧⎪ N ie, x ⎪⎫
⎨ e ⎬ = J ⎨ e ⎬. (3.19)
⎩⎪ N i ,η ⎪⎭ ⎪⎩ N i , y ⎭⎪

Inverting this relation gives

⎧⎪ N ie, x ⎫⎪ ⎧N e ⎫
−1 ⎪ i ,ξ ⎪
⎨ e ⎬=J ⎨ e ⎬, (3.20)
⎩⎪ N i , y ⎭⎪ ⎩⎪ N i ,η ⎭⎪

where J −1 is the inverse of the Jacobian matrix. The 2 × 2 Jacobian matrix J can easily be
inverted and the result is

1 ⎡ y,η − y,ξ ⎤
J −1 = ⎢ x,ξ ⎥⎦
, (3.21)
det J ⎣ − x,η

where the Jacobian determinant is

x,ξ y,ξ
det J = = x,ξ y,η − x,η y,ξ . (3.22)
x,η y,η

Relations (3.20) can now be written in final form as

1
N ie, x = ( y,η N ie,ξ − y,ξ N ie,η ),
det J
(3.23)
1
N e
i, y = (− x,η N ie,ξ + x,ξ N ie,η ).
det J
Finite element analysis of stresses in beam structures 13

3.3 NUMERICAL INTEGRATION

If typical two dimensional problems are analyzed using the finite element method, following
line and surface integrals over the element are needed

I1e = ∫ F1 (s)ds ,
e
I 2e = ∫ F2 ( x, y)dA .
e
(3.24)
s A

These integrals can be expressed in natural coordinates, by using equations (3.6) and (3.15)
and changing the integration limits

+1 +1 +1
I1e = ∫ f1 (ξ )d ξ , I 2e = ∫∫ f 2 (ξ ,η )dξ dη , (3.25)
−1 −1 −1

where

f1 (ξ ) = F1 (ξ ) x,ξ2 (ξ ) + y,ξ2 (ξ ), f 2 (ξ ,η ) = F2 [ x(ξ ,η ), y (ξ ,η )]det J (ξ ,η ) . (3.26)

Integrands (3.26) are, in general case, so complicated functions, that analytical integration is
impossible. Therefore numerical integration is needed. Approximate values of integrals (3.26)
are obtained numerically using formulas

n n
I1e ≈ ∑ wk f1 (ξ k ) , I 2e ≈ ∑ wk f 2 (ξ k ,ηk ) , (3.27)
k =1 k =1

where n and n are the number of integration points of the line element and the surface
element, respectively, wk is the weight and ξ k and ηk are the coordinates of the integration

point k . These quantities for one and two dimensional integration are listed in Tables 3.1
and 3.2 respectively.
Finite
e element analysis
a of stresses
s in beam
b structures 14
4

Table
e 3.1: Integrration weights, coordinates of integrration pointss and maxim
mum integer exponent p,,
which
h allows exa
act integratio
on of power ξp, for one
e dimensiona
al Gauss nu
umerical inte
egration overr
intervval −1 ≤ ξ ≤ 1 .
n Figure k wk ξk p
1 1 1 2 0 1
ξ
2 1 2 1 1 −1/ 3 3
ξ
2 1 +1/ 3
3
1 2 3
1 5/9 − 3/5 5
ξ 2 8/9 0
3 5/9 + 3/5

Table
e 3.2: Integrration weights, coordinates of integrration pointss and maxim
mum integer exponent p,,
h allows exact integration
which n of power te
erm ξ pη p , for two dimeensional Gauuss numericaal integrationn
over square −1 ≤ ξ ≤ 1, − 1 ≤ η ≤ 1 .
n Figure k wk ξk ηk p
1 1 4 0 0 1

4 1 1 −1/ 3 −1/ 3 3
2 1 +1/ 3 −1/ 3
3 1 −1/ 3 +1/ 3
4 1 +1/ 3 +1/ 3

9 1 25 / 81 − 3 / 5 − 3/5 5
2 40 / 81 0 − 3/5
3 25 / 81 + 3 / 5 − 3/5
4 40 / 81 − 3 / 5 0
5
64 / 81 0 0
6 + 3/5 0
40 / 81 + 3/5
7 − 3/5
25 / 81 + 3/5
8 0
40 / 81 + 3 / 5 + 3/5
9
25 / 81
Finite element analysis of stresses in beam structures 15

4 METHOD DEVELOPMENT

In order to achieve the stress distribution, first of all an analysis of the frame must be realised
to obtain stress resultants that act in each cross-section, and later, the stress values are
calculated with this information.

4.1 ASSUMPTIONS

When doing the analysis, some hypothesis will be taken into account: material is elastic and
isotropic and the analysis will be linear geometrically and materially. Cross-sections can be
made of different materials and the frame can be composed of different cross-sections. Thus,
plasticity, effect of joints and buckling will not be considered.

The frame can be loaded in different ways: distributed axial forces, transverse forces and
twisting moments; and concentrated nodal forces and moments.

4.2 FRAME ANALYSIS

The first use of the finite element method is made with the frame analysis. The entire frame
must be discretized into several bar elements. Each element provides a set of equations to
form the general equation set and solve the problem [17], [18].

4.2.1 FINITE ELEMENT EQUATIONS OF A BAR ELEMENT

The basic unknown functions of a bar element under tension/compression, torsion and
biaxial bending are axial displacement, u ( x ) , deflections v ( x ) and w( x ) and angle of twist

ϕt ( x) . The axial displacement and angle of twist within an element e are approximated using
C 0 − continuous linear finite element approximations
uˆ ( x) = N1l ( x)u1 + N 2l ( x)u2 ,
(4.1)
ϕˆt ( x) = N1l ( x)ϕt1 + N 2l ( x)ϕt 2 ,

where

N1l = 1 − ξ , N 2l = ξ , (4.2)

and
Finite
e element analysis
a of stresses
s in beam
b structures 16
6

x − x1e
ξ= . (4.3))
Le

oximated ussing C 1 − ccontinuous cubic finite

The deflectionss within an element e are appro e
ment approximations
elem

vˆ( x) = N1c ( x)v1 + N 2c ( x)ϕ z1 + N3c ( x)v2 + N 4c ( x)ϕ z 2 ,

(4.4))
wˆ ( x) = N1c ( x) w1 − N 2c ( x)ϕ y1 + N3c ( x) w2 − N 4c ( x)ϕ y 2 ,

wherre

N1c = 1 − 3ξ 2 + 2ξ 3 , N 2c = Le (ξ − 2ξ 2 + ξ 3 ), N 3c = 3ξ 2 − 2ξ 3 , N 4c = Le ( −ξ 2 + ξ 3 ). (4.5))

Fig
g. 4.1: (a) Nu
umbering of degrees
d of frreedom, (b) generalized
g d
displacemen
nts and (c) ge
eneralized
forcess

Num
mbering of the
t degrees of freedo
om of the beam elem
ment is sho
own in Fig. 4.1a, the
e
eralized dissplacementss aie , i = 1,
gene ,12 (see b) are the nodal disp
e Fig. 4.1b placementss

ui , vi ,wi (i = 1, 2)) and rotations ϕti ,ϕ yi ,ϕ zi , (i = 1, 2)

2 and the correspond
c ing generallized forcess
Finite element analysis of stresses in beam structures 17

Fi e , i = 1, ,12 (see Fig. 4.1c) are the nodal forces U i , Vi ,Wi (i = 1, 2) and moments

M ti ,M yi ,M zi , (i = 1, 2) . In the following, relationships between generalized forces and

generalized displacements of the beam elements will be derived.

The approximations (4.1) and (4.4) of the basic unknowns can be written as

⎧ uˆ ( x) ⎫
⎪ vˆ( x) ⎪
⎪ ⎪
uˆ ( x) ≡ ⎨ ⎬ = N ( x)a
e e
(4.6)
4×1 ˆ
w ( x )
⎪ ⎪
⎪⎩ϕˆt ( x) ⎪⎭

where

⎡ N1l 0 0 0 0 0 N 2l 0 0 0 0 0 ⎤
⎢ ⎥
0 N1c 0 0 0 N 2c 0 N 3c 0 0 0 N 4c ⎥
N =⎢
e
(4.7)
4×12 ⎢0 0 N c
0 −N c
0 0 0 N c
0 − N 4c 0 ⎥
⎢ ⎥
1 2 3

⎢⎣ 0 0 0 N1l 0 0 0 0 0 N 2l 0 0 ⎥⎦

and

⎧ a1e ⎫
⎪ ⎪
a e = ⎨ ⎬ ≡ [u1e v1e w1e ϕte1 ϕ ye1 ϕ ze1 u2e v2e w2e ϕte2 ϕ ye 2 ϕ ze2 ]T . (4.8)
12×1
⎪a e ⎪
⎩ 12 ⎭

The approximations of the axial strain, twist per unit length and curvatures are

duˆ
εˆx ( x) ≡ ( x) = N1,l x ( x)u1 + N 2,l x ( x)u2 ,
dx
dϕˆ
θˆ( x) ≡ t ( x) = N1,l x ( x)ϕt1 + N 2,l x ( x)ϕt 2 ,
dx
(4.9)
d 2 wˆ
κˆ y ( x) ≡ − 2 ( x) = − N1,c xx ( x) w1 + N 2,c xx ( x)ϕ y1 − N 3,c xx ( x) w2 + N 4,c xx ( x)ϕ y 2 ,
dx
d 2 vˆ
κˆ z ( x) ≡ − 2 ( x) = − N1,c xx ( x)v1 − N 2,c xx ( x)ϕ z1 − N 3,c xx ( x)v2 − N 4,c xx ( x)ϕ z 2
dx

and they can be written as

Εˆ ( x) = Be ( x)ae , (4.10)
Finite element analysis of stresses in beam structures 18

where

⎧ εˆx ( x) ⎫
⎪ ˆ ⎪
ˆΕ ( x) = ⎪⎨ θ ( x) ⎪⎬ (4.11)
⎪κˆ y ( x) ⎪
4×1

⎪κˆ ( x) ⎪
⎩ z ⎭

and

⎡ N1,l x 0 0 0 0 0 N 2,l x 0 0 0 0 0 ⎤
⎢ l l ⎥
0 0 0 N 0 0 0 0 0 N 0 0 ⎥
B =⎢
e 1, x 2, x
4×12 ⎢ 0 0 − N1,c xx 0 N 2,c xx 0 0 0 − N3,c xx 0 N 4,c xx 0 ⎥
⎢ ⎥
⎢⎣ 0 − N1,c xx 0 0 0 − N2,c xx 0 − N3,c xx 0 0 0 − N 4,c xx ⎥⎦
(4.12)

In the following, the finite element equations of a beam element are derived using the
principle of virtual work. In this context, the virtual generalized displacements δ u and the
virtual generalized strains δε are approximated similarly to

⎧ δ uˆ ( x) ⎫
⎪ δ vˆ( x) ⎪
⎪ ⎪
δ uˆ ( x) ≡ ⎨ ⎬ = N ( x)δ a
e e
(4.13)
⎪δ wˆ ( x) ⎪
⎪⎩δϕˆt ( x) ⎪⎭

and

⎧δε x ( x) ⎫
⎪ δθ ( x) ⎪
⎪ ⎪
δΕ( x) ≡ ⎨
ˆ ⎬ = B ( x)δ a ,
e e
(4.14)
⎪δκ y ( x) ⎪
⎪⎩δκ z ( x) ⎪⎭

where the virtual generalized displacements are

⎧ δ a1e ⎫
⎪ ⎪
δ ae = ⎨ ⎬ = [δ u1 δ v1 δ w1 δϕt1 δϕ y1 δϕ z1 δ u2 δ v2 δ w2 δϕt 2 δϕ y 2 δϕ z 2 ] .
e e e e e e e e e e e e T
(4.15)
⎪δ a e ⎪
⎩ 12 ⎭

Now, the principle of virtual work to element e is used. That is

Finite element analysis of stresses in beam structures 19

δ W ≡ δ Wint + δ Wext = 0 , (4.16)

where the internal virtual work of the element is

δ Wint = − ∫ ( N δε x + M tδθ + M yδκ y + M zδκ z )dx = − ∫ δΕT Σdx , (4.17)

e e
L L

where δΕ is the vector of virtual generalized strains and

⎧ N ( x) ⎫
⎪ M ( x) ⎪
⎪ ⎪
Σ ( x) = ⎨ t ⎬ (4.18)
4×1
⎪ M y ( x) ⎪
⎪⎩ M z ( x) ⎪⎭

is the vector of generalized stresses.

The relation between the generalized stresses and generalized strains can be written as

Σ = D(Ε − Ε 0 ) , (4.19)

where

⎡( EA) 0 0 0 ⎤
⎢ 0 (GI )t 0 0 ⎥
D=⎢ ⎥ (4.20)
⎢ 0 0 ( EI ) y ( EI ) yz ⎥
⎢ ⎥
⎣ 0 0 ( EI ) yz ( EI ) z ⎦

is the generalized stress-strain matrix, where its terms are described in chapter 4.3.1 and
4.3.4, and

⎧ε x0 ⎫
⎪θ ⎪
⎪ ⎪
Ε0 = ⎨ 0 ⎬ (4.21)
⎪κ y 0 ⎪
⎪⎩κ z 0 ⎪⎭

is the vector of initial generalized strains. By inserting expression (4.19) of generalized

stresses into the expression (4.17) of internal virtual work, one gets
Finite element analysis of stresses in beam structures 20

δ Wint = − ∫ δΕT DΕdx + ∫ δΕT DΕ 0 dx . (4.22)

e e
L L

For the external virtual work of the element,

δ Wext = ∫ (qxδ u + q yδ v + qzδ w + mδϕt )dx

Le

+ U1δ u ( x1e ) + V1δ v( x1e ) + W1δ w( x1e ) + M t1δϕt ( x1e ) + M y1δϕ y ( x1e ) + M z1δϕ z ( x1e ) (4.23)
+ U 2δ u ( x2e ) + V2δ v( x2e ) + W2δ w( x2e ) + M t 2δϕt ( x2e ) + M y 2δϕ y ( x2e ) + M z 2δϕ z ( x2e ),

where qx ( x ) , q y ( x ) and qz ( x ) are axial and transverse distributed loads and mt ( x) is

distributed twisting moment load. Now, the total virtual work of the element can be obtained
as

δ W ≡ δ Wint + δ Wext
= − ∫ δΕT DΕdx + ∫ δΕT DΕ 0 dx + ∫ δ uT qdx
Le Le Le (4.24)
+ U1δ u ( x1e ) + V1δ v( x1e ) + W1δ w( x1e ) + M t1δϕt ( x1e ) + M y1δϕ y ( x1e ) + M z1δϕ z ( x1e )
+ U 2δ u ( x2e ) + V2δ v( x2e ) + W2δ w( x2e ) + M t 2δϕt ( x2e ) + M y 2δϕ y ( x2e ) + M z 2δϕ z ( x2e )

where

⎧ qx ⎫
⎪q ⎪
⎪ ⎪
q = ⎨ y⎬ (4.25)
⎪ qz ⎪
⎪⎩ m ⎪⎭

is the vector of distributed generalized loads. Based on the principle of virtual work, the total
virtual work (4.24) should disappear with arbitrary virtual generalized displacements δ u( x ) .
Substituting the approximations (4.9), (4.10), (4.12) and (4.13) into the expression (4.24) and
taking further into account that

δ u ( x1e ) = δ u1 = δ a1e , δ v( x1e ) = δ v1 = δ a2e , δ w( x1e ) = δ w1 = δ a3e ,

δϕt ( x1e ) = δϕt1 = δ a4e , δϕ y ( x1e ) = δϕ y1 = δ a5e , δϕ z ( x1e ) = δϕ z1 = δ a6e , (4.26)

δ u ( x2e ) = δ u2 = δ a7e , δ v( x2e ) = δ v2 = δ a8e , δ w( x2e ) = δ w2 = δ a9e ,

δϕt ( x2e ) = δϕt 2 = δ a10e , δϕ y ( x2e ) = δϕ y 2 = δ a11e , δϕ z ( x2e ) = δϕ z 2 = δ a12e

Finite element analysis of stresses in beam structures 21

the expression (4.24) gets the form

δ Wˆ = δ aeT (− ∫ B eT DB e dx a e + ∫ B eT DΕ 0 dx + ∫ N eT qdx + F e ) . (4.27)

Le Le Le

The principle of virtual work δ W = 0 with arbitrary virtual generalized displacements δ a e now
results to the element equation of the beam element

F e = K ea e − R e , (4.28)

where

⎧ F1e ⎫
⎪ ⎪
F e = ⎨ ⎬ ≡ [U1e V1e W1e M te1 M ye1 M ze1 U 2e V2e W2e M te2 M ye 2 M ze2 ]T , (4.29)
⎪F e ⎪
⎩ 12 ⎭

is the generalized nodal force vector,

K e = ∫ B eT DB e dx (4.30)
e
L

is the element stiffness matrix and

R e = ∫ N eT qdx + ∫ B eT DΕ 0 dx (4.31)
e e
L L

is the element load vector.

In the general case, when the generalized stress-strain matrix D, the initial generalized
strain vector Ε 0 and distributed load vector q are functions of x , the element stiffness

matrix and load vector can be obtained using numerical integration. The integration variable
is first changed to ξ resulting to

1
K e = ∫ B eT (ξ )D(ξ )B e (ξ ) Le d ξ (4.32)
0

and

1
R = ∫ [N eT (ξ )q(ξ ) + B eT (ξ )D(ξ )Ε 0 (ξ )]Le d ξ
e
(4.33)
0
Finite element analysis of stresses in beam structures 22

Numerical integration is then applied resulting to

n
K e = ∑ wk B eT (k )D( k )B e (k ) Le (4.34)
k =1

and

n
R e = ∑ wk [N eT (k )q( k ) + B eT ( k )D( k )Ε 0 ( k )]Le , (4.35)
k =1

where the notation (k ) = (ξ k ) for the value of a function at integration point k is used.

Here the problem is restricted to the case in which the generalized stress strain matrix D
and the generalized initial strain vector Ε0 are constants, De and Ε e0 , within the element.

The distributed load q is supposed to be a linear function of x and can be written as

q(ξ ) = q1 (1 − ξ ) + q 2 ξ . (4.36)

It is easy to see that the element stiffness matrix and the element load vector get the form of

1
K e = ∫ BeT (ξ )De Be (ξ ) Le dξ (4.37)
0

and

1 1 1
R = L ∫ (1 − ξ )N (ξ )dξ q1 +L ∫ ξ N (ξ )dξ q 2 + L ∫ BeT (ξ )dξ De Ε 0e
e e eT e eT e
(4.38)
0 0 0

The integrations in equations (4.37) and (4.38) can be performed analytically resulting to final
expressions (4.39) for the element stiffness matrix and (4.40) for the element load vector.
Finite element analysis of stresses in beam structures 23

⎡ ( EA)e ( EA)e ⎤
⎢ e
0 0 0 0 0 − 0 0 0 0 0 ⎥
⎢ L Le ⎥
⎢ 12( EI )ez 12( EI )eyz 6( EI )eyz 6( EI )ez 12( EI )ez 12( EI )eyz 6( EI )eyz 6( EI ) z ⎥
e

⎢ 0 0 − 0 − − 0 − ⎥
⎢ Le3 L e3
L e2
Le 2 Le3 L e3
Le 2 Le 2 ⎥
⎢ 12( EI )eyz 12( EI )ey 6( EI )ey 6( EI )eyz 12( EI )eyz 12( EI )ey 6( EI )ey 6( EI ) yz ⎥
e

⎢ 0 0 − 0 − − 0 − ⎥
⎢ Le3 Le3 Le 2 Le 2 Le3 Le3 Le 2 Le 2 ⎥
⎢ (GI ) e
(GI ) e ⎥
⎢ 0 0 0 t
0 0 0 0 0 − t
0 0 ⎥
⎢ Le Le ⎥
⎢ 6( EI ) e
6( EI ) e
4( EI ) e
4( EI ) e
6( EI ) e
6( EI ) e
2( EI )ey 2( EI ) yze ⎥

⎢ 0 − yz
− y
0 y
− yz
0 yz y
0 − ⎥
⎢ Le ⎥
e2 e2 e e e2 e2
L L L L L L Le
⎢ 6( EI ) e
6( EI )eyz 4( EI )eyz 4( EI ) e
6( EI ) e
6( EI )eyz 2( EI )eyz 2( EI ) ze ⎥
⎢ 0 z
0 − z
0 − z
− 0 − ⎥
K e = ⎢⎢ Le 2 Le 2 Le Le Le 2 Le 2 Le Le ⎥
⎥
( EA)e ( EA) e
⎢− e 0 0 0 0 0 0 0 0 0 0 ⎥
⎢ L Le ⎥
⎢ 12( EI )eyz 6( EI )eyz 12( EI )eyz 6( EI )eyz ⎥
⎢ 0 12( EI )ez 6( EI )ez 12( EI )ez 6( EI )ez ⎥
− − 0 − 0 0 − e2
⎢ Le3 L e3
L e2
Le 2 Le3 L e3
Le 2 L ⎥
⎢ 12( EI )eyz 12( EI )ey 6( EI )ey 6( EI )eyz 12( EI )eyz 12( EI )ey 6( EI )ey 6( EI )eyz ⎥⎥
⎢ 0 − − 0 − 0 0 −
⎢ Le3 Le3 Le 2 Le 2 Le3 Le3 Le 2 Le 2 ⎥
⎢ e e ⎥
⎢ 0 (GI ) (GI ) ⎥
0 0 − t
0 0 0 0 0 t
0 0
⎢ Le Le ⎥
⎢ ⎥
⎢ 0 6( EI )eyz 6( EI )ey 2( EI )ey 2( EI )eyz 6( EI )eyz 6( EI )ey 4( EI )ey 4( EI )eyz ⎥
− − 0 − 0 + + 0 −
⎢ L e2
L e2
L e
L e
L e2
L e2
Le L ⎥
e

⎢ ⎥
⎢ 0 6( EI ) e
6( EI )eyz 2( EI )eyz 2( EI ) e
6( EI ) e
6( EI )eyz 4( EI )eyz 4( EI )ez ⎥
⎢⎣
z
0 − z
0 − z
− 0 − ⎥⎦
Le 2 Le 2 Le Le Le 2 Le 2 Le Le
(4.39)

⎧ 2qxe1 + qxe 2 e ⎫
⎪ L − ( EA)e ε x 0 ⎪
⎪ 6 ⎪
⎪ 7q ey1 + 3q ey 2 e ⎪
⎪ L ⎪
⎪ 20 ⎪
⎪ 7qze1 + 3qze2 e ⎪
⎪ L ⎪
⎪ 20 ⎪
⎪ 2m1e + m2e e ⎪
⎪ L − (GI t ) θ 0
e
⎪
⎪ 6 ⎪
⎪ ⎛ qe qe ⎞ e2 ⎪
⎪ − ⎜ z1 + z 2 ⎟ L − ( EI )ey κ ye 0 − ( EI )eyz κ ze0 ⎪
⎪ ⎝ 20 30 ⎠ ⎪
⎪ e ⎪
⎪ ⎛ q y1 q y 2 ⎞ e 2
e
e e ⎪
⎜ +
⎪ ⎜ 20 30 ⎟ ⎟ L + ( EI ) e
κ e
yz y 0 + ( EI ) κ
z z0 ⎪
⎪⎝ ⎠ ⎪
R =⎨
e
⎬
⎪ qx1 + 2qx 2 e
e e
⎪
L + ( EA)e ε x 0
⎪ 6 ⎪
⎪ ⎪
⎪ 3q y1 + 7q y 2 e
e e
⎪
⎪ L ⎪
20
⎪ ⎪
⎪ 3qz1 + 7qz 2 e
e e
⎪
⎪ L ⎪
20
⎪ ⎪
⎪ m1 + 2m2 e
e e
⎪
⎪ L + (GI t ) θ 0
e
⎪
6
⎪ ⎪
⎪ ⎛ qze1 qze2 ⎞ e 2 e ⎪ (4.40)
⎪ ⎜ 30 + 20 ⎟ L + ( EI ) y κ y 0 + ( EI ) yz κ z 0 ⎪
e e e

⎪⎝ ⎠ ⎪
⎪ ⎛ qe qe ⎞ ⎪
⎪− ⎜ y1 + y 2 ⎟ Le 2 − ( EI )eyz κ ye 0 − ( EI )ez κ ze0 ⎪
⎪ ⎜⎝ 30 20 ⎟⎠ ⎪
⎩ ⎭
Finite element analysis of stresses in beam structures 24

4.2.2 COORDINATE CHANGE

When assembling the global stiffness matrix and the global load vector, it is compulsory that
elementary components of both are expressed in the same coordinate system, which is
arbitrary and will be called the global coordinate system.

The degrees of freedom and element equations of a three dimensional frame element in
element coordinate system x′, y ′, z ′ (see Fig. 4.2) are identical to those of a beam element
under axial loading, twisting and biaxial bending (see Eq. 4.28). The element equations in
element coordinate system can thus be written as

F ′e = K ′e a′e − R ′e , (4.41)

where K ′e and R′e are the element stiffness matrix and the element load vector of a beam
element under axial loading, twisting and biaxial bending.

z
k
i
x
z′
n
j Global coordinate 1
system l Element coordinate

m system

y y′ 2

x′

Fig. 4.2: Global coordinate system and element coordinate system

Essential part of forming the element equations of a three dimensional frame element is
finding out the coordinate change between the global coordinate system and the element
coordinate system (Fig. 4.2). The base vectors l, m, n of the element coordinate system in
the global coordinate system are written as
Finite element analysis of stresses in beam structures 25

l = lx i + l y j + lz k ,
m = mx i + m y j + m z k , (4.42)
n = nx i + n y j + n z k .

In the following, the components of these base vectors will be chosen in a reasonable way.
Because the base vector l is coinciding to the axis of the element, it is obtained for its
components

x2e − x1e y2e − y1e z2e − z1e

lx = , l y = l z = , (4.43)
Le Le Le

where

Le = ( x2e − x1e )2 + ( y2e − y1e ) 2 + ( z2e − z1e )2 (4.44)

is the length of the element. The components of base vectors m and n can be selected using
different ways. Here, the global y − axis has been chosen to be vertical and directing
downwards. Consequently, the x − and z − axes are in a horizontal plane. Because beams in
civil engineering frame structures usually consist of vertical side planes and are often
symmetric with respect to a vertical plane, a reasonable choice could be to demand that the
x ', y ′ − plane should be vertical. This means, that z ′ − axis should lie in a horizontal plane or
the component nz of the base vector n should vanish. Because the length of the base vector

n is one and the base vectors n and l are orthogonal, one gets

| n |≡ nx 2 + ny2 = 1 ⇒ ny = ± 1 − nx2 . (4.45)

and

ly
nil ≡ nx lx + ny l y = 0 ⇒ nx lx ± 1 − nx2 l y = 0 ⇒ nx = . (4.46)
lx2 + l y2

Thus we have for the components of the base vector n

ly lx
nx = , ny = − , nz = 0 . (4.47)
lx2 + l y2 lx2 + l y2

The base vector m can be obtained as a cross-product

Finite element analysis of stresses in beam structures 26

m = n × l = (nx i + n y j) × (l x i + l y j + l z k ) = n y lz i − nx l z j + ( nx l y − n y l x )k . (4.48)

Thus its components are

m x = n y l z , m y = − n x l z , m z = nx l y − n y l x . (4.49)

Equations (4.43), (4.47) and (4.49) in this order, can now be used for calculating the
components of the base vectors l , m and n .

In case that the frame contains vertical beams, base vector n of these beams must be
obtained in a different way. It has been decided that in this case, vector n is going to be the
same as its counterpart in the adjacent non-vertical beam. Vector m can be obtained using
Eq. (4.49).

A typical displacement vector u and rotation vector ϕ in element coordinates and global
coordinates are written as

u = u ′l + v′m + w′n = ui + vj + wk (4.50)

and

ϕ = ϕ x′ l + ϕ ′y m + ϕ z′n = ϕ x i + ϕ y j + ϕ z k (4.51)

Multiplying equations (4.50) and (4.51) by base vectors l , m and n results to

⎫ ϕ x′ = lxϕ x + l yϕ y + lzϕ z
u ′ = lx u + l y v + lz w ⎫
⎪ ⎪
v′ = mx u + my v + mz w⎬ , ϕ ′y = mxϕ x + m yϕ y + mzϕ z ⎬ . (4.52)
⎪ ⎪
w′ = nx u + ny v + nz w ⎭ ϕ z′ = nxϕ x + n yϕ y + nzϕ z ⎭

Applying these relations to the nodal deflections and rotations at both nodes of the element
results to equation

a ′e = T e a e , (4.53)

where

⎧a1′e ⎫ ⎧ a1e ⎫
⎪ ⎪ ⎪ ⎪
a′ e = ⎨ ⎬ , a e = ⎨ ⎬ (4.54)
⎪a′e ⎪ ⎪a e ⎪
⎩ 12 ⎭ ⎩ 12 ⎭
Finite element analysis of stresses in beam structures 27

are the element nodal displacements degrees of freedom of in element coordinates and
global coordinates, respectively, and

⎡ lx ly lz 0 0 0 0 0 0 0 0 0⎤
⎢m my mz 0 0 0 0 0 0 0 0 0 ⎥⎥
⎢ x
⎢ nx ny nz 0 0 0 0 0 0 0 0 0⎥
⎢ ⎥
⎢0 0 0 lx ly lz 0 0 0 0 0 0⎥
⎢0 0 0 mx my mz 0 0 0 0 0 0⎥
⎢ ⎥
⎢0 0 0 nx ny nz 0 0 0 0 0 0⎥
T =⎢
e
(4.55)
0 0 0 0 0 0 lx ly lz 0 0 0⎥
⎢ ⎥
⎢0 0 0 0 0 0 mx my mz 0 0 0⎥
⎢0 0 0 0 0 0 nx ny nz 0 0 0⎥
⎢ ⎥
⎢0 0 0 0 0 0 0 0 0 lx ly lz ⎥
⎢ ⎥
⎢0 0 0 0 0 0 0 0 0 mx my mz ⎥
⎢0 0 0 0 0 0 0 0 0 nx ny nz ⎥⎦
⎣

is the corresponding transformation matrix.

The expression of the virtual work of a three dimensional frame element in element
coordinate system is

δ Wˆ = δ a′eT (−K ′ea′e + R ′e + F′e ) . (4.56)

Using equation (4.53) for both real and virtual displacement degrees of freedom, the
expression (4.56) of the virtual work of the element can be expressed in global coordinates

δ Wˆ = δ a eT (−K ea e + R e + F e ) , (4.57)

where

K e = TeT K ′e Te (4.58)

and

R e = TeT R′e (4.59)

are the stiffness matrix and load vector, respectively, in global coordinate system. These are
the final element stiffness matrix and element load vector of a three dimensional frame
element.
Finite element analysis of stresses in beam structures 28

4.2.3 GLOBAL SYSTEM

Once element stiffness matrices (Eq. 4.39) and element load vectors (Eq. 4.40) are
assembled using the corresponding element transformation matrix (Eq. 4.55), the equation to
solve is

Ka = R , (4.60)

where K is the generalized stiffness matrix, a is the generalized displacement vector and
R the generalized load vector.

4.2.4 STRESS RESULTANTS

After the generalized displacements aie , i = 1, ,12 of the element have been determined,

the corresponding generalized forces Fie , i = 1, ,12 can be calculated using the element
stiffness equation (4.28). In terms of these generalized forces the stress resultants at the
element nodes are

N1e = − F1e , Qye1 = − F2e , Qze1 = − F3e , M te1 = − F4e , M ye1 = − F5e , M ze1 = F6e ,
(4.61)
N 2e = F7e , Qye 2 = F8e , Qze2 = F9e , M te2 = F10e , M ye 2 = F11e , M ze2 = − F12e .

It can be shown that, in connection with elements of uniform bending stiffness (constant De ),
these nodal values are exact. Using these values and standard equilibrium consideration it is
possible to construct stress resultant distributions within each element, which are also exact.
Here, however, it is sufficient with these element nodal values because it is possible to
choose so dense finite element grid that by connecting these nodal values by a straight line
one is able to describe the stress resultant distribution of any beam with sufficient accuracy.

4.3 STRESS DISTRIBUTION

Using the superposition principle, stress values in any cross-section can be obtained from
the sum of stress values due to each stress resultant separately.

Some basic cross-sectional properties are needed to solve the problem, which are going to
be calculated first [19].
Finite element analysis of stresses in beam structures 29

4.3.1 CROSS-SECTIONAL PROPERTIES

Typical cross-sectional properties of a composite beam are defined as follows. Weighted

area of the cross-section is

( EA) = ∫ EdA . (4.62)

A

Weighted first moments of area are

( ES ) y = ∫ EzdA, (ES ) z = ∫ EydA. (4.63)

A A

Bending stiffness’s or weighted second moments of area are

( EI ) y = ∫ Ez 2 dA, (EI ) z = ∫ Ey 2 dA, (EI ) yz = ∫ EyzdA . (4.64)

A A A

The coordinates of the centre of weighted area of the cross-section are obtained from

( ES ) z ( ES ) y
yC = , zC = . (4.65)
( EA) ( EA)

The centre of cross-section of a composite beam is this centre C of weighted area. The origin
of the cross-sectional coordinate system x, y , z is located in this point and the axis of the
beam is the locus of these centre points.

If the geometry of the cross-section is complicated, one possibility of determining the cross-
sectional properties (4.62), (4.63) and (4.64), is to use finite element methodology. The
cross-section is divided into N finite elements. The procedure is described here briefly in
connection with the weighted product moment (Eq. 4.64c). One can write

N N
( EI ) yz = ∫ EyzdA = ∑ ∫ E e yzdA = ∑ ( EI )eyz , (4.66)
A e=1 Ae e=1

e
where E e is the modulus of elasticity and ( EI ) yz is the weighted product moment of element

e . Thus the weighted product moment ( EI ) yz is obtained as a sum of the weighted product
Finite element analysis of stresses in beam structures 30

e
moments ( EI ) yz of the elements. By changing the integration variables one gets, for the

e
weighted product moment ( EI ) yz ,

1 1
( EI )eyz = ∫E yzdA = ∫ ∫E y (ξ ,η ) z (ξ ,η ) det J (ξ ,η )d ξ dη
e e
(4.67)
Ae −1 − 1

and applying numerical integration further

n
( EI )eyz ≈ ∑ wk E e y (k ) z (k ) det J (k ), (4.68)
k =1

where the notation (k ) = (ξ k ,ηk ) for the value of a function at integration point k is used. This

e
is the final formula for the weighted product moment ( EI ) yz of element e . In order to use this

formula one must first calculate the coordinates of the integration points y ( k ) , z ( k ) (Eq.

3.12), the derivatives y ,ξ ( k ) , z ,ξ ( k ) , y ,η ( k ) , z ,ξ ( k ) (Eq. 3.17) and Jacobian determinants

det J (k ) (Eq. 3.22) at the integration points.

The rest of properties into finite element terms can be obtained in a similar manner.

4.3.2 TENSION / COMPRESSION

In beam theory, a beam under tension or compression only produces normal stress in the
axial direction, σ x . Using Hooke’s law, this stress is

σ x = Eε x , (4.69)

where ε x is the axial strain of the beam.

Using the definition of the axial force N x and (Eq. 4.62) one obtains

N x = ∫ σ x dA = ∫ Eε x dA = ( EA)ε x , (4.70)
A A

which results to
Finite element analysis of stresses in beam structures 31

Nx
εx = . (4.71)
( EA)

Substituting (Eq. 4.71) into (Eq. 4.69) gives the standard expression of the normal stress

EN x
σx = . (4.72)
( EA)

Element nodal values of this, σ xie , are obtained simply using

Ee Nx
σ =
e
xi i = 1,..., m , (4.73)
( EA)

where E e is the modulus of elasticity of the element and m is the number of nodes in the
element.

4.3.3 BIAXIAL BENDING

Fig. 4.3: Non-symmetric cross-section of a beam

Consider pure biaxial bending of a straight uniform beam with non-symmetric cross-section
(Fig. 4.3). A part of the beam will be considered, where it is assumed that the rotations ϕ y ( x )

and ϕ z ( x) are quadratic function of the axial coordinate x. Thus the curvatures

κ y ( x) = −ϕ ′y ( x) and κ z ( x) = −ϕ z′ ( x) of the beam are linear and their derivatives κ ′y ( x) and

κ z′ ( x) , which are denoted k y and kz , are constants.

For the axial displacement of the beam, the expression (4.74) is assumed
Finite element analysis of stresses in beam structures 32

u ( x, y, z ) = −ϕ y ( x) z − ϕ z ( x) y + k yψ y ( y, z ) + k zψ z ( y, z ) , (4.74)

where the functions ψ y ( y, z ) and ψ z ( y, z ) are called warping functions.

In this case, the axial strain of the beam is

∂u
εx ≡ = −ϕ z′ ( x) y − ϕ ′y ( x) z = κ z ( x) y + κ y ( x) z . (4.75)
∂x

Because in beam theory it is assumed that the transverse stress components σ y and σ z

disappear, one gets, based on the generalized Hooke’s law, for the transverse strains of the
beam the results

ε y = ε z ≡ −νε x = −νκ z ( x) y −νκ y ( x) z . (4.76)

Using the strain and displacement relations, one gets, for the transverse displacements v
and w , differential equations

∂v ∂w
= ε y = −ν (κ z y + κ y z ), = ε z = −ν (κ z y + κ y z ) . (4.77)
∂y ∂z

In these equations the dependency of the curvature on coordinate x has been left out for
clarity. In the following derivation it is assumed that the Poisson’s ratio v of the cross-section
is constant. This assumption naturally holds if the beam is made of one material. In a
composite beam, however, the Poisson’s ratios of the materials of the cross-section must be
equal. Here it is expected that the derived equations could, however, be used in composite
beams, if the Poisson’s ratios of the materials do not deviate too much. These equations can
be solved for the transverse displacements resulting to

v( x, y, z ) = v0 ( x) − νκ z ( x) ( y 2 − z 2 ) −νκ y ( x) yz ,
1
2 (4.78)
w( x, y, z ) = w0 ( x) −νκ z ( x) yz − νκ y ( x) ( z 2 − y 2 ) .
1
2

The expressions (4.74) and (4.78) now form the displacement assumption of the beam.

For the shear strains one obtains

Finite element analysis of stresses in beam structures 33

⎡ ∂ψ y 1 ⎤ ⎛ ∂ψ z ⎞
γ xy = k z ⎢ − ν ( y 2 − z 2 )⎥ + k y ⎜ −ν yz ⎟ ,
⎣ ∂y 2 ⎦ ⎝ ∂y ⎠
(4.79)
⎡ ∂ψ ⎤ ⎛ ∂ψ ⎞
γ xz = k y ⎢ z − ν ( z 2 − y 2 ) ⎥ + k z ⎜ y −ν yz ⎟ .
1
⎣ ∂z 2 ⎦ ⎝ ∂z ⎠

For the normal stress one obtains

σ x = Eε x = E (κ z y + κ y z ) = 2G (1 + ν )(κ z y + κ y z ) (4.80)

and for the shear stresses

⎡ ∂ψ y 1 ⎤ ⎛ ∂ψ z ⎞
τ xy ≡ Gγ xy = k z G ⎢ − ν ( y 2 − z 2 )⎥ + k yG ⎜ −ν yz ⎟ ,
⎣ ∂y 2 ⎦ ⎝ ∂y ⎠
(4.81)
⎡ ∂ψ ⎤ ⎛ ∂ψ ⎞
τ xz ≡ Gγ xz = k y G ⎢ z − ν ( z 2 − y 2 ) ⎥ + k z G ⎜ y −ν yz ⎟ .
1
⎣ ∂z 2 ⎦ ⎝ ∂z ⎠

The axial equilibrium equation of the stress components, if volume forces are omitted, is

∂σ x ∂τ xy ∂τ xz
+ + =0. (4.82)
∂x ∂y ∂z

Using the expressions (4.80) and (4.81) and the relation

E = 2G (1 + ν ) , (4.83)

the equilibrium equation (4.82) gets first the form

∂ ⎪⎧ ⎡ ∂ψ y 1 ⎤ ⎛ ∂ψ z ⎞ ⎫⎪
⎨k z G ⎢ − ν ( y 2 − z 2 )⎥ + k yG ⎜ −ν yz ⎟ ⎬
∂y ⎪⎩ ⎣ ∂y 2 ⎦ ⎝ ∂y ⎠ ⎪⎭
∂ ⎪⎧ ⎡ ∂ψ z 1 ⎤ ⎛ ∂ψ y ⎞ ⎫⎪
+ ⎨k y G ⎢ − ν ( z 2 − y 2 )⎥ + kzG ⎜ −ν yz ⎟ ⎬ (4.84)
∂z ⎪⎩ ⎣ ∂z 2 ⎦ ⎝ ∂z ⎠ ⎪⎭
+ k z 2G (1 +ν ) y + k y 2G (1 + ν ) z} = 0 in A

and further
Finite element analysis of stresses in beam structures 34

⎧⎪ ∂ ⎪⎧ ⎡ ∂ψ ⎤ ⎫⎪ ∂ ⎡ ⎛ ∂ψ y ⎞⎤ ⎪⎫
k z ⎨ ⎨G ⎢ y − ν ( y 2 − z 2 ) ⎥ ⎬ + ⎢G ⎜
1
−ν yz ⎟ ⎥ + 2G (1 + ν ) y ⎬
⎩⎪ ∂y ⎪⎩ ⎣ ∂y 2 ⎦ ⎪⎭ ∂z ⎣ ⎝ ∂z ⎠⎦ ⎭⎪
(4.85)
⎪⎧ ∂ ⎧ ⎡ ∂ψ ⎤ ⎫ ∂ ⎡ ⎛ ∂ψ z ⎞⎤ ⎪⎫
+ k y ⎨ ⎨G ⎢ z − ν ( z 2 − y 2 ) ⎥ ⎬ + ⎢G ⎜
1
−ν yz ⎟ ⎥ + 2G (1 + ν ) z ⎬ = 0 in A.
⎪⎩ ∂z ⎩ ⎣ ∂z 2 ⎦ ⎭ ∂y ⎣ ⎝ ∂y ⎠⎦ ⎪⎭

Equation (4.85) holds if the coefficients of constants k z and k y disappear. Thus, two

separate differential equations for the warping functions ψ y ( y, z ) and ψ z ( y, z ) are obtained

∂ ⎧⎪ ⎡ ∂ψ y 1 ⎤ ⎫⎪ ∂ ⎡ ⎛ ∂ψ y ⎞⎤
⎨G ⎢ − ν ( y 2 − z 2 ) ⎥ ⎬ + ⎢G ⎜ −ν yz ⎟ ⎥ + 2G (1 + ν ) y = 0 in A, (4.86a)
∂y ⎪⎩ ⎣ ∂y 2 ⎦ ⎪⎭ ∂z ⎣ ⎝ ∂z ⎠⎦

∂ ⎧ ⎡ ∂ψ z 1 ⎤ ⎫ ∂ ⎡ ⎛ ∂ψ z ⎞⎤
⎨G ⎢ − ν ( z 2 − y 2 ) ⎥ ⎬ + ⎢G ⎜ −ν yz ⎟ ⎥ + 2G (1 +ν ) z = 0 in A. (4.86b)
∂z ⎩ ⎣ ∂z 2 ⎦ ⎭ ∂y ⎣ ⎝ ∂y ⎠⎦

The boundary condition of the shear stresses on the unloaded boundary of the cross-section
is

n yτ xy + nzτ xz = 0 . (4.87)

Using the expressions (4.81) the boundary condition gets first the form

⎧⎪ ⎡ ∂ψ ⎤ ⎛ ∂ψ z ⎞ ⎫⎪
n y ⎨k z G ⎢ y − ν ( y 2 − z 2 ) ⎥ + k y G ⎜
1
−ν yz ⎟ ⎬
⎩⎪ ⎣ ∂y 2 ⎦ ⎝ ∂y ⎠ ⎭⎪
(4.88)
⎪⎧ ⎡ ∂ψ ⎤ ⎛ ∂ψ y ⎞ ⎫⎪
+ nz ⎨ k y G ⎢ z − ν ( z 2 − y 2 ) ⎥ + k z G ⎜
1
−ν yz ⎟ ⎬ = 0 on s,
⎪⎩ ⎣ ∂z 2 ⎦ ⎝ ∂z ⎠ ⎪⎭

and further

⎪⎧ ⎡ ∂ψ ⎤ ⎛ ∂ψ y ⎞ ⎫⎪
k z ⎨ n y G ⎢ y − ν ( y 2 − z 2 ) ⎥ + nz G ⎜
1
−ν yz ⎟ ⎬
⎩⎪ ⎣ ∂y 2 ⎦ ⎝ ∂z ⎠ ⎭⎪
(4.89)
⎧ ⎛ ∂ψ z ⎞ ⎡ ∂ψ ⎤⎫
−ν yz ⎟ + nz G ⎢ z − ν ( z 2 − y 2 ) ⎥ ⎬ = 0 on s.
1
+ k y ⎨n y G ⎜
⎩ ⎝ ∂y ⎠ ⎣ ∂z 2 ⎦⎭
Finite element analysis of stresses in beam structures 35

Equation (4.89) holds if the coefficients of constants k z and k y disappear. Thus, two

separate boundary condition equations for the warping functions ψ y ( y, z ) and ψ z ( y, z ) are

obtained

⎡ ∂ψ ⎤ ⎛ ∂ψ y ⎞
n y G ⎢ y − ν ( y 2 − z 2 ) ⎥ + nz G ⎜
1
−ν yz ⎟ = 0 on s, (4.90a)
⎣ ∂y 2 ⎦ ⎝ ∂z ⎠

⎛ ∂ψ z ⎞ ⎡ ∂ψ ⎤
−ν yz ⎟ + nz G ⎢ z − ν ( z 2 − y 2 ) ⎥ = 0 on s.
1
ny G ⎜ (4.90b)
⎝ ∂y ⎠ ⎣ ∂z 2 ⎦

Equations (4.86a) and (4.90a), and (4.86b) and (4.90b) form boundary value problems for
solving the warping functions ψ y ( y, z ) and ψ z ( y, z ) , respectively. They can be solved using

the finite element method.

On the other hand, using the definition of the bending moments one obtains

M y = ∫ σ x zdA = ∫ E (κ y z 2 + κ z yz )dA =( EI ) y κ y + ( EI ) yz κ z ,
A A
(4.91)
M z = ∫ σ x ydA = ∫ E (κ y yz + κ z y 2 )dA =( EI ) yz κ y + ( EI ) z κ z ,
A A

Equations (4.91) hold also when the cross-section is non-homogeneous. These can be
solved for the curvatures resulting to

( EI ) z M y − ( EI ) yz M z ( EI ) y M z − ( EI ) yz M y
κy = , κz = . (4.92)
( EI ) y ( EI ) z − ( EI ) 2
yz ( EI ) y ( EI ) z − ( EI ) 2yz

Differentiating these equations with respect to x and using the equilibrium equations
Qy = M z′ and Qz = M ′y gives

( EI ) z Qz − ( EI ) yz Qy ( EI ) y Qy − ( EI ) yz Qz
ky = , kz = . (4.93)
( EI ) y ( EI ) z − ( EI ) 2
yz ( EI ) y ( EI ) z − ( EI ) 2yz

Equations (4.92) and (4.93) can be used for determining the curvatures and their derivatives,
when the bending moments and shearing forces of the cross-section are known.
Finite element analysis of stresses in beam structures 36

Weak form of the boundary value problem of equations on ψ y (4.86a) and (4.90a) is derived

in the following. Multiplying both sides of equation (4.86a) by test function δψ y and

integrating over the domain A (cross-section) gives

⎧⎪ ∂ ⎪⎧ ⎡ ∂ψ y 1 ⎤ ⎪⎫ ∂ ⎡ ⎛ ∂ψ y ⎞⎤ ⎪⎫
∫ δψ ⎨ ⎨G ⎢ − ν ( y 2 − z 2 ) ⎥ ⎬ + ⎢G ⎜ −ν yz ⎟ ⎥ + 2G (1 + ν ) y ⎬dA = 0 (4.94)
⎪⎩ ∂y ⎩⎪ ⎣ ∂y 2 ⎦ ⎭⎪ ∂z ⎣ ⎝ ∂z
y
A ⎠⎦ ⎪⎭

and

⎧⎪ ∂ ⎪⎧ ⎡ ∂ψ y 1 ⎤ ⎪⎫ ∂ ⎡ ⎛ ∂ψ y ⎞ ⎤ ⎪⎫
∫ δψ y ⎨ ⎨G ⎢ − ν ( y 2 − z 2 ) ⎥ ⎬ + ⎢G ⎜ −ν yz ⎟ ⎥ ⎬dA + ∫ δψ y 2G (1 +ν ) ydA = 0 .(4.95)
A ⎩⎪ ∂y ⎩⎪ ⎣ ∂y 2 ⎦ ⎭⎪ ∂z ⎣ ⎝ ∂z ⎠ ⎦ ⎭⎪ A

Applying integration by parts gives further

⎧⎪ ⎡ ∂ψ y 1 ⎤ ⎛ ∂ψ y ⎞ ⎫⎪
∫ δψ y ⎨nyG ⎢ − ν ( y 2 − z 2 ) ⎥ + nz G ⎜ −ν yz ⎟ ⎬ ds
s ⎪⎩ ⎣ ∂y 2 ⎦ ⎝ ∂z ⎠ ⎪⎭
(4.96)
⎪ δψ y ⎡ ∂ψ y 1
⎧∂ ⎤ ∂δψ y ⎛ ∂ψ y ⎞ ⎫⎪
−∫ ⎨ G⎢ − ν ( y 2 − z 2 )⎥ + G⎜ −ν yz ⎟ ⎬dA + ∫ δψ y 2G (1 + ν ) ydA = 0.
A⎪⎩ ∂y ⎣ ∂y 2 ⎦ ∂z ⎝ ∂z ⎠ ⎪⎭ A

Based on the boundary condition (4.90a), the first term of equation (4.96) vanishes and it
gets the form

⎛ ∂δψ y ∂ψ y ∂δψ y ∂ψ y ⎞ ⎡ ∂δψ y 1 ∂δψ y ⎤

∫A ⎝ ∂y ∂y ∂z ∂z ⎠ ∫A ⎣
G ⎜ + ⎟dA = G ⎢δψ y (1 + ν ) y +
∂y 2
ν ( y 2
− z 2
) +
∂z
ν yz ⎥dA. (4.97)
⎦

This is the weak form of the boundary value problem of differential equations (4.86a) and
(4.90a). The weak form corresponding to differential equations (4.86b) and (4.90b) can be
obtained similarly.

The functions ψ y and δψ y can be expressed using finite element approximations

M M
ψˆ y = ∑ N iψ yi , δψˆ y = ∑ N iδψ yi . (4.98)
i =1 i =1

By substituting these into the weak form (4.97) gives

Finite element analysis of stresses in beam structures 37

M ⎧⎪ M ⎛ ∂N i ∂N j ∂N i ∂N j ⎞
∑ δψ ⎨⎪∑ ∫ ⎜ G
yi
∂y ∂y
+
∂z ∂z ⎠
⎟dAψ y j
i =1 ⎩ ⎝ j =1 A
(4.99)
⎡ ∂N ∂N ⎤ ⎫
− ∫ ⎢ 2G (1 +ν ) yN i + G ν ( y 2 − z 2 ) i + Gν yz i ⎥dA⎬ = 0
1
A⎣
2 ∂y ∂z ⎦ ⎭

or

M ⎛ M ⎞
∑ δψ ⎜ ∑ K ψyi ij yj − Ryi ⎟ = 0, (4.100)
i =1 ⎝ j =1 ⎠

where

Kij = ∫ G( Ni , y N j , y + Ni , z N j , z )dA , (4.101)

A

and

⎡ ⎤
Ryi = ∫ G ⎢ 2(1 +ν ) yNi + ν ( y 2 − z 2 ) Ni , y +ν yzN i , z ⎥dA ,
1
(4.102)
A ⎣ 2 ⎦

where typical notation Ni , y = ∂Ni / ∂y and N i , z = ∂N i / ∂z is used for the derivatives of the

shape functions. Equation (4.100) holds with arbitrary nodal values δψ yi of the test function,

only if the terms in parentheses will vanish with all values of i. Thus we get, corresponding to
the weak form (4.97), the finite element equations

∑K ψ
j =1
ij yj = Ryi , i = 1, ,M. (4.103)

Proceeding in the same manner with equations on ψ z (4.86b) and (4.90b), one obtains

similar finite element equations

∑K ψ
j =1
ij zj = Rzi , i = 1, ,M, (4.104)

where

⎡ ∂N 1 ∂N ⎤
Rzi = ∫ G ⎢ 2(1 + ν ) zN i + ν yz i + ν ( z 2 − y 2 ) i ⎥dA . (4.105)
A ⎣ ∂y 2 ∂z ⎦
Finite element analysis of stresses in beam structures 38

Equations (4.103) and (4.104) are, in matrix form,

Ka y = R y , Ka z = R z , (4.106)

where

⎧ ψ y1 ⎫ ⎧ ψ z1 ⎫
⎪ ⎪ ⎪ ⎪
ay = ⎨ ⎬ , az = ⎨ ⎬ (4.107)
⎪ψ ⎪ ⎪ψ ⎪
⎩ yM ⎭ ⎩ zM ⎭

are the vectors of unknown nodal values of the warping functions.

e
The elements of the element matrix K e and element vectors R y and R ez corresponding to

the elements of the system matrix (4.101) and system vectors (4.102) and (4.105) are

K ije = ∫ G e ( N ie, y N ej , y + N ie, z N ej , z )dA (4.108)

e
A

and

⎡ ⎤
Ryie = ∫ G e ⎢ 2(1 + ν ) yN ie + ν ( y 2 − z 2 ) N ie, y + ν yzN ie, z ⎥dA,
1
(4.109a)
Ae ⎣ 2 ⎦

⎡ ⎤
Rzie = ∫ G e ⎢ 2(1 + ν ) zN ie + ν yzN ie, y + ν ( z 2 − y 2 ) N ie, z ⎥dA.
1
(4.109b)
Ae ⎣ 2 ⎦

e
In these expressions, the shear modulus has been assumed to be constant G within the
element, which is a reasonable assumption. The Poisson’s ratio should be constant within
the whole cross-section.

By changing integration variable to natural coordinates and applying numerical integration,

expressions (4.108) and (4.109) can be reduced to

n
K ije = ∑ wk G e ⎡⎣ N ie, y (k ) N ej , y (k ) + N ie, z (k ) N ej , z (k ) ⎤⎦ det J (k ), (4.110)
k =1

and
Finite element analysis of stresses in beam structures 39

n
⎧ 1 ⎫
Ryie = ∑ wk G e ⎨2(1 + ν ) y (k ) N ie (k ) + ν ⎡⎣ y (k ) 2 − z (k ) 2 ⎤⎦ N ie, y (k ) + ν y (k ) z (k ) N ie, z ( k ) ⎬ det J (k )
k =1 ⎩ 2 ⎭
,
e ⎧ ⎫
n
1
Rzi = ∑ wk G ⎨2(1 + ν ) z (k ) N i (k ) + ν ⎡⎣ z (k ) − y (k ) ⎤⎦ N i , z (k ) + ν y (k ) z (k ) N i , y (k ) ⎬ det J (k )
e e 2 2 e e

k =1 ⎩ 2 ⎭
(4.111)

where n is the number of integration points. These equations can be solved with properties
described before in chapter 3.

At this point, stresses due to biaxial bending in a cross-section can be obtained. Substituting
the curvature of equation (4.92) into equation (4.80) gives the standard expression of the
normal stress. Its nodal values are obtained simply using

σ xie = E e (κ z yie + κ y zie ) i = 1,..., m, (4.112)

where curvatures κ y and κ z can be found in (Eq. 4.92) and yie and zie are the y − and z −

coordinates of the element nodes, respectively.

Substituting the derivative of the curvature of equation (4.93) into equation (4.81) gives
improved expressions for the shear stresses. The element nodal values of the shear stresses
are obtained by first calculating their values at the integration points and then using

⎧ 1 ⎫
τ xy (k ) = k z G e ⎨ψ y , y (k ) − ν ⎣⎡ y (k ) 2 − z (k )2 ⎦⎤ ⎬ + k y G e ⎡⎣ψ z , y (k ) −ν y (k ) z (k ) ⎤⎦ ,
⎩ 2 ⎭
(4.113)
⎧ 1 ⎫
τ xz (k ) = k y G ⎨ψ z , z (k ) − ν ⎣⎡ z (k ) 2 − y (k ) 2 ⎦⎤ ⎬ + k z G e ⎡⎣ψ y , z (k ) −ν y (k ) z (k ) ⎤⎦ .
e

⎩ 2 ⎭

The derivative of the warping function at the integration points are calculated using

m m
⎫
ψˆ y , y (k ) = ∑ N i , y (k )ψ yie , ψˆ y , z (k ) = ∑ N i , z (k )ψ yie , ⎪
i =1 i =1 ⎪
m m ⎬ , k = 1, , n. (4.114)
ψˆ z , y (k ) = ∑ N i , y (k )ψ , ψˆ z , z (k ) = ∑ N i , z (k )ψ
e e ⎪
i =1
zi
i =1
zi
⎪⎭

The element nodal values of the shear stresses τ xyi , τ xzi , i = 1,

e e
, m are finally obtained by

extrapolating shear stresses at the integration point τ xy (k ), τ xz (k ), k = 1, , n from the

integration points to the element nodes. This procedure is explained in the appendix.
Finite element analysis of stresses in beam structures 40

4.3.3.1 S HEAR CENTRE

Shear centre S (see Fig. 4.3) is the point of action of the resultant of the shear stresses
caused by pure bending. For determining its coordinates yS and zS one can write

−Qy zS + Qz yS = ∫ (−τ xy z + τ xz y )dA , (4.115)

A

Substituting the shear stresses (4.81) into equation (4.115) first gives

−Qy zS + Qz yS = k y (GJ )ψ z − k z (GJ )ψ y (4.116)

where

⎡ ∂ψ ∂ψ z ⎤
z + ν ( yz 2 + y 3 ) ⎥ dA,
1
(GJ )ψ z = ∫ G ⎢ z y −
A ⎣ ∂z ∂y 2 ⎦
(4.117)
⎡ ∂ψ ∂ψ y ⎤
y + ν ( y 2 z + z 3 ) ⎥ dA.
1
(GJ )ψ y = ∫ G ⎢ y z −
A ⎣ ∂y ∂z 2 ⎦

Substituting further the curvature derivatives (4.93) into equations (4.116) gives

( EI ) y (GJ )ψ y + ( EI ) yz (GJ )ψ z ( EI ) z (GJ )ψ z + ( EI ) yz (GJ )ψ y

−Qy zS + Qz yS = −Qy + Qz . (4.118)
( EI ) y ( EI ) z − ( EI ) 2yz ( EI ) y ( EI ) z − ( EI ) 2yz

Because these equations must hold for all values of the shear forces Qy and Qz , we get for

the coordinates of the shear centre the equations

( EI ) z (GJ )ψ z + ( EI ) yz (GJ )ψ y ( EI ) y (GJ )ψ y + ( EI ) yz (GJ )ψ t

yS = , zS = . (4.119)
( EI ) y ( EI ) z − ( EI ) 2
yz ( EI ) y ( EI ) z − ( EI ) 2yz

The quantities (GJ )ψ y and (GJ )ψ z are cross-sectional properties like bending stiffness. In

connection with isoparametric elements the element contributions of (GJ )ψ y and (GJ )ψ z are

calculated using formulas

n
⎡ 1 1 ⎤
(GJ )ψe y = ∑ wk G e ⎢ψˆ y , y (k ) z (k ) −ψˆ y , z ( k ) y (k ) + ν y (k ) 2 z ( k ) + ν z (k )3 ⎥ det J (k ),
k =1 ⎣ 2 2 ⎦
(4.120)
n
⎡ 1 1 ⎤
(GJ )ψe z = ∑ wk G e ⎢ψˆ z , y (k ) y (k ) −ψˆ z , z (k ) z (k ) + ν y ( k ) z (k ) 2 + ν y ( k )3 ⎥ det J (k ).
k =1 ⎣ 2 2 ⎦
Finite element analysis of stresses in beam structures 41

Quantities (GJ )ψ y and (GJ )ψ z are obtained by summing their element contributions

E E
(GJ )ψ y = ∑ (GJ )ψe y , (GJ )ψ z = ∑ (GJ )ψe z . (4.121)
e =1 e =1

The product of transverse external load and the distance between its application point and
the shear centre must be included into the frame analysis as an additional external twisting
moment.

4.3.4 TORSION

Fig. 4.4: Twisted cross-section

Consider pure torsion of a straight uniform beam (Fig. 4.4). A part of the beam, where it is
assumed that the angle of twist ϕt ( x ) is a linear function of the axial coordinate x, is

considered. Thus rate of twist per unit length of the beam θ = ϕt′ is a constant. For the axial

displacement of the beam, it can be assumed the expression

u ( x, y, z ) = θψ t ( y, z ) , (4.122)

where the function ψ t ( y, z ) is called warping function. Further it is assumed that the

projection of the cross-section in the y , z − plane rotates like a rigid plate. Thus we have for
the transverse displacement components the expressions

v( x, y, z ) = −ϕt ( x) z, w( x, y, z ) = ϕt ( x) y. (4.123)

Expressions (4.122) and (4.123) form the displacement assumptions of so-called Saint
Venant’s torsion or free torsion.
Finite element analysis of stresses in beam structures 42

For the shear strains one obtains

∂u ∂v ∂ψ t ⎛ ∂ψ ⎞
γ xy ≡ + =θ − ϕt′z = θ ⎜ t − z ⎟ ,
∂y ∂x ∂y ⎝ ∂y ⎠ (4.124)
∂u ∂w ∂ψ t ⎛ ∂ψ ⎞
γ xz ≡ + =θ + ϕt′ y = θ ⎜ t + y ⎟
∂z ∂x ∂z ⎝ ∂z ⎠

and the other strain components are zero. For the shear stresses one obtains

⎛ ∂ψ t ⎞ ⎛ ∂ψ ⎞
τ xy ≡ Gγ xy = Gθ ⎜ − z ⎟ , τ xz ≡ Gγ xz = Gθ ⎜ t + y ⎟ (4.125)
⎝ ∂y ⎠ ⎝ ∂z ⎠

and the other stress components are zero.

The axial equilibrium equation of the stress components, if volume forces are omitted, is

∂σ x ∂τ xy ∂τ xz
+ + =0. (4.126)
∂x ∂y ∂z

Using the expressions (4.125) and noting that σ x = 0 , the equilibrium equation (4.126) gets

the form

∂ ⎡ ⎛ ∂ψ t ⎞ ⎤ ∂ ⎡ ⎛ ∂ψ ⎞⎤
⎢G ⎜ − z ⎟ ⎥ + ⎢G ⎜ t + y ⎟ ⎥ = 0 in A, (4.127)
∂y ⎣ ⎝ ∂y ⎠ ⎦ ∂z ⎣ ⎝ ∂z ⎠⎦

where the area A of the cross-section has been added to remark that this equation must
hold at every point on the cross-section.

The boundary condition of the shear stresses on the unloaded boundary of the cross-section
is

n yτ xy + nzτ xz = 0 , (4.128)

where n y and nz are the components of the unit normal vector n of the boundary line of the

cross-section (Fig. 4.4). Using the expressions (4.125), it can be written as

⎛ ∂ψ ⎞ ⎛ ∂ψ ⎞
n y G ⎜ t − z ⎟ + nz G ⎜ t + y ⎟ = 0 on s , (4.129)
⎝ ∂y ⎠ ⎝ ∂z ⎠
Finite element analysis of stresses in beam structures 43

where s has been added to remark, that this equation must hold at every point on the
boundary line of the cross-section. Equations (4.127) and (4.129) form a simple elliptic
boundary value problem for determining the warping function ψ t ( y , z ) .

With the help of Fig. 4.5, one obtains for the torque

M t = ∫ (τ xz y − τ xy z )dA . (4.130)
A

Fig. 4.5: Defining the torque

Substituting the expressions (4.125) of the shear stresses into equation (4.130) gives

M t = (GI )t θ , (4.131)

which relates the torque and twisting angle per unit length. Here

∂ψ t ∂ψ t
(GI )t = ∫ G ( y− z + y 2 + z 2 )dA (4.132)
A
∂z ∂y

is the torsional stiffness of Saint Venant’s or so called free torsion. Equations (4.131) and
(4.132) apply also to the case where the cross-section is non-homogeneous (composite
beam). In the special case of a homogeneous cross-section, the torsional stiffness is
(GI )t = GI t , where

∂ψ t ∂ψ t
It = ∫ ( y− z + y 2 + z 2 )dA (4.133)
A
∂z ∂y
Finite element analysis of stresses in beam structures 44

is the torsional constant. The torsional stiffness (4.132) or torsional constant (4.133) can be
determined after the warping function ψ t ( y, z ) has been solved from the boundary value

problem of equations (4.127) and (4.129).

Equation (4.131) can now be solved for the twisting angle per unit length resulting to

Mt
θ= . (4.134)
(GI )t

Substituting this into equations (4.125) results to

GM t ⎛ ∂ψ t ⎞ GM t ⎛ ∂ψ t ⎞
τ xy = ⎜ − z ⎟ , τ xz = ⎜ + y⎟. (4.135)
(GI )t ⎝ ∂y ⎠ (GI )t ⎝ ∂z ⎠

These are final expressions of the shear stresses corresponding to a given torque. In the
case of a homogeneous cross-section (GI )t = GI t these expressions get slightly simpler form

M t ⎛ ∂ψ t ⎞ M ⎛ ∂ψ ⎞
τ xy = ⎜ − z ⎟ , τ xz = t ⎜ t + y ⎟ . (4.136)
I t ⎝ ∂y ⎠ I t ⎝ ∂z ⎠

Using the same methodology as in chapter 4.3.3 for the case of biaxial bending, weak form
of the boundary value problem of equations (4.127) and (4.129) is

⎛ ∂δψ t ∂ψ t ∂δψ t ∂ψ t ⎞ ⎛ ∂δψ t ∂δψ t ⎞

∫ G ⎜⎝
A
∂y ∂y
+
∂z ∂z ⎠
⎟dA = ∫ G ⎜
A ⎝ ∂y
z−
∂z
y ⎟dA
⎠
(4.137)

The finite element equations corresponding to the weak form of equation (4.137) are

∑K ψ
j =1
ij tj = Ri , (4.138)

where

⎛ ∂N ∂N j ∂Ni ∂N j ⎞
Kij = ∫ G ⎜ i + ⎟dA (4.139)
A ⎝ ∂y ∂y ∂z ∂z ⎠

and
Finite element analysis of stresses in beam structures 45

⎛ ∂N ∂N ⎞
Ri = ∫ G ⎜ i z − i y ⎟dA . (4.140)
A ⎝ ∂y ∂z ⎠

Equations (4.138) in matrix form are

Ka = R . (4.141)

Equations (4.139) and (4.140) are expressions of the elements of the system matrix K and
system vector R and

⎧ ψ t1 ⎫
⎪ ⎪
a =⎨ ⎬ (4.141)
M ×1 ⎪ ⎪
⎩ψ tM ⎭

is the vector of unknown nodal values of the warping function.

e e
The elements of the element matrix K and element vector R corresponding to the system
matrix and system vector are

K ije = ∫ G e ( N ie, y N ej , y + N ie, z N ej , z )dA, (4.142)

Ae

and

Rie = ∫ G (N z − N ie, z y )dA .

e e
i, y (4.143)
e
A

Final equations for calculating the elements of the element matrix and element vector using
finite element methodology are

n
K ije = ∑ wk G e [( N ie, y (k ) N ej , y ( k ) + N ie, z ( k ) N ej , z ( k )]det J ( k ), (4.144)
k =1

and

n
Rie = ∑ wk G e [ N ie, y (k ) z (k ) − N ie, z (k ) y (k )]det J (k ) . (4.145)
k =1
Finite element analysis of stresses in beam structures 46

e e
In these formulas, wk is the weight of integration point k, Ni , y (k ) and Ni , z (k ) are the values

of the shape function derivatives on node i at integration point k and n is the number of
integration points.

The torsional stiffness of the cross-section is obtained as a sum of the torsional stiffnesses of
the elements, or

E
(GI )t = ∑ (GI )te (4.146)
e =1

where

(GI )te = ∫ G e (ψ t , z y −ψ t , y z + y 2 + z 2 )dA . (4.147)

A

The torsional stiffness (4.147) of element e can be calculated using equation

n
(GI )te = ∑ wk G e [ψ t , z (k ) y (k ) −ψ t , y (k ) z (k ) + y (k ) 2 + z (k ) 2 ]det J (k ). (4.148)
k =1

In equations (4.147) and (4.148), the shear modulus of the element G e is assumed to be
constant, which is a reasonable assumption in connection with composite beams.

The shear stresses (4.135) at integration point k of element e corresponding to a given

torque M t can be calculated using

GeM t GeM t
τ xy (k ) = [ψ t , y (k ) − z (k )], τ xz (k ) = [ψ t , z (k ) + y (k )]. (4.149)
(GI )t (GI )t

The element nodal values of the shear stresses τ xyi , τ xzi , i = 1,

e e
, m are finally obtained by

extrapolating shear stresses at the integration point τ xy (k ), τ xz (k ), k = 1, , n from the

integration points to the element nodes.

Finite element analysis of stresses in beam structures 47

4.3.5 GENERAL CASE

Finally, to handle the case of a cross-section under tension/compression, biaxial bending and
torsion at the same time, one just has to add each independent result to each other. Thus,
the normal stress is the sum of one obtained from tension/compression and bending
moments. It can be expressed as

EN x
σx = + E (κ z y + κ y z ) , (4.150)
( EA)

and its nodal values as

Ee Nx
σ xie = + E e (κ z yie + κ y zie ) . i = 1,..., m , (4.151)
( EA)

On the other hand, shear stresses are the sum of the results obtained from shear forces and
twisting moments, and can be expressed as

⎡ ∂ψ y 1 ⎤ ⎛ ∂ψ z ⎞ GM t ⎛ ∂ψ t ⎞
τ xy = k z G ⎢ − ν ( y 2 − z 2 )⎥ + k yG ⎜ −ν yz ⎟ + ⎜ − z ⎟,
⎣ ∂y 2 ⎦ ⎝ ∂y ⎠ (GI )t ⎝ ∂y ⎠
(4.152)
⎡ ∂ψ ⎤ ⎛ ∂ψ ⎞ GM t ⎛ ∂ψ t ⎞
τ xz = k y G ⎢ z − ν ( z 2 − y 2 ) ⎥ + k z G ⎜ y −ν yz ⎟ +
1
⎜ + y ⎟.
⎣ ∂z 2 ⎦ ⎝ ∂z ⎠ (GI )t ⎝ ∂z ⎠

Their values at integration points are

⎧ 1 ⎫
τ xy (k ) = k z G e ⎨ψ y , y (k ) − ν ⎡⎣ y (k ) 2 − z (k )2 ⎤⎦ ⎬ + k y G e ⎡⎣ψ z , y (k ) −ν y (k ) z (k ) ⎤⎦
⎩ 2 ⎭
GeM t
+ [ψ t , y (k ) − z (k )],
(GI )t
(4.153)
⎧ 1 ⎫
τ xz (k ) = k y G ⎨ψ z , z (k ) − ν ⎣⎡ z (k ) 2 − y (k ) 2 ⎦⎤ ⎬ + k z G e ⎡⎣ψ y , z (k ) −ν y (k ) z (k ) ⎤⎦
e

⎩ 2 ⎭
e
G Mt
+ [ψ t , z (k ) + y (k )].
(GI )t

The element nodal values of the shear stresses τ xyi , τ xzi , i = 1,

e e
, m are finally obtained by

extrapolating shear stresses at the integration point τ xy (k ), τ xz (k ), k = 1, , n from the

integration points to the element nodes.

Finite element analysis of stresses in beam structures 48

5 BIQUADRATIC FINITE PRISM METHOD

As said before, the methodology described until now to obtain stresses in a frame will be
compared with another one more refined: the finite prism method.

5.1 ASSUMPTIONS

This methodology will only be developed for a simply supported straight beam with uniform
cross-section. Thus the comparison between both methods must be realised in the same
conditions.

Similarly, material is elastic and isotropic, and the analysis will be linear geometrically and
materially.

Trigonometric series will be used for describing the displacements and the load can be
applied in a more realistic way, i.e., as a surface load, which is piecewise constant in axial
direction and quadratically distributed in transverse direction, volume force, which is
piecewise constant in axial direction, or distributed nodal line load, which is piecewise
constant in axial direction.

5.2 METHOD DESCRIPTION

Consider a simply supported straight beam with uniform cross-section. Let the axis of the
beam be parallel to x − axis and the cross-sectional plane parallel to yz − plane. The
displacements of the beam within a typical prism e (Fig. 5.1) are expressed using
trigonometric series as

mπ x
u ( x, y, z ) = ∑ um ( y, z ) cos ,
m L
mπ x
v( x, y, z ) = ∑ vm ( y, z ) sin , (5.1)
m L
mπ x
w( x, y, z ) = ∑ wm ( y, z ) sin ,
m L

where the coefficients um ( y, z ) , vm ( y, z ) and wm ( y, z ) are unknown functions of

coordinates y and z .
Finite element analysis of stresses in beam structures 49

Fig. 5.1: A simply supported straight beam: biquadratic finite element e on the cross-section and the
corresponding finite prism element inside the beam and a quadratic finite element e on the boundary
of the cross-section and the corresponding finite strip

Equation (5.1) can be expressed in matrix form as

u( x, y, z ) = ∑ Xum ( x)u m ( y, z ) (5.2)

m

where

⎧u ⎫ ⎧ um ⎫
⎪ ⎪ ⎪ ⎪
u = ⎨ v ⎬ , u m = ⎨ vm ⎬ (5.3)
⎪ w⎪ ⎪w ⎪
⎩ ⎭ ⎩ m⎭

and matrix

⎡ mπ x ⎤
⎢ cos L 0 0 ⎥
⎢ ⎥
mπ x
Xm ( x) = ⎢
u
0 sin 0 ⎥ (5.4)
3×3
⎢ L ⎥
⎢ mπ x ⎥⎥
⎢ 0 0 sin
⎣⎢ L ⎦⎥

contains the trigonometric terms. These unknown coefficients functions u m ( y, z ) are

e
approximated within the cross-sectional area A of the prism using finite element
approximation

uˆ m ( y, z ) = N e ( y, z )aem , (5.5)
Finite element analysis of stresses in beam structures 50

where

⎡ N1e 0 0 N9e 0 0 ⎤
⎢ ⎥
Ne = ⎢ 0 N1e 0 0 N9e 0 ⎥, (5.6)
3×27 ⎢ ⎥
⎢⎣ 0 0 N1e 0 0 N9e ⎥
⎦

Nie ( y, z ), i = 1, ,9 are biquadratic Lagrange shape functions and

⎧ ume 1 ⎫
⎪ ⎪
⎪ vme 1 ⎪
⎪ e ⎪
⎪ wm1 ⎪
⎪ ⎪
aem = ⎨ ⎬ (5.7)
27×1 ⎪ e ⎪
⎪ um 9 ⎪
⎪ ve ⎪
⎪ m9 ⎪
⎪ we ⎪
⎩ m9 ⎭

is the vector of nodal parameters corresponding to m -th term of the series. Substituting the
approximation (5.5) of the coefficient functions into the series expansion (5.2) of the
displacements gives the expression

uˆ ( x, y, z ) = ∑ Xum ( x)N e ( y, z )a em (5.8)

m

for the finite prism approximation of the displacements. In component form, these
expressions are

mπ x 9 e
uˆ ( x, y, z ) = ∑ cos ∑ e
Ni ( y, z )umi ,
m L i =1
mπ x 9 e
vˆ( x, y, z ) = ∑ sin ∑ e
Ni ( y, z )vmi , (5.9)
m L i =1
mπ x 9 e
wˆ ( x, y, z ) = ∑ sin ∑ e
Ni ( y, z ) wmi .
m L i =1

With the help of these expressions and the definitions of infinitesimal strains, one obtains
following expressions for the strain components
Finite element analysis of stresses in beam structures 51

mπ x 9 mπ e e
εˆx ≡ uˆ, x = ∑ sin ∑ − N i um ,
m L i =1 L
mπ x 9 e e
εˆ y ≡ vˆ, y = ∑ sin ∑ Ni, y vmi ,
m L i =1
mπ x 9 e e
εˆz ≡ wˆ , z = ∑ sin ∑ Ni, z wmi ,
m L i =1
(5.10)
mπ x 9
γˆ yz ≡ vˆ, z + wˆ , y = ∑ sin ∑ ( Nie, z vmi
e
+ Nie, y wmi
e
),
m L i =1
mπ x 9 mπ e e
γˆzx ≡ wˆ , x +uˆ, z = ∑ cos ∑ ( Ni wmi + Nie, z umi
e
),
m L i =1 L
mπ x 9 mπ e e
γˆxy ≡ uˆ, y +vˆ, x = ∑ cos ∑ ( Nie, y umi
e
+ Ni vmi ),
m L i =1 L

which can be expressed in matrix form as

εˆ ( x, y, z ) = ∑ Xεm ( x)Bem ( y, z ) aem , (5.11)

m

where

⎧ εˆx ⎫
⎪ εˆ ⎪
⎪ y⎪
⎪⎪ εˆz ⎪⎪
εˆ = ⎨ ⎬ , (5.12)
6×1 ⎪γˆ yz ⎪
⎪γˆ ⎪
⎪ zx ⎪
⎪⎩γˆxy ⎪⎭
Finite element analysis of stresses in beam structures 52

⎡ mπ x ⎤
⎢sin L 0 0 0 0 0 ⎥
⎢ ⎥
⎢ 0 mπ x ⎥
sin 0 0 0 0
⎢ L ⎥
⎢ mπ x ⎥
⎢ 0 0 sin 0 0 0 ⎥
X m ( x) = ⎢⎢
ε ⎥
L
(5.13)
mπ x ⎥
6×6
⎢ 0 0 0 sin 0 0 ⎥
⎢ L ⎥
⎢ mπ x ⎥
⎢ 0 0 0 0 cos 0 ⎥
⎢ L ⎥
⎢ mπ x ⎥
⎢⎣ 0 0 0 0 0 cos
L ⎥⎦

and

⎡ mπ e mπ e ⎤
⎢ − L N1 0 0 −
L
N9 0 0 ⎥
⎢ ⎥
⎢ 0 N1,e y 0 0 N9,e y 0 ⎥
⎢ ⎥
⎢ 0 0 N1,e z 0 0 e
N91, z ⎥
Bm =
e ⎢ ⎥ (5.14)
6×27 ⎢
0 N1,e z N1,e y 0 N9,e z N9,e y ⎥
⎢ ⎥
⎢ Ne mπ e mπ e ⎥
0 N1 N9,e z 0 N9
⎢ 1, z
L L ⎥
⎢ mπ e mπ e ⎥
⎢ N1,e y N1 0 N9,e y N9 0 ⎥
⎣ L L ⎦

The generalized Hooke’s law of an isotropic material is

σ = Cε , (5.15)

where

⎧σ x ⎫
⎪σ ⎪
⎪ y⎪
⎪⎪ σ z ⎪⎪
σ = ⎨ ⎬, (5.16)
⎪τ yz ⎪
⎪τ ⎪
⎪ zx ⎪
⎪⎩τ xy ⎪⎭

and
Finite element analysis of stresses in beam structures 53

⎡ 2G + λ λ λ0 0 0⎤
⎢ λ 2G + λ λ0 0 0 ⎥⎥
⎢
⎢ λ λ 2G + λ 0 0 0 ⎥
C=⎢ ⎥ (5.17)
⎢ 0 0 0 G 0 0⎥
⎢ 0 0 0 0 G 0⎥
⎢ ⎥
⎣ 0 0 0 0 0 G⎦

Typical expressions of the shear modulus G and Lame’s parameter λ in terms of the
modulus of elasticity E and Poisson’s ratio ν are

E Eν
G= , λ= . (5.18)
2(1 + ν ) (1 +ν )(1 − 2ν )

The finite prism approximation of the stresses thus gets the form

σˆ ( x, y, z ) = ∑ CXεm ( x)B em ( y, z )aem . (5.19)

m

The expression of the internal virtual work of the finite prism is

δ Wint
e
= − ∫ δεT σdV , (5.20)
e
V

where V e is the volume of the prism. Substituting the approximations of the virtual strains
and stresses into it one obtains

δ Wint
e
= − ∫ δεˆ T σˆ dV = −∑∑ δ aemT T εT ε
∫ Bm Xm CXn B n dVan
e

Ve m n Ve
(5.21)
L
= − ∑∑ δ aemT ∫ BTm ( ∫
εT ε
X m CX n dx)B n dAaen .
m n Ae 0

or

δ Wint
e
= − ∑∑ δ a emT K emn a en (5.22)
m n

where
Finite element analysis of stresses in beam structures 54

L
εT ε
K emn = ∫
e
BTm ( ∫ Xm CXn dx)Bn dA , (5.23)
A 0

e
is the element stiffness matrix of the prism. Here A is the area of the biquadratic finite
element in y , z − plane defining the strip e. The line integral in equation (5.23) gets the form

L
εT ε
∫ Xm CXn dx =
0

⎡ L
mπ x nπ x
L
mπ x nπ x
L
mπ x nπ x
⎢(2G + λ ) ∫ sin sin dx λ ∫ sin sin dx λ ∫ sin sin dx
⎢ 0
L L 0
L L 0
L L
⎢ L L L
⎢ mπ x nπ x mπ x nπ x mπ x nπ x
⎢ λ ∫ sin L sin L dx (2G + λ ) ∫ sin
L
sin
L
dx λ ∫ sin
L
sin
L
dx
⎢ 0 0 0
=⎢ L L L ...
⎢ λ sin mπ x sin nπ x dx mπ x nπ x mπ x nπ x
⎢ ∫ L L
λ ∫ sin
L
sin
L
dx (2G + λ ) ∫ sin
L
sin
L
dx
⎢ 0 0 0
⎢ 0 0 0
⎢ 0 0 0
⎢
⎢⎣ 0 0 0

0 0 0 ⎤
0 0 0 ⎥
⎥
0 0 0 ⎥
⎥
mπ x nπ x
L
⎥
G ∫ sin sin dx 0 0 ⎥
L L (5.24)
... 0 ⎥
L ⎥
mπ x nπ x ⎥
0 G ∫ cos cos dx 0
L L ⎥
0 ⎥
nπ x ⎥
L
mπ x
0 0 G ∫ cos cos dx ⎥
0
L L ⎥⎦

Because the integrals have the properties

L L
mπ x nπ x mπ x nπ x L
∫ sin L sin L dx = ∫ cos L cos L dx = 2 δ mn , (5.25)
0 0

one can get

Finite element analysis of stresses in beam structures 55

L L
εT ε εT ε L
∫ Xm CXn dx = 0, if m ≠ n, ∫ Xm CXm dx = 2 C. (5.26)
0 0

Therefore, the expression (5.22) of the internal virtual work can be reduced considerably to
the form

δ Wint
e
= −∑ δ a emT K emma em (5.27)
m

where

L
K emm = ∫ Bm
eT
CBem dA . (5.28)
2
Ae

The expression of the external virtual work of a prism is

L
δ Wext
e
= ∫ δ aeT ( x)F e ( x)dx + ∫ δu
T
( x, s )t ( x, s )dS + ∫ δu
T
( x, y, z )f ( x, y, z ) dV (5.29)
0 Se Ve

where δ a ( x) is the vector of virtual nodal line displacements of the prism, F ( x) is the
e e

vector corresponding to the distributed nodal line forces, t ( x, s ) is traction vector on the

surface of the prism, which coincides with the boundary of the prism, and f ( x, y , z ) is the
volume force vector in the prism.

The virtual displacements are expressed using finite prism approximation (5.8) or

δ uˆ ( x, y, z ) = ∑ Xum ( x)N e ( y, z )δ aem (5.30)

m

Consequently, the vector of virtual nodal line displacements is written as

δ ae ( x) = ∑ X m
a
( x)δ aem , (5.31)
m

where
Finite element analysis of stresses in beam structures 56

⎡ mπ x ⎤
⎢cos L 0 0 0 0 0 ⎥
⎢ ⎥
⎢ mπ x
0 sin 0 0 0 0 ⎥
⎢ L ⎥
⎢ mπ x ⎥
⎢ 0 0 sin 0 0 0 ⎥
⎢ L ⎥
X am ( x) = ⎢ ⎥. (5.32)
27×27 ⎢ ⎥
⎢ mπ x
⎢ 0 0 0 cos 0 0 ⎥⎥
L
⎢ ⎥
⎢ mπ x
0 0 0 0 sin 0 ⎥
⎢ L ⎥
⎢ mπ x ⎥
⎢ 0 0 0 0 0 sin ⎥
⎣ L ⎦

The traction and the volume force vectors are known functions, which are expressed using
Fourier sine and cosine series as

t ( x, s ) = ∑ Xum ( x)t m ( s ) (5.33)

m

and

f ( x, y, z ) = ∑ Xum ( x)f m ( y, z ) , (5.34)

m

where the corresponding coefficient functions vectors have typical expressions

L
2
t m ( s ) = ∫ Xum ( x)t ( x, s )dx (5.35)
L0

and

L
2
f m ( y, z ) = ∫ Xum ( x)f ( x, y, z )dx . (5.36)
L0

With the help of equations (5.30) and (5.31), the expression (5.29) of the external virtual work
of the prism gets the form
Finite element analysis of stresses in beam structures 57

L
δ Wext
e
=∑ δ aeT
m [ ∫ X m ( x )F
a e
( x)dx + ∫N
eT
( s) Xum ( x)t ( x, s )dS
m 0 Se

+ ∫N
eT
( y, z ) Xum ( x)f ( x, y, z )dV ]
Ve
L L
= ∑ δ aeT
m [ ∫ X m ( x)F ( x )dx +
a e
∫ N (s)∫ Xm ( x)t( x, s)dxds
eT u
(5.37)
m 0 s e 0
L
+ ∫ N eT ( y, z ) ∫ Xum ( x)f ( x, y, z )dxdA]
Ae 0
L L
= ∑ δ aeT
m [Fm +
e
∫N
eT
( s )t m ( s )ds + ∫N
eT
( y, z )f m ( y, z )dA]
m 2 e 2 e
s A

where

L
Fme = ∫ Xm
a
( x)F e ( x)dx . (5.38)
0

Finally, the expression of the external virtual work of prism e in matrix form is

δ Wext
e
= ∑ δ aemT (Fme + R te
m + Rm ) ,
fe
(5.39)
m

where

L
m =
R te ∫N
eT
( s )t m ( s)ds (5.40)
2 e
s

and

L
R mfe = ∫N
eT
( y, z )f m ( y, z )dA. (5.41)
2 e
A

The coefficient function t m ( s ) of the traction vector on the boundary of the cross-section is

approximated using quadratic line elements. Within line element e, it is expressed as

3
tˆ m ( s) = ∑ Nie ( s)t emi , (5.42)
i =1
Finite element analysis of stresses in beam structures 58

where Nie ( s ), i = 1, 2,3 are the shape functions of a quadratic line element and

⎧t xmie ⎫
⎪⎪ ⎪⎪
t emi = ⎨t eymi ⎬ , i = 1, 2,3 (5.43)
⎪e ⎪
⎪⎩ t zmi ⎪⎭

are the vectors of nodal values of the tractions corresponding to m-th term of the series. The

coefficient function f m ( y , z ) of the volume force is assumed to be constant f me within each

element. Thus it is

⎧ f xm e ⎫
⎪⎪ ⎪
e ⎪
f me = ⎨ f ym ⎬. (5.44)
⎪ e ⎪
⎪⎩ f zm ⎪⎭

It will be explained later, how the components of (5.43) and (5.44) can be determined in
typical loading cases of a beam.

With the help of equations (5.27) and (5.39), the virtual work equation of the finite prism

δ W e ≡ δ Wint
e
+ δ Wext
e
=0 (5.45)

gets the form

∑ δ aemT (−K emmaem + Fme + R em ) = 0 , (5.46)

m

where

R em = R te
m + Rm .
fe
(5.47)

Equation (5.46) holds with arbitrary δ a em if each bracketed term in the sum disappears. Thus

one gets for each value of m the element equation

Fme = K emmaem − R em . (5.48)

Finite element analysis of stresses in beam structures 59

Equations (5.48) are the element equations of the finite prism e. The corresponding sets of
equations for solving the problem are

K mma m = R m , m = 1,… , M (5.49)

where a m is the vector of system nodal displacements, K mm is the system stiffness matrix

and R m is the system load vector corresponding to m-th term of the series. The latter are

obtained with the help of element stiffness matrices K emm and element load vectors R em of

the finite prisms using standard finite element assembly process. Finally, in equation (5.49),
M is the number of terms used in the trigonometric series.

e e
Fig. 5.2: Traction t ( x, s ) on boundary strip e, which is nonzero and equal to t ( s ) , when
x1 < x < x2 and quadratic in s

Let the distributed traction (Fig. 5.2) within a boundary strip corresponding to a quadratic line
element e be defined as

⎧0, when 0 ≤ x < x1 ,

⎪
t e ( x, s ) = ⎨ t e ( s ), when x1 < x < x2 , (5.50)
⎪0, when x < x < L.
⎩ 2

e
The traction t ( s ) is assumed to have quadratic distribution within the line element

3
t ( s ) = ∑ Nie ( s ) tie ,
e
(5.51)
i =1
Finite element analysis of stresses in beam structures 60

where

⎧ txie ⎫
⎪⎪ ⎪⎪
tie = ⎨ txie ⎬ (5.52)
⎪ e⎪
⎪⎩ txi ⎪⎭

are the nodal tractions of the line element, which are given as data. Let the volume force
within a prism, defined by the biquadratic quadrilateral element e, be defined as

⎧0, when 0 ≤ x < x1 ,

⎪
f e ( x) = ⎨ f e , when x1 < x < x2 , (5.53)
⎪0, when x < x < L.
⎩ 2

The volume forces

⎧ f xe ⎫
⎪⎪ ⎪⎪
f e = ⎨ f ye ⎬ (5.54)
⎪ e⎪
⎪⎩ f z ⎪⎭

are assumed to be constants within the biquadratic element e.

If the tractions are expressed as Fourier sine and cosine series in coordinate x, the
corresponding coefficients are first

L x2
2 2
t em ( s) = ∫ Xum ( x)t e ( x, s)dx = ∫ Xm ( x)dxt
u e
( s)
L0 L x1

⎡ mπ x ⎤
⎢cos L 0 0 ⎥ (5.55)
x ⎢ ⎥
2 2⎢ mπ x 3
= ∫ 0 sin 0 ⎥ dx ⋅ ∑ Nie ( s ) tmi
e
Lx ⎢ L ⎥
i =1
1⎢
⎢ mπ x ⎥⎥
0 0 sin
⎢⎣ L ⎥⎦

and then
Finite element analysis of stresses in beam structures 61

⎡ mπ x2 mπ x1 ⎤
⎢sin L − sin L 0 0 ⎥
⎢ ⎥
3
2 ⎢ mπ x1 mπ x2 ⎥ te
t m ( s ) = ∑ Ni ( s )
e e
0 cos − cos 0
i =1 mπ ⎢ L L ⎥ mi
⎢ mπ x1 mπ x2 ⎥⎥
⎢ 0 0 cos − cos
⎢⎣ L L ⎥⎦
⎧ mπ x2 mπ x1 e ⎫
⎪ (sin L − sin L ) txi ⎪
⎪ ⎪
3
2 ⎪ mπ x1 mπ x2 e ⎪
= ∑ Ni ( s )
e
⎨(cos − cos ) t yi ⎬ .
i =1 mπ ⎪ L L ⎪
⎪ mπ x1 mπ x2 e ⎪
⎪(cos L − cos L ) tzi ⎪
⎩ ⎭
(5.56)

Thus we have for the series coefficients of the tractions

3
t em ( s ) = ∑ Nie ( s )t emi , (5.57)
i =1

where the nodal values are

⎧ mπ x2 mπ x1 e ⎫
⎪ (sin L − sin L ) txi ⎪
⎪ ⎪
2 ⎪ mπ x1 mπ x2 e ⎪
t emi = ⎨(cos − cos ) tyi ⎬ . (5.58)
mπ ⎪ L L ⎪
⎪ mπ x1 mπ x2 e ⎪
⎪(cos L − cos L ) tzi ⎪
⎩ ⎭

Correspondingly, one can obtain for the series coefficients of the volume force

⎧ mπ x2 mπ x1 e ⎫
⎪ (sin L − sin L ) f x ⎪
⎪ ⎪
2 ⎪ mπ x1 mπ x2 e ⎪
fm =
e
⎨(cos − cos ) f y ⎬. (5.59)
mπ ⎪ L L ⎪
⎪ mπ x1 mπ x2 e ⎪
⎪(cos L − cos L ) f z ⎪
⎩ ⎭

With the help of equations (5.57), (5.58) and (5.59), one can calculate the series coefficients,
which are needed in equations (5.40) and (5.41). Once solved the problem (5.49), the
stresses can be obtained using equation (5.19) in a chosen cross-section.
Finite element analysis of stresses in beam structures 62

6 VALIDATION OF THE METHOD

In order to test the validity of the method described in chapter 4, this will be compared with
the finite prism method through simple test examples, which will be analysed with MATLAB®
software to obtain the stress distribution in some cross-sections.

6.1 COMPARISON BETWEEN THE TWO METHODS

As mentioned in chapter 5, to make the comparison, a simply supported straight beam with
uniform cross-section is analysed. It has been decided to load the beam from the midpoint to
the right end by a uniform strip load. Thus some analysis will be done with different position
of the load in z − direction, with different cross-sections but with the same frame (Fig. 6.1).

Fig. 6.1: Example beam

For the frame analysis (see chapter 4.2), the grid shown in Fig. 6.2 will be used. The length
of the beam and Poisson’s ratio are taken as L = 20h andν = 0,3 , respectively, to consider a
realistic case.

Fig. 6.2: Grid with element and nodes

Three sorts of cross-section will be modelled: a solid one, an open thin-walled one and a
closed thin-walled one.

In all cases, the quantities h , E and q0 are taken as parameters and the additional data, for

example L , are expressed in terms of them. In the numerical analysis, the values of these
Finite element analysis of stresses in beam structures 63

parameters are set to unity, i.e. h = 1 , E = 1 , and q0 = 1 , and the corresponding results are

dimensionless. The final results are obtained by multiplying these values by dimensionally
correct products of original parameters ( h , E and q0 ). For the stresses, this product is

simply q0 . One must notice that now it is not so important the exact numerical results than

the comparison between both. In addition, only shear stress distribution will be compared
and not the normal stress, because this one has been obtained using the well-known Navier-
Bernoulli theory, which is widely accepted, and it does not use the warping functions, which
are the interesting point. When the finite prism method is used, 21 terms of the series will be
taken, which is quite sufficient to get good results.

For the first test, rectangular cross-section with 20 elements will be used (see Fig. 6.3), with
b = 2h and e = 0 . The width is bigger than the height because otherwise the case would be
better and the results more similar than the ones that follow.

Fig. 6.3: Grid of the rectangular cross-section

Shear stresses in y-direction at the left quarter of the beam with the method developed in this
thesis can be seen in Fig. 6.4(a); whereas with finite prism method, in Fig. 6.4(b). For shear
stresses in z-direction, see Fig. 6.5. No differences can be appreciated in stress distributions
of this cross-section.
Finite element analysis of stresses in beam structures 64

(a) (b)

Fig. 6.4: Dimensionless shear stress τ xy / q0 distribution in the rectangular cross-section, without
eccentricity, at x = 0, 25 L , using (a) warping functions and (b) the finite prism method

(a) (b)

Fig. 6.5: Dimensionless shear stress τ xz / q0 distribution in the rectangular cross-section, without
eccentricity, at x = 0, 25 L , using (a) warping functions and (b) the finite prism method

Consider now the cross-section where the load starts, i.e., at the midpoint of the beam. Here,
differently from before, the shear stresses values obtained (Fig. 6.6 and 6.7) are a bit higher.

(a) (b)

Fig. 6.6: Dimensionless shear stress τ xy / q0 distribution in the rectangular cross-section, without
eccentricity, at x = 0, 5 L , using (a) warping functions and (b) the finite prism method
Finite element analysis of stresses in beam structures 65

(a) (b)

Fig. 6.7: Dimensionless shear stress τ xz / q0 distribution in the rectangular cross-section, without
eccentricity, at x = 0, 5 L , using (a) warping functions and (b) the finite prism method

Then the cross-section located at the centre of the distributed load will also be submitted to
the analysis (Fig. 6.8 and 6.9). As in the cross-section without load seen before, in this case
the stresses are practically the same using both methods.

(a) (b)

Fig. 6.8: Dimensionless shear stress τ xy / q0 distribution in the rectangular cross-section, without
eccentricity, at x = 0, 75 L , using (a) warping functions and (b) the finite prism method

(a) (b)

Fig. 6.9: Dimensionless shear stress τ xz / q0 distribution in the rectangular cross-section, without
eccentricity, at x = 0, 75 L , using (a) warping functions and (b) the finite prism method
Finite element analysis of stresses in beam structures 66

Adding an eccentricity of e = 0, 4b = 0,8h to the load, the beam comes under, in addition to
the bending, a twisting moment, which must be included when doing the analysis.

It is not interesting to see what occurs in x = 0, 25 L and in x = 0, 75 L because, similar to the

case without eccentricity, the results are really closed to each other. In x = 0,5 L , little
variations can be observed in shear stresses distributions (Fig. 6.10 and 6.11).

(a) (b)

Fig. 6.10: Dimensionless shear stress τ xy / q0 distribution in the rectangular cross-section, loaded with
eccentricity, at x = 0, 5 L , using (a) warping functions and (b) the finite prism method

(a) (b)

Fig. 6.11: Dimensionless shear stress τ xz / q0 distribution in the rectangular cross-section, loaded with
eccentricity, at x = 0, 5 L , using (a) warping functions and (b) the finite prism method

Once seen about using a rectangular cross-section, an open thin-walled one will be
considered. The grid used is shown in Fig. 6.12, with b = 2h .

Firstly, the loading will be considered without eccentricity.

Finite element analysis of stresses in beam structures 67

Fig. 6.12: Grid of the open thin-walled cross-section

Shear stresses in this cross-section at a quarter point of the beam can be seen in Fig. 6.13
and 6.14. Only few differences in values can be noticed.

(a) (b)

Fig. 6.13: Dimensionless shear stress τ xy / q0 distribution in the open thin-walled cross-section,
without eccentricity, at x = 0, 25 L , using (a) warping functions and (b) the finite prism method

(a) (b)

Fig. 6.14: Dimensionless shear stress τ xz / q0 distribution in the open thin-walled cross-section,
without eccentricity, at x = 0, 25 L , using (a) warping functions and (b) the finite prism method
Finite element analysis of stresses in beam structures 68

When x = 0,5 L (Fig. 6.15 and 6.16), it seems not only to be a difference in the stress values
but also in the distribution. However, these colour plots may give an idea that the results are
worse than they really are. This is because the maximum and minimum values of the shear
stresses are located at the internal corners of the cross-section boundary. Theoretically, the
shear stresses are singular at these points and the finite element results are not reliable.

(a) (b)

Fig. 6.15: Dimensionless shear stress τ xy / q0 distribution in the open thin-walled cross-section,
without eccentricity, at x = 0, 5 L , using (a) warping functions and (b) the finite prism method

(a) (b)

Fig. 6.16: Dimensionless shear stress τ xz / q0 distribution in the open thin-walled cross-section,
without eccentricity, at x = 0, 5 L , using (a) warping functions and (b) the finite prism method

Stresses in the cross-section located in the middle of the distributed load do not have
significant differences (Fig. 6.17 and 6.18).
Finite element analysis of stresses in beam structures 69

(a) (b)

Fig. 6.17: Dimensionless shear stress τ xy / q0 distribution in the open thin-walled cross-section,
without eccentricity, at x = 0, 75 L , using (a) warping functions and (b) the finite prism method

(a) (b)

Fig. 6.18: Dimensionless shear stress τ xz / q0 distribution in the open thin-walled cross-section,
without eccentricity, at x = 0, 75 L , using (a) warping functions and (b) the finite prism method

Now, an eccentricity e = 0, 4b = 0,8h to the load will be added.

(a) (b)

Fig. 6.19: Dimensionless shear stress τ xy / q0 distribution in the open thin-walled cross-section,
loaded with eccentricity, at x = 0, 25 L , using (a) warping functions and (b) the finite prism method
Finite element analysis of stresses in beam structures 70

(a) (b)

Fig. 6.20: Dimensionless shear stress τ xz / q0 distribution in the open thin-walled cross-section,
loaded with eccentricity, at x = 0, 25 L , using (a) warping functions and (b) the finite prism method

In this case, at x = 0, 25 L , the use of warping functions gives a bit higher results of stresses
(Fig. 6.19 and 6.20), with the same distribution, approximately.

(a) (b)

Fig. 6.21: Dimensionless shear stress τ xy / q0 distribution in the open thin-walled cross-section,
loaded with eccentricity, at x = 0, 5 L , using (a) warping functions and (b) the finite prism method

(a) (b)

Fig. 6.22: Dimensionless shear stress τ xz / q0 distribution in the open thin-walled cross-section,
loaded with eccentricity, at x = 0, 5 L , using (a) warping functions and (b) the finite prism method
Finite element analysis of stresses in beam structures 71

At the midpoint, even higher values are obtained; both positive and negative shear stresses
(Fig. 6.21 and 6.22).

(a) (b)

Fig. 6.23: Dimensionless shear stress τ xy / q0 distribution in the open thin-walled cross-section,
loaded with eccentricity, at x = 0, 75 L , using (a) warping functions and (b) the finite prism method

(a) (b)

Fig. 6.24: Dimensionless shear stress τ xz / q0 distribution in the open thin-walled cross-section,
loaded with eccentricity, at x = 0, 75 L , using (a) warping functions and (b) the finite prism method

At the quarter point from the right side of the beam, the differences between both methods
(Fig. 6.23 and 6.24) are similar to the ones at the quarter point from the left side.

Finally, a closed thin-walled cross-section will be analysed. The grid used in discretizing it
can be seen in Fig. 6.25, with b = 2h .

No eccentricity in the loading is taken into account now.

Finite element analysis of stresses in beam structures 72

Fig. 6.25: Grid of the closed thin-walled cross-section

(a) (b)

Fig. 6.26: Dimensionless shear stress τ xy / q0 distribution in the closed thin-walled cross-section,
without eccentricity, at x = 0, 25 L , using (a) warping functions and (b) the finite prism method

(a) (b)

Fig. 6.27: Dimensionless shear stress τ xz / q0 distribution in the closed thin-walled cross-section,
without eccentricity, at x = 0, 25 L , using (a) warping functions and (b) the finite prism method

In the cross-section located at the quarter point no differences in shear stresses (Fig. 6.26
and 6.27) can be appreciated.
Finite element analysis of stresses in beam structures 73

(a) (b)

Fig. 6.28: Dimensionless shear stress τ xy / q0 distribution in the closed thin-walled cross-section,
without eccentricity, at x = 0, 5 L , using (a) warping functions and (b) the finite prism method

(a) (b)

Fig. 6.29: Dimensionless shear stress τ xz / q0 distribution in the closed thin-walled cross-section,
without eccentricity, at x = 0, 5 L , using (a) warping functions and (b) the finite prism method

At the midpoint, the most important variations are in the internal corners, where the local
shear stress values obtained with warping functions are lower (Fig. 6.28 and 6.29).

(a) (b)

Fig. 6.30: Dimensionless shear stress τ xy / q0 distribution in the closed thin-walled cross-section,
without eccentricity, at x = 0, 75 L , using (a) warping functions and (b) the finite prism method
Finite element analysis of stresses in beam structures 74

(a) (b)

Fig. 6.31: Dimensionless shear stress τ xz / q0 distribution in the closed thin-walled cross-section,
without eccentricity, at x = 0, 75 L , using (a) warping functions and (b) the finite prism method

At x = 0, 75 L , as in the quarter point, the differences are minimal (Fig. 6.30 and 6.31).

In the following, an eccentricity e = 0, 4b = 0,8h to the load will be added.

(a) (b)

Fig. 6.32: Dimensionless shear stress τ xy / q0 distribution in the closed thin-walled cross-section,
loaded with eccentricity, at x = 0, 5 L , using (a) warping functions and (b) the finite prism method

(a) (b)

Fig. 6.33: Dimensionless shear stress τ xz / q0 distribution in the closed thin-walled cross-section,
loaded with eccentricity, at x = 0, 5 L , using (a) warping functions and (b) the finite prism method
Finite element analysis of stresses in beam structures 75

It is not interesting to see what occurs in x = 0, 25 L and in x = 0, 75 L because, similar to the

case without eccentricity, the results are really closed to each other. In x = 0,5 L , some
variations can be observed in shear stresses distributions (Fig. 6.32 and 6.33), particularly in
the internal corners.

6.2 DISCUSSION OF THE RESULTS

Based on the previous results, some conclusions can be deducted.

The first case analysed, the rectangular solid cross-section, is the one that gives more
accurate results. Both shear stresses in y- and z-direction obtained have nearly the same
distribution and the same values. Thus, probably in many solid cross-sections, the use of
warping functions is valid to obtain the shear stresses distribution.

(a) (b)

Fig. 6.34: Dimensionless shear stress τ xy / q0 distribution in the half-circular cross-section, without
eccentricity, at x = 0, 5 L , using (a) warping functions and (b) the finite prism method

(a) (b)

Fig. 6.35: Dimensionless shear stress τ xz / q0 distribution in the half-circular cross-section, without
eccentricity, at x = 0, 5 L , using (a) warping functions and (b) the finite prism method
Finite element analysis of stresses in beam structures 76

For instance, in Fig. 6.34 and 6.35 the shear stresses are shown using a solid half-circular
cross-section within the same beam used in chapter 6.1 (see Fig. 6.1) and the same load. It
can be noticed that the results, in the midpoint of the beam, where values tend to differ, are
quite similar.

In the same way, not important variations have been observed in the closed thin-walled
cross-section, just in the surroundings of the internal corners and at the point where the load
starts. However, making the grid denser, one can find values closer to the ones obtained with
the finite prism method. To show this, that cross-section has been made denser and the
results are shown in Fig. 6.36.

(a) (b)

Fig. 6.36: Dimensionless shear stress (a) τ xy / q0 and (b) τ xz / q0 distribution in the closed thin-walled
cross-section, without eccentricity, at x = 0, 5 L , using warping functions, with a denser grid

Comparing Fig. 6.36(a) with Fig. 6.28, one can see that the first one gives more precise
values than the other. In a similar way occurs with τ xz comparing Fig. 6.36(b) and Fig. 6.29.

(a) (b)

Fig. 6.37: Dimensionless shear stress (a) τ xy / q0 and (b) τ xz / q0 distribution in the open thin-walled
cross-section, without eccentricity, at x = 0, 5 L , using warping functions, with a denser grid
Finite element analysis of stresses in beam structures 77

(a) (b)

Fig. 6.38: Dimensionless shear stress (a) τ xy / q0 and (b) τ xz / q0 distribution in the open thin-walled
cross-section, loaded with eccentricity, at x = 0, 5 L , using warping functions, with a denser grid

In the case of the open thin-walled cross-section, the differences are a little more significant.
When there is no eccentricity in the load, the values are lower than the ones from the prism
method and higher when the eccentricity exists. Moreover, the distribution is not exactly the
same. Although the grid was denser, the result would be that only the values at singular
points and their surroundings increase, but the distribution does not change. It can be seen
comparing Fig. 6.37(a) with 6.15, 6.37(b) with 6.16, 6.38(a) with 6.21 and 6.38(b) with 6.22.
Finite element analysis of stresses in beam structures 78

7 DEMONSTRATIVE EXAMPLES

Once developed the methodology and studied the suitable cases where the method can be
applied, some illustrative examples will be implemented. These are realistic frames with
possible cross-sections and usual loading cases.

Example 1. Consider the curved beam shown in Fig. 7.1, which is located in the x-z plane
and loaded by a transverse distributed load q0 along the axis of the beam. The cross-section

(Fig. 7.2) is uniform along the beam. A edge can only rotate around x-axis in Fig. 7.2 and B
edge just cannot move vertically (y-axis).

Fig. 7.1: Beam of example 1

The arch is divided into 20 beam elements in the same way as grids made in chapter 6. For
the cross-section, 25 biquadratic elements are used (Fig. 7.2). As earlier, the quantities h ,
E and q0 are taken as parameters and ν = 1 / 6 . Final results of stresses can be obtained

multiplying the dimensionless results by q0 / h .

Fig. 7.2: Grid of the cross-section used in example 1

Finite element analysis of stresses in beam structures 79

It is easy to see that the beam is under bending (shear force and bending moment) and
twisting due to the load and its eccentricity, since the cross-section has no symmetry. Thus a
cross-section that can be of interest is the one located in the middle of the beam, where
shear force and twisting and bending moments are not zero. Therefore, in this cross-section
there are normal stresses (Fig. 7.3), due to the bending moment, and shear stresses (Fig.
7.4), due to the shear force and the twisting moment.

Fig. 7.3: Dimensionless normal stress σ x h / q0 distribution in the middle of the beam of example 1

(a) (b)

Fig. 7.4: Dimensionless shear stress (a) τ xy h / q0 and (b) τ xz h / q0 distribution in the middle of the
beam of example 1

Fig. 7.5: Position of the barycentre, C, and the shear centre, S

Finite element analysis of stresses in beam structures 80

One can deduce that the distributed twisting moment load acts in the positive x-direction
inspecting the position of the shear centre S (see Fig. 7.5) in connection with the barycentre
C.

Example 2. The circular arch shown in Fig. 7.6 is under a uniform transverse load q0 , which

acts in the symmetry plane of the cross-section (Fig. 7.7), so it does not have eccentricity.
The cross-section is uniform along the beam.

Fig. 7.6: Beam of example 2

Fig. 7.7: Half-grid of the cross-section used in example 2

As in the previous example, the arch is divided into 20 frame elements, and biquadratic
elements are used for modelling the cross-section with the grid of Fig. 7.7. Again h , E and
q0 are taken as parameters, ν = 1 / 6 and the other important quantities are shown in Fig.
7.7.

In this case, no twisting moment exists but axial force is present, which causes normal
stresses in addition to the ones due to the bending moment. For instance, analysing one
edge of the arch, where shear force is maximum and bending moment is zero, normal stress
is uniform in the whole cross-section (Fig. 7.8) because it is only owing to axial force.
Finite element analysis of stresses in beam structures 81

Regarding to shear stresses (Fig. 7.9), their distribution is symmetric with respect to the
symmetric axis since just shear force causes them.

Fig. 7.8: Dimensionless normal stress σ x h / q0 distribution at the left end of the beam of example 2

(a) (b)

Fig. 7.9: Dimensionless shear stress (a) τ xy h / q0 and (b) τ xz h / q0 distribution at the left end of the
beam of example 2

Example 3. Consider the straight beam in Fig. 7.10. Its cross-section (Fig. 7.11) is a
composite one made of concrete deck and steel U-shaped thin-walled plate. Approximately,
Young’s modulus of steel is taken as ten times that of concrete.

Fig. 7.10: Beam of example 3

Finite element analysis of stresses in beam structures 82

Fig. 7.11: Half-grid of the cross-section used in example 3

The cross-section is modelled using the grid shown in Fig. 7.11 and biquadratic elements.
The values h , E0 and P are taken as parameters. The rest of quantities can be found in

Fig. 7.11.

By inspecting the beam and its loading, it can be observed that the beam is only under
bending (shear force and bending moment in the x-y plane). The cross-section located at the
quarter point from the left end of the beam is a good one to study the stresses distribution
because it is far enough from the application point of the load. Here, Qy = P / 2 and

M z = 2,5Ph . As can be seen in Fig. 7.12, the most part of compression is handled by the
steel, whereas the concrete works under much lower compression. Similarly, shear stresses
(Fig. 7.13) due to the shear force act essentially on the steel part.

Fig. 7.12: Dimensionless normal stress σ x h2 / P distribution at the quarter point of the beam of
example 3
Finite element analysis of stresses in beam structures 83

(a) (b)

Fig. 7.13: Dimensionless shear stress (a) τ xy h 2 / P and (b) τ xz h 2 / P distribution at the quarter point
of the beam of example 3

Example 4. The cross-section of Fig. 7.11 is now used in the beam of Fig. 7.14, which is
loaded by a concentrated twisting moment in the midpoint.

Fig. 7.14: Beam of example 4

Now, only twisting moment exists along the beam. In the same way as the previous case, the
quarter point will be analysed, where M t = Ph / 2 . In this case, normal stresses are zero and

the concrete must hold shear stresses in z-direction due to the twisting moment, but with less
intensity than the ones which act on the steel because of the difference between Young’s
modulus of both materials.

(a) (b)

Fig. 7.15: Dimensionless shear stress (a) τ xy h2 / P and (b) τ xz h 2 / P distribution at the quarter point
of the beam of example 4
Finite element analysis of stresses in beam structures 84

CONCLUSIONS

The use of warping functions in order to obtain the shear stresses distribution in a cross-
section of a beam often gives accurate enough results. There are some conditions that must
be held when applying this methodology.

On the one hand, having a composite cross-section, made of different materials, does not
affect the results, but the type of the cross-section does. Open thin-walled cross-sections
sometimes give less accurate values, whereas closed thin-walled and solid ones usually can
be handled with no problem.

Special attention was focused to comparing the results at the midpoint and right quadrant of
the beam, where the solution of the enhanced beam theory of this project is not exact. This is
because the theory behind determining the shear stresses is exactly valid only at cross-
sections in an unloaded part of a beam. The technique is expected to work well in the second
uniformly loaded half, but at the middle of the beam it is not expected to work so well. The
main reason for this is the starting point of this thesis, which was classical Saint Venant’s
torsion and Bernoulli-Euler beam theory. Thus warping torsion has not been included in the
formulation for determining the stresses. Therefore, to obtain very precise values, the cross-
section chosen to be analysed must be situated far enough from the points where any load
starts or finishes. This requirement gets less importance as the relation between the length of
the beam and the height of the cross-section gets higher.

On the other hand, as usual, careful treatment should be paid to singular points, such as
internal corners, where the grid should be much denser or special singularity elements (for
instance [20]) should be used. Other more practical way of avoiding this problem is to round
these corners and use dense grid near them. This is, actually, a correct approach in many
steel profiles, where rounded internal corners really exist.

Regarding the finite element method, biquadratic isoparametric Lagrange elements are
helpful enough to model the cross-sections. Probably, other sorts of two-dimensional
elements could also be useful to handle this enhanced theory.

With the help of the finite element method, the methodology studied in this thesis has been
easily implemented in MATLAB® environment. Some advantages of the program developed
are that, as said before, the cross-section can be modelled as a composite one and have any
shape, an eccentricity to the transverse load can be added, different cross-sections can
compose the frame, this one can be loaded in different manners and one can obtain the
Finite element analysis of stresses in beam structures 85

diagrams of the stress resultants to decide which cross-section could be interesting to

analyse.
Finite element analysis of stresses in beam structures 86

ACKNOWLEDGEMENTS

My sincere gratitude to my supervisor, Jukka Aalto, who has believed in me since the
beginning and accepted undertaking my final project. Moreover, he has helped me from start
to finish and I have always counted on his assistance.

I also want to thank all people from the Department of Civil and Structural Engineering for
offering me the opportunity of completing my final project, so allowing me to realise my
Erasmus programme exchange in Finland. Special thanks to all members of this department
which I have been lucky to know for treating me as a one of them and for the good
environment.

Finally, deep thanks to Alba, for her unconditional support, and also to my family. Although I
have not had them near, they have made me feel like I had.
Finite element analysis of stresses in beam structures 87

REFERENCES

[1] TIMOSHENKO, S. P. and GERE, J. M. Mechanics of Materials. Van Nostrand

Reinhold Company, New York, 1972.
[2] BEER, F. P., JOHNSTON, E. R. and DEWOLF, J.T. Mechanics of Materials, 3rd ed. in
SI units. McGraw-Hill, 1992.
[3] TIMOSHENKO, S. P. and GOODIER, J. N. Theory of Elasticity, 3rd ed. McGraw-Hill,
1970, pp. 291-352 and 354-379.
[4] ZIENKIEWICZ, O. C. and TAYLOR, R. L. The Finite Element Method, 3rd ed.
McGraw-Hill, 1979, pp. 381-385 and 432.
[5] LOVE, E. H. A Treatise of the Mathematical Theory of Elasticity, 4th ed. Dower
Publishing, 1944, pp. 329-368.
[6] SOKOLNIKOFF, I. S. Mathematical Theory of Elasticity, 2nd ed. McGraw-Hill, 1956,
pp. 109-134 and 198-215.
[7] HERRMANN, L. R. Elastic torsion analysis of irregular shapes. Journal of the
Engineering Mechanics Division. Vol. 91, No. EM6, 1965.
[8] AALTO, J. On determining stresses in a cross-section of a beam: torsion.
Unpublished report, Aalto University, Department of Civil and Structural Engineering.
[9] MASON, W. E. and HERRMANN, L. R. Elastic shear analysis of general prismatic
beams. Journal of the Engineering Mechanics Division. Vol. 94, No. EM4, 1968.
[10] AALTO, J. On determining stresses in a cross-section of a beam: uniaxial bending.
Unpublished report, Aalto University, Department of Civil and Structural Engineering.
[11] AALTO, J. On determining stresses in a cross-section of a beam: biaxial bending.
Unpublished report, Aalto University, Department of Civil and Structural Engineering.
[12] PRZEMIENIECKI, J. S. Theory of Matrix Structural Analysis. Dover Publications,
1985. Originally published by McGraw-Hill, 1968, pp. 70-82.
[13] CHEUNG, Y. K. Finite Strip Method in Structural Mechanics. Pergamon Press, 1976,
pp. 188-198.
[14] CHEUNG, Y. K. and THAM, L. G. Finite Strip Method. CRC Press, 1998, pp. 138-
147.
[15] AALTO, J. A biquadratic finite prism for simply supported straight beams.
Unpublished report, Aalto University, Department of Civil and Structural Engineering.
[16] AALTO, J. On plane stress analysis of a disk using quadratic isoparametric Lagrange
elements and numerical integration. Unpublished report, Aalto University, Department
of Civil and Structural Engineering.
Finite element analysis of stresses in beam structures 88

[17] AALTO, J. Finite element equations of a bar element under tension/compression,

torsion and biaxial bending. Unpublished report, Aalto University, Department of Civil
and Structural Engineering.
[18] AALTO, J. Finite element equations of a three dimensional frame element.
Unpublished report, Aalto University, Department of Civil and Structural Engineering.
[19] AALTO, J. On calculating cross-sectional properties of a composite beam using
isoparametric finite elements and numerical integration. Unpublished report, Aalto
University, Department of Civil and Structural Engineering.
[20] AALTO, J. Mapping of finite elements around corner singularities of the quasi-
harmonic equation. Communications in numerical methods in engineering. Vol. 9,
1993, pp. 957-974.
Finite element analysis of stresses in beam structures 89

APPENDIX

The shear stress values obtained at the integration points due to the use of numerical
integration in chapters 5.3.3 and 5.3.4 must be extrapolated to the element nodes in order to
have nodal results and be able to plot the shear stress distribution. It will be handled using
biquadratic extrapolation.

1
The principle of the extrapolation is as follows. Imagine that inside the quadrilateral element,
in natural coordinates ξ ,η , there is a square nine node biquadratic element (Fig. A.1) so, that
its nodes coincide to the integration points of the nine point integration scheme. The natural
coordinates of this element are ξ ′ and η ′ .

Fig. A.1: Extrapolation form the integration points k = 1, ,9 to a sampling point k

With the help of Fig. A.1 it is easy to see, that the natural coordinates of these two elements
are related by

ξ = 3 / 5ξ ′, η = 3 / 5η ′ . (A.1)

Thus the imagined natural coordinates ξ ′k and η k′ of point k , whose actual natural

coordinates are ξ k and η k , can be obtained from

1
One might use here the term interpolation as well, because sometimes the points k are located
inside the imagined element.
Finite element analysis of stresses in beam structures 90

ξ k′ = 5 / 3ξ k , ηk′ = 5 / 3ηk . (A.2)

Consider now a function f , whose values f k , k = 1,… ,9 at the integration points are known.

Its value at the extrapolation point k is now obtained from

9 9
f k = ∑ N k (ξ k′ ,η k′ ) f k = ∑ N k ( 5 / 3ξ k , 5 / 3η k ) f k . (A.3)
k =1 k =1

This extrapolation can now be applied separately to each component of the shear stress to
calculate their values at the nodes of the element k n= 1, , n ( n = 9 ). Formally one can write

9
τˆxy ( k ) = ∑ N k ( 5 / 3ξ k , 5 / 3η k )τˆxy (k ), k n= 1, ,n
k =1
(A.4)
9
τˆxz ( k ) = ∑ N k ( 5 / 3ξ k , 5 / 3η k )τˆxz (k ), k n= 1, , n.
k =1