5.

The principle of virtual work

The principle of virtual work is the starting point for the formulation of the
finite element method for solids and structures and we dedicate this chapter
to its study.
We start by introducing the principle of virtual work for a 1-D model
and by exploring the fundamental facts. Then, the principle is given for the
3-D elasticity model and for the remaining mathematical models discussed
in Chapter 4.
In the last section, we introduce some energy concepts in 3-D analy-
sis which provide alternative ways to formulate the mathematical models of
Chapters 3 and 4. These energy based formulations will be useful in Chapters
6 and 8.

5.1 The principle of virtual work for the bar problem

We introduce in this section the principle of virtual work considering a one-

dimensional problem, namely a bar subjected to distributed axial loading
f (x) and a concentrated load R at its end as shown in Figure 5.1a.
The mathematical model for this problem was discussed in Section 4.2.1.
However, for completeness, we recall below the differential formulation and
stress the key steps.

Equilibrium
As usual, we consider a differential element. The equilibrium requirement
gives (see Figure 5.1b)

−τ A + (τ + dτ ) A + f dx = 0 within the bar (5.1)

and so
dτ
A + f = 0. (5.2)
dx
Also, at the right end of the bar we have the equilibrium condition
368 5. The principle of virtual work

Fig. 5.1. a)Bar problem; b) Differential equilibrium for bar problem (plane sections
remain plane)

τ (L)A = R.

Compatibility

The relevant strain for this model is ε = εxx . The strain compatibility
condition is
du
ε= (5.3)
dx
where u is the displacement field1 with the condition u(0) = 0.

Constitutive equation
We are considering a linear elastic material. Therefore

τ = Eε (5.4)

where E is the Young’s is modulus.

Boundary conditions
The boundary conditions at the ends of the bar are listed as

u(0) = 0 (5.5)
1
In this chapter we suppose that the functions such as u(x) are sufficiently smooth
to allow the evaluation of all required integrations and differentiations. In Section
6.2 we will address this issue more rigorously
 5.1 The principle of virtual work for the bar problem 369

and

τ (L)A = R. (5.6)

Considering equations (5.2) to (5.6), which represent the basic equilib-

rium, compatibility and constitutive behavior requirements (including the
boundary conditions at both ends), we can write the differential formulation
for the problem in terms of the displacement field only.

Differential formulation of the bar problem

d2 u
EA +f =0 within the bar (5.7)
dx2
u(0) = 0 (5.8)
du
EA (L) = R (5.9)
dx

Given a particular functional form for f (x), an analytical solution can be

calculated by integrating equation (5.7) and imposing the boundary condi-
tions (5.8) and (5.9).
We would like to derive an alternative form, yet equivalent, to the differ-
ential formulation presented above. Let us consider the equilibrium equation
given in (5.2). We recognize that this equation is valid for any section within
the bar. Therefore we can write
 
dτ
A + f δu = 0 (5.10)
dx
where δu(x) is a continuous function defined in the interval 0  x  L
and zero at the point where the boundary condition in displacement of the
differential formulation is prescribed, i.e., δu(0) = 0. The function δu(x) is
called the virtual displacement field.
Hence, we also have
 L 
dτ
A + f δu dx = 0 (5.11)
0 dx
or equivalently
 L  L
dτ
− A δu dx = f δu dx. (5.12)
0 dx 0
370 5. The principle of virtual work

Integrating the left-hand side of equation (5.12) by parts we arrive at

 L  L
dτ L dδu
− A δu dx = − (Aτ δu)|0 + Aτ dx
0 dx 0 dx

and substituting into equation (5.12), we obtain

 L  L
dδu L
Aτ dx = f δu dx + (Aτ δu)|0 (5.13)
0 dx 0

Calling δε = dδu
dx the virtual strain field, since it is the strain field associated
with the virtual displacement field, and using the force boundary condition
τ (L)A = R we can re-write (5.13) as
 L  L
δε τ A dx = f δu dx + R δu|x=L (5.14)
0 0

where we employed δu|x=0 = 0.

Using the notion of work from basic mechanics let us interpret the terms
of equation (5.14):
• Rδu|x=L is the work done by the force R for the virtual displacement at
%x L= L.
• 0 f δu dx is the work done by the distributed force f for the virtual
displacement field δu.
%L
• 0 δε τ A dx is the internal work done by the stress field τ , solution of
the differential formulation, for the virtual strain field δε which – it is
important to note – corresponds to δu (see also Section 2.3.8).

We denote by δWi the internal virtual work given by

 L
δWi = δε τ A dx
0

and by δWe the external virtual work given by

 L
δWe = f δu dx + R δu|x=L .
0

Of course,

δWi = δWe . (5.15)

Equation (5.14) or equivalently (5.15) is the mathematical statement of

the “Principle of Virtual Work” or the “Principle of Virtual Displacements”
for the bar problem. We note that, for the stress field which satisfies equilib-
rium within the bar and at x = L, the external virtual work is equal to the
 5.1 The principle of virtual work for the bar problem 371

internal virtual work for any virtual displacement field that is zero at x = 0
(where the displacement boundary condition is imposed).
Hence, we have shown that equilibrium implies the virtual work equation,
but not yet the vice-versa. In order to prove the equivalence, we also need to
show that the virtual work statement, equation (5.14), implies equilibrium,
i.e., equations (5.2) and (5.6).
Using integration by parts we have the following identity
 L  L  L
dδu L dτ
δετ Adx = τ Adx = δuτ A|0 − δu Adx. (5.16)
0 0 dx 0 dx

Substituting (5.16) into (5.14) we obtain

 L 
dτ L
A + f δudx − δuτ A|0 + Rδu|x=L = 0
0 dx

and using that δu|x=0 = 0, we arrive at

 L  
dτ
A +f δu dx + [(R − τ A) δu]|x=L = 0. (5.17)
0 dx

Equation (5.17) has to be valid for any δu (x) with δu (0) = 0. Let us take
 
x dτ
δu (x) = x 1 − A +f . (5.18)
L dx
dτ
Although we do not explicitly know the functional form of (A dx + f ),
the above choice for δu (x) is certainly allowed, since this function satisfies
δu (0) = 0. Since for this choice δu (L) = 0, equation (5.17) becomes
  2
L
dτ x
A +f x 1− dx = 0.
0 dx L
   dτ 2
Considering that for (0, L), x > 0, 1 − L x
> 0 and A dx + f ≥ 0, the
only possibility for the above integral to vanish is to have
dτ
A +f =0 within the bar (5.19)
dx
i.e., to satisfy the equilibrium equation (5.2).
Using (5.19), we can now return to equation (5.17) which then reads

[(R − τ A) δu]|x=L = 0.

Since the value of δu (x)|L is arbitrary, we conclude that

R = τ A|x=L .
372 5. The principle of virtual work

In other words, the equilibrium at the boundary, i.e., the force boundary
condition is also implicitly contained in the virtual work statement.
Hence, we proved the equivalence of the equilibrium equations (5.2), (5.6)
and the principle of virtual work equation (5.14). This is a fundamental result
that we emphasize in Figure 5.2. Although we proved this equivalence for the
1-D bar problem, this is a general result which is valid for every mathematical
model in solid and structural mechanics.

Fig. 5.2. Equivalence between equilibrium and the principle of virtual work

The above discussion shows that we may interpret the principle of virtual
work to be an integral form of the equilibrium equations, and it is also re-
ferred to as a variational formulation because the virtual displacements can
be thought of as a variation of the real displacements.
We summarize what we have discussed so far in Table 5.1. The repeti-
tion shown in this table is intentional, because we want to emphasize which
equations are the same and which are not the same for the two formulations.
In equations (5.7), (5.8) and (5.9) we wrote the differential formulation
in terms of the displacement field only. We write below the variational for-
mulation in terms of displacements only.

Variational formulation of the bar problem

Find u (x), u(0) = 0 such that

 L  L
dδu du
EA dx = f δudx + Rδu|x=L (5.20)
0 dx dx 0

for any δu (x), δu (0) = 0.

The relation (5.20) is the principle of virtual work in terms of displace-

ments only, and also contains the compatibility and constitutive require-
ments.
 5.1 The principle of virtual work for the bar problem 373

Table 5.1. Summary of differential and variational formulations for bar problem

Differential Formulation Variational Formulation

Find τ (x), ε(x) and continuous u(x) such that the Find τ (x), ε(x) and continuous u(x) such that the
following holds: following holds:

Displacement boundary condition Displacement boundary condition

u(0) = 0 u(0) = 0

Differential equilibrium Principle of virtual work

%L %L
A dτ
dx
+f =0 0
δετ A dx = 0
f δu dx + Rδu|x=L

Equilibrium at the boundary for any δu(x), δu(0) = 0

dδu
τ A = R for x = L and δε = dx

Compatibility Compatibility

du du
ε= dx
ε= dx

Constitutive equation Constitutive equation

τ = Eε τ = Eε
374 5. The principle of virtual work

Example 5.1

Considering the bar problem of Figure 5.1, show that the principle of virtual
work contains the condition of global equilibrium of the bar.

Solution

In Figure 5.3, we show the reaction at the support represented by F , which

is given by

Fig. 5.3. Bar problem with reaction explicitly represented

F = −τ (0) A.

In this case, the reaction can be calculated directly by imposing equilibrium

in the x direction to obtain
 
L
F =− f dx + R .
0

Consider the problem described in Figure 5.3 with no support and let us
apply the principle of virtual work taking as a virtual displacement field a
rigid body translation in the x direction of magnitude Δ. Since it is a rigid
body motion, the virtual strain field associated with it is identically zero (in
fact, ε = dδu d
dx = dx Δ = 0). Therefore, the right-hand side of the virtual work
statement is also zero. Then, the principle of virtual work yields
 L  L
δε τ A dx = 0 = f Δdx + F Δ + RΔ
0 0
 
L
0=Δ f dx + F + R
0

Since Δ is arbitrary
 L
f dx + F + R = 0
0

and therefore
 5.1 The principle of virtual work for the bar problem 375

 
L
F =− f dx + R
0

which is the global equilibrium equation of the bar.


Hence, we note that the principle of virtual work contains not only the
local, i.e., the differential equilibrium written in integral form as previously
shown, but also the global equilibrium. Although we explored this fact in this
very particular setting, this result is general and it will be detailed further
for a 2-D case (see Example 5.6).
In other words, if we introduce the support reactions as external loads,
suppress the kinematic restraints provided by the supports, and use as virtual
displacement fields rigid body modes, the principle of virtual work yields the
global equilibrium equations.

The condition that the equilibrium of every part of a solid implies global
equilibrium was also exemplified for a truss structure (see Section 2.2). Hence,
the above result that the principle of virtual work contains the global equi-
librium condition must be expected, since this principle is equivalent to dif-
ferential equilibrium, that is, equilibrium of every part of the bar.

Example 5.2

Consider the problem of a steel bar which supports a weight as described in

Figure 5.4.

Fig. 5.4. Steel bar subjected to gravity. The density ρ[kg/m3 ], the area A[m2 ] and
the mass M [kg]
376 5. The principle of virtual work

(i) State and solve the differential formulation of this problem.

(ii) Show explicitly, considering the exact solution derived in (i), that the
principle of virtual work is satisfied for the following virtual displacement
patterns: δu(x) = ax and δu(x) = ax2 , where a is a constant.

Solution

This problem is clearly a particular case of the bar problem studied above
and its differential formulation can be written as
d2 u
EA + ρgA = 0 (5.21)
dx2
u(0) = 0 (5.22)


du 
EA = Mg (5.23)
dx x=L

where L = 1 m is the bar length.

The solution can be found by substitution and integration and is given
by

!   "
g ρx2 M
u(x) = − + + ρL x .
E 2 A

(ii) The exact solution for the stress field is

!  "
du M
τ (x) = E = g −ρx + + ρL .
dx A

Considering the virtual displacement δu(x) = ax, the associated virtual

strain field is
dδu(x)
δε(x) = = a.
dx
For the principle of virtual work to be satisfied for this virtual displace-
ment field, we need to have that
 L !  "  L
M
ag −ρx + + ρL Adx = ρgAaxdx + M gaL
0 A 0

which indeed is satisfied.

 5.1 The principle of virtual work for the bar problem 377

For the second virtual displacement field δu(x) = ax2 , we obtain

δε(x) = 2ax

and the equation to be verified is

 L !  "  L
M
2axg −ρx + + ρL Adx = ρgAax2 dx + M gaL2
0 A 0

which also holds.



Example 5.3

Consider the bar problem with varying cross-section described in Figure 5.5.

Fig. 5.5. Bar problem with varying bar cross-sectional area

(i) Establish the differential formulation and the principle of virtual work
for this problem.

(ii) Find the exact solution for this problem.

(iii) Show that the principle of virtual work is satisfied for the exact stress
field and a virtual displacement field δu(x) = ax.

Solution

(i) Since the cross-sectional area is not constant along the bar, we need to
generalize equation (5.1) as,

−τ A + (τ + dτ )(A + dA) + f dx = 0

dτ A + τ dA + dτ dA + f dx = 0
378 5. The principle of virtual work

and hence obtain

d
(τ A) + f = 0. (5.24)
dx
Substituting the strain and constitutive equations into (5.24) we obtain
the differential formulation for this problem.
Find u(x) such that
 
d du
EA +f = 0 within the bar (5.25)
dx dx
u(0) = 0 (5.26)
du
EA |x=L = R. (5.27)
dx
(ii) For this example, f = 0, and so (5.25) becomes

! "
E d 2 du
(1 + x) = 0
104 dx dx

2 d2 u du
(1 + x) 2
+ 2 (1 + x) = 0
dx dx
or
d2 u 2 du
2
+ =0
dx (1 + x) dx
leading to
 
R · 104 x
u(x) = .
E 1+x
(iii) To verify that the principle of virtual work holds for δu(x) = ax we use

δε(x) = a.
The stress field is given by
du 1
τ (x) = E = R · 104
dx (1 + x)2
and the principle obviously holds
 L
R · 104
a (1 + x)2 · 10−4 dx = RaL.
0 (1 + x)2


 5.1 The principle of virtual work for the bar problem 379

Example 5.4

Consider the problem described in Figure 5.6. The Young’s modulus is con-
stant and equal to E1 for 0 < x < L/2 and constant and equal to E2 for
L/2 < x < L with E1 = E2 .

Fig. 5.6. Bar problem with varying material properties

(i) Establish the differential formulation for this problem.

(ii) Establish the principle of virtual work formulation for this problem.

Solution

(i) The differential equilibrium equation of the bar in terms of displacements

when both E and A may vary was derived in Example 5.3
 
d du
EA +f =0
dx dx
L
However in this problem, E changes discontinuously at x = 2 therefore
we need to use the differential formulation for subdomain 1
d 2 u1 L
E1 A +f =0 0<x<
dx2 2
and for subdomain 2
d 2 u2 L
E2 A +f =0 < x < L.
dx2 2
where u1 (x) and u2 (x) represent the axial displacement for subdomains
1 and 2 respectively. The displacement boundary conditions are

u1 (0) = 0
380 5. The principle of virtual work

u2 (L) = 0
u1 |( L ) = u2 |( L ) .
2 2

Also, the normal force N at x = L2 should be the same when we approach

x = L2 from the right side or from the left side of the section. From the
left side
du1
N |( L ) = E1 A | L
2 dx ( 2 )
and from the right side
du2
N |( L ) = E2 A | L .
2 dx ( 2 )
Hence we have the force boundary condition
du1 du2
E1 A |( L ) = E 2 A | L .
dx 2 dx ( 2 )
(ii) The principle of virtual work can however be directly written for the
whole bar
 L  L
δετ Adx = δuf dx
0 0
or
 L/2  L  L
dδu du dδu du
E1 Adx + E2 Adx = δuf dx
0 dx dx L/2 dx dx 0

because the discontinuous change of E does not introduce a difficulty in

this equation.

5.2 The principle of virtual work in 2-D and 3-D

analyses
In the previous section we introduced the principle of virtual work for the
one-dimensional bar problem. That setting, due to its simplicity, allowed us
to concentrate on the fundamental facts.
We now plan to state the principle for the other mathematical models
discussed in Chapters 3 and 4. We choose to first present the principle of
virtual work for the three-dimensional elasticity mathematical model and
then discuss its specialization to the remaining mathematical models.
Recalling the discussion of the principle of virtual work for the bar prob-
lem, we concluded that satisfying the principle of virtual work is equivalent
to satisfying local (and therefore also global) equilibrium:
 5.2 The principle of virtual work in 2-D and 3-D analyses 381

(I) If a stress field τ for which the virtual work of the stresses is equal to the
virtual work of the externally applied forces for any virtual displacement
field δu, zero at the fixed boundary, then this stress field also satisfies
the differential equilibrium (equation (5.2)) and the stress equilibrium at
the free boundary, i.e., the force boundary condition (equation (5.6)).

(II) If a stress field τ satisfies differential equilibrium (equation (5.2)) and the
force boundary condition (equation (5.6)) then for any virtual displace-
ments δu, zero at the fixed boundary, the internal virtual work of this
stress field is equal to the virtual work of the externally applied forces
(the reverse of (I)).

5.2.1 The principle of virtual work for 3-D elasticity

In Figure 3.59 of Chapter 3, we described the 3-D elasticity problem. For a

given choice of body forces f B , surface tractions f S on Sf and displacement
boundary conditions u = û on Su , let τxx , τyy , τzz , τxy , τxz , τyz be the stress
field that corresponds to the solution of this elasticity problem. Therefore,
this stress field satisfies differential equilibrium in the volume of the body
and at the boundary, i.e., the force boundary conditions on Sf .
The above statements (I) and (II), which were proved in the context of
the one-dimensional bar problem, are also valid for 2-D and 3-D problems.
However, we need to precisely define, in each case, the internal virtual work.
The internal virtual work for the 3-D problem is a generalization of the
internal virtual work for the 1-D bar problem for which the stress and strain
states were fully represented by only one stress (τ ) and one strain component
(ε). The virtual work of the 3-D stress field τxx , τyy , τzz , τxy , τxz , τyz con-
sidering a virtual strain field δεxx , δεyy , δεzz , δγxy , δγxz , δγyz (derived from
the virtual displacements) is given by

δWi = (δεxx τxx + δεyy τyy +

+δεzz τzz + δγxy τxy + δγxz τxz + δγyz τyz ) dV (5.28)

which can be re-written using the stress and strain column matrices as
 
T
δWi = δε τ dV = ε̄T τ dV.
V V

As in the expression above, we now denote the virtual quantities by an over-

bar.
The external virtual work is given by
 
T B
δWe = ū f dV + ūT f S dS
V Sf
382 5. The principle of virtual work

where

⎡ ⎤
ū(x, y, z)
⎢ ⎥
⎢ ⎥
ū = ⎢ v̄(x, y, z) ⎥
⎣ ⎦
w̄(x, y, z)

is the column matrix which represents the virtual displacement field. Note
that by the definition of the virtual displacement field ū = 0 on Su .
Therefore the virtual work statement reads2
  
T T B
ε̄ τ dV = ū f dV + ūT f S dS. (5.29)
V V Sf

In the following example we demonstrate an important result which was

already shown for the 1-D bar problem in Section 5.1.

Example 5.5
In the context of a 3-D body, show that if a stress field satisfies differ-
ential equilibrium and the force boundary conditions then the virtual work
statement is satisfied by this stress field.

Solution
The differential equilibrium equations for a 3-D body are given by equa-
tions (3.114) and the force boundary conditions by equations (3.117) which
are assumed to be satisfied by the stress field τxx , τyy , τzz , τxy , τxz , τyz
considered.
Let ū(x, y, z), v̄(x, y, z) and w̄(x, y, z) be an arbitrarily chosen displace-
ment field such that

ū = 0, v̄ = 0, w̄ = 0 on Su . (5.30)

Hence ū, v̄ , w̄ are a virtual displacement field.

Using the equilibrium equations, we write
 
∂τxx ∂τxy ∂τxz
+ + + fxB ū = 0 (5.31)
∂x ∂y ∂z
 
∂τyx ∂τyy ∂τyz B
+ + + fy v̄ = 0 (5.32)
∂x ∂y ∂z
 
∂τzx ∂τzy ∂τzz
+ + + fzB w̄ = 0. (5.33)
∂x ∂y ∂z
2
It is implicitly understood that for the integral over Sf the virtual displacements
ū in the integrand represent the virtual displacements evaluated on Sf . We could
have used ūSf to emphasize this fact
 5.2 The principle of virtual work in 2-D and 3-D analyses 383

Using the derivative product rule, we have

∂ ∂τxx ∂τyx ∂τzx
(τxx ū + τyx v̄ + τzx w̄) = ū + v̄ + w̄
∂x ∂x ∂x ∂x
∂ ū ∂v̄ ∂ w̄
+τxx + τyx + τzx
∂x ∂x ∂x
and re-arranging terms, we obtain
∂τxx ∂τyx ∂τzx ∂
ū + v̄ + w̄ = (τxx ū + τyx v̄ + τzx w̄)
∂x ∂x ∂x ∂x
 
∂ ū ∂v̄ ∂ w̄
− τxx + τyx + τzx .
∂x ∂x ∂x
Now, using analogous expressions for the derivatives with respect to y and z,
and equations (5.31) to (5.33) we can write

 !
∂ ∂
(τxx ū + τyx v̄ + τzx w̄) + (τxy ū + τyy v̄ + τzy w̄) +
V ∂x ∂y
∂
+ (τxz ū + τyz v̄ + τzz w̄) +
∂z

∂ ū ∂v̄ ∂ w̄ ∂v̄ ∂ ū
− τxx + τyy + τzz + τyx + τxy +
∂x ∂y ∂z ∂x ∂y

∂ w̄ ∂ ū ∂ w̄ ∂v̄
τzx + τxz + τzy + τyz
∂x ∂z ∂y ∂z
+fxB ū + fyB v̄ + fzB w̄ dV = 0. (5.34)
The following mathematical identity follows from the use of the divergence
theorem3
 !
∂ ∂
(τxx ū + τyx v̄ + τzx w̄) + (τxy ū + τyy v̄ + τzy w̄) +
V ∂x ∂y
"
∂
+ (τxz ū + τyz v̄ + τzz w̄) dV
∂z

= [(τxx ū + τyx v̄ + τzx w̄) nx + (τxy ū + τyy v̄ + τzy w̄) ny +
S

(τxz ū + τyz v̄ + τzz w̄) nz ] dS (5.35)

where nx , ny and nz are the components of the outward unit normal to the
surface S. But
3
The divergence theorem states: Let F be a vector field in the volume V ; then
% ∂Fy %
V
∂Fx
∂x
+ ∂y
+ ∂Fz
∂z
dV = S
F · n dS where n is the unit outward normal on
the surface S of V
384 5. The principle of virtual work


[(τxx nx + τxy ny + τxz nz ) ū + (τyx nx + τyy ny + τyz nz ) v̄+
S

(τzx nx + τzy ny + τzz nz ) w̄] dS


 S
= ūfx + v̄fyS + w̄fzS dS (5.36)
Sf

where we used (3.117) and (5.30) .

Next, using the symmetry of the stress tensor and substituting (5.35) and
(5.36) into (5.34), we arrive at

(ε̄xx τxx + ε̄yy τyy + ε̄zz τzz + γ̄xy τxy + γ̄yz τyz + γ̄xz τxz ) dV =
V
 
 B  S
ūfx + v̄fyB + w̄fzB dV + ūfx + v̄fyS + w̄fzS dS. (5.37)
V Sf

We note that the strains with the overbar are obtained from the virtual dis-
placement field using the strain-displacement relations (3.115) and therefore
are virtual strains. Re-writing (5.37) using matrix quantities gives
  
T T B
ε̄ τ dV = ū f dV + ūT f S dS (5.38)
V V Sf

which is the virtual work statement (5.29).


Hence we proved statement (II) for the 3-D problem. The proof of state-
ment (I) is obtained in the same way as for the 1-D case, by starting from
equation (5.38) − the virtual work statement − and perform the mathemat-
ical operations backwards.
Regarding the principle of virtual work equation, we emphasize once more:
• The virtual strains ε̄ are calculated by the differentiations given in (3.115)
from the assumed virtual displacements ū.
• The virtual displacements ū must be smooth enough to allow the evaluation
of the integral in (5.29), with ū equal to zero on Su .
• All integrations are performed over the original volume and surface area of
the body, unaffected by the imposed virtual displacements.
The principle of virtual work in terms of the displacement variables only
is the variational formulation which contains all requirements of equilibrium,
compatibility and constitutive behavior.
 5.2 The principle of virtual work in 2-D and 3-D analyses 385

Variational formulation of the 3-D elasticity model

Find u, u = û on Su such that4

  
( ∂ ū) T C ∂  ū dV = ūT f B dV + ūT f S dS (5.39)
V V Sf

for any ū which is zero on Su

Note that we have used equations (4.330) and (4.331) to enforce the strain-
displacement and constitutive requirements.
Since all field and boundary conditions of the differential formulation
are represented in the above variational formulation and vice-versa, these
formulations are equivalent.
Of course, we can similarly derive the principle of virtual work for all
differential formulations considered in Chapter 4 and all the results above
are valid for these models. We detail below the derivation of the principle of
virtual work for some of these models.

5.2.2 The principle of virtual work for the plane stress model

Figure 4.5 describes a plane stress situation. The non-zero stress components
are τxx , τyy and τxy , and the non-zero strain components are εxx , εyy , γxy , εzz .
Starting from the general expression of the internal virtual work of the 3-D
elasticity problem and introducing the plane stress assumptions we have
   +h/2
ε̄T τ dV = (ε̄xx τxx + ε̄yy τyy + ε̄zz · 0 + γ̄xy τxy ) dz dA
V A −h/2

=h (ε̄xx τxx + ε̄yy τyy + γ̄xy τxy ) dA
A
where we used h to be constant.
Since both the body forces and surface tractions are assumed to have non-
zero components only on the xy plane and do not depend on z, the external
virtual work reads
 
T B
u f dV + uT f S dS =
V Sf

4
Throughout the discussion, we assume that u and ū are continuous
386 5. The principle of virtual work

   
+h/2   +h/2  
= ū fxB + v̄ fyB dz dA + ū fxS + v̄ fyS dz dL
A −h/2 Lf −h/2
 
   
=h ū fxB + v̄ fyB dA + h ū fxS + v̄ fyS dL.
A Lf

with V = h × A and Sf = h × Lf .
Using the column matrix quantities defined in Chapter 4 for the plane
stress and plane strain models, we are led to the principle of virtual work for
plane stress
  
T T B
h ε̄ τ dA = h ū f dA + h ūT f S dL (5.40)
A A Lf

for all ū, ū = 0 on Lu and ε̄ = ∂ε ū .

Formally, the same kind of expression for the virtual work principle as for
the 3-D case is obtained, that is,
  
ε̄T τ dA = ūT f B dA + ūT f S dL (5.41)
A A Lf

for all ū, ū = 0 on Su , with Su = h × Lu and ε̄ = ∂ε ū .

We note that in the above expressions f B is given per unit of volume and
f S is given per unit of surface area.
Sometimes, it is more convenient to use (5.40) instead of (5.41). For ex-
ample, when considering concentrated loads applied at the boundary (see the
example below).

Example 5.6

Consider the plane stress problem defined in Figure 5.7.

Calculate the reactions by simple statics, and calculate the reactions using
the principle of virtual work.

Solution

Referring to Figure 5.7, the equilibrium equations of statics are

  
Fx = 0, Fy = 0, MA = 0

which leads to

  L
XA = 0, YA + YB + t fyB dA + t fyS dx = 0
A 0

and
 5.2 The principle of virtual work in 2-D and 3-D analyses 387

Fig. 5.7. a) Definition of the plane stress problem, fxB = 0, fyB = −2400 N/m3 ,
t = 0.1 m (thickness); b) Equilibrium of the plate including reactions

  L
YB L + t fyB xdA + t fyS xdx = 0
A 0

from which we obtain YA = 273.33 N , YB = 306.67 N.

We use equation (5.40) since we have an extra term on the right-hand
side corresponding to the virtual work of the concentrated forces given by


Nc
Fi δdi
i=1

where Fi , i = 1, · · · , NC are the magnitudes of the NC concentrated forces

and δdi are the virtual displacements in the direction and the same orientation
as the concentrated loads Fi .
Note that if we considered equation (5.41) , we would need to divide the
magnitude of the concentrated load by the thickness t since for this equation
we divided by the constant thickness.
To calculate the reactions we use 3 independent virtual displacement fields
corresponding to virtual rigid body mode displacements. Then, the internal
virtual work is zero.
Choosing first, as virtual displacement, a translation in the direction x of
magnitude Δx , we obtain
⎡ ⎤ ⎡ ⎤
    L 
0 0
0 = t Δx 0 ⎣ ⎦ dA + t Δx 0 ⎣ ⎦ dL + XA Δx
A fyB 0 fyS

0 = XA
which is the equilibrium equation in the x direction.
Considering now a translation in the y direction of magnitude Δy , we
obtain
388 5. The principle of virtual work

⎡ ⎤ ⎡ ⎤
    L  
0 0
0=t 0 Δy ⎣ ⎦ dA+t 0 Δy ⎣ ⎦ dL+YA Δy +YB Δy
A fyB 0 fyS

or
  L
0 = YA + YB + t fyB dA + t fyS dx
A 0

which is the equilibrium equation in the y direction.

As the last virtual displacement field, we choose an infinitesimal rigid
rotation about point A which is given by

⎡ ⎤
−dθ y
ū = ⎣ ⎦.
dθ x

The virtual work expression becomes

⎡ ⎤ ⎡ ⎤
  L
0 0
0 = YB (Ldθ) + t [−dθ y dθ x] ⎣ ⎦ dA + t [−dθ y dθ x]⎣ ⎦ dx
A fyB 0 fyS
  L
0 = YB L + t fyB xdA + t fyS xdx
A 0
which corresponds to the moment equilibrium condition. Since we obtained
exactly the same equilibrium equations, the reactions are those already calcu-
lated. 

5.2.3 The principle of virtual work for the plane strain model

The plane strain model was discussed in Section 4.1.1. Due to the kinematic
assumptions the strains εzz = γyz = γzx = 0 and the non-zero components
are, therefore, εxx , εyy and γxy . The stresses τxx , τyy , τxy and τzz are, in
general, non-zero.
Since in plane strain we have for every section the same stress and strain
fields, and loading, we can take a unit thickness to write the principle of
virtual work. Analogously to what we did for the plane stress model, we can
use the matrix quantities defined in Chapter 4. Following the derivations of
the plane stress case and considering a unit thickness, the principle of virtual
work reads
  
T T B
ε̄ τ dA = ū f dA + ūT f S dL. (5.42)
A A Lf

for all ū, ū = 0 in Su (i.e. Lu ) and ε̄ = ∂ε ū .

 5.2 The principle of virtual work in 2-D and 3-D analyses 389

5.2.4 The principle of virtual work for the axisymmetric model

Considering the discussion of the axisymmetric model in Section 4.1.3, besides

the usual stress and strain components τxx , τyy , τxy and εxx , εyy , γxy for the
plane models, we have the hoop stress and strain components τzz and εzz
which are also non-zero.
In Figure 4.12 a generic cross-section of the solid of revolution is shown.
The y axis is the symmetry axis and the x axis corresponds to the radial
direction. We need to define a volume to apply the principle of virtual work.
It is usual to take the volume corresponding to an angle of one radian centered
at the symmetry axis. Then, for the evaluation of the internal virtual work,
we have
   1  
ε̄T τ dV = ε̄T τ x dθ dA = ε̄T τ x dA
V A 0 A

where θ is the circumferential angle coordinate of the cylindrical system used

(refer to Section 4.1.3 for details) and we adopt the column matrices of the
axisymmetric model. Using the same integration procedure for the external
work terms, we arrive at the principle of virtual work
  
T T B
ε̄ τ x dA = ū f x dA + ūT f S x dL.
A A Lf

for all ū, ū = 0 in Su (i.e. Lu ) and ε̄ = ∂ε ū .

5.2.5 The principle of virtual work for the Bernoulli-Euler beam

model

Let us derive the principle of virtual work for a structural mathematical

model. We choose as a typical model, the Bernoulli-Euler beam model which
we discussed in Section 4.2.2. Recognizing that the stress and strain compo-
nents that do work are τxx and εxx , the internal virtual work reads

δWi = ε̄xx τxx dV.
V

Referring to equation (4.145) re-written below

M
τxx = − z
I
and to equations (4.124) and (4.141) leading to

d2 w
εxx = −z
dx2
we obtain
390 5. The principle of virtual work

 L      L 2
d2 w̄ M d w̄
δWi = −z 2 − z dA dx = 2
M dx
0 A dx I 0 dx

2
Recall that κ = ddxw2 is the curvature of the axis of the beam for infinitesimally
small displacements. Considering only a transverse distributed force p as the
external load, the principle of virtual work for a beam, for example, clamped
at both ends reads
 L  L
κ̄M dx = w̄p dx
0 0

2
for all w̄, w̄(0) = w̄(L) = 0 = ddxw̄ (0) = ddxw̄ (L) and κ = ddxw̄2 .
This result was anticipated since for this model τ = [M ] and ε = [κ] (note
that, if one end of the beam is free and subjected to an externally applied
force or moment, their virtual work contribution should be included in δWe ).

5.3 Strain and potential energy in 3-D

In this section we discuss the concept of strain and potential energy for 3-D
elastic solids.

5.3.1 Strain energy

Let us extend, to a 3-D linear elastic solid, the concept of strain energy
introduced in Chapter 2 for a truss bar, i.e., for a one-dimensional state of
stress and strain.
Recall that the internal work per unit of volume in a 1-D state of stress
and strain for a linear elastic material is given by
 τ
1 1
W (ε) = τ (ε) dε = τ ε = Eε2
0 2 2

which gives the total work per unit of volume at a point performed by the
stress throughout the deformation, that is, from a stress-free state to the
current state. Then, the total strain energy is defined as

U(ε) = W (ε) dV.
V

Also recall that the strain energy depends only on the current state of
deformation and gives the energy stored as elastic deformation.
Considering a 3-D elastic solid, we can generalize the above concepts
defining the internal work per unit of volume or the strain energy density by
 5.3 Strain and potential energy in 3-D 391

1
W (εxx , εyy , εzz , γxy , γxz , γyz ) = (τxx εxx + τyy εyy
2
1 T
+τzz εzz + τxy γxy + τyz γyz + τzx γzx ) = ε τ
2
1 T
= ε Cε
2
and the total strain energy by
 
1
U(ε) = W (ε) dV = εT Cε dV.
V 2 V
Sometimes we use the notation

a(v, v) = 2 U(ε(v)) = ( ∂ε v )T C∂ε v dV (5.43)
V

which gives twice the strain energy associated with the displacement field v.

5.3.2 The total potential energy

The total potential of an elastic solid subjected to a field of body forces f B

in V and a field of surface tractions f S on Sf is defined as

Π(v) = U (ε (v)) − P (v)

where v is a compatible displacement field, that is, a sufficiently smooth dis-

placement field (the strains ε (v) = ∂ε v can be evaluated) which satisfies
the kinematic boundary conditions on Su , and P (v) is the potential of the
external loads, that is,
 
P (v) = vT f B dV + vT f S dS.
V Sf

There are two important results regarding the total potential energy which
we discuss below.
(i) The stationarity condition δΠ = 0 is equivalent to the principle of vir-
tual work and the displacement field u which makes Π stationary is the
solution of the elasticity problem.

Proof
Let us consider a variation δu from a compatible displacement field u.
Let δε be the strain corresponding to δu. Then

1  T
δΠ = δε C ε (u) + εT (u) CT δε dV
2 V
392 5. The principle of virtual work

 
− δu f dV −
T B
δuT f S dS = 0.
V Sf

Since C is symmetric, we obtain from the above equation

  
δεT C ε (u) dV = δuT f B dV + δuT f S dS. (5.44)
V V Sf

The variation δu is arbitrary provided that u+δu is a compatible dis-

placement field. Hence, δu = 0 on Su , since u satisfies the displacement
boundary conditions on Su .
The variation δu can be interpreted as a virtual displacement field since
it is an arbitrary displacement field which is zero on Su . Since (5.44) holds
for an arbitrary virtual displacement, it corresponds to the principle of
virtual work in terms of displacements and, hence, the displacement u is
the solution of the elasticity problem.

(ii) The total potential energy attains its minimum value for the solution u
of the elasticity problem.

Proof
Let us evaluate the potential energy for the displacement field u+δu
where u is the solution of the elasticity problem and δu an arbitrary
variation. Also, let ε and δε be the strain corresponding to the displace-
ments u and δu, respectively.

1 T
Π(u+δu) = (ε+δε) C (ε+δε) dV
2 V
 
 T   T 
− u +δuT f B dV − u +δuT f S dS
V Sf

  
1
= εT C ε dV − uT f B dV − uT f S dS (5.45)
2 V V Sf
  
+ δε C ε dV −
T
δu f dV −
T B
δuT f S dS
V V Sf

1
+ δεT C δε dV
2 V

Considering that δΠ = 0, equation (5.45) becomes


1
Π(u+δu) = Π (u) + δεT C δε dV.
2 V

Note that
 5.3 Strain and potential energy in 3-D 393


1
U (δu) = δεT C δε dV
2 V

is the strain energy corresponding to the displacement variation. The

strain energy is always greater than or equal to zero and when the body is
properly supported the strain energy is strictly positive for any compati-
ble displacement field. Therefore, we can conclude that the total potential
energy attains its minimum at the solution of the elasticity problem.


We note that the definition of the total potential energy
  
1
Π(u) = ( ∂ε u )T C∂ε u dV − uT f B dV − uT f S dS
2 V V Sf

is quite general and is valid for all mathematical models discussed in Chapters
3 and 4 provided, of course, that the variables are those defined in Tables
4.3 and 4.4. Hence the above results (i) and (ii) are also valid for all these
models.