Equations of Motion, Momentum and Energy For Deformable Solids

Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

Home

Chapter 6

Equations of Motion, Momentum and Energy for Deformable Solids

6.1 Linear and angular momentum balance equations for a deformable solid

Deformable solids are governed by the same physical laws (Newton’s laws) as a system of particles.   You will
hopefully recall that, for a particle system in which particles can only interact by exerting forces on one another
(and cannot exert moments on one another):
1.           The net external force acting on the system is equal to the time derivative of the total linear
momentum of the system
2.           The total external moment acting on the system is equal to the rate of change of its total angular
momentum
3.      The rate of work done by external forces is equal to the sum of the rate of change of kinetic energy of
the system and the rate of work done by internal forces.

A deformable solid can be thought of as an infinite number of infinitesimal particles.   For a system of this
kind, the balance laws can be re-written as a set of partial differential equations, as outlined in the sections to
follow.

6.1.1 Linear momentum balance in terms of


Cauchy stress

Let σij   denote the Cauchy stress distribution


within a deformed solid.  Assume that the solid
is subjected to a body force bi , and let
ui , vi  and ai  denote the displacement, velocity
and acceleration of a material particle at
position yi  in the deformed solid.

Newton’s third law of motion (F=ma) can be expressed as


∂σij
∇y ⋅ σ + ρb = ρa     or          ∂ y + ρ bj = ρ aj
i
Written out in full
∂σ11 ∂σ21 ∂σ31
∂y1
+ ∂y2
+ ∂y3
+ ρb1 = ρa1
∂σ12 ∂σ22 ∂σ32
∂y1
+ ∂y2
+ ∂y3
+ ρb2 = ρa2
∂σ13 ∂σ23 ∂σ33
∂y1
+ ∂y2
+ ∂y3
+ ρb3 = ρa3
Note that the derivative is taken with respect to position in the actual, deformed solid. For the special (but
rather common) case of a solid in static equilibrium in the absence of body forces
∂σij
∂yi
=0

Derivation: Recall that the resultant force acting on an arbitrary volume of material V within a solid is

Pi = ∫ Ti (n)dA+ ∫ ρ bi dV
A V
where T(n) is the internal traction acting on the surface A with normal n that bounds V.

The linear momentum of the volume V is


Λi = ∫ ρ vi dV
V
where v is the velocity vector of a material particle in the deformed solid. Express T in terms of σij and set
Pi = dΛi /dt

⎪ ⎫

∫ σji nj dA+ ∫ ρ bi dV = ⎨∫
dt ⎩
ρ vi dV ⎬

d
⎪ ⎪
A V V
Apply the divergence theorem to convert the first integral into a volume integral, and note that one can show
(see Appendix D) that

⎪ ⎫

⎨∫
dt ⎩
ρ vi dV ⎬ = ∫ ρai dV

d
⎪ ⎪
V V
so
∂σji ∂σji
∫ dV + ∫ ρ bi dV = ∫ ρai dV ⇒ ∫ ( + ρ bi − ρ ai ) dV = 0
∂yj ∂yj
V V V V
Since this must hold for every volume of material within a solid, it follows that
∂σji
∂yj
+ ρbi = ρai
as stated.

6.1.2 Angular momentum balance in terms of Cauchy stress

Conservation of angular momentum for a continuum requires that the Cauchy stress satisfy
σji = σij
i.e. the stress tensor must be symmetric.

Derivation: write down the equation for balance of angular momentum for the region V within the  deformed
solid

⎪ ⎫

∫ y × T dA + ∫ y × ρ bdV = ⎨∫
dt ⎩
y × ρ vdV ⎬

d
⎪ ⎪
A V V
Here, the left hand side is the resultant moment (about the origin) exerted by tractions and body forces acting
on a general region within a solid.  The right hand side is the total angular momentum of the solid about the
origin.

We can write the same expression using index notation



⎪ ⎫

∫ ∈ijk yj Tk dA + ∫ ∈ijk yj bk ρ dV = ⎨∫
dt ⎩
yj vk ρ dV ⎬

d
⎪ ⎪
∈ijk
A V V
Express T in terms of σij and re-write the first integral as a volume integral using the divergence theorem

∫ ∈ijk yj Tk dA = ∫ ∈ijk yj σmk nm dA = ∫ (∈ijk yj σmk ) dV
∂xm
A A V
∂σmk
= ∫ ∈ijk (δjm σmk + yj ) dV
∂xm
V
We may also show (see Appendix D) that

⎪ ⎫

⎨∫
dt ⎩
∈ijk yj vk ρ dV ⎬ = ∫ ∈ijk yj ak ρ dV

d
⎪ ⎪
V V
Substitute the last two results into the angular momentum balance equation to see that
∂σmk
∫ ∈ijk (δjm σmk + yj ) dV + ∫ ∈ijk yj bk ρ dV = ∫ ∈ijk yj ak ρ dV
∂xm
V V V
∂σmk
⇒ ∫ ∈ijk δjm σmk = − ∫ ∈ijk yj ( + ρbk − ρak ) dV
∂ym
V V
The integral on the right hand side of this expression is zero, because the stresses must satisfy the linear
momentum balance equation.  Since this holds for any volume V, we conclude that
∈ijk δjm σmk =∈ijk σjk = 0
⇒∈imn ∈ijk σjk = 0
⇒ (δjm δkn − δmk δnj ) σjk = 0
⇒ σmn − σnm = 0
which is the result we wanted.

6.1.3 Equations of motion in terms of other stress measures

In terms of nominal and material stress the balance of linear momentum is


Sij
∇ ⋅ S + ρ0 b = ρ0 a ∂xi
+ ρ0 bj = ρ0 aj
∂(Σik Fjk )
∇ ⋅ [Σ ⋅ F T ] + ρ0 b = ρ0 a ∂xi
+ ρ0 bj = ρ0 aj
Note that the derivatives are taken with respect to position in the undeformed solid.

The angular momentum balance equation is


⥄⥄⥄⥄⥄⥄⥄⥄ F ⋅ S = [F ⋅ S]T ⥄⥄⥄⥄⥄⥄⥄ Σ = Σ T

To derive these results, you can start with the integral form of the linear momentum balance in terms of
Cauchy stress

⎪ ⎫

∫ σji nj dA+ ∫ ρ bi dV = ⎨∫
dt ⎩
ρ vi dV ⎬

d
⎪ ⎪
A V V
Recall (or see Appendix D for a reminder) that area elements in the deformed and undeformed solids are
related through
dAni = J Fki−1 n0k dA0
In addition, volume elements are related by dV = JdV0 .  We can use these results to re-write the integrals as
integrals over a volume in the undeformed solid as

⎪ ⎫

∫ σji J Fkj−1 n0k dA0 + ∫ ρ bi JdV0 = ⎨∫
dt ⎩
ρ vi JdV0 ⎬

d
⎪ ⎪
A0 V0 V0
Finally, recall that Sij = J Fik−1 σkj  and that Jρ = ρ0  to see that

⎪ ⎫

∫ S ki n0k dA0 + ∫ ρ 0 bi dV0 = ⎨∫
dt ⎩
ρ 0 vi d V 0 ⎬

d
⎪ ⎪
A0 V0 V0
Apply the divergence theorem to the first term and rearrange
∂Sji
∫ ( ) d V0 = 0
dvi
+ ρ0 bi − ρ0
∂xj dt
V
Once again, since this must hold for any material volume, we conclude that
Sij
∂xi
+ ρ0 bj = ρ0 aj
The linear momentum balance equation in terms of material stress follows directly, by substituting into this
equation for Sij  in terms of Σij

The angular momentum balance equation can be derived simply by substituting into the momentum balance
equation in terms of Cauchy stress σij = σji

6.1.4 Equations of motion and equilibrium equations for small deformations

The general equations of motion for a deformable solid are hard to solve,
because the shape of the solid must be calculated as part of the solution.  
In many engineering applications, the deformation is small enough to be
neglected.   If this is the case, we can simplify the calculations as follows
1. We neglect the differences between the various stress measures
described in Section 6.1.3.  
2. We neglect the difference between the density of the deformed and
undeformed solid
3. We satisfy the equations of motion (or equilibrium) on the undeformed
solid, instead of the deformed solid. 

Accordingly, let xi  denote the position of a material particle in the undeformed solid, and let ρ0  denote the
mass density of the undeformed material.   Assume that the solid is subjected to a body force b per unit
mass.   Newton’s third law of motion (F=ma) can be expressed as
∂σij
∇ ⋅ σ + ρ0 b = ρ0 a     or          ∂ x + ρ0 bj = ρ0 aj
i
Written out in full
∂σ11 ∂σ21 ∂σ31
∂x1
+ ∂x2
+ ∂x3
+ ρ0 b1 = ρ0 a1
∂σ12 ∂σ22 ∂σ32
∂x1
+ ∂x2
+ ∂x3
+ ρ0 b2 = ρ0 a2
∂σ13 ∂σ23 ∂σ33
∂x1
+ ∂x2
+ ∂x3
+ ρ0 b3 = ρ0 a3
For the special (but rather common) case of a solid in static equilibrium in the absence of body forces
∂σij
∂xi
=0

Example: The stress field


−3Pk yk yi yj
σij = R = √−
y−−−
k yk
4πR5
represents the stress in an infinite, incompressible linear elastic solid that is subjected to a point force with
components Pk   acting at the origin (you can visualize a point force as a very large body force which is
concentrated in a very small region around the origin).

(a)    Verify that the stress field is in static equilibrium

This is a tedious exercise in index notation

( )
∂σij 3Pk δi k yi yj yk δii yj yk yi δij yk yi yj yi
∂yi
=− 4π
+ + −5 =0
R5 R5 R5 R6 R

(b)    Consider a spherical region of material centered at the origin.  This region is subjected to (1) the body
force acting at the origin; and (2) a force exerted by the stress field on the outer surface of the sphere.  
Calculate the resultant force exerted on the outer surface of the sphere by the stress, and show that it is
equal in magnitude and opposite in direction to the body force.

The traction acting on the exterior surface is ti = σij ni  and a unit vector normal vector to a sphere radius
R centered at the origin is ni = yi /R.   The resultant force is thus
yk yj
Fj = −3
P
4π k
∫ dA
R3
S
The integral clearly vanishes for k ≠ j by symmetry.   Choosing k=j=3 without loss of generality we can
evaluate the remaining integral in spherical-polar coordinates as
2π π
(R cos θ)2 2
F3 = −3
P
4π 3
∫ ∫ R sin θdθdϕ = −P3
R3
0 0

Example Let ϕ  be a twice differentiable, scalar function of position.  Derive a plane stress field from
ϕ  by setting
∂2 ϕ ∂2 ϕ ∂2 ϕ
σ11 = σ22 = σ12 = σ21 = − ∂y
∂y22 ∂y12 1 ∂y2

Show that this stress field satisfies the equations of stress equilibrium with zero body force.

The equilibrium formulas (with zero body force) are


∂σ11 ∂σ21 ∂σ31
∂y1
+ ∂y2
+ ∂y3
=0
∂σ12 ∂σ22 ∂σ32
∂y1
+ ∂y2
+ ∂y3
=0
∂σ13 ∂σ23 ∂σ33
∂y1
+ ∂y2
+ ∂y3
=0
The third equation is satisfied trivially.   The others can be verified by substitution
= ∂y∂ ( ) + ∂y∂ (− ∂y )=0
∂σ11 ∂σ21 ∂σ31 ∂2 ϕ ∂2 ϕ
∂y1
+ ∂y2
+ ∂y3 1 ∂y22 2 1 ∂y2

= ∂y∂ (− ∂y ) + ∂y∂ ( )=0


2 2
∂σ12 ∂σ22 ∂σ32 ∂ ϕ ∂ ϕ
∂y1
+ ∂y2
+ ∂y3 1 1 ∂y2 2 ∂y12

Example: A prismatic concrete column of mass density ρ  supports its


own weight, as shown below.  (Assume that the solid is subjected to a
uniform gravitational body force of magnitude g per unit mass). Show
that the stress distribution
σ22 = −ρg(H − x2 )
satisfies the equations of static equilibrium
∂σij
∂xi
+ ρ bj = 0
and also satisfies the boundary conditions σij ni =0
on all free boundaries.

The body force (per unit mass) is b = −ge2  .   For this case, the equilibrium equations (written out in
full) are
∂σ11 ∂σ21 ∂σ31
∂y1
+ ∂y2
+ ∂y3
=0
∂σ12 ∂σ22 ∂σ32
∂y1
+ ∂y2
+ ∂y3
− ρg = 0
∂σ13 ∂σ23 ∂σ33
∂y1
+ ∂y2
+ ∂y3
=0
The first and last equatinos are satisfied automatically, and substituting for the stress into the second
and evaluating the derivative shows that that, too, is satisfied.

On the sides of the column, the normal vectors point in the ±e1 , ±e2   directions.   Since
σ1i = 0, σ3i = 0, σ12 = σ32 = 0  it follows that nj σji = 0 on the sides.     The stress is zero at
x2 = H  , so the tractions on the top face are also zero.

6.2 Work done by stresses

In this section, we derive formulas that enable you to


calculate the work done by stresses acting on a solid. 

6.2.1 Work done by Cauchy stresses

Consider a solid with mass density ρ0   in its initial


configuration, and density ρ   in the deformed solid. Let
σij   denote the Cauchy stress distribution within the
solid.  Assume that the solid is subjected to a body force bi (per unit mass), and let ui , vi  and ai  denote the
displacement, velocity and acceleration of a material particle at position yi  in the deformed solid. In addition,
let
Dij = 12 ( ∂yi + )
∂v ∂vj
j ∂yi
denote the stretch rate in the solid.

The rate of work done by Cauchy stresses per unit deformed volume is then σij Dij .  This energy is either
dissipated as heat or stored as internal energy in the solid, depending on the material behavior.

We shall show that the rate of work done by internal forces acting on any sub-volume V bounded by a surface
A in the deformed solid can be calculated from

⎪ ⎫

⎨∫ ρvi vi dV ⎬
1
ṙ = ∫ Ti vi dA + ∫ ρbi vi dV = ∫ σij Dij dV +
(n)
dt ⎩ ⎭
d
⎪ 2 ⎪
A V V V
Here, the two terms on the left hand side represent the rate of work done by tractions and body forces acting
on the solid (work done = force x velocity).  The first term on the right-hand side can be interpreted as the work
done by Cauchy stresses; the second term is the rate of change of kinetic energy. 
(n)
Derivation: Substitute for Ti  in terms of Cauchy stress to see that

ṙ = ∫ + ∫ ρbi vi dV = ∫ nj σji vi dA + ∫ ρbi vi dV


(n)
Ti vi dA
A V A V
Now, apply the divergence theorem to the first term on the right hand side

ṙ = ∫ (σji vi ) dV + ∫ ρbi vi dV
∂yj
V V
Evaluate the derivative and collect together the terms involving body force and stress divergence
∂vi ∂σji
ṙ = ∫ {σji +( + ρbi ) vi } dV
∂yj ∂yj
V
Recall the equation of motion
∂σji
∂yj
+ ρbi = ρai
and note that since the stress is symmetric σij = σji
= 12 σij ( ∂yi + ) = σij Dij
∂v 1 ∂vi ∂v ∂vj
σji ∂yi = 2
(σij + σji ) ∂yj ∂yi
j j
to see that

ṙ = ∫ {σij Dij + ρai vi } dV


V
Finally, note that
dvi 1 d
∫ ρai vi dV = ∫ ρo vi dV0 = ∫ ρo ( vi vi ) d V 0
dt 2 dt
V V0 V0

⎛ ⎞ ⎛ ⎞
dt ⎜
ρ0 (vi vi ) dV0 ⎟ = dt ⎜
ρ0 (vi vi ) dV0 ⎟ =
1 1 1
= d
∫ d
∫ d
∫ ρ (vi vi ) dV
⎝V 2 ⎠ ⎝V 2 ⎠ 2
dt
0 0 V
Finally, substitution leads to

⎪ ⎫

⎨∫ ρvi vi dV ⎬
1
ṙ = ∫ Ti vi dA + ∫ ρbi vi dV = ∫ σij Dij dV +
(n)
dt ⎩ ⎭
d
⎪ 2 ⎪
A V V V
as required.

6.2.2 Rate of mechanical work in terms of other stress measures

 The rate of work done per unit undeformed volume by Kirchhoff stress is τij Dij
 The rate of work done per unit undeformed volume by Nominal stress is Sij F˙ji
 The rate of work done per unit undeformed volume by Material stress is Σij Ė ij

This shows that nominal stress and deformation gradient are work conjugate, as are material stress and
Lagrange strain.

In addition, the rate of work done on a volume V0  of the undeformed solid can be expressed as

⎪ ⎫

⎨∫ ρ0 vi vi dV0 ⎬
1
ṙ = ∫ + ∫ ρbi vi dV = ∫ τij Dij dV0 +
(n)
⎩ ⎭
d
⎪ ⎪
Ti vi dA dt 2

⎧ ⎫
A V V0 V0

⎪ ⎪
ṙ = ∫ Ti vi dA + ∫ ρbi vi dV = ∫ Sij F˙ji dV0 + dtd ⎨∫ ρ0 vi vi dV0 ⎬
(n) 1

⎪ 2 ⎭

⎧ ⎫
A V V0 V0

⎪ 1 ⎪
ṙ = ∫ Ti vi dA + ∫ ρbi vi dV = ∫ Σij Ėij dV0 + dt ⎨∫ ρ0 vi vi dV0 ⎬
(n)
⎩ 2 ⎭
d
⎪ ⎪
A V V0 V0

Derivations: The proof of the first result (and the stress power of Kirchhoff stress) is straightforward and is left
(n)
as an exercise.  To show the second result, note that Ti dA = n0j Sji dA0  and dV = JdV0  to re-write the
integrals over the undeformed solid; then and apply the divergence theorem to see that

ṙ = ∫ (Sji vi ) dV0 + ∫ ρbi vi JdV0
∂xj
V0 V0
Evaluate the derivative, recall that Jρ = ρ0  and use the equation of motion
Sij dvj
∂xi
+ ρ0 bj = ρ0 dt
to see that
∂vi dvi
ṙ = ∫ Sji dV0 + ∫ ρ0 vi d V 0
∂xj dt
V0 V0
Finally, note that ∂vi /∂xj = (∂u̇i /∂xj ) = F˙ij  and re-write the second integral as a kinetic energy term as
before to obtain the required result.

The third result follows by straightforward algebraic manipulations – note that by definition
Sij F˙ji = Σik Fjk F˙ji
Since Σij  is symmetric it follows that

Σik Fjk F˙ji = 12 (Σik + Σki ) Fjk F˙ji = Σik 12 (Fjk F˙ji + Fji F˙jk ) = Σik Ėik

6.2.3 Rate of mechanical work for infinitesimal deformations

For infintesimal motions all stress measures are equal; and all strain rate measures can be approximated by
the infinitesimal strain tensor ε.  The rate of work done by stresses per unit volume of either deformed or
undeformed solid (the difference is neglected) can be expressed as σij ε̇ ij , and the work done on a volume
V0  of the solid is

⎪ ⎫

⎨∫ ρ0 vi vi dV0 ⎬
1
ṙ = ∫ + ∫ ρbi vi dV = ∫ σij ε̇ ij dV0 +
(n)
⎩ ⎭
d
⎪ ⎪
Ti vi dA dt 2
A V V0 V0

6.3 The principle of Virtual Work

The principle of virtual work is an alternative way of expressing the equations of motion and equilibrium
derived in Section 6.1.  At first sight it appears to be fairly useless, but it turns out to be an extremely useful
way of deriving equilibrium equations or equations of motion for a solid in which the deformation is
approximated in some way – for example, in a beam, plate or shell.  In addition, the principle of virtual work is
also used as the starting point for finite element analysis for nonlinear solids, and so is a particularly important
result.
Some definitions are needed before the principle
can be stated.  Suppose that a deformable solid is
subjected to loading that induces a displacement
field u(x), and a velocity field v(x) .  The loading
consists of a prescribed displacement on part of
the boundary (denoted by S1   ), together with a
traction t (which may be zero in places) applied to
the rest of the boundary (denoted by S2   ).   The
loading induces a Cauchy stress σij .  The stress
field satisfies the angular momentum balance
equation σij = σji .

The principle of virtual work is a different way of re-writing partial differential equation for linear moment
balance
∂σji dvi
∂yj
+ ρbi = ρ dt
in an equivalent integral form, which is much better suited for computer solution.

To express the principle, we define a kinematically admissible virtual velocity field δv(y) , satisfying  
δv = 0  on S1 .  You can visualize this field as a small change in the velocity of the solid, if you like, but it is
really just an arbitrary differentiable vector field.  The term `kinematically admissible’ is just a complicated way
of saying that the field is continuous, differentiable, and satisfies δv = 0  on S1  - that is to say, if you perturb
the velocity by δv(y) , the boundary conditions on displacement are still satisfied.
In addition, we define an associated virtual velocity gradient, and virtual stretch rate as
δDij = 12 ( )
∂δvi ∂δvi ∂δvj
δLij = ∂yj ∂yj
+ ∂yi

The principal of virtual work may be stated in two ways.

6.3.1 First version of the principle of virtual work

The first is not very interesting, but we will state it anyway.  Suppose that the Cauchy stress satisfies:
1.      The boundary condition ni σij = tj  on S2
2.      The linear momentum balance equation
∂σji dvi
∂yj
+ ρbi = ρ dt
Then the virtual work equation
dv
∫ σ : δDdV + ∫ ρ ⋅ δvdV − ∫ ρb ⋅ δvdV − ∫ t ⋅ δvdA = 0
dt
V V V S
Or equivalently, with index notation
dvi
∫ σij δDij dV + ∫ ρ δvi dV − ∫ ρbi δvi dV − ∫ ti δvi dA = 0
dt
V V V S2
is satisfied for all virtual velocity fields.

Proof:  Observe that since the Cauchy stress is symmetric


σij δDij = 12 σij ( ) = 12 (σji ) = σji
∂δvi ∂δvj ∂δvi ∂δvj ∂δvi
∂yj
+ ∂yi ∂yj
+ σij ∂yi ∂yj
Next, note that
∂v ∂σji
σji ∂yi = ∂y∂ (σji δvi ) − ∂yj
δ vi
j j
Finally, substituting the latter identity into the virtual work equation, applying the divergence theorem, using the
linear momentum balance equation and boundary conditions on σ and δv(y)  we obtain the required result.

6.3.2 Second version of the principle of virtual work

The converse of this statement is much more interesting and useful.  Suppose that σij  satisfies the virtual
work equation
dv
∫ σ : δDdV + ∫ ρ ⋅ δvdV − ∫ ρb ⋅ δvdV − ∫ t ⋅ δvdA = 0
dt
V V V S
Or, using index notation
dvi
∫ σij δDij dV + ∫ ρ δvi dV − ∫ ρbi δvi dV − ∫ ti δvi dA = 0
dt
V V V S2
for all virtual velocity fields δv(y) .  Then the stress field must satisfy
3.      The boundary condition ni σij = tj  on S2
4.      The linear momentum balance equation
∂σji dvi
∂yj
+ ρbi = ρ dt
The significance of this result is that it gives us an alternative way to solve for a stress field that satisfies the
linear momentum balance equation, which avoids having to differentiate the stress.   It is not easy to
differentiate functions accurately in the computer, but it is easy to integrate them.  The virtual work statement
is the starting point for any finite element solution involving deformable solids.

Proof: Follow the same preliminary steps as before, i..e.


σij δDij = 12 σij ( + ∂y ) = 12 (σji ∂y i + σij ∂y ) = σji
∂δvi ∂δvj ∂δv ∂δvj ∂δvi
∂yj i j i ∂yj
∂vi ∂ ∂σji
σji ∂y = ∂y (σji δvi ) − ∂y δvi
j j j
and substitute into the virtual work equation
∂ ∂σji
∫ { δvi } dV + ∫ ρ
dvi
(σji δvi ) − δvi dV − ∫ ρbi δvi dV − ∫ ti δvi dA = 0
∂yj ∂yj dt
V V V S2
Apply the divergence theorem to the first term in the first integral, and recall that δv = 0  on S1 , we see that
∂σji
−∫ { } δvi dV + ∫ (σji nj − ti ) δvi dA = 0
dvi
+ ρbi − ρ
∂yj dt
V S2
Since this must hold for all virtual velocity fields we could choose
δvi = f(y) { ∂y + ρbi − ρ }
∂σji dvi
j dt
where f(y) = 0 is an arbitrary function that is positive everywhere inside the solid, but is equal to zero on S . 
For this choice, the virtual work equation reduces to
∂σji ∂σki
− ∫ f (y) { }{ } dV = 0
dvi dvi
+ ρbi − ρ + ρbi − ρ
∂yj dt ∂yk dt
V
and since the integrand is positive everywhere the only way the equation can be satisfied is if
∂σji dvi
∂yj
+ ρbi = ρ dt
Given this, we can next choose a virtual velocity field that satisfies
δvi = (σji nj − ti )
on S2 .  For this choice (and noting that the volume integral is zero) the virtual work equation reduces to

+ ∫ (σji nj − ti ) (σki nk − ti ) dA = 0
S2
Again, the integrand is positive everywhere (it is a perfect square) and so can vanish only if
σji nj = ti
as stated.

6.3.3 The Virtual Work equation in terms of other stress measures.

It is often convenient to implement the virtual work equation in a finite element code using different stress
measures. 

To do so, we define
∂ui
1.      The actual deformation gradient in the solid Fij = δij + ∂xj
2.      The virtual rate of change of deformation gradient  δF˙ij =
∂δvi ∂δv
F = ∂x i
∂yk kj j

3.      The virtual rate of change of Lagrange strain δĖ ij = 2 (Fki δF˙kj + δF˙ki Fkj )
1

In addition, we define (in the usual way)


1.      Kirchhoff stress  τij = J σij
2.      Nominal (First Piola-Kirchhoff) stress Sij = J Fik−1 σkj
−1
3.      Material (Second Piola-Kirchhoff) stress Σij = J Fik σkl Fjl−1

In terms of these quantities, the virtual work equation may be expressed as


dvi
∫ τij δDij dV0 + ∫ ρ0 δvi dV0 − ∫ ρ0 bi δvi dV0 − ∫ ti δvi dA = 0
dt
V0 V0 V0 S2
dvi
∫ Sij δF˙ji dV0 + ∫ ρ0 δvi dV0 − ∫ ρ0 bi δvi dV0 − ∫ ti δvi dA = 0
dt
V0 V0 V0 S2
dvi
∫ Σij δĖij dV0 + ∫ ρ0 δvi dV0 − ∫ ρ0 bi δvi dV0 − ∫ ti δvi dA = 0
dt
V0 V0 V0 S2
Note that all the volume integrals are now taken over the undeformed solid –  this is convenient for computer
applications, because the shape of the undeformed solid is known.  The area integral is evaluated over the
deformed solid, unfortunately.  It can be expressed as an equivalent integral over the undeformed solid, but the
result is messy and will be deferred until we actually need to do it.

6.3.4 The Virtual Work equation for infinitesimal deformations.

For infintesimal motions, the Cauchy, Nominal, and Material stress tensors are equal; and the virtual stretch
rate can be replaced by the virtual infinitesimal strain rate
δε̇ ij = 12 ( ∂x i + )
∂δv ∂δvj
j ∂xi
There is no need to distinguish between the volume or surface area of the deformed and undeformed solid. 
The virtual work equation can thus be expressed as
dvi
∫ σij δε̇ ij dV0 + ∫ ρ0 δvi dV0 − ∫ ρ0 bi δvi dV0 − ∫ ti δvi dA0 = 0
dt
V0 V0 V0 S2
for all kinematically admissible velocity fields.
As a special case, this expression can be applied to a quasi-static state with vi = 0. Then, for a stress state
σij  satisfying the static equilibrium equation σij /dxi + ρ0 bj = 0  and boundary conditions σij nj = ti  on
S2 , the virtual work equation reduces to
∫ σij δεij dV0 = ∫ ρ0 bi δui dV0 + ∫ ti δui dA
V0 V0 S2
In which δui  are kinematically admissible displacements components (δui = 0 on S2) and
δεij = (∂δui /xj + ∂δuj /xi ) /2 .

Conversely, if  the stress state σij  satisfies ∫ σij δεij dV0 = ∫ ρ0 bi δui dV0 + ∫ ti δui dA for every set of
V0 V0 S2
kinematically admissible virtual displacements, then the stress state σij  satisfies the static equilibrium
equation σij /dxi + ρ0 bj = 0 and boundary conditions σij nj = ti  on S2 . 

Example: The shell shown in the figure is subjected to a radial body force
b = ρb(R)eR , and a radial pressure pa , pb   acting on the surfaces at
R = a  and R = b. The loading induces a spherically symmetric state of
stress in the shell, which can be expressed in terms of its components in a
spherical-polar coordinate system as
σRR eR ⊗ eR + σθθ eθ ⊗ eθ + σϕϕ eϕ ⊗ eϕ   (or if you prefer matrix
notation
⎡ σRR 0 0 ⎤
σ = ⎢ 0 σθθ 0 ⎥
⎣ 0 0 σϕϕ ⎦

(a) By considering a virtual velocity of the form δv = w(R)eR , (i.e. all points in the sphere can only move in
the radial direction, but they can move by an arbitrary displacement) show that the stress state is in static
equilibrium if

b b

∫ {σRR + (σθθ + σϕϕ ) } 4πR2 dR − ∫ b(R)w(R)4πR2 dR − 4πa2 pa w(a)+4πb2 pb w(b) = 0


dw w
dR R
a a
for all w(R).

(b) Hence, show that the stress state must satisfy


dσRR 1
dR
+ R
(2σRR − σθθ − σϕϕ ) + b = 0 σRR = −pa (R = a) σRR = −pb (R = b)

The principle of virtual work will usually magically give you a simplified form of the general equilibrium
equation for any simplified kind of deformation.   There is a standard process to follow, which involves two
steps: (i) substitute a virtual displacement field that has the same form as the simplified kind of motion;
and then (usually) (ii) integrate the virtual work equation by parts to remove any derivatives of the
displacement field.  

We will use this method here.   To evaluate the virtual work equation, we must first calculate the virtual
stretch rate δD .  By definition D = {∇v + (∇v)T } /2 .   We can use the formula from Section 3.5.8 to
calculate the gradient in spherical polar coordinates
⎡ ⎤
∂vR 1 ∂vR ∂vR vϕ
⎡ dR 0⎤
vθ 1
− −
⎢ ⎥
dw
∂R R ∂θ R R sin θ ∂ϕ R 0
∇v ≡ ⎢
⎢ ⎥
⎥ =⎢
⎢ 0 0⎥⎥
∂vθ 1 ∂vθ ∂vθ
⎢ ⎥
vR 1 vϕ
+ − cot θ R w

⎢ ⎥ ⎣ w ⎦
∂R R ∂θ R R sin θ ∂ϕ R

⎣ ⎦ 0 0
∂vϕ 1 ∂vϕ 1 ∂vϕ v vR
∂R R ∂θ R sin θ ∂ϕ
+ cot θ Rθ + R
R

This is symmetric already, so D = ∇v in this case.   It follows that


dw w
σ : δD = σRR dR + (σθθ + σϕϕ ) R

The remaining terms in the virtual work equation are standard dot products, and the volume integral can be
re-written as an integral with respect to the radial coordinate R, so the virtual work principle therefore
reduces to
∫ σ : δDdV − ∫ ρb ⋅ δvdV − ∫ t∗ ⋅ δvdA =
V V S
b b

∫ {σRR + (σθθ + σϕϕ ) } 4πR2 dR − ∫ b(R)w(R)4πR2 dR − 4πa2 pa w(a)+4πb2 pb w(b) = 0


dw w
dR R
a a

In this form, the equation does not tell us anything interesting.   However, if we integrate the first term
by parts, we can re-write the virtual work equation as
b
1 d(R2 σRR ) 1
∫ { + (σθθ + σϕϕ ) − b(R)} w4πR2 dR
R2 dR R
a
−4πa2 (σRR (a) + pa )w(a) + 4πb2 (σRR (b) + pb )w(b) = 0
This must vanish for all w.   Since this includes any w with w=0 at R=a,R=B, it follows that
b
1 d(R2 σRR ) 1
∫ { 2
+ (σθθ + σϕϕ ) − b(R)} w4πR2 dR = 0
R dR R
a
This must vanish even if we happen to pick
2
1 d(R σRR ) 1
w= + (σθθ + σϕϕ ) − b(R)
R2 dR R
which requires that
b 2
1 d(R2 σRR ) 1
∫ { + (σθθ + σϕϕ ) − b(R)} 4πR2 dR = 0
R2 dR R
a
The integrand is now positive or zero everywhere, so the integral can only be zero if the integrand is
zero.   Therefore
dσRR 1
dR
+ R
(2σRR − σθθ − σϕϕ ) + b = 0
Since the integrand vanishes, we now are left with
−4πa2 (σRR (a) + pa )w(a) + 4πb2 (σRR (b) + pb )w(b) = 0
Again, this condition must be satisfied for all w, which is only possible if
σRR = −pa (R = a) σRR = −pb (R = b)

You might also like