Nonlinear Solid Mechanics

Andrew Hazel
Introduction

Typically, want to determine the response of a solid body to

an applied load.

If a solid body is not rigid, then it can deform.

Introduction

Typically, want to determine the response of a solid body to

an applied load.

If a solid body is not rigid, then it can deform.

@
@

Introduction

Typically, want to determine the response of a solid body to

an applied load.

If a solid body is not rigid, then it can deform.

@
@

Question: How do we measure the (nite) deformation of a

deformable body?
Lagrangian coordinates

We parametrise the position vector to any material point

within the undeformed body, r by Lagrangian coordinates,
i
.
Tangent vector
g
i
=
r

i
Lagrangian coordinates

As the body deforms, the Lagrangian coordinates remain

attached to the same material points in the body.
Tangent vector
g
i
=
r

i
Deformed position, R(
i
)
Tangent vector
G
i
=
R

i
The metric tensor

A vector line segment dr = g

i
d
i
has length (squared)
(ds)
2
= drdr = g
i
d
i
g
j
d
j
= g
ij
d
i
d
j
,
where
g
ij
= g
i
g
j
,
is called the (covariant) metric tensor.

g
ij
expresses information about the length of material lines in
the undeformed body.

If the global and Lagrangian coordinates coincide g

ij
=
ij
.
The metric tensor

A vector line segment dr = g

i
d
i
has length (squared)
(ds)
2
= drdr = g
i
d
i
g
j
d
j
= g
ij
d
i
d
j
,
where
g
ij
= g
i
g
j
,
is called the (covariant) metric tensor.

g
ij
expresses information about the length of material lines in
the undeformed body.

If the global and Lagrangian coordinates coincide g

ij
=
ij
.
The strain tensor

An objective measure of the deformation (strain) is given by

the Green strain tensor

ij
=
1
2
(G
ij
g
ij
) .

g
ij
is the metric tensor of the undeformed conguration.

G
ij
is the metric tensor of the deformed conguration.

Decompose the deformed position into

R(
i
) = r(
i
) + u(
i
),
The strain tensor

An objective measure of the deformation (strain) is given by

the Green strain tensor

ij
=
1
2
(G
ij
g
ij
) .

Decompose the deformed position into

R(
i
) = r(
i
) + u(
i
),
where u(
i
) is the displacement eld and 1, then
G
i
= g
i
+
u

i
G
ij
= g
ij
+
_
g
i

j
+g
j

i
_
+O(
2
).
The strain tensor

An objective measure of the deformation (strain) is given by

the Green strain tensor

ij
=
1
2
(G
ij
g
ij
) .

Decompose the deformed position into

R(
i
) = r(
i
) + u(
i
),

ij
=
1
2
_
g
i

j
+g
j

i
_
The strain tensor

An objective measure of the deformation (strain) is given by

the Green strain tensor

ij
=
1
2
(G
ij
g
ij
) .

Decompose the deformed position into

R(
i
) = r(
i
) + u(
i
),

ij
=
1
2
_
g
i

j
+g
j

i
_

If
i
are chosen to be global Cartesian coordinates

ij

1
2
_
u
i
x
j
+
u
j
x
i
_
, the innitessimal strain tensor.
Example

Consider the simple radial expansion of a unit cube

0
i
= x
i
1, with deformed position given by
R = 2 r = 2 x.

Lagrangian coordinates are global Cartesian coordinates so

g
i
=
x
x
i
= e
i
(a unit vector) g
ij
=
ij
.

The deformed metric tensor is

G
i
=
2x
x
i
= 2e
i
G
ij
= 4
ij
.

Hence the strain tensor is

ij
=
3
2

ij
,

Note that the innitessimal strain tensor (with = 1) would

give
ij

ij
An aside: Non-Cartesian tensors

Starting from the position vector as a function of the

Lagrangian coordinates r(
i
), we found the tangent vectors
g
i
=
r

i
.

For a general set of coordinates,

i
, these vectors are not
necessarily orthonormal,
g
i
g
j
=
_
1 i = j
0 otherwise

We dene another set of vectors g

j
so that
g
i
g
j
=
j
i
=
_
1 i = j
0 otherwise

The up-down index notation is used to simplify notation.

An aside: Non-Cartesian tensors

We decompose r into global Cartesian base vectors

r =

k
r
k
e
k
,
where e
k
is a unit vector in the k-th global Cartesian direction.

Hence,
g
i
=
r
k

i
e
k
,

It follows that
g
j
=

j
r
n
e
n
,
... from which we deduce that
g
i
= g
ij
g
j
and g
j
= g
ji
g
i
, where g
ji
= g
j
g
i
.
Forces and loads

A deformable body is typically subject to surface loads (or

tractions), t, and body forces, F.

The stress vector t on a surface S within the strained body

is dened by
t = lim
S0
F
S
,
where F is the (statically equivalent) force acting on the
surface.

t represents the force per unit area exerted by the material

located to one side of the surface on that to the other.
The stress tensor

We consider a force balance on an innitessimal tetrahedron

in the deformed conguration.

Three faces are aligned with the planes

i
= const and
spanned by the other two covariant (lower index) tangent
vectors G
j
.

The vector representation of the face

i
= const is
G
i
S
i
2

G
ii
,
where G
i
/
_
(G
ii
) is a unit vector normal to the face and the
area of the face is S
i
/2.

Note that we have had to use the contravariant (upper index)

vector to ensure orthogonality.
The stress tensor

Vector representation of the remaining face is nS/2 and so

nS =

i
G
i
S
i
_
(G
ii
)
.

Decomposing the normal n = n

i
G
i
, then
n
i
_
(G
ii
)S = S
i
.

If the tetrahedron is in equilibrium then

tS =

i
t
i
S
i
,
The stress tensor

Vector representation of the remaining face is nS/2 and so

nS =

i
G
i
S
i
_
(G
ii
)
.

Decomposing the normal n = n

i
G
i
, then
n
i
_
(G
ii
)S = S
i
.

If the tetrahedron is in equilibrium then

tS =

i
t
i
S
i
,
The stress tensor

Vector representation of the remaining face is nS/2 and so

nS =

i
G
i
S
i
_
(G
ii
)
.

Decomposing the normal n = n

i
G
i
, then
n
i
_
(G
ii
)S = S
i
.

If the tetrahedron is in equilibrium then

tS =

i
t
i
S
i
,
t =

i
n
i
_
(G
ii
) t
i
,
The stress tensor
t =

i
n
i
_
(G
ii
)t
i
,

t is invariant if n remains the same (Cauchys principle).

However, n
i
are components of a covariant vector so t
i
_
(G
ii
)
must be contravariant.

In other words
t
i
_
(G
ii
) =
ij
G
j
.

The quantity
ij
is called the stress tensor.

Physical components of the stress tensor are obtained by

expressing the stress vectors in terms of unit tangent vectors
t
i
=

(ij )
G
j
/
_
(G
jj
)
(ij )
=
_
(G
jj
)/(G
ii
)
ij
.
The stress tensor
t =

i
n
i
_
(G
ii
)t
i
,

t is invariant if n remains the same (Cauchys principle).

However, n
i
are components of a covariant vector so t
i
_
(G
ii
)
must be contravariant.

In other words
t
i
_
(G
ii
) =
ij
G
j
.

The quantity
ij
is called the stress tensor.

Physical components of the stress tensor are obtained by

expressing the stress vectors in terms of unit tangent vectors
t
i
=

(ij )
G
j
/
_
(G
jj
)
(ij )
=
_
(G
jj
)/(G
ii
)
ij
.
Rate of Work

In the strained body, the total rate of work is

RW =
__
t

RdS +
___
_
F

R
_

RdV,
where the

R is the velocity of the material and

R is its the
acceleration.
Rate of Work

In the strained body, the total rate of work is

RW =
__
t

RdS +
___
_
F

R
_

RdV,
where the

R is the velocity of the material and

R is its the
acceleration.
RW =
__
T
i

R
n
i

G
dS +
___
_
F

R
_

RdV,
where T
i
= t
i
_
(GG
ii
) and G = det G
ij
.
Rate of Work

In the strained body, the total rate of work is

RW =
__
t

RdS +
___
_
F

R
_

RdV,
where the

R is the velocity of the material and

R is its the
acceleration.
RW =
__
T
i

R
n
i

G
dS +
___
_
F

R
_

RdV,
where T
i
= t
i
_
(GG
ii
) and G = det G
ij
.

Now use the divergence theorem (see Green & Zerna)

__
a
i
n
i
dS =
___
1

i
(a
i

G) dV.
Rate of Work

The rate of work becomes

RW =
___
1

i
_
T
i


R
_
+
_
F

R
_

RdV,
Rate of Work

The rate of work becomes

RW =
___
1

i
_
T
i


R
_
+
_
F

R
_

RdV,
=
___
1

G
T
i

i
+
1

G
_
T
i

i
+

G
_
F

R
_
_

RdV,
Rate of Work

The rate of work becomes

RW =
___

ij
G
j

i
dV
Rate of Work

The rate of work becomes

RW =
___

ij
G
j

i
dV

By symmetry of the stress tensor

RW =
1
2
___

ij
_
G
j

i
+G
i

j
_
dV
Rate of Work

The rate of work becomes

By symmetry of the stress tensor

RW =
1
2
___

ij
_
G
j

i
+G
i

j
_
dV

But recall

ij
=
1
2
(G
ij
g
ij
) ,
so

ij
=
1
2

t
(G
i
G
j
) =
1
2
_
G
i

j
+


R

i
G
j
_
.

Hence
RW =
___

ij

ij
dV.
Rate of Work

The rate of work becomes

By symmetry of the stress tensor

RW =
1
2
___

ij
_
G
j

i
+G
i

j
_
dV

But recall

ij
=
1
2
(G
ij
g
ij
) ,
so

ij
=
1
2

t
(G
i
G
j
) =
1
2
_
G
i

j
+


R

i
G
j
_
.

Hence
RW =
___

ij

ij
dV.

ij
and
ij
are a work conjugate pair.
The principle of virtual diplacements

Consider a deformable body that is load by a surface traction

T and a body force F

The body is subject to a virtual displacement R

The principle of virtual diplacements

Consider a deformable body that is load by a surface traction

T and a body force F

The body is subject to a virtual displacement R

The virtual work induced by the displacement is

___ _
F

2
R
t
2
_
RdV +
__
TRdS
The principle of virtual diplacements

Consider a deformable body that is load by a surface traction

T and a body force F

The body is subject to a virtual displacement R

The virtual work induced by the displacement is

___ _
F

2
R
t
2
_
RdV +
__
TRdS
which must be balanced by a change of internal energy
___

ij

ij
dV
The principle of virtual diplacements

Consider a deformable body that is load by a surface traction

T and a body force F

The governing variational principle is

___

ij

ij
dV
___ _
F

2
R
t
2
_
RdV
__
TRdS = 0.
The principle of virtual displacements

The variation of the strain is given by

ij
=
1
2
_
G
i

j
+
R

i
G
j
_
.

Using the symmetry of the stress tensor we can write the

variational principle as
___

ij
R

i

R

j

_
F

2
R
t
2
_
RdV
__
TRdS = 0.
Working with the variational principle

The integral is over the deformed domain, which is unknown.

It is much more convenient to integrate over the undeformed

domain.

We already know the mapping from undeformed to deformed

domain R(
i
).

The Jacobian of the mapping is

_
G/g.

Hence dV =
_
G/g dV
0
and
Working with the variational principle

The integral is over the deformed domain, which is unknown.

It is much more convenient to integrate over the undeformed

domain.

We already know the mapping from undeformed to deformed

domain R(
i
).

The Jacobian of the mapping is

_
G/g.

Hence dV =
_
G/g dV
0
and
___

G
g

ij
R

i

R

G
g
_
F

2
R
t
2
_
RdV
0

__
TRdS = 0.
Working with the variational principle

The integral is over the deformed domain, which is unknown.

It is much more convenient to integrate over the undeformed

domain.

We already know the mapping from undeformed to deformed

domain R(
i
).

The Jacobian of the mapping is

_
G/g.

Hence dV =
_
G/g dV
0
and
___

ij
R

i

R

j

_
f
0

2
R
t
2
_
RdV
0

__
TRdS = 0.
where
ij
=
_
G/g
ij
is the second PiolaKirchho stress tensor,
f is the body force per unit undeformed volume,

0
is the undeformed density.
Finite element approximation of the Largange coordinates

The undeformed position is given by

r(
i
) = r
k
(
i
)e
k
.

If there are no special symmetries, choose Lagrangian

coordinates as the global Cartesian coordinates.

Then components of undeformed position are r

i
(
i
) =
i
.

The tangent vectors are Cartesian base vectors

g
i
= e
i

Undeformed metric tensor is Kronecker delta g

ij
=
ij
.

Approximate the Lagrangian coordinates by a nite element

basis

i
=

i
l

l
,

i
l
is the i -th Lagrangian coordinate at the l -th node.
Finite element approximation of the Largange coordinates

The undeformed position is given by

r(
i
) = r
k
(
i
)e
k
.

If there are no special symmetries, choose Lagrangian

coordinates as the global Cartesian coordinates.

Then components of undeformed position are r

i
(
i
) =
i
.

The tangent vectors are Cartesian base vectors

g
i
= e
i

Undeformed metric tensor is Kronecker delta g

ij
=
ij
.

Approximate the Lagrangian coordinates by a nite element

basis

i
=

i
l

l
,

i
l
is the i -th Lagrangian coordinate at the l -th node.
Finite element approximation of the Largange coordinates

The undeformed position is given by

r(
i
) = r
k
(
i
)e
k
.

If there are no special symmetries, choose Lagrangian

coordinates as the global Cartesian coordinates.

Then components of undeformed position are r

i
(
i
) =
i
.

The tangent vectors are Cartesian base vectors

g
i
= e
i

Undeformed metric tensor is Kronecker delta g

ij
=
ij
.

Approximate the Lagrangian coordinates by a nite element

basis

i
=

i
l

l
,

i
l
is the i -th Lagrangian coordinate at the l -th node.
Finite element approximation of the Largange coordinates

The undeformed position is given by

r(
i
) = r
k
(
i
)e
k
.

If there are no special symmetries, choose Lagrangian

coordinates as the global Cartesian coordinates.

Then components of undeformed position are r

i
(
i
) =
i
.

The tangent vectors are Cartesian base vectors

g
i
= e
i

Undeformed metric tensor is Kronecker delta g

ij
=
ij
.

Approximate the Lagrangian coordinates by a nite element

basis

i
=

i
l

l
,

i
l
is the i -th Lagrangian coordinate at the l -th node.
Finite element approximation of the Largange coordinates

The undeformed position is given by

r(
i
) = r
k
(
i
)e
k
.

If there are no special symmetries, choose Lagrangian

coordinates as the global Cartesian coordinates.

Then components of undeformed position are r

i
(
i
) =
i
.

The tangent vectors are Cartesian base vectors

g
i
= e
i

Undeformed metric tensor is Kronecker delta g

ij
=
ij
.

Approximate the Lagrangian coordinates by a nite element

basis

i
=

i
l

l
,

i
l
is the i -th Lagrangian coordinate at the l -th node.
Finite element approximation of the variational principle

Use the same basis function for the unkown positions

(isoparametric mapping)
R
k
=

R
k
l

l
.

The basis functions are xed so

R =

R
k
l

l
e
k
and
R

j
=

R
k
l

j
e
k
.

Thus the principle of virtual displacements becomes

l
___ _

ij
R
k

j

_
f
k

2
R
k
t
2
_

l
_

R
k
l
d
1
d
2
d
3

__
T
k

R
k
l
dS = 0.
Finite element approximation of the variational principle

Use the same basis function for the unkown positions

(isoparametric mapping)
R
k
=

R
k
l

l
.

The basis functions are xed so

R =

R
k
l

l
e
k
and
R

j
=

R
k
l

j
e
k
.

Thus the principle of virtual displacements becomes

l
___ _

ij
R
k

j

_
f
k

2
R
k
t
2
_

l
_

R
k
l
d
1
d
2
d
3

__
T
k

R
k
l
dS = 0.
Finite element approximation of the variational principle

The discrete variations may be taken outside the integrals

l
____ _

ij
R
k

j

_
f
k

2
R
k
t
2
_

l
_
d
1
d
2
d
3

__
[T
k

l
] dS
_

R
k
l
= 0.

The variations of the nodes are independent, so the terms in

braces give one discrete equation for each nodal unknown.
___ _

ij
R
k

j

_
f
k

2
R
k
t
2
_

l
_
d
1
d
2
d
3

__
[T
k

l
] dS = 0.

These may be assembled in an element-by-element manner.

Finite element approximation of the variational principle

The discrete variations may be taken outside the integrals

l
____ _

ij
R
k

j

_
f
k

2
R
k
t
2
_

l
_
d
1
d
2
d
3

__
[T
k

l
] dS
_

R
k
l
= 0.

The variations of the nodes are independent, so the terms in

braces give one discrete equation for each nodal unknown.
___ _

ij
R
k

j

_
f
k

2
R
k
t
2
_

l
_
d
1
d
2
d
3

__
[T
k

l
] dS = 0.

These may be assembled in an element-by-element manner.

Summary of the method

Divide the undeformed domain into elements.

For each element compute the contribution to the discrete

volume residual
R
kl
=
___ _

ij
R
k

j

_
f
k

2
R
k
t
2
_

l
_
d
1
d
2
d
3

Note that the integral is over the Lagrangian coordinates

(undeformed domain).

Loop over the surfaces to add any tractions

__
[T
k

l
] dS.

Note that this integral is over the deformed surface

Assemble the contributions into a global residuals vector.

Compute the Jacobian (by nite dierences if necessary).

Solve the linear system.

Summary of the method

Divide the undeformed domain into elements.

For each element compute the contribution to the discrete

volume residual
R
kl
=
___ _

ij
R
k

j

_
f
k

2
R
k
t
2
_

l
_
d
1
d
2
d
3

Note that the integral is over the Lagrangian coordinates

(undeformed domain).

Loop over the surfaces to add any tractions

__
[T
k

l
] dS.

Note that this integral is over the deformed surface

Assemble the contributions into a global residuals vector.

Compute the Jacobian (by nite dierences if necessary).

Solve the linear system.

Summary of the method

Divide the undeformed domain into elements.

For each element compute the contribution to the discrete

volume residual
R
kl
=
___ _

ij
R
k

j

_
f
k

2
R
k
t
2
_

l
_
d
1
d
2
d
3

Note that the integral is over the Lagrangian coordinates

(undeformed domain).

Loop over the surfaces to add any tractions

__
[T
k

l
] dS.

Note that this integral is over the deformed surface

Assemble the contributions into a global residuals vector.

Compute the Jacobian (by nite dierences if necessary).

Solve the linear system.

Constitutive Laws

Assembling the residuals requires knowledge of the stress

tensor.

For an elastic material, the stress depends only on the current

state of strain

ij
(
jk
).

The specic relationship between stress and strain is known as

a constitutive law.

Given a constitutive law and a compressible material then we

simply compute

ij
_
1
2
(G
ij
g
ij
)
_
,
at all the integration points within the element.

What about incompressible materials?

Incompressible Solid Mechanics

If a solid material is incompressible, its volume cannot change

det G
ij
= det g
ij
(1)

Enforce the condition (1) by a Lagrange multiplier that plays

the role of a pressure so that

ij
= p G
ij
+
ij
(
kl
),
where
ij
is the deviatoric part of the stress tensor.

The resulting linear system is again of saddle-point type and

so the combinations of pressure/velocity nite elements
appropriate for NavierStokes equations can be reused.

If material is nearly incompressible it is also advantageous

to use a mixed formulation.
Incompressible Solid Mechanics

If a solid material is incompressible, its volume cannot change

det G
ij
= det g
ij
(1)

Enforce the condition (1) by a Lagrange multiplier that plays

the role of a pressure so that

ij
= p G
ij
+
ij
(
kl
),
where
ij
is the deviatoric part of the stress tensor.

The resulting linear system is again of saddle-point type and

so the combinations of pressure/velocity nite elements
appropriate for NavierStokes equations can be reused.

If material is nearly incompressible it is also advantageous

to use a mixed formulation.
Incompressible Solid Mechanics

If a solid material is incompressible, its volume cannot change

det G
ij
= det g
ij
(1)

Enforce the condition (1) by a Lagrange multiplier that plays

the role of a pressure so that

ij
= p G
ij
+
ij
(
kl
),
where
ij
is the deviatoric part of the stress tensor.

The resulting linear system is again of saddle-point type and

so the combinations of pressure/velocity nite elements
appropriate for NavierStokes equations can be reused.

If material is nearly incompressible it is also advantageous

to use a mixed formulation.
Example problem: Compression of a block

A square block of material is compressed.

What are appropriate boundary conditions?

u = 0

xx
= 0

xy
= 0

xx
= 0

xy
= 0

Fix the base vertically (and horizontally?)

Leave the sides traction free (do nothing).

Example problem: Compression of a block

A square block of material is compressed.

What are appropriate boundary conditions?

u = 0

xx
= 0

xy
= 0

xx
= 0

xy
= 0
u = Ke
y

Fix the base vertically (and horizontally?)

Leave the sides traction free (do nothing).

Impose a displacement on the top.

Example problem: Compression of a block

A square block of material is compressed.

What are appropriate boundary conditions?

u = 0

xx
= 0

xy
= 0

xx
= 0

xy
= 0

ij
n
j
= pn
i

Fix the base vertically (and horizontally?)

Leave the sides traction free (do nothing).

Impose a pressure load on the top.

Example problem: Compression of a block

A square block of material is compressed.

What are appropriate boundary conditions?

u = 0

xx
= 0

xy
= 0

xx
= 0

xy
= 0

ij
n
j
= (0, L)

Fix the base vertically (and horizontally?)

Leave the sides traction free (do nothing).

Impose a vertical load on the top.

Example problem: Compression of a block

A square block of material is compressed.

What are appropriate boundary conditions?

u = 0
u
x
= 0

xy
= 0
u
x
= 0

xy
= 0

ij
n
j
= (0, L)

Fix the base vertically (and horizontally?)

Constrain sides horizontally

Impose a vertical load on the top.

Incompressible material?
Example problem: Compression of block

Incompressible Mooney-Rivlin material

Loaded on top by P cos(x)N

Example problem: Compression of block

Incompressible Mooney-Rivlin material

Loaded on top by P cos(x)N

Example problem: Compression of block

Incompressible Mooney-Rivlin material

Loaded on top by P cos(x)N

Example problem: Compression of block

Incompressible Mooney-Rivlin material

Loaded on top by P cos(x)N

Example problem: Compression of block

Incompressible Mooney-Rivlin material

Loaded on top by P cos(x)N

Example problem: Compression of block

Incompressible Mooney-Rivlin material

Loaded on top by P cos(x)N

u
x
= 0,
xy
= 0
u
y
= 0,
xy
= 0

xx
= 0

xy
= 0
Summary of the method

Choose type of element (pick an LBB stable one if

incompressible problem).

Choose a timestepper (must compute second derivatives).

Generate mesh in undeformed solid domain.

Specify boundary and initial conditions

For Dirichlet conditions, replace the discrete weak form.

For traction conditions, assemble the surface integral.

Loop over elements and assemble the global residuals and

Jacobian matrix for each time step

Solve the (non)linear residual equations using Newtons

method.

Repeat for as many timesteps as desired.