Handout_L2_MECH6341

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 58

Contents

1 Stress analysis
1.1 Stress vector at a material point M 5
1.2 Stress tensor at a material point M 7
1.3 Governing equations 10
1.3.1 Static equilibrium resultant force . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.3.2 Static equilibrium resultant moment and symmetry . . . . . . . . . . . . . . . . . . . . . . . . 11
1.4 Spectral analysis of the stress tensor 12
1.4.1 Plane stress assumption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.4.2 Full three-dimensional stress state . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

2 Continuum kinematics and strain analysis


2.1 Body transformation 22
2.1.1 Gradient of the transformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.2 Cauchy-Green and Green-Lagrange tensors 24
2.2.1 Physical interpretation of the Green-Lagrange tensor . . . . . . . . . . . . . . . . . . . . . . 25
2.2.2 Expression with respect to the displacement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.2.3 Polar decomposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.2.4 Matrix form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.2.5 Spectral analysis of the strain tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.3 Infinitesimal strain theory and small deformation 28
2.3.1 Physical interpretation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.3.2 Polar decomposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.3.3 Strain compatibility conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

3 Constitutive equations and isotropic linear elasticity


3.1 Material Characterization and linear elasticity 35
3.2 Stiffness and compliance tensors 37
3.2.1 Voigt notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.2.2 Change of basis and material symmetries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.3 Formulation of the isotropic elasticity problem 42
3.3.1 Boundary conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.3.2 Superposition principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.4 Solution methods 44
3.4.1 Displacement approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.4.2 Stress approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.5 Plane elasticity 46
3.5.1 Plane strain assumption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.5.2 Plane stress assumption . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.5.3 Airy stress function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4 Table des matières

A Tensor algebra
A.1 Points and vectors 53
A.2 Index notation and Einstein convention 53
A.3 Tensors 55
A.3.1 Invariants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
A.3.2 Matrix form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
1
Stress analysis

1.1 Stress vector at a material point M


A general body denoted by  N D  [ @ where  D C [  is subjected to external
forces acting on its surface as depicted in figure 1.1. A fictitious plane … cuts the

linearly
distributed load … fs
f

force per
s
a unit area
M n
@ fn
a

C


idealized
concentrated force

Figure 1.1 – Structure of interested subjected to external forces and fictioulsy separated
by a plane …

body along surface a. One side of plane … is considered as positive, C and the
other side as negative,  . The portion of the body on the positive side of … exerts a
force on the portion of the body on the negative side. This force is transmitted through
the hypothetical plane … by direct contact forces of the parts of the body on the two
sides of …. In the current configuration, the force transmitted through a small area
a defined around a material point M.x/ of cartesian coordinates x D .x1 ; x2 ; x3 / by
the part on the positive side of … is denoted by f. The force f may be resolved
into components fn and fs , along unit normal n and unit tangent s, respectively, to
the plane …. The force fn is called the normal force on area a and fs is called
the shear force on a. The forces f, fn , and fs depend on the point M and the
orientation of plane … expressed by n. The concept of stress at a point is obtained
6 1. Stress analysis

by letting a become infinitesimal. The limiting ratio of f=a as a goes to zero


defines the stress vector t.M; n/ arising at point M in direction n:

f.M; n/
t.M; n/ D lim (1.1)
a!0 a
The stress vector t (also called the traction vector) always lies along the limiting
direction of the force vector f, which in general is neither normal nor tangent to
the plane …. These considerations imply that the balancing action of internal contact
forces generates a contact force density t.M; n/ that represents a distribution of internal
contact forces throughout the volume of the body in a particular configuration of the
body.
Cauchy’s postulate The stress vector t.M; n/ depends only on the position of the
material point M and on the local orientation of the surface element as defined by its
normal vector n.

Similarly, the limiting ratios of fn =a and fs =a define the normal stress vector
tn .M; n/ and the shear stress vector ts .M; n/ that act at a point M in the plane …:

fn .M; n/ fs .M; n/


tn .M; n/ D lim I ts .M; n/ D lim (1.2)
a!0 a a!0 a

Cauchy’s lemma Traction vectors acting on opposite sides of a surface are equal and
opposite. This can be expressed in vector form as:

t.M; n/ D t.M; n/ (1.3)

This is equivalent to Newton’s third law of action and reaction illustrated in figure 1.2.
In other words, it says that the action of  on C through plane … is opposite in

t.M; e2 / e3

dv

t.M; e2 / e2
dv C
e1

Figure 1.2 – Tractions acting on conceptually separated small portions of material


particles: t.M; e2 / D t.M; e2 /

direction and equal in amplitude to the action of C on  .


1.2 Stress tensor at a material point M 7

Notes

 The Cauchy stress refers to the current configuration, that is, it is a measure of force per
unit area acting on a surface in the current configuration.
 It should be noted that formula (1.1) is empirical, i.e. confirmed by experimental findings.
It formally defines the stress as a pointwise property.

As such, the stress vector is not a very attractive object since it depends on space through
point M as well as on the normal to the cutting plane, n. A counterpart depending
on M only is effectively more suitable.

1.2 Stress tensor at a material point M


At this point, the fundamental concept of stress vector is defined from a mathematical,
physical and continuum mechanics point of view. It characterizes the internal forces
acting within the investigated structure due to external loading. However, its main
drawback lies in the fact that it depends on a direction n, and in that sense, is not a field.
At any given point M , the traction will, in general, be different on different planes that
pass through the point and the internal forces will be completely characterised only if
all the possible directions n are considered. This yields practical issues and the stress
vector shall advantageously be replaced by the concept of stress tensor which eliminates
the difficulty that t is a function of two vectors, namely f and n.
To derive the full expression of this mathematical object, an infinitesimal tetrahedron
is introduced as illustrated in figure 1.3. The arbitrary slicing plane is chosen to be
e3

t.M; e1 /
t.M 0 ; n/

n
dz M0 e2

t.M; e2 / dy

dx

t.M; e3 /

e1
Figure 1.3 – Free body diagram of a Cauchy’s infinitesimal tetrahedral portion of
material
8 1. Stress analysis

a small inclined triangle close to the point M at which the state of stress needs to
be known. The static equilibrium of the tetrahedron can be derived from balancing
forces acting on it. Three faces of the tetrahedron have outward unit normal vectors
that coincide with e1 D . 1; 0; 0/, e2 D .0; 1; 0/, and e3 D .0; 0; 1/. The
inclined face of the tetrahedron has an outward unit normal vector n D .n1 ; n2 ; n3 / in
.e1 ; e2 ; e3 /. The area of the face with orthogonal unit vector n is taken to be da. The
areas of the three other faces with outward unit normal vectors n D ei equal:

dai D ni da (1.4)
 Exercise 1.1 Prove equality (1.4). 

Accordingly, the total force acting on one of these faces is:

t.M; ei / dai D t.M; ei / da ni (1.5)

Balancing forces acting on the tetrahedron leads to:

t.M 0 ; n/ da C t.M; e1 / da n1 C t.M; e2 / da n2 C t.M; e3 / da n3 D 0 (1.6)

Cancelling out the common area da, taking the limit M 0 ! M , and utilizing Cauchy’s
lemma implies:

t.M; n/ D t.M; e1 /n1 C t.M; e2 /n2 C t.M; e3 /n3 (1.7)

which is the essence of Cauchy’s theorem:


Theorem 1.1 — Cauchy’s theorem. The stress vector acting on each oriented surface
through a given point depends linearly upon the unit normal to the surface.

This theorem is a cornerstone to the subject of continuum mechanics for solids because
it yields the existence of the Cauchy stress tensor, central to the development of local
balance equations and to the formulation of constitutive theories.
 Exercise 1.2 In the above derivations involving Cauchy’s tetrahedron, body forces as well as
inertial forces were omitted. Show that expression (1.7) is still valid when these quantities are
incorporated into the formulation. 

A consequence of (1.7) is that there exists a tensor field commonly denoted by  .M /,


called the stress, such that:

t.M; n/ D  .M /  n (1.8)

Accordingly, the stress vector t can be written as the vector dot product of the stress
tensor,  , and the unit vector normal to the surface, n. Whereas t depends on n,  does
not; it is a function only of position in the body. This generalizes the familiar notion of
pressure, for which t.M; n/ D p.M /n to incorporate the possibility that the traction
acting at a point on a surface may include components perpendicular the the orientation
of the surface at that point.
1.2 Stress tensor at a material point M 9

Note In a basis of representation, the stress tensor can be written as a matrix, but a tensor
has specific physical properties that are more important compared to that of a regular matrix.
These properties relate to the manner in which the components of a tensor transform when the
coordinate system is changed.

When projected on the .e1 ; e2 ; e3 / basis, equality (1.7) reads:

t1 .M; n/ D t.M; n/  e1 D t.M; e1 /  e1 n1 C t.M; e2 /  e1 n2 C t.M; e3 /  e1 n3


t2 .M; n/ D t.M; n/  e2 D t.M; e1 /  e2 n1 C t.M; e2 /  e2 n2 C t.M; e3 /  e2 n3 (1.9)
t3 .M; n/ D t.M; n/  e3 D t.M; e1 /  e3 n1 C t.M; e2 /  e3 n2 C t.M; e3 /  e3 n3

where the notation ij D t.M; ei /  ej , i D 1; 2; 3 and j D 1; 2; 3 is adopted; this


finally yields the expression of the stress tensor in .e1 ; e2 ; e3 / as a matrix since:

t1 .M; n/
0 1 2 30 1
11 .M / 12 .M / 13 .M / n1
@t2 .M; n/A D 421 .M / 22 .M / 23 .M /5 @n2 A (1.10)
t3 .M; n/ 31 .M / 32 .M / 33 .M / n3

The M -dependence of  is kept in the notation to emphasized this fact. It will be


omitted in the sequel. The stress tensor displayed by its components stored, illustrated
in figure 1.4, in the stress matrix unequivocally defines the state of stress at an arbitrary
point within a deformable body.
e3
33
t.e3 /

23
13
32
31 t.e2 /

22
21 12
e2
11
t.e1 /
e1

Figure 1.4 – Tractions and stress components acting on surfaces of an infinitesimal


parallelepiped with normals in the coordinate directions. The center of the cube is
point M . Stresses are forces per unit area and are positive as drawn

 Exercise 1.3 The state of stress at a point M is given in the matrix form:
2 3
2 1 3
 D 41 2 25 (1.11)
3 2 1

Determine:
10 1. Stress analysis

1. the traction vector acting on a plane through the point whose unit normal in basis
.e1 ; e2 ; e3 / is n D .1; 2; 2/=3.
2. the component of this traction acting perpendicular to the plane.
3. the shear component of traction. 

Note The unit of stress is pascal Pa D N/m2 . The stress magnitude 1 Pa is produced by the
force 1 N which acts normal or parallel to a square metre large surface. As 1 N is a small force
and 1 m2 a large surface, 1 Pa is a very small stress. The old unit atmospheric pressure (1 bar)
corresponds to 100 kPa.

1.3 Governing equations


It is understood that neither body couples nor surface torques are acting on the explored
system. Only surface forces, ie the stress components times the surface area, and body
forces written in the form fb dxdydz where fb D .fb1 ; fb2 ; fb3 /, are present. Densities
of moments are rarely observed in practical problems since they stem from magnetic
fields.
The previous section was devoted to proving/showing the necessary existence of
internal forces within flexible structures. It is now time to derive their governing
equations, that will arise in the form of partial differential equations. To this end, the
infinitesimal cube in figure 1.4 is now looked at in the .y; z/ coordinate system with
associated basis .e2 ; e3 / as displayed in figure 1.5.

@33 dz
33 C
@z 2

@23 dz
23 C
@z 2

.z; e3 / @32 dy
32 C
@y 2
@22 dy @22 dy
22 dz .y; e2 / 22 C
@y 2 M.x; y; z/ @y 2
@32 dy
32
@y 2
dy

@23 dz
23
@z 2

@33 dz
33
@z 2

Figure 1.5 – Stress components and their first variation in the .y; z/ coordinate system

Note Notice that the continuous differentiability of the stress tensor field is now taken advantage
of. The first variation of the stress tensor will participate into the governing equations while the
1.3 Governing equations 11

main terms 11 , 22 , and 33 will cancel out. In the previous section, on the existence of the
stress field, only the main terms remained while the variation could be neglected.

1.3.1 Static equilibrium resultant force


The resultant force acting along e2 on the infinitesimal domain of interest reads:
 @22 dy  @22 dy 
22 C 22 dxdz
@y 2 @y 2
 @23 dz  @23 dz  (1.12)
C 23 C 23 dxdy
@z 2 @z 2
 @21 dx  @21 dx 
C 21 C 21 dydz C fb2 dxdydz D 0
@x 2 @x 2
Two similar equalities can be derived along e1 and e3 , respectively. They can all be
simplified and gathered together in the following system of partial differential equations
in ij :

@11 @12 @13


C C C fb1 D 0
@x @y @z
@21 @22 @23
C C C fb2 D 0 (1.13)
@x @y @z
@31 @32 @33
C C C fb3 D 0
@x @y @z

which is also written in the more compact form:

div  C fb D 0 (1.14)

 Exercise 1.4 For a system in static equilibrium, isolate a small volume of material  bounded
by surface †. Consider M 2  and P 2 † and provide a proof of equality (1.13) based on the
divergence theorem. 

1.3.2 Static equilibrium resultant moment and symmetry


From the above derivations provided in a fully three-dimensional context, the state
of stress at a point M is fully described by nine components in a given basis of
representation. With the adopted notation, the columns of the stress tensor are the
traction vectors along the coordinate axes.
Let us now derive the expression of the rotational equilibrium of the infinitesimal
element with respect to the point M.x; y; z/ along e1 :

dy  @32 dy  dy  @32 dy 
32 C dxdz 32 C dxdz
2 @y 2 2 @y 2
(1.15)
dz  @23 dz  dz  @23 dz 
23 C dxdy C 23 C dxdy D 0
2 @z 2 2 @z 2
12 1. Stress analysis

which yields 23 D 32 . Counterpart expressions along e2 and e3 leads to the symmetry
of the stress tensor in the absence of external distributed torques:

21 D 12
32 D 23 (1.16)
13 D 31

also known as the shear stress reciprocity property. The relation obtained shows
that the e2 component of the shearing stress exerted on a face perpendicular to e1 is
equal to the e1 component of the shearing stress exerted on a face perpendicular to e2 .
Accordingly, we note that, at a given point, shear cannot take place in one plane only;
an equal shearing stress must be exerted on another plane perpendicular to the first
one. Consequently the three-dimensional stress state depends on only six independent

e2 t.M; e2 / e2
t.M; e1 /
21

M
M 12 e1 e1

Figure 1.6 – Shear stress reciprocity property

components: three normal stresses and three shear stresses.


 Exercise 1.5 Prove the symmetry of the stress tensor by applying the mechanical law of
conservation of angular momentum. 

 Exercise 1.6 Prove the symmetry of the stress tensor by deriving the expression of the
rotational equilibrium of an infinitesimal element with respect to any point. 

1.4 Spectral analysis of the stress tensor


Suppose the basis .e1 ; e2 ; e3 / is changed to another one .e10 ; e20 ; e30 /. Can the primed
components be expressed in terms of the original ones? The answer is yes. All primed
stress components can be expressed in terms of the unprimed ones and of the direction
cosines of .e10 ; e20 ; e30 / with respect to .e1 ; e2 ; e3 /. This operation is called a stress
transformation.
Why should we look at this class of transformations? One important reason is
that material failure may depend on the maximum normal tensile stress (for brittle
materials) or the maximum absolute shear stress (for ductile materials). These stresses
have a simple expression with respect to the eigenvalues of the stress tensor. Once such
dangerous stress maxima are found for critical points of a given structure, the engineer
can determine strength safety factors.
1.4 Spectral analysis of the stress tensor 13

1.4.1 Plane stress assumption


For a general three-dimensional state, this operation is complicated because there are
three direction cosines. The transformations are simpler and more explicit in two
dimensions since changing axes depends on only one rotation angle. As a matter of
fact, in a large class of important problems, certain approximations may be applied
by analyzing the stresses that occur in a thin flat plate subjected to in-plane forces
only. A thin plate, as depicted in Fig. 1.7, is a prismatic member of a very small
thickness h. The middle surface of the plate, located halfway between its ends (faces)

z; e3
x; e1

y; e2

Figure 1.7 – Plane stress state: external forces belong to the .e1 ; e2 /-plane and all
stress components along e3 are assumed to be negligible

and parallel to them, may be taken as the .x; y/ plane. The thickness direction is then
coincident with the direction of the z axis. If the plate is not loaded on its faces, then
33 D 23 D 13 D 0 on its lateral surfaces z D ˙h=2. Since the plate is thin, as a
first approximation, it may be assumed that:

33 D 23 D 13 D 0 (1.17)

throughout the plate thickness z 2 Œ h=2; h=2. Also assumed is that the remaining
stress components 11 , 12 , and 22 are independent of z. With these approximations,
the stress tensor reduces to a function of the two variables .x; y/. It is called the tensor
of plane stress which has the following matrix form is .e1 ; e2 /:
 
11 .x; y/ 12 .x; y/
 .x; y/ D (1.18)
12 .x; y/ 22 .x; y/
The question is to obtain its components 10 10 , 10 20 , and 20 20 in .e10 ; e20 / expressed
with respect to its components 11 , 12 and 22 . Since the stress tensor reduces here
to a linear mapping from R2 to R2 , it is sufficient to express vectors t0 and n0 versus t
and n, respectively. Consider a vector v D .v1 ; v2 / in .e1 ; e2 /. Its components .v10 ; v20 /
in a new basis .e01 ; e02 / rotated by an angle  with respect to .e1 ; e2 / are expressed as:

cos  sin 
    
v10 v1
D , v0 D Pv , v D P> v0 (1.19)
v20 sin  cos  v2
14 1. Stress analysis

e20 e2

v2 v

v20 e10
v10


e1
v1

Figure 1.8 – Basis transformation

where, for rotation matrices, P 1


D P> . Accordingly, t0 D Pt and n0 D Pn and:

t D n , P> t0 D  P> n0 , t0 D P P> n0 , t0 D  0 n0 (1.20)

which equivalently means:

 0 D P P> (1.21)

A direct term-by-term identification of the stress components:

cos  sin  11 12 cos  sin 


     
10 10 10 20
D (1.22)
20 20 sin  cos  22 sin  cos 

leads to:
11 C 22 11 22
1 0 1 0 D C cos 2 C 12 sin 2
2 2
11 C 22 11 22
2 0 2 0 D cos 2 12 sin 2 (1.23)
2 2
22 11
1 0 2 0 D sin 2 C 12 cos 2
2

Principle stresses Tensor  will be diagonal if .e01 ; e02 / is rotated with respect to
.e1 ; e2 / by an angle ns satisfying:

212
10 20 .ns / D 0 ) tan 2ns D (1.24)
11 22

In this basis, the stress tensor takes the following matrix form:
 
 0
 D 1
0
(1.25)
0 2

where it is assumed by definition that 1 > 2 .


1.4 Spectral analysis of the stress tensor 15

Maximum shear Component 10 20 will be an extremum if .e01 ; e02 / is rotated with
respect to .e1 ; e2 / by an angle ms satisfying:

d10 20 22 11


.ms / D 0 ) tan 2ms D (1.26)
d 212

By definition, the maximum shear stress is denoted ms D 10 20 .ms /.


 Exercise 1.7 Show that the normal stresses generally do not vanish in the sections with
extreme shear stresses. 

 Exercise 1.8 Show that for ns , 1 and 2 are extrema. 

 Exercise 1.9 Show that ms D ns ˙ =4. Properly discuss the two possible configurations. 

 Exercise 1.10 Derive the expressions of 1 , 2 , and ms with respect to 11 , 12 , and 22 only.
What is the relationship between 1 , 2 , and ms ? 

 Exercise 1.11 The two principal stresses 1 D 40 MPa and 2 D 20 MPa of a plane stress
state are known. Determine the orientation of a system of axes with respect to the principal
axes for which 11 D 0 and 12 > 0. Calculate the stresses 22 and 12 . 

22 2
ms
12 1
12
ms
ns
M 11 M M

(a) plane stress in ba- (b) principal stress pa- (c) maximum shear
sis .e1 ; e2 / rameterized by ns parameterized by ms

Figure 1.9 – Components of the stress tensor in various bases

1.4.2 Full three-dimensional stress state


As quickly shown in a two-dimensional context, the principal stresses and principal
directions can be found by asking whether or not there are planes on which the traction
vector is purely normal, with no shear component, or equivalently, a basis where the
matrix of the stress tensor is diagonal. On such planes, the traction vectors are aligned
parallel to the outward unit vectors. As a consequence, such unit vectors are solutions
of an eigenvalue problem:

 n D n , t D n (1.27)
16 1. Stress analysis

This is the fundamental equation for determining eigenvalues (principal stress magni-
tudes) and eigenvectors (principal stress directions) of the stress tensor. Splitting it into
components results in the following linear system of equations:
2 30 1 0 1
11  12 13 n1 0
4 12 22  23 5 @ n2 D 0A
A @ (1.28)
13 23 33  n3 0
Non-zero admissible solutions of this linear, homogeneous set of equations can be
found only if the determinant of the matrix equals zero, i.e.:
ˇ ˇ
ˇ11   12  13
ˇ
(1.29)
ˇ ˇ
ˇ 21
ˇ 22  23 ˇˇ D 0
ˇ 31 32 33  ˇ

When the determinant is expanded out, it takes the form of a cubic equation in :

3 I1 2 C I2  I3 D 0 (1.30)

where .I1 ; I2 ; I3 / are called tensor invariants. Values of stress invariants are independent
of the coordinate system used.
 Exercise 1.12 Based on the above comment, explain the terminology tensor invariants for I1 ,
I2 , and I3 . 

The physical content of a stress tensor is reflected exclusively in the stress invariants.
For example, pressure in all directions, as is the case in the hydrostatic state of stress,
results from I1 . The three invariants of the stress tensor are given by:
I1 D 11 C 22 C 33
ˇ ˇ ˇ ˇ ˇ ˇ
ˇ11 12 ˇ ˇ11 13 ˇ ˇ22 23 ˇ
I2 D ˇˇ ˇ C ˇ ˇ C ˇ ˇ
21 22 ˇ ˇ31 33 ˇ ˇ32 33 ˇ
ˇ ˇ (1.31)
ˇ11 12 13 ˇ
ˇ ˇ
I3 D ˇˇ21 22 23 ˇˇ
ˇ31 32 33 ˇ

I1 is called the trace of the stress tensor and I2;3 can be computed from the coeffi-
cient determinants. The three solutions of the characteristic equation (1.30) are called
principal normal stresses .1 ; 2 ; 3 / and commonly denoted .1 ; 2 ; 3 /. The double
subscripts of the normal stress components can be reduced to single suffixes, since
the shear components per definition become zero. With the help of the fundamental
theorem of algebra, equation (1.30) can be written as:

. 1 /. 2 /. 3 / D 0 (1.32)

and after the transformation, the stress matrix has the following diagonal form:
2 3
1 0 0
 D 4 0 2 0 5 (1.33)
0 0 3
1.4 Spectral analysis of the stress tensor 17

The corresponding eigenvectors, commonly called principal axes of stress or prin-


cipal directions are calculated by successively replacing the found eigenvalues in
system (1.28) which becomes singular by definition. Hence, sequentially solving the
following singular system for i D 1; 2; 3:
2 30 1 0 1
11 i 12 13 n1 0
4 12 22 i 23 5 @ n2 D 0 A
A @ (1.34)
13 23 33 i n3 0

with a non-zero .n1 ; n2 ; n3 / gives access to the principal stress directions. There are
infinitely many solutions and once one is picked it is usually normalized. Orthogonal
planes on which shearing stresses vanish are called principal planes of stress. The
principal directions are chosen in such a way that the magnitudes 1 > 2 > 3 are
correctly organized.
This is all summarized in the spectral theorem in finite dimensions.
Theorem 1.2 — Spectral theorem. Any real and symmetric matrix M can be diago-
nalized by an orthogonal matrix P satisfying P 1
D P> such that:

D D PMP> (1.35)

where matrix D is diagonal.

Notes
 Other equivalent formulations of the same theorem exist where M D PDP> . The
difference lies in the definition of the matrix P. Here, we keep the convention used in the
plane stress study.
 Matrix P is essentially a rotation matrix that can be parameterized through Euler angles
in R3 .

The principal normal stress components .1 ; 2 ; 3 / can be visualized in three dimen-
sion using a rotated cube as depicted in Fig. 1.10 where principal stress directions
.e10 ; e20 ; e30 / are rotated with respect to the initial coordinate system .e1 ; e2 ; e3 /. While
six components of stress are necessary to define the state of stress in an arbitrary ori-
ented cube by virtue of the stress tensor symmetry, only three components of stress are
required in the rotated cube of principal stresses.
 Exercise 1.13 — Transformation of a stress matrix. Consider the stress matrix:
2 p 3
13 3 0
1 p
 D 4 3 15 0 5 (1.36)
4
0 0 12

1. Find the three principal stresses .1 ; 2 ; 3 / by solving the eigenvalue problem.
2. Compute the stress invariants .I1 ; I2 ; I3 / from principal stresses.
3. What are the direction cosines of the planes on which the principal stresses act?
4. The matrix formed by the nine components of the three eigenvectors describes what
kind of geometrical operation in space? 

 Exercise 1.14 — Diagonalization of a plane stress matrix. The state of plane stress in a
metal sheet is given by 11 D 64 MPa, 22 D 32 MPa and 12 D 20 MPa. Determine:
18 1. Stress analysis

(a) reference basis (b) principal directions

Figure 1.10 – Components of the three-dimensional stress tensor. Note that to ensure
an ordering of the eigenvalues with increasing magnitudes, a non-direct basis of
principal directions may be required.

1. the stresses in a section which is inclined at an angle of 60ı to e1 ;


2. the principal stresses and principal directions through the rotation matrix (subsec-
tion 1.4.1) and by direct diagonalisation (subsection 1.4.2);
3. the maximum shear stress and the associated directions of the sections.
Display the stresses at an element for each case. 

It is useful to have a way of presenting the stress tensor that clearly shows whether
or not there are any shear stresses acting at the point in question. To do so, the stress
tensor is decomposed into an isotropic contribution and a deviatoric contribution. The
isotropic part of the stress tensor is defined as:
1
ijiso D kk ıij (1.37)
3
while the deviatoric stress becomes:
1
ijdev D ij kk ıij (1.38)
3
Note that the total stress is then simply the sum:
ij D ijiso C ijdev (1.39)
The isotropic contribution is the same in all coordinate systems. It can also be shown
that the principal directions of the deviatoric stress are the same as those of the stress
tensor.
Summary A specific stress component acting on a specific slicing plane inside a
deformable body can be described by a stress vector. Three traction vectors are needed
to unequivocally define the state of stress at a point inside the body resulting in nine
components of the physical quantity stress tensor. Mathematically, the stress tensor
1.4 Spectral analysis of the stress tensor 19

can be written as a stress matrix representing all stress components acting on three
orthogonal slicing planes through a single, arbitrarily chosen body point M . Due to
the symmetry, only six stress components remain independent in the stress tensor:
three normal and three shear stresses. The stress unit is force per area N/m2 D Pa.
This tensor shall be advantageously diagonalized to express the maximum normal as
well as shear magnitudes acting at point M .
2
Continuum kinematics and strain
analysis

The previous chapter on stress analysis showed that a surface density of internal forces,
mathematically described through the stress tensor, arises within a flexible mechanical
component undergoing external forces. The six independent components of the stress
tensor should satisfy equilibrium equations in the form of three partial differential
equations with appropriate force boundary conditions. As such, this problem cannot be
solved and extra information is required. This is partially achieved in this chapter by
defining the strain tensor.
The theory of stress of a continuous medium rests solely on Newton’s laws. Simi-
larly, the theory of strain rests solely on geometric concepts. Both the theories of stress
and strain are, therefore, independent of material behaviour and, as such, are applicable
to the study of all materials. The word strain is used about local deformation in a
material, ie deformation in the neighbourhood of a particle. Strains represent change in
material lines, angles, and volume and are primarily due to mechanical stress and tem-
perature changes in the material. A physical visualization is displayed in Fig. 2.1 from
the reference to the current shapes which result from the deformation of a rectangle, a
triangle and a rectangular parallelepiped.
In this chapter we learn how to describe the motion of continuously distributed
matter, independent of what is causing the motion. Central to this discussion is the
displacement vector, u.X; t/ D x.X; t/ X. Since x is the position of a parcel whose
position at time t D 0 is X, the displacement u is simply the movement of the parcel
since the initial time1 .

Figure 2.1 – Deformation of an infinitesimal region

1
Note that in statics, time t is usually ignored and only transformations from the initial configuration
to the final one are investigated.
22 2. Continuum kinematics and strain analysis

2.1 Body transformation


In continuum mechanics, body transformations are characterized by considering two
configurations of a member: the initial configuration, or reference configuration, where
the body is assumed to be unloaded (undeformed and unstressed) and occupies a domain
denoted , and the current configuration where it is loaded (stressed and deformed) and
occupies domain  . As illustrated in figure 2.2, let .x; y; z/ be a rectangular coordinate
system. A particle M.X/ is located at the general coordinate point X  .X; Y; Z/

 ˆ


M M

dX2 dx2
M  .x/ dx1
M.X/ y
dX1
x

Figure 2.2 – General three-dimensional transformation

in the undeformed body. Under a deformation, the particle moves to a point M  .x/
located at the general coordinate point x  .x; y; z/ in the deformed state defined by
the equations:

x D x.X; Y; Z/
y D y.X; Y; Z/ (2.1)
z D z.X; Y; Z/

Equivalently, we assume the existence of a continuous and differentiable mapping ˆ


from the initial to the current configurations of the body2 :

x D ˆ.X; t/ (2.2)

which is invertible:

XDˆ 1
.x; t/ (2.3)

This physically means that:


2
A discontinuity of these functions would imply a rupture of the member.
2.1 Body transformation 23

 each initial point M has a unique image M  ;


 two different points M1 and M2 in the initial configuration cannot have the same
image.

Example 1 — Two-dimensional transformation. A two-dimensional and time-dependent


transformation is defined as:
1
x.X; Y; t / D .18t C 4X C 6tY /
4 (2.4)
1
y.X; Y; t / D .14t C .4 C 2t/Y /
4
Each initial point M.X; Y / is transformed to an image point M  .x; y/ through these two
functions.

Of particular interest is the variation in the displacement of neighboring particles. If all


particles move together, all displacements are identical. This corresponds to a uniform
translation of the entire body. The differences in displacement between neighboring
particles are related to more interesting things like rotation and deformation of the body.
These properties are properly described by the gradient of the transformation.
 Exercise 2.1 — Uniform translation. In two dimensions, provide the general mathematical
expression of a uniform translation. 

2.1.1 Gradient of the transformation


A small segment dX D .dX; dY; dZ/ separating the two points M.X; Y; Z/ and P .X C
dX; Y C dY; Z C dZ/ in the initial configuration is transformed through ˆ into a small
segment dx D .dx; dy; dz/ separating the two images M  .x; y; z/ and P  .x Cdx; y C
dy; z C dz/ in the deformed configuration. The definition of the mapping ˆ yields:

dx D ˆ.P / ˆ.M / D ˆ.X C dX/ ˆ.X/ (2.5)

which can be linearized through a Taylor’s expansion up to first order as:

@ˆ @ˆ
dx D ˆ.X/ C  dX ˆ.X/ D  dX (2.6)
@X @X
which leads us to the definition of the gradient tensor of the transformation at X:


F.X/ D .X/ (2.7)
@X
This tensor linearly transforms any infinitesimal vector dX in the undeformed config-
uration to another infinitesimal vector dx in the deformed configuration of the body.
This is illustrated in Fig. 2.2 for vectors dX1 and dx1 or vectors dX2 and dx2 , respec-
tively.
24 2. Continuum kinematics and strain analysis

Example 2 — Gradient tensor of a uniform transformation. Consider the transformation


of example 1. Its gradient tensor expressed in .e1 ; e2 / is:
 
1 2 3t
F.t / D (2.8)
2 0 2Ct

The transformation is uniform, or homogeneous, because the gradient is space-independent.

Example 3 — Rigid body transformation. Consider the transformation:

x.X; t/ D R.t/X C c.t/ (2.9)

where R.t/ denotes a rotation with respect to the origin and c.t/, a translation. We have:

F.t / D R.t/ (2.10)

 Exercise 2.2 — Pure shear. A three-dimensional deformation of the form:

x1 .X1 ; X2 ; X3 / D X1 C X2 ; 2R
x2 .X1 ; X2 ; X3 / D X2 (2.11)
x3 .X1 ; X2 ; X3 / D X3

is called a pure shear. For this transformation:


1. sketch a unit square and its image;
2. compute the matrices of F and C;
3. compute the list of principal invariants of C;
4. compute the principal stretches. 

 Exercise 2.3 — Plane strain. A three-dimensional deformation of the form:

x1 .X1 ; X2 ; X3 / D f .X1 ; X2 /
x2 .X1 ; X2 ; X3 / D g.X1 ; X2 / (2.12)
x3 .X1 ; X2 ; X3 / D X3

is called a plane strain. Show that for such a deformation the principal stretch 3 is unity. 

2.2 Cauchy-Green and Green-Lagrange tensors


Example 3 shows that the gradient tensor is not a proper measure of the deformation of a
mechanical body since it is not zero for a rigid body transformation. Instead, the correct
way to reflect the distorsion within a body is to analyze how the scalar product between
two initial infinitesimal vectors dX1 and dX2 emanating from point M is modified. By
definition, these two quantities are mapped through ˆ to dx1 and dx2 emanating from
M  respectively, hence3 :

dx>
1  dx2 D .F dX1 /  .F dX2 / D dX1  .F F/dX2 D dX1  CdX2
> > > >
(2.13)
3
Note that the transpose for vectors is theoretically useless if we see vectors as tensors of first order
since they are single-index quantities. It is kept here to ease the reading.
2.2 Cauchy-Green and Green-Lagrange tensors 25

where C D F>  F is called the right Cauchy-Green tensor. The difference between the
initial and modified scalar products gives rise to a new tensor E:

dx>
1  dx2 dX>
1  dX2 D dX1  CdX2
>
dX>
1  IdX2
D dX>
1  .C I/dX2 (2.14)
D dX>
1  2EdX2

This tensor E defined as:


1 1
E D .F> F I/ D .C I/ (2.15)
2 2
is the Green-Lagrange strain tensor and is a measure of deformation.

Example 4 For a rigid-body transformation, E D O.

Note Other strain measures are used in continuum mechanics. They all must predict zero strains
for arbitrary rigid-body motions, and must reduce to the infinitesimal strains if the nonlinear
terms are neglected.

2.2.1 Physical interpretation of the Green-Lagrange tensor


Diagonal terms We first consider two identical initial infinitesimal vectors defined
along direction N, dX1 D dX2 D dX D dL N, they will be mapped through ˆ to two
identical transformed vectors dx1 D dx2 D dx D d` n defined along direction n and
Eq. (2.14) becomes:

dx>  dx dX>  dX D dX> 2EdX (2.16)

which can be recast in:

d`2 dL2
D N> 2EN (2.17)
dL2
The diagonal terms of E will be isolated when direction N is aligned with the basis
vectors e1 , e2 , and e3 , respectively.

Example 5 — Term E11 . The choice N D e1 yields:

d`2 dL2
E11 D (2.18)
2dL2
which measures the relative length variation of an infinitesimal vector initially aligned with e1 .

Considering a vector N as a combination of the basis vectors is still acceptable but it


will combine several components of tensor E in expression (2.17), whose analysis will
thus not be straight forward.
26 2. Continuum kinematics and strain analysis

Off-diagonal terms Consider two infinitesimal vectors that are initially orthogonal
that is dX1 D dL1 N1 and dX2 D dL2 N2 with N> 1  N2 D 0. Once again, by definition
of the transformation undergone by the system of interest, they will be mapped to vectors
dx1 D d`1 n1 and dx2 D d`2 n2 , respectively, that are not necessarily orthogonal. Thus,
there is an angle such that n>1  n2 D sin and expression (2.17) simplifies to:

d`1 d`2
sin D N>
1 2EN2 (2.19)
dL1 dL2

The off-diagonal terms of E can be extracted by considering two orthogonal directions


that are initially aligned with the basis vectors.

Example 6 — Term E12 . The choice N1 D e1 and N2 D e2 yields:

d`1 d`2
E12 D sin 12 (2.20)
2dL1 dL2
which measures the angle variation between two infinitesimal vectors initially aligned with e1
and e2 . Note that angle variations are coupled to length variations as explained in exercise 2.4.

Fe2

12

e2 2 Fe1
=

=2
e1

Figure 2.3 – Shear term E12

The expressions of the remaining diagonal and off-diagonal terms of E are derived by
choosing the appropriate basis vectors ei , i D 1; 2; 3.

 Exercise 2.4 When N1 D e1 and N2 D e2 , show that:

d`1 d`2 p p
sin D sin 2E11 C 1 2E22 C 1 (2.21)
dL1 dL2

Note It is very important to notice that strains are dimensionless quantities.

2.2.2 Expression with respect to the displacement


By definition, the gradient tensor is:

@x @.X C u/ @u
FD D DIC (2.22)
@X @X @X
2.2 Cauchy-Green and Green-Lagrange tensors 27

and:

1
E D .F> F I/
2 
1  @u >  @u 
D IC IC I (2.23)
2 @X @X
 
1 @u  @u >  @u >  @u 
D C C
2 @X @X @X @X

 Exercise 2.5 — Symmetry of the Green-Lagrange tensor. Show that tensor E is a symmetric
tensor. 

2.2.3 Polar decomposition


From the polar-decomposition theorem A.4, we deduce that any transformed vector dX
can be decomposed as:

dx D F dX D R U dX D V R dX (2.24)

which physically means that dX first undergoes a rotation through R then an extension
through V, or vice versa through U and R, respectively.

2.2.4 Matrix form


In basis .e1 ; e2 ; e3 /, the Green Lagrange has the following matrix form:
2 3
E11 E12 E13
ŒE D 4 E22 E23 5 (2.25)
sym E33

with:
 
@u 1  @u 2  @v 2 @w 2
E11 D C C C
@X 2 @X @X @X
  
@v 1 @u 2  @v  2 @w 2
E22 D C C C
@Y 2 @Y @Y @Y
  
@w 1 @u 2  @v  2 @w 2
E33 D C C C
@Z 2 @Z @Z @Z (2.26)
1  @u @v @u @u @v @v @w @w 
E12 D C C C C
2 @Y @X @X @Y @X @Y @X @Y
1  @v @w @u @u @v @v @w @w 
E23 D C C C C
2 @Z @Y @Y @Z @Y @Z @Y @Z
1  @u @w @u @u @v @v @w @w 
E13 D C C C C
2 @Z @X @X @Z @X @Z @X @Z
28 2. Continuum kinematics and strain analysis

2.2.5 Spectral analysis of the strain tensor


By analogy with stress theory, the strain tensor can be diagonalized and fully investigated
at any point M . This is not detailed here. Still, the strain tensor shares similarities
with the stress tensor since both are symmetric. Accordingly, through any point M
in an undeformed member, there exist line elements defined along three mutually
perpendicular principal directions that remain perpendicular under the deformation.
The strains of these three line elements are called the principal strains at the point.

2.3 Infinitesimal strain theory and small deformation


The deformation theory developed in the previous section is purely geometrical and the
associated equations are exact. Nevertheless, in most applications of structural materials
like steel, aluminum, concrete, and wood, the strains, displacements and rotations are
small. For instance, elastic strains in mild steel under uniaxial stress are less than 0:001.
For instance, the displacement will be considerably less than a characteristic length
of the structure. This means that the change of the geometry of the structure, due to
the deformation, need not to be taken into account when the action of the forces is
considered. This framework, called small deformation or infinitesimal strain, is valid
when both the strains and rotations are small.
Note Within the small deformation approximation, initial and current coordinates shall not be
distinguished anymore, that is .X; Y; Z/  .x; y; z/, and @=@xi  @=@Xi .

The analysis of small deformation is significantly simpler than that for large strains
since the quadratic term in the Green-Lagrange tensor E is neglected, thus:
 
1 @u  @u >
ED C (2.27)
2 @x @x
which becomes in an expanded matrix form:

@u 1  @u @v 
11 D 12 D C
@x 2 @y @x
@v 1 @v
 @w 
22 D 23 D C (2.28)
@y 2 @z @y
@w 1 @u @w 

33 D 31 D C
@z 2 @z @x
 Exercise 2.6 — Finite versus small strains. A bar of length L originally along the X  x
axis (the reference configuration ) is rigidly rotated 90ı to lie along the Y  y axis while
retaining the same length (the current configuration  ). The origin X D Y D 0 stays at the
same location:
1. Sketch the transformation and display one point M.X; Y; Z/ and its image M  .x; y; z/.
2. Verify that the transformation from  to  is given by x D Y , y D X, and z D Z.
3. Obtain the displacement field u, the deformation gradient matrix F and the Green-
Lagrange axial strain E11 . Show that the Green-Lagrange measure correctly predicts
zero axial strain whereas the infinitesimal strain measure 11 predicts the absurd value
of 100% strain. 
2.3 Infinitesimal strain theory and small deformation 29

2.3.1 Physical interpretation


The physical interpretation of this new tensor  is almost identical to the interpretation
of E. The assumption that the strains are small allows for simplifying linearizations that
are detailed below.

Diagonal terms In the context of infinitesimal strains, we assume that dL  d` in


such a way that Eq. (2.17) simplifies to:

d`2 dL2 .d` dL/.d` C dL/ 2.d` dL/


 D D N>  2N (2.29)
dL2 dL2 dL
hence:
d` dL
D N>  N (2.30)
dL
from which we deduce that the diagonal terms of  measure the relative extensions
of elementary vectors initially aligned with the basis vectors e1 , e2 , and e3 respec-
tively.

Example 7 — Interpretation of 11 . The choice N D e1 yields:

d` dL
11 D (2.31)
dL
which measures the relative length variation of an infinitesimal vector initially aligned with e1 .

Off-diagonal terms We now consider two infinitesimal vectors that are initially
orthogonal that is dX1 D dL1 N1 and dX2 D dL2 N2 with N> 1  N2 D 0. Using
dL1  d`1 , dL2  d`2 , and sin  , the linearization of expression (2.19) leads to:

D N>
1  N2 (2.32)
2
Hence, in the context of small deformations, the off-diagonal terms of  capture half
of the angle reduction between basis vectors. This is much simpler than in finite
strains where elongations along the basis vectors are involved in the counterpart expres-
sion.

Example 8 — Interpretation of 12 . The choice N1 D e1 and N2 D e2 yields:


12
12 D (2.33)
2

The above approximations imply that the strains and rotations are small compared to
unity. The latter condition is not necessarily satisfied in the deformation of thin flexible
bodies, such as rods, plates, and shells. For these bodies the rotations may be large.
Consequently, the small-displacement theory must be used with caution: it is usually
applicable for massive (thick) bodies, but it may give results that may be seriously in
error when applied to thin flexible bodies.
30 2. Continuum kinematics and strain analysis

 Exercise 2.7 — Homogeneous transformation. A transformation is said to be homogeneous


when the associated gradient tensor is not space-dependent. Show that for such a deformation,
material planes and straight lines in the reference configuration deform into planes and straight
lines in the present configuration. 

 Exercise 2.8 Displacements were measured in a deformed body, which may be approximated
by the expressions:

u.x; y; z/ D 5x 2 C 3xy C 4 C 4y 2 C 3yz


v.x; y; z/ D 6xy C 4y 2 C 5 C 2z 2 (2.34)
w.x; y; z/ D 4xz C 2y 2 C 3y C 6z 2

Knowing that both deformations and rotations are sufficiently small to be considered as infinites-
imal, determine the functions describing the strains and the rotations in the body. 

2.3.2 Polar decomposition


The polar decomposition of the gradient tensor F in small deformation is also simplified:

@u
FDIC DICC! (2.35)
@x
where:
    @u > 
1 @u  @u > 1 @u
D C and ! D (2.36)
2 @x @x 2 @x @x
 Exercise 2.9 Expression (2.35) can be approximated as follows:

F D I C  C !  .I C /.I C !/ (2.37)

which can be read as the polar decomposition of F in small strains. Explain why the tensor I C 
is associated to an extension, ie U D V D I C  and why the tensor I C ! is associated to a
rotation, ie R D I C !. 

While the rotation tensor R is orthogonal in the general case of finite strains, it is
orthogonal only up to the first order in small strains:

.I C !/> .I C !/ D .I !/.I C !/ D I !!  I (2.38)


 Exercise 2.10 — Polar decomposition of tensors. A homogeneous deformation is given by:

x1 .X1 ; X2 ; X3 / D 2X1 =3 2X2 C 2X3


x2 .X1 ; X2 ; X3 / D 4X1 =3 C X2 (2.39)
x3 .X1 ; X2 ; X3 / D 4X1 =3 C X3

Determine:
1. The deformation gradient F and the deformation tensor C.
2. The principal values 2k and the principal directions nk of C.
3. The stretch tensor U and its inverse U 1 from the formulas:
3 3
p X X 1
UD CD k nk ˝ nk I U 1
D nk ˝ nk (2.40)
k
kD1 kD1
2.3 Infinitesimal strain theory and small deformation 31

4. The rotation tensor R from the relation R D FU 1 , the axis of rotation and the angle of
rotation.
5. The shear strain 23 and the extremal values of longitudinal strain. 

 Exercise 2.11 — Homogeneous deformation. A homogeneous deformation is given by:

x1 .X1 ; X2 ; X3 / D X1 2˛X2 C 2˛X3


x2 .X1 ; X2 ; X3 / D 2˛X1 C X2 (2.41)
x3 .X1 ; X2 ; X3 / D 2˛X1 C X3

with ˛ D ˛.t/:
1. Determine the deformation gradient F and the deformation tensor C.
2. Compute the principal values and principal directions of C. 

2.3.3 Strain compatibility conditions


Some problems are such that one can guess the form ij .x/ D j i .x/ of the strain field.
The stresses can be computed from the constitutive equations and equilibrium equations
and force boundary conditions can then be checked. However, there remains a question:
is it possible to find a displacement field u.x/  .u.x; y; z/; v.x; y; z/; w.x; y; z// such
that the strain-displacement relations (2.28) are satisfied? The answer to this question
is yes if compatibility equations are satisfied (the six components of a symmetric
strain tensor cannot be specified arbitrarily since a displacement vector has only three
components). In Cartesian coordinates, these (81) equations read:

@2 ij @2 kl @2 ik @2 j l


C D 0; i; j; k; l D 1; 2; 3 (2.42)
@xk @xl @xi @xj @xj @xl @xi @xk

It can be shown that only the following six equations are independent:

@2 11 @2 23 @2 12 @2 13


C D0
@x2 @x3 @x1 @x1 @x1 @x3 @x1 @x1
@2 22 @2 13 @2 21 @2 23
C D0
@x1 @x3 @x2 @x2 @x2 @x3 @x2 @x1
@2 33 @2 12 @2 31 @2 32
C D0
@x1 @x2 @x3 @x3 @x3 @x2 @x3 @x1
(2.43)
@2 11 @2 22 @2 12
C D2
@x2 @x2 @x1 @x1 @x1 @x2
@2 11 2
@ 33 @2 13
C D2
@x3 @x3 @x1 @x1 @x1 @x3
@2 22 2
@ 33 @2 23
C D2
@x3 @x3 @x2 @x2 @x1 @x2
These conditions are necessary and sufficient for a simply connected domain, otherwise
additional conditions must be written in order to ensure that a single-valued displacement
field is found. A simply connected domain is such that any closed curve contained in
the domain can be continuously shrunk to a point.
32 2. Continuum kinematics and strain analysis

Theorem 2.1 — Schwarz’ theorem. Consider a function f W  2 Rn ! R that has


continuous second partial derivatives at any point x  .x1 ; : : : ; xn / in , then:

@2 f @2 f
.x/ D .x/; 8i; j 2 N ; i; i 6 n (2.44)
@xi @xj @xj @xi

2.3.3.1 Plane strains


In most engineering structures, stress and strain are correctly reflected through tensors
embedded in the three-dimensional space. However, in prismatic structures, the length
of the structure is much greater than the other two dimensions. If they are uniformly
loaded along their length, they are recognized to exhibit a simplified state of strains,
commonly called plane strain, for which the normal strain "33 and the shear strains
"13 and "23 are assumed to be small compared to the cross-sectional strains. The
corresponding displacement becomes two-dimensional, u.x/  .u.x; y/; v.x; y// and
the strain tensor for plane strain is written as:

 D ij .x; y/ei ˝ ej (2.45)

Under these assumptions, the compatibility conditions reduce to a single one:

@2 11 @2 22 @12


2
C 2
D2 (2.46)
@y @x @x@y

Example 9 — Integration of a compatible strain field. Consider a strain field represented


by the following matrix:

.1 C /.c 2 y 2 /
 
k 2xy
Œ D (2.47)
2 .1 C /.c 2 y 2 / xy

where  stands for Poisson’s ratio and k and c are constants. We want to find the two
components of the displacement field u.x; y/ and v.x; y/ from which is derived the strain
field. The latter is compatible since:

@2 11 @2 22 @12


2
D 2
D D0 (2.48)
@y @x @x@y
From the definition of the components of , we obtain:

@u k
11 D D Kxy ) u.x; y/ D x 2 y C f1 .y/
@x 2
(2.49)
@v k 2
22 D D Kxy ) v.x; y/ D xy C f2 .x/
@y 2
2.3 Infinitesimal strain theory and small deformation 33

expressions which lead to:

1  @u @v  11 2 1 
12 D C D kx C f1;y .y/ ky 2 C f2;x .x/
2 @y @x 2 2 2
(2.50)
k
D .1 C /.c 2 y 2 /
2
From this equality, it is possible to separate the x-only dependent terms from the y-only
dependent terms yielding:

1 2 k 1
kx C f2;x .x/ D .1 C /.c 2 y 2 / C ky 2 f1;y .y/ D K1 (2.51)
2 2 2
The two functions f1 .y/ and f2 .x/ can be retrieved by direct integration of the previous
equation:
1 2 1 3
f2;x .x/ D K1 kx ) f2 .x/ D K1 x C K2 k
2 6
1
f1;y .y/ D K1 C k.1 C /.c 2 y 2 / C ky 2 (2.52)
2
1
) f1 .y/ D K1 y C K3 C ky 3 C k.1 C /.c 2 y y 3 =3/
6
The three constants K1 , K2 , and K3 are uniquely defined through extra boundary conditions
for instance. Finally:
1 1
u.x; y/ D kx 2 y C ky 3 C k.1 C /.c 2 y y 3 =3/ K1 y C K3
2 6 (2.53)
1 1 3
v.x; y/ D kxy 2 kx C K1 x C K2
2 6
 Exercise 2.12 For the strain tensor:
2 3
2 1 0
ij D 41 3 35 (2.54)
0 3 1

where  1:
1. Find the unit extension in the direction defined by the vector .3 1 1/.
2. Find the change caused by the strain in the angle between the two vectors with initial
directions .3 1 1/ and .1 0 3/. 

 Exercise 2.13 — Compatibility conditions in plane strains. Show that the strain tensor:
 
2˛Y ˛X
ij D (2.55)
˛X 2ˇY

satisfies the compatibility conditions, and then integrate compatibility conditions to obtain a
displacement field consistent with ij . 

Summary The state of deformation, as the state of stress, may be described by a


symmetric second order tensor with six components. This allows a full analogy
between the purely mathematical tensor transformations in the two cases. While
the six components of the strain tensor are completely independent, they must obey
34 2. Continuum kinematics and strain analysis

functional compatibility conditions to be integrable. The mathematical expressions


for the deformation state are developed by restricting the analysis to an infinitesimal
neighbourhood around a point so that the use of integral and differential calculus is
natural. The consideration of infinitesimal deformations and rotations has advanta-
geous simplifying consequences from the mathematical viewpoint, yet at the expense
of a less accurate description of the problem’s physics.
3
Constitutive equations and
isotropic linear elasticity

The previous two chapters established independent field equations related to the equi-
librium of the internal stress field and the associated kinematics of deformation theory.
Based on these physical concepts, three strain-displacement relations, six compatibil-
ity equations, and three equilibrium equations were developed for the general three-
dimensional case. It is found that we have now developed nine field equations. The
unknowns in these equations include three (3) displacement components, six (6) strain
components, and six (6) stress components, yielding a total of fifteen (15) unknowns.
Since these nine (9) equations are not sufficient to solve for the fifteen unknowns,
additional field equations are needed. These new equations form the constitutive law of
the investigated material and are developed in this chapter. It is a mathematical model
that relates stress and strain:

 D f ./ (3.1)

and is characterized through dedicated experimental investigations.

3.1 Material Characterization and linear elasticity


In their simplest form, relations (3.1) only express the stress tensor components as
functions of the strain tensor components. More sophisticated models also involve
strain rates, strain history, and temperature for instance: they are not reviewed in this
chapter. Linear elasticity is a particular case for which these relations are linear. It
properly describes a deformable continuum that recovers its original configuration when
the loadings causing the deformation are removed. Many structural materials including
metals, plastics, ceramics, wood, rock, concrete, and so forth exhibit linear elastic
behavior under small deformations.
The tension test, in which a specimen is loaded axially in a testing machine as
illustrated in Fig. 3.1 is commolyn employed to characterize the mechanical behavior
of real materials. Strain is determined by the change in length between prescribed
reference marks on the sample and is usually measured by a clip gage. As displayed
in Fig. 3.2, it is observed that each material exhibits an initial stress-strain response
for small deformation that is approximately linear (elastic domain). This is followed
by a change to nonlinear behavior that can lead to large deformation (yieldCstrain-
hardeningCnecking), finally ending with sample failure. For each material, the initial
36 3. Constitutive equations and isotropic linear elasticity

Figure 3.1 – Testing machine and specimen

linear response ends at a point normally referred to as the proportional limit. At some
point on the stress-strain curve unloading does not bring the sample back to zero strain
and some permanent plastic deformation results. The point at which this nonelastic
behavior begins is called the elastic limit. The stress Y at which yield is initiated is
called the yield strength of the material, the stress U corresponding to the maximum
load applied to the specimen is known as the ultimate strength, and the stress B
corresponding to rupture is called the breaking strength.

U rupture
rupture U D B

Y
B

yield strain-hardening necking

elastic lim.
(a) ductile material (b) brittle material

Figure 3.2 – Typical stress-strain diagrams

Ductile materials The stress-strain response indicates extensive plastic deformation,


and during this period the sample dimensions will be changing. In particular the
sample’s cross-sectional area undergoes significant reduction.
Brittle materials The stress-strain response does not show large plastic deformation.
Very little nonelastic or nonlinear behavior is observed.
3.2 Stiffness and compliance tensors 37

Accordingly, a large variety of real materials exhibits linear elastic behavior under small
deformations. This would lead to a linear constitutive model for the one-dimensional ax-
ial loading case given by the relation  D E, where E is the slope of the uniaxial stress-
strain curve. This simple concept is extended to the general three-dimensional forms of
the linear elastic constitutive model as described in the remainder.

Example 10 — Constitutive laws.


1. Linear isotropic elasticity
2. Thermoelasticity
3. Viscoelasticity
4. Hyperelasticity
5. Plasticity
6. Damage model

3.2 Stiffness and compliance tensors


In order to construct a general three-dimensional constitutive law for linear elastic
materials, we assume that each stress component is linearly related to each strain
component, which is usually expressed as1 :
 DCW (3.2)
or in the index notation:
ij D Cij k` k` (3.3)
where Cij k` is the fourth-order elasticity tensor whose components include all the
material parameters necessary to mathematically characterize the material. As such,
it has 81 components. Based on the symmetry of the stress and strain tensors, the
elasticity tensor must have the following properties:
Cij k` D Cj ik`
(3.4)
Cij k` D Cij `k
thus reducing the number of independent components to 36. The concept of strain
energy, which is not detailed at this point, leads to a further reduction to 21 independent
elastic components.

3.2.1 Voigt notation


These relations can be cast into a symmetric matrix format storing the 21 remaining
components of the stiffness tensor as follows:
0 1 2 30 1
11 C1111 C1122 C1133 C1123 C1113 C1112 11
B22 C 6 C2222 C2233 C2223 C2213 C2212 7 B22 C
7 B
B C 6 C
B33 C 6 C3333 C3323 C3313 C3312 7 7 B33 C (3.5)
B C
B CD6
B23 C 6 C2323 C2313 C2312 77 B23 C
B C
sym.
B C 6
@13 A 4 C1313 C1312 5 @13 A
12 C1212 12
1
The “W” operation is called double contraction in tensor analysis.
38 3. Constitutive equations and isotropic linear elasticity

The components of Cij k` are called elastic moduli and have units of stress [N/m2 ]. They
are commonly renumbered as follows for the sake of simplicity:
0 1 2 30 1
11 C11 C12 C13 C14 C15 C16 11
B22 C 6 C22 C23 C24 C25 C26 7 B22 C 7 B
B C 6 C
B33 C 6 C 33 C 34 C35 C 36
7 B33 C
B CD6
B23 C 6
7B C (3.6)
C44 C45 C46 7 7 B23 C
B C
sym.
B C 6
@13 A 4 C55 C56 5 @13 A
12 C66 12
If the material is homogenous, the elastic behavior does not vary spatially, and thus all
elastic moduli are constant over the domain of the structure.

3.2.2 Change of basis and material symmetries


Many materials including crystalline minerals, wood, and fiber-reinforced composites
have different elastic moduli in different directions: they are said to be anisotropic.
For most real materials, there exist particular directions where the properties are the
same, which indicate material symmetries. For many engineering materials, the elastic
properties are found to be essentially identical in all directions. Such materials with
complete symmetry are called isotropic.
Note In a way, anisotropic elasticity attempts to reflect at the macroscopic scale the inhomo-
geneity of real materials. This simplifies the overall constitutive law and corresponding solution
methods to be developed later.
Consider two orthogonal bases .xI ; xII ; xIII / and .e1 ; e2 ; e3 / of the vector space R3 . The
basis .x/ is associated to roman indices to clearly distinguish between the original basis
and the new basis. The matrix T describing the change of basis has components Tij ,
.i; j / D 1; 2; 3.

Vectors In a change of basis from .xI ; xII ; xIII / to .e1 ; e2 ; e3 /, the components of any
vector v  .VI ; VII ; VIII /  .v1 ; v2 ; v3 / satisfy the following transformation:
VI D TIj vj ; j D 1; 2; 3; I D I; II; III (3.7)

Second order tensors In a change of basis from .xI ; xII ; xIII / to .e1 ; e2 ; e3 /, the com-
ponents of any second order tensor S  .SIJ /I;J DI;II;III  .sij /i;j D1;2;3 satisfy the
following transformation:
SIJ D TI i TJj sij ; i; j D 1; 2; 3; I; J D I; II; III (3.8)

Fourth order tensors In a change of basis from .xI ; xII ; xIII / to .e1 ; e2 ; e3 /, the com-
ponents of any fourth order tensor C  .CIJKL /I;J;K;LDI;II;III  .cij k` /i;j;k;`D1;2;3
satisfy the following transformation:
CIJKL D TI i TJj TKk TL` cij k` ; i; j; k; ` D 1; 2; 3; I; J; K; L D I; II; III (3.9)
The tensorial form (3.9) provides a convenient way to establish the stress-strain relations
for a variety of materials featuring symmetries.
3.2 Stiffness and compliance tensors 39

 Exercise 3.1 — Transformation of fourth-order tensors. The transformation rules for sec-
ond order tensors are derived through the transformation rules for vectors. Similarly, the
transformation rules of fourth-order tensors are derived through the transformation rules for
second-order tensors. By generalizing Eq. (1.20), prove equality (3.9). 

This material symmetry relation will provide a system of equations that allows reduction
in the number of independent elastic moduli.

3.2.2.1 Monoclinic materials


A material with one plane of symmetry is referred to as a monoclinic material. For a
symmetry with respect to the .e1 ; e2 /-plane as shown in Fig. 3.3, the required transfor-
mation is simply a mirror reflection about this plane, that is:
2 3
1 0 0
T D 40 1 0 5 (3.10)
0 0 1

Using this specific transformation in relation (3.9) gives Cij k` D Cij k` if the index 3

Figure 3.3 – Monoclinic materials and plane of symmetry

appears an odd number of times: the corresponding moduli would have to vanish and
the stiffness tensor simplifies to:
2 3
C11 C12 C13 0 0 C16
6
6 C22 C23 0 0 C26 7 7
C 33 0 0 C 36
(3.11)
6 7
ŒC D 6 7
6 C44 C45 0 7
sym.
6 7
4 C55 0 5
C66

where only 13 independent elastic moduli are needed.


40 3. Constitutive equations and isotropic linear elasticity

3.2.2.2 Orthotropic materials


A material with three mutually perpendicular planes of symmetry is called orthotropic.
Common examples of such materials include wood and fiber-reinforced composites.
The symmetry relations can be determined starting from the reduced form from the

Figure 3.4 – Orthotropic materials and corresponding planes of symmetry

previous monoclinic case, and reapply the same transformation with respect to the
.e2 ; e3 / and .e1 ; e3 /-planes: this is illustrated in Fig. 3.4. The corresponding stiffness
matrix reduces to having only nine (9) independent components given by:
2 3
C11 C12 C13 0 0 0
6
6 C22 C23 0 0 0 77
C33 0 0 0 7
(3.12)
6
ŒC D 6 6 7
C 44 0 0 7
sym.
6 7
4 C55 0 5
C66
Actually, only two transformations were needed to develop the final reduced constitutive
form (3.12).
 Exercise 3.2 Using material symmetry through 180ı rotations about each of the three coor-
dinate axes, explicitly show the reduction of the elastic stiffness matrix to nine independent
components for orthotropic materials. Also demonstrate that after two reflections, the third
transformation is actually already satisfied. 

3.2.2.3 Transverse isotropic materials


Another common form of material symmetry is with respect to rotations about an axis.
Such materials are said to feature transverse isotropy. For instance, the transformation
for arbitrary rotations about the e3 -axis depicted in Fig. 3.5 is given by:
cos  sin  0
2 3

ŒT D 4 sin  cos  05 (3.13)


0 0 1
3.2 Stiffness and compliance tensors 41

Figure 3.5 – Transverse isotropic materials materials and axis of symmetry

The elasticity stiffness matrix for this case reduces to:


2 3
C11 C12 C13 0 0 0
6
6 C11 C13 0 0 0 77
C 33 0 0 0 7 with (3.14)
6 7
ŒC D 6 C66 D C11 C12
6 C 44 0 0 7
sym.
6 7
4 C44 0 5
C66
Thus, transverse isotropy necessitates only five independent elastic constants.

3.2.2.4 Isotropic materials


An isotropic behavior implies that the elasticity tensor must be the same under all
rotations of the coordinate system. This can be visualized as a spherical symmetry,
Fig. 3.6, and the fourth-order elasticity tensor has the following matrix form:
2 3
C11 C12 C12 0 0 0
6
6 C11 C12 0 0 0 7
7
6 C11 0 0 0 7
7 with C44 D C11 C12 (3.15)
ŒC D 6
6
C 44 0 0 7
sym.
6 7
4 C44 0 5
C44
Only two independent elastic constants exist for isotropic materials. and two versions
of the corresponding constitutive law are available for isotropic materials:
1. the two independent coefficients are called Lamé’s coefficients  and :  is
called Lamé’s constant, and  is referred to as the shear modulus. The stress and
strain tensors are linearly related to each other in the stiffness form by:
 D 2 C  tr./I , ij D 2ij C kk ıij (3.16)
42 3. Constitutive equations and isotropic linear elasticity

Figure 3.6 – Isotropic materials materials and spherical symmetry

or in the compliance form (inverse of the stiffness form) by:


  tr. /I ij kk ıij
D , ij D (3.17)
2 2.3 C 2/ 2 2.3 C 2/
2. the two independent coefficients are the modulus of elasticity, also called Young’s
modulus, E and , Poisson’s ratio. The stress and strain tensors are linearly
related to each other in the stiffness form by:
E E
 D C tr./I (3.18)
1C .1 C /.1 2/
or in the compliance form by:
1C 
D  tr. /I (3.19)
E E
Stiffness relations (3.16) or (3.18) (or their compliance counterpart) are called the gen-
eralized Hooke’s law for linear isotropic elastic solids. The two families of coefficients
are obviously not independent and:
E E
D I D (3.20)
.1 C /.1 2/ 2.1 C /
and conversely:
.3 C 2/ 
ED I D (3.21)
C 2. C /

3.3 Formulation of the isotropic elasticity problem


These governing equations of the isotropic linear elasticity problem comprise a system
of differential and algebraic relations among the stresses, strains, and displacements
that express particular physics at all points within the body under investigation:
3.3 Formulation of the isotropic elasticity problem 43

1. Strain-displacement relations:
1
 D .grad u C grad> u/ (3.22a)
2
2. Compatibility relations:

ij;kl C kl;ij ik;j l j l;ik D 0; i; j; k; l D 1; 2; 3 (3.22b)

3. Equilibrium equations:

div  C fb D 0 (3.22c)

4. Elastic constitutive law:


1C 
 D 2 C  tr./I or D  tr. /I (3.22d)
E E
The compatibility relations (3.22b) ensure that the displacements are continuous and
single-valued and are necessary only when the strains are arbitrarily specified. If, how-
ever, the displacements are included in the problem formulation as the main unknown,
the solution normally generates single-valued displacements and strain compatibility
is automatically satisfied. Thus, in discussing general elasticity, these relations are
normally set aside, to be used only with the stress approach.
Therefore, the general system of elasticity field equations refers to the 15 rela-
tions (3.22a), (3.22c), and (3.22d). This system involves 15 unknowns including 3
displacements ui , 6 strains ij , and 6 stresses ij . It depends on two elastic material
constants for isotropic materials and on the body forces, given a priori with the for-
mulation. This system is of such complexity that general solutions by using analytical
methods are almost always impossible. Simplified formulations are still possible.

3.3.1 Boundary conditions


Boundary conditions for elasticity applications normally include specification of how
the body is being supported or loaded. These conditions specify the physics that occur
on the boundary of body and provide the loading inputs that physically create the
interior stress, strain, and displacement fields. The domain of interest is denoted by 
and its boundary, @. The later is divided into three subsets @u , @ , and @u?
such that @ D @u [ @ [ @u? and @u \ @ \ @u? D ¿: this essentially
means that every point on the boundary has to be specified a boundary condition, either
in displacement @u or in stress on @ . Fig. 3.7 illustrates this general idea. The
boundary conditions are usually written as follows:
 Boundary conditions in displacement on @u :

8M 2 @u ; u.M / D up (3.23)

 Boundary conditions in stress on @ :

8M 2 @ ;  .M /  n.M / D tp (3.24)


44 3. Constitutive equations and isotropic linear elasticity

where up and tp are known quantities.


Although it is generally not possible to specify completely both the displacements
and tractions at the same boundary point, it is possible to prescribe part of the dis-
placement and part of the traction. Such mixed condition involves the specification of
a traction and displacement in two different orthogonal directions on @u? where
displacement and stresses are prescribed along orthogonal directions.
  n D tp

@ @u
@ @u

  

u D up
Traction conditions Displacement conditions Both conditions

Figure 3.7 – Typical boundary conditions

3.3.2 Superposition principle


A very useful tool for the solution to many problems in engineering is the principle of su-
perposition. This technique applies to any problem that is governed by linear equations.
Under the assumption of small deformations and linear elastic constitutive behavior, all
elasticity field equations are linear. The usual boundary conditions are also linear. Thus,
under these conditions the superposition concept can be applied:
Theorem 3.1 — Principle of Superposition. For a given problem domain, if the state
. .1/ ; .1/ ; u.1/ / is a solution to the fundamental elasticity equations with prescribed
body forces fb.1/ and boundary conditions BC.1/ , and the state . .2/ ; .2/ ; u.2/ / is a
solution to the fundamental equations with prescribed body forces fb.2/ and boundary
conditions BC.2/ , then the state . .1/ C ˛ .2/ ; .1/ C ˛.2/ ; u.1/ C ˛u.2/ / will be
a solution to the problem with body forces fb.1/ C ˛fb.2/ and boundary conditions
BC.1/ C ˛BC.2/ where ˛ is an arbitrary real number.

3.4 Solution methods


3.4.1 Displacement approach
We now wish to develop the reduced set of field equations solely in terms of the dis-
placements by eliminating the strains and stresses from the fundamental system (3.22).
This is easily accomplished by using the strain-displacement relations in Hooke’s law
to give:
ij D uk;k ıij C .ui;j C uj;i / (3.25)
3.4 Solution methods 45

Using these relations in the equilibrium equations (3.22c) leads to:


. C /uk;ki C ui;kk C fbi D 0 (3.26)
which are the equilibrium equations in terms of the displacements and are referred to as
Navier-Lamé’s equations. This system can be expressed in an intrinsic vector form as:
. C / grad.div u/ C u C fb D 0 (3.27)
Navier’s equations are the desired formulation for the displacement problem, and the
system represents three equations for the three unknown displacement components.
 Exercise 3.3 — Functionally graded materials. There has been recent interest in nonho-
mogeneous material behavior related to functionally graded materials. For a two-dimensional
model in which the Lamé’s coefficients vary as .x/ D 0 .1 C ax/ and .x/ D k.x/, where,
0 , a, and k are constants, derive the two dimensional Lamé-Navier’s equations. 

The general strategy to solve a problem in displacement is:


 postulate the form of the displacement solution;
 check the boundary conditions in displacement;
 check Navier-Lamé’s equations;
 calculate the associated strains and stress;
 check the boundary condition in stress.

3.4.2 Stress approach


In order to develop solution methods when the boundary conditions are to be given
only in terms of the stress components, it is helpful to reformulate system (3.22) by
eliminating the displacements and strains and thereby cast a new system solely in terms
of the stresses. By eliminating the displacements, we must now include the compatibility
equations in the fundamental system of field equations. We therefore start by using
Hooke’s law (3.22d) and eliminate the strains in the compatibility relations (2.43) to
get:
ij;kk C kk;ij ik;j k j k;i k
 (3.28)
D .mm;kk ıij C mm;ij ıkk mm;j k ıi k mm;i k ıj k /
1C
A more useful result is found by incorporating the equilibrium equations into the system.
Substituting (3.22c) and ıkk D 3 into (3.28) gives:
1 
ij;kk C kk;ij D mm;kk ıij fbi;j fbj;i (3.29)
1C 1C
When i D j , the previous relation reduces to:
1C
i i;kk D fbi;i (3.30)
1 
Substituting this result back into (3.29) gives the desired compatibility relations:
1 
ij;kk C kk;ij D ıij fbk;k fbi;j fbj;i (3.31)
1C 1 
46 3. Constitutive equations and isotropic linear elasticity

expressed in terms of the stress: they are called the Beltrami-Michell compatibility
conditions. They also take the following intrinsic tensor form:
1 
 C grad.grad.tr  // C div fb I C .grad fb C grad> fb / D 0 (3.32)
1C 1 
 Exercise 3.4 — Beltrami-Michell conditions. Go through the details and explicitly develop
the Beltrami-Michell compatibility equations (3.31). 

The general strategy to solve a problem in stress is:


 postulate the form of the stress solution;
 check the boundary conditions in stress;
 check equilibrium equations;
 check Beltrami-Michell’s equations;
 integrate the associated displacement;
 check the boundary condition in displacement.

3.5 Plane elasticity


Because of the complexity of the elasticity field equations, closed-form solutions to
fully three-dimensional problems are very difficult to accomplish. Thus, most analytical
solutions are developed for reduced problems that typically include axi-symmetry or
two-dimensionality to simplify particular aspects of the formulation and solution. Is now
examined in detail the formulation of two-dimensional problems in elasticity resulting
in a boundary-value problem cast within a two-dimensional domain in the .x; y/-plane
using Cartesian coordinates.
All real elastic structures are three-dimensional, the coming plane elasticity develop-
ments are approximate models. The nature and accuracy of the approximation depend
on problem and loading geometry. The two basic theories of plane strain and plane
stress represent the fundamental plane problem in elasticity. These two theories apply
to significantly different types of two-dimensional bodies; however, their formulations
yield very similar field equations. These two theories can be reduced to one governing
equation in terms of a single unknown stress function which allows many solutions to
be generated.

3.5.1 Plane strain assumption


Consider an infinitely long prismatic body. If the body forces and tractions on the lateral
boundaries are independent of the z coordinate and have no e3 component, then the
deformation field within the body can be taken in the reduced two-dimensional form:
u.x; y/ D u.x; y/e1 C v.x; y/e2 (3.33)
This deformation is referred to as a state of plane strain where all cross-sections have
identical displacements. The strains corresponding to this plane problem become:
2 3
11 .x; y/ 12 .x; y/ 0
Œ.x; y/ D 412 .x; y/ 22 .x; y/ 05 (3.34)
0 0 0
3.5 Plane elasticity 47

(a) tunnel (b) dam

Figure 3.8 – Plane strain conditions in long prismatic structures

From Hooke’s law (3.22d), the allowable stresses reduce to:


0 1 2 30 1
11 1   0 11
E
@22 A D 4  1  0 5 @22 A (3.35)
 .1 C /.1 2/ 0 0 1 2 
12 12

together with:

13 D 23 D 0 and 33 D .11 C 22 / (3.36)

Once 11 and 22 are determined, 33 is found from Hooke’s law. Although 33 D 0,
the corresponding normal stress 33 will not in general vanish. The inversion of
expression (3.35) leads to:
0 1 2 30 1
11 1   0 11
@22 A D 1 C  4  1  05 @22 A (3.37)
 E 0 0 1 
12 12

 Exercise 3.5 — Compatibility conditions in plane strains. Show that the strain tensor:
 
2˛Y ˛X
ij D (3.38)
˛X 2ˇY

satisfies the compatibility conditions, and then integrate compatibility conditions to obtain a
displacement field consistent with ij . 

3.5.2 Plane stress assumption


The second type of two-dimensional theory applies to domains bounded by two parallel
planes separated by a distance that is small in comparison to other dimensions in the
problem. Again, choosing the .x; y/-plane to describe the problem, the domain is
bounded by two planes z D ˙h=2, as shown in Fig. 1.7. The theory further assumes
that these planes are stress free, ie 33 D 13 D 23 D 0 on each face z D ˙h=2.
Because the region is thin in the z direction, there can be little variation in these stress
components through the thickness, and thus they will be approximately zero throughout
48 3. Constitutive equations and isotropic linear elasticity

the entire domain. Finally, because the region is thin in the z direction it can be
argued that the other nonzero stress components will have little variation with z. These
arguments can then be summarized by the stress state:
2 3
11 .x; y/ 12 .x; y/ 0
Œ .x; y/ D 412 .x; y/ 22 .x; y/ 05 (3.39)
0 0 0

and this form constitutes a state of plane stress in an elastic solid. In order to maintain a
stress field independent of z, there can be no body forces or surface tractions in the z
direction. Furthermore, the nonzero body forces and tractions must be independent of z.
The corresponding strain field follows from Hooke’s law is:
0 1 2 30 1
11 1  0 11
@22 A D 1 4  1 0 5 @ 22 A (3.40)
 E 0 0 1C 
12 12

together with:

13 D 23 D 0 and 33 D .11 C 22 / (3.41)
1 
Once inverted, expression (3.40) becomes:
0 1 2 30 1
11 1  0 11
@22 A D E 4 1 0 5 @ 22 A (3.42)
1  2
 12 0 0 1   12

Beltrami-Michell conditions are generally not satisfied in plane-stress assumptions: full


three-dimensional elasticity is violated but plane-stress is a very good approximation
for thin structures.
 Exercise 3.6 — In-plane elasticity in polar coordinates. In polar coordinates, the displace-
ment is u.r;  / D ur .r; /er C u .r; /e and the corresponding strain components take the form:

rr D ur;r I   D .ur C u; /=r I r D .ur; C ru;r u /=.2r/ (3.43)

For rigid-body motion, the strains will vanish. Under these conditions, integrate the previ-
ous strain-displacement relations to show that the most general form of a rigid-body motion
displacement field in polar coordinates is given by:

ur .r; / D a sin  C b cos 


(3.44)
u .r;  / D a cos  b sin  C cr

where a, b, and c are constants. 

3.5.3 Airy stress function


Numerous solutions to plane elasticity problems can be determined through the use of a
particular stress function called the Airy stress function: this reduces the formulation
3.5 Plane elasticity 49

to a single governing equation in terms of a single unknown. The method is started by


reviewing the equilibrium equations for the plane elasticity problem:
11;1 C 12;2 C fb1 D 0
(3.45)
21;1 C 22;2 C fb2 D 0
The body forces are assumed to derive from a potential function V .x; y/ such that:
fb1 .x; y/ D V;1 .x; y/ .D V;x .x; y//
(3.46)
fb2 .x; y/ D V;2 .x; y/ .D V;y .x; y//
and the equilibrium equations become:

.11 V /;1 C 12;2 D 0 (3.47)


12;1 C .22 V /;2 D 0 (3.48)

Many body forces found in applications fall into this category.


Consider two new functions A.x; y/ and B.x; y/ such that:
11 V D A;2
(3.49)
22 V D B;1
then equations (3.47) and (3.48) are satisfied as soon as 12 D A;1 and 12 D B;2 ,
respectively, that is A;1 D B;2 . Accordingly, there is an arbitrary function .x; y/, the
so-called Airy stress function such that A D ;2 and B D ;1 such that:

11 V D ;22
22 V D ;11 (3.50)
12 D ;12 D ;21
and immediately yielding the symmetry of the stress tensor.
With equilibrium now satisfied, we focus attention on the remaining field equations
in the stress formulation, that is, the compatibility relations expressed in terms of stress.
In plane strain, plugging system (3.37) in Eq. (2.46) yields:
1
.11 C 22 /;11 C .11 C 22 /;22 C .fb1;1 C fb2;2 / D 0 (3.51)
1 
Counterpart derivations in plane stress are quite more involved and not detailed here;
they produce:

.11 C 22 /;11 C .11 C 22 /;22 C .1 C /.fb1;1 C fb2;2 / D 0 (3.52)

It is noted that they differ only by the coefficient in front of the body force terms.
Substituting the stress function form (3.50) into these compatibility relations gives the
following pair:
1 2
;1111 C ;2222 C 2;1122 D .V;11 C V;22 / (plane strain) (3.53)
1 
;1111 C ;2222 C 2;1122 D .1 /.V;11 C V;22 / (plane stress) (3.54)
50 3. Constitutive equations and isotropic linear elasticity

which can also be written as:


1 2 1 2 2
2  D V , r 4  D r V (3.55)
1  1 
2  D .1 /V , r 4  D .1 /r 2 V (3.56)
The form 2  r 4 is called the biharmonic operator. If the body forces derive from
a potential function V .x; y/ itself solution to V D 0, then both the plane strain and
plane stress forms reduce to:
r 4  D ;1111 C ;2222 C 2;1122 D 0 (3.57)
This relation is called the biharmonic equation. The plane problem of elasticity has
been reduced to a single equation in terms of the Airy stress function . This function
is to be determined in the two-dimensional region. Appropriate boundary conditions
are necessary to complete a solution.
 Exercise 3.7 Explicitly show that the fourth-order polynomial Airy stress function .x; y/ D
ax 4 C bx 2 y 2 C cy 4 will not satisfy the biharmonic equation unless 3a C b C 3c D 0. 

 Exercise 3.8 Show that the following stress components satisfy the equations of equilibrium
with zero body forces, but are not the solution to a problem in elasticity:
11 D c.y 2 C .x 2 y 2 //
22 D c.x 2 C .x 2 C y 2 //
33 D c.x 2 C y 2 / (3.58)
12 D 2cxy
23 D 13 D 0;
and c is a non-zero constant. 

 Exercise 3.9 — Polynomial Airy function and cantilever beam. A two dimensional can-
tilever beam of length ` and height 2h is subjected to a distributed shear stress 0 x=` on its
upper face. It is clamped at x D 0:
1. Sketch the system of interest with the appropriate boundary conditions.
2. A polynomial Airy stress function is assumed:
1 X
X 1
.x; y/ D cij x i y j (3.59)
iD0 j D0

Determine the general expression of the system of algebraic equations that the cij
coefficients should solve.
3. Provide the expressions of c02 , c03 , and c20 .
4. Consider terms up to order 4, x 4 ; x 3 y; x 2 y 2 ; : : : , in the proposed expansion and plot 11 ,
12 and 22 . 

Summary In this chapter, the fundamental linear elasticity formulation was estab-
lished by relating the stress field equations to the strain field equations through the
fourth-order stiffness tensor. Additional boundary conditions both in stress and dis-
placement specify the physics that occur on the boundary of body, and generally
provide the loading inputs that physically create the interior stress, strain, and dis-
placement fields. Although the field equations are the same for all problems, boundary
3.5 Plane elasticity 51

conditions are different for each problem. This eventually leads to two different formu-
lations: one in terms of displacements and the other in terms of stresses. Because these
boundary value problems are difficult to solve, simplifying strategies shall be invoked
to aid in problem solution: this typically includes the approximate formulation of
two-dimensional problems in elasticity. The nature and accuracy of the approximation
depend on problem and loading geometry. The two basic theories of plane strain
and plane stress represent the fundamental plane problem in elasticity. Both theories
can be reduced to one governing equation in terms of a single unknown Airy stress
function. This reduction then allows many solutions to be generated to problems of
engineering interest.
A
Tensor algebra

A.1 Points and vectors


The space under consideration is a three-dimensional Euclidean point space E also
known as the three-dimensional Euclidean affine space. The term point is reserved for
elements of E: they are designated by their coordinates, x  .x1 ; x2 ; x3 /  .x; y; z/
along the corresponding axes. From these axes are defined unit vectors .e1 ; e2 ; e3 / 
.ex ; ey ; ez /  .i; j; k/ which form the canonical basis of the vector space V, associated
to E, in which live vectors defined by their components u  .u1 ; u2 ; u3 /  .ux ; uy ; uz /,
that is:
3
X
uD D ui  ei (A.1)
i D1

Also, we note that:


 the difference v D y x of two points denoted y and x, respectively, is a vector;
 the sum y D x C v of a point x and a vector v is a point;
 the sum of two points is not a meaningful concept.
A point o in E is chosen as the origin and each coordinate xi , i D 1; : : : ; n of a point x
is fully characterized through the usual inner product as:

xi D .x o/  ei (A.2)

where the quantity x o is commonly named the position-vector. In cartesian coordi-


nates, the scalar (or inner) product is defined as:
3
X
uvD ui vi (A.3)
i D1

The scalar product plays an important role in vector and tensor algebra. The properties
of the vector space essentially depend on whether and how the scalar product is defined
in this space.

A.2 Index notation and Einstein convention


Consider the following notation “ˇi , i D 1; : : : ; n”: ˇi is said to be an indexed quantity,
i , the index, and 1; : : : ; n, the index range. Einstein convention for index notation is
54 A. Tensor algebra

adopted for its conciseness and compactness. Its basic goal is to make the use of the
summation sign transparent. It applies to generic expressions of the following form:

aik bkj;j C ci D 0 (A.4)

where the ranges are i D 1; : : : ; n, j D 1; : : : ; m, and k D 1; : : : ; p for the sake of


generality. In mechanical engineering, n D m D p D 3. Both isolated terms ai k bkj;j
and ci are referred to as individual terms. The convention stipulates what follows:

Rule 1 The fact that an index appears only once in individual terms means that the
equation is valid for each index within the index range. Such index is called a
free index. In Eq. (A.4), rule 1 applies to index i which appears only once in each
of the individual terms.
Rule 2 The fact that an index appears twice in individual terms means that the term
must be summed over the index range. Such index is called a dummy index. The
letter used for this index is not of importance since it is only devoted to reflect the
summation. For example:

Aik Bkj D Ai` B`j (A.5)

Also, in Eq. (A.4), rule 2 applies to indices k and j in the first term.
Rule 3 An index cannot appear more than twice within an individual term.

Also, attention should be paid to the notations for (partial) derivatives. The coordinate
along which the differentiation is conducted is indexed:

@./ @./ @./ @./


  ./;1 or   ./;2 (A.6)
@x @x1 @y @x2
which becomes in a general form:

@./
 ./;i (A.7)
@xi
Such indices denoting (partial) differentiation should not be ignored in index notation.
It follows from these rules that Eq. (A.4) is the concise version of:
p
m X
X
aik bkj;j C ci D 0; i D 1; : : : ; n (A.8)
kD1 j D1

that is, for n D m D p D 2:

a11 b11;1 C a11 b12;2 C a12 b21;1 C a12 b22;2 C c1 D 0


(A.9)
a21 b11;1 C a21 b12;2 C a22 b21;1 C a22 b22;2 C c2 D 0

where the partial derivatives on the b’s reflect the fact that quantities a, b, and c are
possibly constant functions in at least two variables.
A.3 Tensors 55

Example 11 — Gradient of a scalar function. Consider a function in two variables u.x; y/.
Its gradient in index notation is:

grad u D u;i ei (A.10)

Geometrically the gradient of a scalar field represents a normal vector to the level surface
u.x; y/ D u0 .

 Exercise A.1 Express in index notation the divergence of the gradient of a scalar field u.x/. 

Kronecker symbol The Kronecker delta symbol ıij is defined as follows:


(
1 for i D j
ıij D (A.11)
0 for i ¤ j

Example 12 — Factoring with the Kronecker symbol. Given an orthonormal basis of the
Euclidean space, by viewing nj D ıij nj , the equation:

ij nj D nj ; 2R (A.12)

whall be written as:

.ij ıij /nj D 0 (A.13)

Unless stated otherwise, the Einstein convention in index notation is used throughout
this document. This convention has to be ignored as soon as the sum sign is used.

A.3 Tensors
The term tensor is understood as a synonym for linear transformation from V into V.
Accordingly, a tensor S is a linear map that assigns to each vector u a vector v D Su.
By linear map, we mean that the following conditions are satisfied:
S.u C ˛v/ D Su C ˛Sv (A.14)
Also, consider two tensors S and T; by definition, they can be added and multiplied:
.S C T/u D Su C Tu and .˛S/u D ˛.Su/ (A.15)

Important tensors The zero tensor O maps every vector v into the zero vector:
Ov D 0; 8v (A.16)
The identity tensor I is defined by:
Iv D v; 8v (A.17)
56 A. Tensor algebra

Example 13 — Index notation for the divergence of a tensor field. In three dimensions, the
divergence of a continuously differentiable vector field u D ui  ei (more exactly, u.x; y; z/ D
ui .x; y; z/  ei ) is equal to the scalar-valued function:

div u D ui;i (A.18)

When extended row-wise to tensor fields, ie tensors depending on space for instance, this
convention yields the usual index version of the static equilibrium equation:

div  C f D 0 , ij;j C fi D 0 (A.19)

that is, each row i of  is seen as a vector on which the divergence operator acts.

Important properties The product ST of two tensors is the tensor ST satisfying:

.ST/v D S.Tv/; 8v (A.20)

We write S> for the transpose of S; S> is the unique tensor with the property for all
vectors u and v:

.Su/  v D u  .S> v/; 8v (A.21)


 Exercise A.2 show that .S C T/> D S> C T> , .ST/> D T> S> , and .S> /> D S. 

A tensor S is symmetric if:

S> D S (A.22)

and skew-symmetric if:

S> D S (A.23)

Every tensor S can be expressed uniquely as the sum of a symmetric tensor E and a
skew-symmetric tensor W:

1 1
S D E C W D .S C S> / C .S S> / (A.24)
2 2
The tensor product a ˝ b of two vectors a and b is the tensor that assigns to each
vector v the vector .b  v/a:

.a ˝ b/v D .b  v/a (A.25)

where  is the usual inner product.


 Exercise A.3 Show that:
1. a ˝ b is actually a tensor.
2. .a ˝ b/> D .b ˝ a/
3. .a ˝ b/.c ˝ d/ D .b  c/a ˝ d
4. I D ıij ei ˝ ej 
A.3 Tensors 57

Interpretation Let e be a unit vector. Then e ˝ e applied to a vector v gives .e  v/e,


which is the participation of v in the direction of e, while I e ˝ e applied to v gives
v .e  v/e, which is the projection of v onto the plane perpendicular to e (Fig. 2).
The components Tij of a tensor T are defined by:

.T/ij D Tij D ei  Tej (A.26)

With this definition, v D Su is equivalent to:

vi D Sij uj (A.27)

where we notice that the usual matrix product is retrieved. Also, the definitions of the
tensor product and the components of T yield:

T D Tij ei ˝ ej (A.28)

Theorem A.1 — Tensor space. The set of all tensors forms a vector (or linear) space
whose canonical basis is formed by the terms ei ˝ ej .

 Exercise A.4
1. Based on Eqs. (A.25) and (A.26), prove relation (A.28).
2. Prove theorem A.1.
3. Show that .a ˝ b/ij D ai bj 

Example 14 — Cauchy stress tensor. It is as a linear mapping of the unit surface normal
into the Cauchy stress vector. Hence, it is a tensor.

Trace operator The trace is the linear operation that assigns to each tensor S a scalar
tr S and satisfies:

tr.u ˝ v/ D u  v (A.29)

for all vectors u and v. By (3) and the linearity of the trace:

tr S D tr.Sij ei ˝ ej / D Sij tr.ei ˝ ej / D Sij ei  ej D Si i (A.30)

This operation has the following properties:

tr S> D tr S
(A.31)
tr.ST/ D tr.TS/

 Exercise A.5 Let Sij D Wij C Wj i and Aij D Wij Wj i , where W is a second-order tensor.
Show that the product of the symmetric tensor S and the antisymmetric tensor A vanishes; that
is, show that Sij Aij D 0.
58 A. Tensor algebra

Orthogonal tensor A tensor Q is orthogonal if it preserves inner products:

Qu  Qv D u  v (A.32)

for all vectors u and v. A necessary and sufficient condition that Q be orthogonal is
that:

QQ> D Q> Q D I (A.33)

or equivalently:

Q> D Q 1
(A.34)

An orthogonal tensor with positive determinant is called a rotation. (Rotations are


sometimes called proper orthogonal tensors.) Every orthogonal tensor is either a
rotation or the product of a rotation with I.
 Exercise A.6 Write the matrix in basis .ei /, i D 1; 2; 3 corresponding to the following tensor
3ei ˝ ei C 2ei ˝ ej 4ej ˝ ei C ek ˝ ek .

Important theorems A scalar  is an eigenvalue of a tensor S if there exists a unit


vector e such that Se D e, in which case e is an eigenvector. The characteristic space
for S corresponding to  is the subspace of V consisting of all vectors v that satisfy the
equation Sv D v.
If this space has dimension n, then  is said to have multiplicity n. The spectrum
of S is the list .1 ; 2 ; : : : /, where 1 > 2 > : : : are the eigenvalues of S with each
eigenvalue repeated a number of times equal to its multiplicity.
Theorem A.2

1. The eigenvalues of a positive definite tensor are strictly positive.


2. The characteristic spaces of a symmetric tensor are mutually orthogonal.

The next result is one of the central theorems of linear algebra.


Theorem A.3 — Spectral Theorem. Let S be symmetric. Then there is an orthonormal
basis for V consisting entirely of eigenvectors of S. Moreover, for any such basis
.e1 ; e2 ; e3 / the corresponding eigenvalues 1 ; 2 ; 3 when ordered, form the entire
spectrum of S and:
3
X
SD i ei ˝ ei (A.35)
i

Conversely, if S has the form (A.35) with .ei /, i D 1; 2; 3 orthonormal, then 1 , 2 ,


3 are eigenvalues of S with e1 , e2 , e3 as corresponding eigenvectors. Further:

1. S has exactly three distinct eigenvalues if and only if the characteristic spaces
of S are three mutually perpendicular lines through 0.
2. S has exactly two distinct eigenvalues if and only if S admits the representation:
A.3 Tensors 59

S D 1 e ˝ e C 2 .I e ˝ e/; jej D 1; 1 ¤ 2 (A.36)

3. S has exactly one eigenvalue if and only if:

S D I (A.37)

Theorem A.4 — Polar Decomposition. Let F be a tensor. Then, there exist positive
definite, symmetric tensors U, V and a rotation R such that:

F D RU D VR (A.38)

Basis transformation At this point, we need to remember is that in order for a


“mathematical object with indices” to represent tensor components, it has to obey
precise transformation rules. In a change of coordinates from one orthonormal basis
.el ; e2 ; e3 / to another .e0l ; e02 ; e03 /, a vector:

u D ui ei D u0i e0i (A.39)

transforms according to:

u0i D Qij uj ; ui D Qj i u0i (A.40)

where Q is an orthogonal tensor, that is:

Qik Qj k D Qki Qkj D ıij ; det Q D 1 (A.41)

Row .i/ of the matrix of Q contains the components of vector e0i in basis .el ; e2 ; e3 /.
Similarly, a second-order tensor:

A D Aij ei ˝ ej D A0kl e0k ˝ e0l (A.42)

transforms according to:

A0ij D Qik Akl Qj l ; Aij D Qki A0kl Qlj (A.43)

A.3.1 Invariants
Given a tensor S, the determinant of S I admits the representation:

det.S I/ D 3 C I1 .S/2 I2 .S/2 C I3 .S/ (A.44)

for every  2 R, where:

I1 .S/ D tr S
I2 .S/ D ..tr S/2 tr.S2 //=2 (A.45)
I3 .S/ D det.S/
60 A. Tensor algebra

are called the invariants of S. They are completely characterized by the spectrum
.1 ; 2 ; 3 / of S since:

I1 .S/ D 1 C 2 C 3
I2 .S/ D 1 2 C 2 3 C 1 3 (A.46)
I3 .S/ D 1 2 3
 Exercise A.7 This exercise depends on exercise A.3. Determine the spectrum and a spectral
decomposition for each of the following tensors:

A D ˛I C ˇm ˝ m
(A.47)
BDm˝nCn˝m

Here, ˛ and ˇ are scalars, while m and n are orthogonal unit vectors.

 Exercise A.8 A tensor P is a perpendicular projection if P is symmetric and P2 D P. Let n be a


unit vector. Show that each of the following tensors is a perpendicular projection:

I; O; n ˝ n; I n˝n (A.48)

A.3.2 Matrix form


The matrix form of S, in other words, the expression of S in a basis is commonly
expressed as an array storing its components1 :
2 3
S11 S12 S13
ŒS D S.e1 ;e2 ;e3 / D 4S21 S22 S23 5 (A.49)
S31 S32 S33

Example 15 — Identity matrix. The identity tensor has the following matrix form:
2 3
1 0 0
ŒI D I.e1 ;e2 ;e3 / D 40 1 05 (A.50)
0 0 1

Relation (A.35) is called a spectral decomposition of S. Note that the matrix of S


relative to a basis .e1 ; e2 ; e3 / of eigenvectors is diagonal:
2 3
1 0 0
ŒS D 4 0 2 0 5 (A.51)
0 0 3

1
The notations ŒS or S.e1 ;e2 ;e3 / are here to distinguish the tensor from its matrix representation. They
are author-dependent.

You might also like