Solution of St.-Venant's and

SOLUTION OF ST.
-VENANTS AND
ALMANSI-MICHELLS PROBLEMS
BY
LUCA PLACIDI
A Thesis submitted to the Graduate School
in partial fulllment of the requirements
for the Degree
Master of Science
in
Engineering Mechanics
at
Virginia Polytechnic Institute
and
State University
Dr. R. C. Batra
Dr. E. G. Henneke
Dr. D. H. Morris
May 2, 2002
Blacksburg, Virginia
Keywords: Non Linear Elasticity, Linear Elasticity, Saint-Venants
Problem, Polynomial hypothesis, Clebsch hypothesis, Stressed
Reference Conguration
SOLUTION OF ST.-VENANTS AND ALMANSI-MICHELLS
PROBLEMS
LUCA PLACIDI
(ABSTRACT)
We use the semi-inverse method to solve a St. Venant and an Almansi-
Michell problem for a prismatic body made of a homogeneous and
isotropic elastic material that is stress free in the reference congura-
tion. In the St. Venant problem, only the end faces of the prismatic
body are loaded by a set of self-equilibrated forces. In the Almansi-
Michell problem self equilibrated surface tractions are also applied on
the mantle of the body. The St. Venant problem is also analyzed for
the following two cases: (i) the reference conguration is subjected to a
hydrostatic pressure, and (ii) stress-strain relations contain terms that
are quadratic in displacement gradients. The Signorini method is also
used to analyze the St. Venant problem. Both for the St. Venant and
the Almansi-Michell problems, the solution of the three dimensional
problem is reduced to that of solving a sequence of two dimensional
problems. For the St. Venant problem involving a second-order elastic
material, the rst order deformation is assumed to be an innitesi-
mal twist. In the solution of the Almansi-Michell problem, surface
tractions on the mantle of the cylindrical body are expressed as a
polynomial in the axial coordinate. When solving the problem by the
semi-inverse method, displacements are also expressed as a polynomial
in the axial coordinate. An explicit solution is obtained for a hollow
circular cylindrical body with surface tractions on the mantle given by
an ane function of the axial coordinate
iii
Contents
List of Figures vii
Chapter
1 Introduction 1
Suggested future work . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2 Theory of Elasticity 4
Theory of deformation for a continuous media . . . . . . . . . . . . . . 4
Deformations of line, area, and volume elements . . . . . . . . . . 4
Innitesimal deformations . . . . . . . . . . . . . . . . . . . . . . 9
Equilibrium laws . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Assumptions on Forces . . . . . . . . . . . . . . . . . . . . . . . . 9
The Conservation of Mass . . . . . . . . . . . . . . . . . . . . . . 11
Balance of Linear Momentum . . . . . . . . . . . . . . . . . . . . 11
Balance of Momentum of Momentum . . . . . . . . . . . . . . . . 13
Constitutive Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
Equations of elastostatics . . . . . . . . . . . . . . . . . . . . . . . . . 20
Signorinis Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
Semi-inverse method . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
iv
3 St.-Venants Problem 31
Governing equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
Solution of St.-Venants problem by a semi-inverse method . . . . . . . 42
Generalized Axisymmetric Problems . . . . . . . . . . . . . . . . . . . 45
Uniform pressure in the reference conguration . . . . . . . . . . . . . 47
Formulation of the problem . . . . . . . . . . . . . . . . . . . . . 47
Global balance of forces and moments . . . . . . . . . . . . . . . . 49
Solution of St.-Venants Problem by Signorinis Method . . . . . . . . . 53
4 The Almansi-Michells problem 57
An example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
5 Conclusions 78
6 Appendix 80
Decomposition of a 2
nd
order tensor . . . . . . . . . . . . . . . . . . . . 80
The star of Hodge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
Dierential operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
An Integral formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
Solution of four boundary value problems . . . . . . . . . . . . . . . . . 92
Problem A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
Problem B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
v
Problem C . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
Problem D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
Cylindrical coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
vi
List of Figures
2.1 A system of linear springs . . . . . . . . . . . . . . . . . . . . . . 21
3.1 A cylindrical body . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.2 Positive direction of paths on the boundary and direction of the
outward normal . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
vii
Chapter 1
Introduction
Our goal is to nd the deformed shape of an elastic cylindrical body when certain
loads are applied to its boundary
1
. We can start from the basic properties of
structure of the matter like atoms; thus use the Quantum Mechanics approach
2
.
However, in most applications we do not need this level of precision.
Accordingly, we neglect the microstructure of matter and regard the body as
a three-dimensional continua
3
. We make other simplifying assumptions in order
to solve the problem; the validity of these assumptions is veried a posteriori
through a comparison of computed results with those observed experimentally.
The rst simplifying assumption is on the shape of the body. We consider
prismatic body of uniform cross-section loaded either only at the end faces and/or
on the mantle by tractions that can be expressed as a polynomial in the axial co-
ordinate
4
.
The material of the body is assumed to be elastic
5
.
1
Usually known as "The Problem of Saint-Venant". Many scientists have worked upon it; e.
g. see [14], [17], [16], [18], [19], [20], [27], [28], [29], [30], [36], [39] and [40].
2
A Quantum Mechanic approach for this kind of problem does not exist; however one could
start from the principles of Quantum and Statistical Mechanics ([13], [42]).
3
The theory of Continuum Mechanics is avaliable in many texts: [21], [22], [35] and [38].
4
Usually known as the Almansi-Michells problem. The few references found are: [2], [3] [10],
[11] and [12].
5
Many texts concerning the theory of elasticity are available: [23], [34], [41] and [43].
1
The rst problem is usually referred to as the St.-Venants Problem and the
second one as Almansi-Michells. It is dicult to solve these problems in com-
plete generality. Accordingly, we assume that the deformations of the body are
innitesimal and the stresses depend upon strains at most quadratically.
We study four problems and in one of them we assume that the reference
conguration is not stress free. These problems are summarized in the following
table.
List of Problems studied:
Problem Mantle traction-free Ref. Conf. stress-free Order
6
St.-Venant Yes Yes 1
St.-Venant Yes No 1
St.-Venant Yes Yes 2
Alm.-Michell No Yes 1
We will use the semi-inverse method to solve these problems. That is, a part
of the solution is assumed a priori and the remaining unknowns are found by
satisfying the equilibrium equations and the boundary conditions.
6
The order refers to both the Balance and Constitutive Equations.
2
Suggested future work
We have analyzed the case when the state of stress in the reference congu-
ration is that of hydrostatic pressure. The next step is to consider a general state
of stress in the reference conguration. An other extension of the present work is
the consideration of body forces in the solution of the Almansi-Michells problem.
The third possibility is to use the complete Almansi-Michell solution to nd the
general non-linear solution for the St.-Venants Problem.
3
Chapter 2
Theory of Elasticity
Theory of deformation for a continuous media
Deformations of line, area, and volume elements
Let Sbe a continuous body. Consider the reference conguration C
. The goal
of this section is to describe the deformation of S from the reference conguration
C
to the present conguration C . Let X be a generic point in the reference

conguration C
and x its position in the present conguration C . The nite

deformation from C
to C is described by the following vector valued function:

x = x(X) , (2.1)
which associates to each point X C
its corresponding position in C.

For the sake of simplicity, we do not study tearing and cavitation; so x is
assumed to be a single valued function of X. This means that a point does not
deform into more then one point. Besides we consider the impenetrability of bodies
and so assume the injective property for x; i.e., two points do not deform into
one. The function x is thus invertible. Furthermore we require that x C
2
(C
),
4
and
J det (F) det
x
X
> 0. (2.2)
Taking the dierential of both sides of (2.1), we get
dx = FdX, F =
x
X
, (2.3)
where F is a linear transformation called the deformation gradient dened X C
.
It associates to each innitesimal vector dX at X C
the corresponding dx at
x C.
The polar decomposition theorem of Cauchy applied to the matrix F gives
F = RU , F = VR; (2.4)
where R is a rotation and U = U
T
and V = V
T
describe the stretch. R is an
orthogonal matrix and U and V are symmetric and positive denite matrices.
The right and left Cauchy-Green tensors are dened as
_
_
C F
T
F = U
T
R
T
RU = U
2
,
B = FF
T
= VRR
T
V
T
= V
2
.
(2.5)
Proposition 1 The tensors B and C have the same eigenvalues; the eigenvalues
are real and positive.
The deformation eld and the deformation gradient can be described equiva-
lently by the displacement eld
u(X) x(X) X, (2.6)
5
and the displacement gradient
H
u
X
. (2.7)
From (2.6) we have:
F = I +H. (2.8)
The Green-Saint-Venant strain tensor is dened as
G
1
2
(CI) . (2.9)
G is symmetric and in terms of the displacement eld it is given by
G =
1
2
H+H
T
+H
T
H
. (2.10)
For the description of the deformation of a continuous body it is important
to know how the elements of length, surface and volume change. Let dX = dl
M
be an innitesimal vector at the point X; M is the unit vector along dX. Let the
corresponding innitesimal vector at x be dx = dl m; m is the unit vector along
dx. Then
dl =
m Cmdl
. (2.11)
Let dS
= N dS
be the innitesimal element of area at X; N is the outward

unit normal. Let the corresponding innitesimal element of area at x be dS =
ndS; n is its outward unit normal. Then
dS = JF
T
dS
(2.12)
6
Finally let dc
be the innitesimal volume at X, and dc be the corresponding

innitesimal volume at x. Then
dc = Jdc
(2.13)
From (2.12) we have
dS
2
= |dS|
2
= (dS)
T
(dS) = dS
T
F
1
JJF
T
dS
, (2.14)
= dS
N
T
F
1
JJF
T
NdS
= J
2
dS
2
N
T
F
1
F
T
N,
= J
2
dS
2
N
T
C
1
N;
and so
dS = J
N
T
C
1
NdS
. (2.15)
Let the rst deformation deform the body S from the reference conguration
C
into the intermediate conguration C

1
; the corresponding deformation eld,
the deformation gradient and the displacement gradient are
x
1
= x
1
(X) X+u
1
(X) ,
F
1

x
1
X
,
H
1

u
1
X
.
(2.16)
Thus
C
1
F
T
1
F
1
= I +H
1
+H
T
1
+H
T
1
H
1
,
G
1

1
2
(C
1
I) =
1
2
H
1
+H
T
1
+H
T
1
H
1
.
(2.17)
The second deformation deforms the systemS from the intermediate congu-
ration C
1
into the present conguration C = C
2
. We can write the present position
7
in two ways: x = x(X) or equivalently x
2
= x
2
(x
1
); so we have:
x = x(X) = x
2
= x
2
(x
1
) . (2.18)
Accordingly
_
_
u(X) = x(X) X,
H
u
X
;
(2.19)
or
_
_
u
2
(x
1
) = x
2
(x
1
) x
1
,
H
2

u
2
x
1
.
(2.20)
However it is important to note that displacement elds u and u
2
are not the
same. Analogous to (2.17) we have
_
_
C
2
I +H
2
+H
T
2
+H
T
2
H
2
,
G
2

1
2
(C
2
I) =
1
2
H
2
+H
T
2
+H
T
2
H
2
.
(2.21)
The deformation gradient for the total deformation from C
to C is given by
F =
x
X
=
x
2
x
1
x
1
X
=

x
1
x
1
+
u
2
x
1
x
1
X
,
= (I +H
2
) F
1
= F
1
+H
2
F
1
. (2.22)
From (2.9), (2.5)
1
and (2.22) we obtain
G = G
1
+F
T
1
G
2
F
1
(2.23)
for Green-Saint-Venant strain tensor.
8
Innitesimal deformations
A deformation is said to be innitesimal if the components of u and of H are
small so that |u|
2
and |H|
2
are negligible as compared to |u| and |H| respectively.
Note that:
lim
u0
{F, C, B} = {I, I, I} , (2.24)
and
lim
u0
{H, G} = {0, 0} . (2.25)
For two successive innitesimal deformations we have from (2.23)
G w G
1
+G
2
. (2.26)
If we dene the innitesimal strain E by
E = Sym(H) , (2.27)
we obtain
G w E = E
1
+E
2
(2.28)
from (2.10) and (2.26). It can be proved that
_
_
F = I +H,
C = I + 2H,
J = det (I +H) = 1 +tr (H) .
(2.29)
Equilibrium laws
Assumptions on Forces
Let c C be a subregion of the body in the present conguration. In
order to simplify the problem, in continuum mechanics it is usual to make some
9
assumptions on forces acting upon c. The forces acting on c are divided into two
kinds: body forces and contact forces. The body forces are distributed within the
region c and the contact forces upon its boundary c. We can thus write the total
force and the total momentum with respect to the pole O as
F(c, c
e
) =
Z
c
bdc +
Z
c
t dS ,
M
O
(c, c
e
) =
Z
c
(x x
O
) bdc +
Z
c
(x x
O
) t dS ,
(2.30)
where x
O
is the position vector of O and is the mass density in the present
conguration; b and t are respectively the specic body force dened in c and
the contact force per unit surface area dened on c. F and M
O
depend not only
upon the region c but also upon its external world c
e
. We suppose that the body
force can not depend upon changes in the positions of particles of C that belong
to c
e
i.e. that are outside c. We thus neglect, for example, the mutual attractive
or repulsive force that particles feel because of other particles of the same body.
The functional dependence of b is: b = b(x). For the density of contact force,
we adopt Cauchys hypothesis; t = t (x, n), where n is the outward unit normal
to c at x. We state without proof
Theorem 2 (The fundamental theorem of Cauchy.) If t is regular enough
, then it is a linear function of n. That is,
t (x, n) = T(x) n. (2.31)
10
The Conservation of Mass
The mass m(c
) of a generic volume c
in the reference conguration is

given by
m(c
) =
Z
c
(X) dc
. (2.32)
We assume that the mass, in the present conguration, of the corresponding
part c is the same. Thus
m(c) =
Z
c
(x) dc = m(c
) . (2.33)
We have:
Z
c
(x) dc =
Z
c
(x(X)) J dc
=
Z
c
(X) dc
(2.34)
and the relation between the mass densities in the present and the reference con-
gurations is
J (x) =
(X(x)) , (2.35)
or
J (x(X)) =
(X) . (2.36)
Balance of Linear Momentum
We study only static deformations of a body, and assume that in the present
conguration the total force on a generic part c C is null:
F(c, c
e
) =
Z
c
(x) b(x) dc +
Z
c
t (x, n) dS = 0, c C. (2.37)
Using the fundamental theorem of Cauchy (2.31) we have
11
Z
c
(x) b(x)dc +
Z
c
T(x) ndS = 0, c C. (2.38)
Using the Divergence theorem, we get the local form
DivT(x) + (x) b(x) = 0, x C, (2.39)
of the equilibrium equation in the Eulerian description of deformation.
Note that this is a system of partial dierential equations and in order to solve
them we also need boundary conditions.
At every point x C, either the surface traction t
t (x, n) = T(x) n, (2.40)
or the displacement u,
u(x) = u(x) (2.41)
or a combination of t and u is prescribed. If we have only surface traction condi-
tions
_
_
DivT(x) + (x) b(x) = 0, x C,
t (x, n) = T(x) n, x C.
(2.42)
We transform the integral in (2.38) over c to that on the region c
in the
reference conguration that corresponds to c. Using (2.35) and (2.12), we get
Z
c
(X) b(x(X))dc
+
Z
c
T(x(X)) JF
T
NdS
= 0. (2.43)
12
In terms of the rst Piola-Kirchho stress tensor
T
(X) = JT(x(X)) F
T
, (2.44)
(2.43) becomes
Z
c
bdc
+
Z
c
NdS
= 0, c
. (2.45)
Using the Divergence theorem we write its local form and the traction bound-
ary condition as
DivT
b = 0, X C
,
T
= t
, X C
.
(2.46)
The Lagrangian form (2.46) of the equilibrium equations has the following
advantages over the Eulerian form (2.42). The mass density
is known and
elds are dened on the known shape C
of the body.
Balance of Momentum of Momentum
We assume that in the present conguration, the total momentum of forces
acting on a generic part c C is null. Using the fundamental theorem of Cauchy
we have
M
O
(c, c
e
) =
Z
c
(x x
O
) (x) b(x)dc +
Z
c
(x x
O
) T(x) ndS = 0.
Applying the Divergence theorem we get the local form of the balance of
momentum of momentum in the Eulerian form:
Div [(x x
O
) T(x)] + (x x
O
) (x) b(x) = 0, x C. (2.47)
13
In a rectangular Cartesian co-ordinate system (x
O
, e
1
, e
2
, e
3
),
(Div [(x x
O
) T(x)])
i
=
ihk
T
hk
+ ((x x
O
) (DivT))
i
,
where
ihk
is the permutation tensor. Because of the balance of linear momentum,
(2.47) is equivalent to
ihk
T
hk
= 0, i = 1, 2, 3.
The skew symmetry of implies the symmetry of the Cauchy stress tensor,
T = T
T
. (2.48)
Using (2.44), (2.48) becomes
T
F
T
= FT
T
.
Constitutive Equations
Dierent materials react in dierent ways to the same set of forces. In this
section we characterize an elastic material by equations called Constitutive Equa-
tions.
An unconstrained elastic body is characterized by the following constitutive
equation:
T = T(X) = T(X, F(X)) , X C
, (2.49)
where T(X, F(X)) is a C
1
function. A body is said to be homogeneous if in the
reference conguration C
the mass density does not depend upon the material

14
point X, i.e.,
= const, X C
,
and the function T on the right-hand side of the (2.49) does not depend explicitly
upon X. Note that for a homogeneous elastic body,
T
= T
(F) . (2.50)
An unconstrained material point is called hyperelastic if there exists a function
(F), called the specic energy, such that
T
(F) =

F
, (2.51)
or
T =

F
F
T
. (2.52)
The principle of Material Frame indierence requires that
= (G) . (2.53)
Thus
T
= F
G
(2.54)
and
T(F) =

G
F
T
=

(I +H)

G
I +H
T
. (2.55)
It can be proved that for an isotropic material
=(G) = (I
G
, II
G
, III
G
) ,
15
where I
G
,II
G
and III
G
are respectively the rst, the second and the third prin-
cipal invariants of G.
Assuming that is an analytical function of G, we obtain
(G) = (0) +

G
|
G=0
G+
1
2

2
G
2
|
G=0
G
2
+...,
and therefore
G
=

G
|
G=0
+
G
2
|
G=0
G+... .
From (2.35) and (2.29)
2
we have
=
1
J
= 1 tr (H) +... , (2.56)
and (2.55) can be written as
T = (1 tr (H)) (I +H)
G
|
G=0
+
G
2
|
G=0
G
I +H
T
+... . (2.57)
We set
T
0
=

G
|
G=0
,
C =

2
G
2
|
G=0
,
(2.58)
where T
0
equals the stress in the reference conguration, and C = C
T
is the
fourth-order elasticity tensor. In rectangular Cartesian coordinates, C satises
C
ijkl
= C
klij
= C
ijlk
.
Equation (2.57) now becomes
T = (1 tr (H)) (I +H) [T
0
+CG]
I +H
T
+... . (2.59)
16
For innitesimal deformations, we neglect terms in H of order greater than
one. Equation (2.59) becomes
T = T
0
+HT
0
+T
0
H
T
+CET
0
tr (H) . (2.60)
If T
0
= 0 the reference conguration is called natural and we have
T = CE. (2.61)
The values of the components of the elasticity tensor C depend upon the choice
of the reference conguration. For innitesimal deformations, because of (2.44),
(2.56), (2.60) and (2.8), the rst Piola-Kirchho stress tensor is given by
T
= JTF
T
' (1 +tr (H))
T
0
+HT
0
+T
0
H
T
+CET
0
tr (H)

I H
T
,
' T
0
+HT
0
+CE. (2.62)
For a stress free reference conguration, T
= T in linear elasticity. When

|T
0
| << |C|, then HT
0
is negligible as compared to the other two terms on the
right-hand side of (2.62) and we obtain
T
= T
0
+CE.
For the specic energy of deformation given by
(G) =
+ 2
2
I
2
G
2II
G
, (2.63)
we conclude from (2.55) that
T = (trE) I + 2E. (2.64)
17
For an innitesimal deformation with displacement gradient H
2
superimposed
upon a nite deformation of deformation gradient F
1
, we have from (2.23) and
(2.53),
= (G) =
G
1
+F
T
1
E
2
F
1
,
= (G
1
) +

G
|
G=G
1
(GG
1
) +
1
2

2
G
2
|
G=G
1
(GG
1
)
2
+... .
Neglecting terms of order greater than one in E
2
, we get
G
=

G
|
G=G
1
+
G
2
|
G=G
1
(GG
1
) , (2.65)
and

1
(1 tr (H
2
)) . (2.66)
From (2.55), (2.66), (2.22) and (2.65) we conclude that
T =

1
(1 tr (H
2
)) (F
1
+H
2
F
1
)
G
|
G=G
1
+
G
2
|
G=G
1
(GG
1
)
F
T
1
+F
T
1
H
T
2
.
Expanding these terms and recalling that the Cauchy stress tensor in the
intermediate conguration is given by
T
1
=

1
F
1

G
1
F
T
1
,
we get
T = T
1
+T
1
H
T
2
+

1
F
1
G
2
1
: F
1
E
2
F
T
1
F
T
1
+H
2
T
1
T
1
tr (H
2
) . (2.67)
For the specic deformation energy given by (2.63), (2.67) becomes
T = T
1
+T
1
H
T
2
+CE
2
+H
2
T
1
T
1
tr (H
2
) ,
18
where in rectangular Cartesian coordinates
1
,
C
ijkl
=

1
2B
1
il
B
1
kj
+B
1
ij
B
1
kl
.
When considering second order eects in the constitutive relation for inni-
tesimal deformations, we expand the specic deformation energy until the third
order terms in G. That is,
' () +

I
G
I
G
+

II
G
II
G
+

III
G
III
G
+
1
2
I
2
G
I
2
G
+
1
2
II
G
I
G
I
G
II
G
+
1
6
I
3
G
I
3
G
, (2.68)
where terms in () are evaluated at G = 0.
Substitution from (2.68) into (2.54) and retaining terms up to and including
H
2
yields
T
(H) ' I
E
I + 2E
+
I
HH
T + 2I
2
E
+
3
I
2
E
+
4
II
E
I
+(2 +
5
) I
E
EI
E
H
T
H
T
2
+
6
E
2
; (2.69)
where
3
,
4
,
5
and
6
, are the second order elasticities for the isotropic material,
and the reference conguration has been assumed to be stress free.
1
B
1
ij
are the components in rectangular coordinates of the left Cauchy-Green stress tensor
B
1
dened in (2.5)
2
.
19
Equations of elastostatics
The equilibrium equations (2.46) with the constitutive equation (2.50) give
the following partial dierential equation for a homogeneous elastic body.
T
iL
X
L
+b
i
(x(X)) =
T
iL
F
jM
F
jM
X
L
+b
i
(x(X)) = 0. (2.70)
Here we have used rectangular Cartesian coordinates. With the denitions
A
ijLM
(F) =
T
iL
F
jM
, b
i
(X) = b
i
(x(X)) ,
we write (2.70) as
A
ijLM
(F)

2
x
j
X
L
X
M
+
i
(X) = 0. (2.71)
This gives a system of 3 partial dierential equations in the 3 unknown defor-
mation elds x
i
(X) dened on the regular and known domain C
.
The relevant boundary conditions is (2.46)
2
. More general boundary condition
that account for the interaction between the body and its surroundings have been
discussed by Batra [4].
When a system of linear springs (g. 2.1) with spring constant k act on a part
of the boundary and springs are constrained to stay vertical, then
t(x, n) = ku
2
(x)e
2
.
However when prescribed surface traction on C
are such that they do not

depend upon the deformation of C
, then they are called dead loads.

20
Reference Configuration
Present Configuration
Figure 2.1: A system of linear springs
Traction boundary value problems involving dead load are easier to analyze
than those in which surface tractions depend upon the deformation of the surface
of the body. An exception may be the pressure loading.
A boundary value problem for a homogenous body, that is stress free in the
reference conguration, in the Lagrangian description of motion can be written as
A
ijLM
(F)

2
x
j
X
L
X
M
+
b
i
(X) = 0, X C
,
x(X) = x(X) , X C
0
,
T
(F) N
= t
(X, x, F) , X C
00
,
(2.72)
where C
= C
0
C
00
and C
0
C
00
=
If C
0
= then the problem is said to be a traction-value problem; if C

00
=
then the problem is said to be a place-value problem. In other cases the problem
21
is a mixed boundary-value problem. If t
= t
(X) then the load is called dead

load. x(X) is an assigned function of X.
We study only static problems and are not interested in its rigid motion. If
nal positions of three non-colinear points are prescribed, then the rigid motion
is not possible. For a traction value problem, the applied traction must be such
that the body is in equilibrium. That is
R
C
TndS +
R
C
(x) b(x) dc = 0,
R
C
(x x
O
) TndS +
R
C
(x x
O
) (x) b(x) dc = 0,
(2.73)
or
R
C
(F) N
dS
+
R
C
bdc
= 0,
R
C
(x x
O
) T
(F) N
dS
+
R
C
(x x
O
)
bdc
= 0.
(2.74)
We see that it is not possible to verify a priori if loads (t
, b) satisfy (2.74)
because the dependence of T
on F is unknown. Even for dead loads we have the

same problem because of the presence of x in (2.74)
2
. For this reason (2.74) is
used as a compatibility condition.
Truesdell and Noll [38] have given examples illustrating the non uniqueness of
solutions in non-linear elasticity.
We now consider the case when the applied surface tractions and body forces
depend continuously upon a small parameter . We hypothesize that displace-
ments resulting from the application of these loads to an elastic body also depend
22
continuously upon . We further assume that
b
(X, u
) = b
1
(X, u
) , X C
,
u
(X) = u
1
(X) , X C
0
,
T
= t
(X, u
, H
) = t
1
(X, u
, H
) , X C
00
.
(2.75)
Thus, X C
lim
0
u
(X) = 0,
and
u
(X) = u
1
(X) +o () ,
H
(X) = H
1
(X) +o () ,
E
(X) = E
1
(X) +o () ,
F
= I +H
1
(X) +o () ,
C
= I + 2E
T
1
+o () ,
C
1
= I 2E
T
1
+o () ,
T
= T
(H
) = CE
1
+o () ,
J
= 1 +tr (H
1
) +o () .
(2.76)
From (2.44), the rst Piola-Kirchho stress tensor is given by
T
= J
F
T
= (1 +tr (H
1
)) (CE
1
)
1 +H
T
1
+o () , (2.77)
= CE
1
+o () = T
.
Thus T
equals the Cauchy stress tensor T
.
Let
t
(x, u
, H
) = t
1
(x, u
, H
) +o () ,
23
be the load in the present conguration. Since the force acting on the same set of
material particles has to be the same, therefore
t
(x, u
, H
) dS
= t
(x, u
, H
) dS.
Using (2.15), (2.29)
3,2
, we obtain
t
(X, u
, H
) = J
(x, u
, H
)
p
N
T
C
1
N,
= [1 +tr (H
1
)] t
(x, u
, H
1 N
T
E
1
(X) N
+o () ,
= t
(x, u
, H
) +o () .
Expanding t
(x, u
, H
) in Taylor series around the point (X, 0, 0), we get

t
(x, u
, H
) = t
(X, 0, 0)
+
(
(X,0,0)
(x X) +
(X,0,0)
u
(X,0,0)
H
)
+o (k(x, u
, H
) (X, 0, 0)k) = t
1
(X, 0, 0)
+
2
(
t
1
x
(X,0,0)
u
1
(X) +
t
1
u
(X,0,0)
u
1
(X) +
t
1
H
(X,0,0)
H
1
(X)
)
+o
.
Thus
t
(X, u
, H
) = t
1
(X, 0, 0) +o () =
t
1
(X) +o () . (2.78)
Similarly, for the body force we have
b
(X, u
) = b
(X, 0) +
(X,0)
u
+o (k(X, u
) (X, 0)k) , (2.79)

= b
1
(X, 0) +
2
b
1
u
(X,0)
u
1
+o
= b
1
(X, 0) +o () =
b
1
(X) +o () .
24
We conclude from (2.78) and (2.79) that in Linear Elasticity only dead loads
can be considered.
From (2.46), (2.77), (2.78), (2.79) and (2.61), we conclude that
Div (CE
1
) +
b
1
(X) +o() = 0, X C
,
u
1
(X) +o() = u
1
(X) , X C
0
,
(CE
1
) N
+o() =
t
1
(X) , X C
00
.
We divide by , take the limit as 0, and arrive at the following equations
governing static deformations of a linear elastic body that is stress-free in the
reference conguration.
Div (CE
1
) +
b
1
= 0, X C
,
u
1
= u
1
, X C
0
,
(CE
1
) N
t
1
, X C
00
.
(2.80)
For a traction-value problem, loads must satisfy the following equations for
the problem to have a solution.
R
C
t
1
dS
+
R
C

b
1
dc
= 0,
R
C
(XX
O
)
t
1
dS
+
R
C
(XX
O
)
b
1
dc
= 0.
(2.81)
Signorinis Method
Non-linear problems that satisfy the following conditions can be solved by the
Signorini method.
1. The rst Piola-Kirchho stress tensor T
depends analytically upon the

25
displacement eld H:
T
= A(H) =

X
n=1
A
(n)
(H) , A(0) = 0,
where A
n
(H) is a homogeneous polynomial of degree n in H. In component
form,
T
iL
= A
iL
(H) = A
(1)
iL
(H) +A
(2)
iL
(H) +A
(3)
iL
(H) +..., (2.82)
= C
(1)
ijLM
H
jM
+C
(2)
ijhLMN
H
jM
H
hN
+C
(3)
ijhkLMNO
H
jM
H
hN
H
kO
+... .
2. The surface traction is a known function of the position X in the reference
conguration, so the loads are dead:
t
= t
(X) .
3. There exists a non-dimensional parameter such that loads (b, t
) are an-
alytical functions of . That is,
b(, X) =

X
n=1
n
b
n
(X) , t
(, X) =

X
n=1
n
t
n
(X) .
Under these hypotheses, Poincarre, as cited by Romano [35], has demon-
strated that the displacement eld is an analytical function of and can be rep-
resented as
u(, X) =

X
n=1
n
u
n
(X) . (2.83)
Thus
H =

X
n=1
n
H
(n)
. (2.84)
26
Substituting (2.84) into (2.82), we obtain
T
iL
= C
(1)
ijLM
H
(1)jM
+
2
H
(2)jM
+...
+C
(2)
ijhLMN
H
(1)jM
+
2
H
(2)jM
+...

H
(1)hN
+
2
H
(2)hN
+...
+... ,
which can be written as
T
=

X
n=1
C
(1)
H
(n)
+B
n
(H
1
, ..., H
n1
)
, (2.85)
where
B
1
= 0,
and
C
(1)
= C
is the fourth-order tensor of linear elasticity. Because of the symmetry properties
of this tensor, (2.85) can be written as
T
=

X
n=1
CE
(n)
+B
n
H
(1)
, ..., H
(n1)
, (2.86)
where
E
(n)
= Sym
H
(n)
=
1
2
H
(n)
+H
T
(n)
.
Substitution from (2.86) into (2.46)
1
and (2.72)
2,3
yields the following equa-
tions for a mixed boundary value problem.
DivCE
n
+
b
n
= 0, X C
,
u
n
= u
n
, X C
0
,
(CE
n
) N
t
n
, X C
00
,
(2.87)
27
where
b
n

b
n
+DivB
n
H
(1)
, ..., H
(n1)
t
n
= t
n
B
n
H
(1)
, ..., H
(n1)
.
(2.88)
For n = 1, we have a linear problem which can be solved. Then we can solve
the linear problem for n = 2. Note that loads for this linear problem depend
upon the solution of the problem for n = 1. We can continue this procedure for
n = 3, 4, ... . In order for the n-th order problem to have a solution, loads (2.88)
must satisfy the overall balance of forces and moments given by (2.74).
There are two ways to accomplish this. We rst note that the loads are
dead. Thus (2.74)
2
is a compatibility condition because of the presence of x in
it. Equation (2.74)
1
can be considered as a priori restriction on the loads. We
can either use Da Silvas theorem, as done by Romano [35], or follow Green and
Adkins [21] and set at X = 0:
u = 0, H = H
T
. (2.89)
Green and Adkins have demonstrated that (2.89) can be substituted for the
compatibility condition (2.74).
Semi-inverse method
Techniques to solve a given boundary-value problem can be classied into the
following three categories.
1. Direct method.
28
2. Inverse method.
3. Semi-inverse method.
In the direct method, governing equations are integrated without any a priori
assumptions on the displacement eld, and the constants of integration are de-
termined from the boundary conditions. In the other two methods, either all or
a part of the solution is assumed and loads required to produce the solution are
computed. If the computed loads equal the prescribed ones, we have a solution of
the given boundary- value problem; otherwise, we need to redo the problem. In
the semi-inverse method, a part of solution is presumed and unknowns in it are
determined by satisfying the pertinent eld equations and boundary conditions.
Each of these three methods is suitable for analyzing a certain class of problems.
We often invoke the St.-Venant principle to nd an approximate solution of a
traction-value problem; it may be stated as follows.
If on a given portion of a body or on a part of its surface, small with respect
to a typical dimension of the body, act respectively a system of body or contact
forces and the body is in equilibrium, then in the region that is far from that
where forces act, the deformation and the state of stress are determined from the
resultant forces and the resultant moment of these forces. This means that the
details of the distribution of these forces are important only in the neighborhood
of the region of application.
29
This principle was enunciated by Saint Venant for a cylindrically-body. Boussi-
nesq extended it to the general three dimensional case.
Mathematically precise versions of this principle have been given by Knowles
and Sternberg [31], Toupin [37] and others; Batra ([6], [7] and [8] ) proved it even
for a helical body ([5] and [9]). Estimates of the rate of decay of the solution with
the distance from the loaded surface have been given by Horgan ([24] and [26]).
This principle is important in applications because it enables us to substitute
a particular distribution of forces by one statically equivalent to it and simpler to
use.
We will use the St.-Venant principle in solving the St.-Venant problem in the
sense that displacement elds within a cylindrical body loaded only at the end
faces will be expressed in terms of the resultant of these loads. In the solution
of the problem, we will assume that the displacement eld is a polynomial in the
axial co-ordinate.
30
Chapter 3
St.-Venants Problem
Governing equations
We analyze innitesimal deformations of a homogeneous elastic cylindrical
body like the one shown in Fig. 3.1. The cross-section of the body is arbitrary
except that we require it to be smooth enough so that the divergence theorem in
the plane is applicable.
We point out that the Euclidean space E is not vectorial, so we introduce
a xed point O E and measure positions of points in E relative to O; the
point O is called the origin. Thus points become vectors belonging to the space
of translations U V E . We introduce the unit vector e U to characterize
the direction of generators of a cylinder; the axis W of the cylinder is dened as
follows
1
:
W span{e} . (3.1)
The vector e is called the axis of cylinder. The sections D of the cylinder
1
The span
S span{v
1
, v
2
, ..., v
n
}
of a set of vectors {v
1
, v
2
, ..., v
n
} is the vectorial space of all linear combinations of vectors
v
1
, v
2
, ..., v
n
. That is,
a S (a
1
, a
2
, ..., a
n
) R
n
a = a
i
v
i
.
31
eee
Figure 3.1: A cylindrical body
belong to the space orthogonal to W that is denoted by V:
D V W
= {e}
. (3.2)
So the space of translation can be written as
2
U = W V. (3.3)
D represents all sections of the cylinder. Often it is necessary to specify only
one particular section which can be done by giving the distance of the section from
the origin O. The section distant Z from O is denoted by D {Z}, or D
Z
. To
simplify the notation we locate the origin in the bottom section of the cylinder.
2
is the direct sum when the vectorial spaces being summed are orthogonal.
32
We can dene the cylinder as
C

n
X =

X+Ze :

X D , Z [0, L]
o
, (3.4)
and write
C
= D[0, L],
C
= (D[0, L]) (D{0}) (D{L}) .

(3.5)
We require that D is closed, linearly connected and
D = (D
0
) \
=1
D
,
where
3
D
0
is the smallest simply connected set that contains D,
_
_
{1, ..., } , D
{(D
0
) \ (D
0
)} ,
, {1, ..., } and 6= , D
= .
(3.6)
Thus a cross-section can have at most simply connected holes.
The boundary of D is a one-dimensional set, its connected parts are the ex-
ternal boundary D
0
and the boundaries of the holes D
; so we write
D =

=0
D
. (3.7)
In order to evaluate a line integral it is important to x the orientation of the
boundary D. Let e
t
be a unit tangent vector of D that is coherent with the
orientation, and the vector N be an outward unit normal. The orientation of D
is taken such that (e
t
, N, e) follows the right-hand rule. Hence the orientation
of D
0
is anticlockwise and that of D
( = 1, ..., ) clockwise; these are shown

in Fig. 3.2.
3
\ is the usual symmetric dierence.
33
N
N
e
t
e
t
N
N
N
N
e
t
e
t
Figure 3.2: Positive direction of paths on the boundary and direction of the
outward normal
34
The St.-Venant problem involves nding the displacement eld in a cylindrical
body loaded only at the end faces D
0
and D
L
. We rst consider innitesimal
deformations of a stress-free reference conguration, and assume that the body
is made of a homogeneous and isotropic linear elastic material that obeys the
constitutive equation
T
= (trE)1 + 2E, X C
. (3.8)
Let
u = we +v, X C
,
T
= e e +t e +e
t +

T, X C
,
E = SymGradu = e e + e +e +

E, X C
.
(3.9)
Because of (3.8)
t = t, X C
.
Here w and v are, respectively, the axial and the in-plane components of the
displacement eld u; and t are respectively the axial and the transverse shear
components of the rst Piola-Kirchho stress tensor T
;

T is the in-plane part of
T
; and are, respectively, the axial and the transverse shear components of
the innitesimal strain tensor E;

E is the in-plane part of E. This decomposition
can be represented in the matrix notation as
T
=
_
_
_
_
T t
t
_
_
_
_
, E =
_
_
_
_
E

_
_
_
_
.
35
In rectangular Cartesian coordinates with X
1
and X
2
axes in D
0
, we have
T
=
_
_
_
_
_
_
_
_
T
11

T
12
t
1
T
12

T
22
t
2
t
1
t
2

_
_
_
_
_
_
_
_
, E =
_
_
_
_
_
_
_
_
E
11

E
12

1
E
12

E
22

2
1

2

_
_
_
_
_
_
_
_
.
The components of the innitesimal strain tensor are related to the displace-
ment eld by
= w
0
,
=
1
2
(v
0
+gradw) ,
E = Symgradv,
(3.10)
where a prime denotes dierentiation with respect to Z, and grad is the 2-
dimensional gradient operator. These can be easily veried by noting that
Gradu = w
0
e e +v
0
e +e (gradw) +gradv, X C
.
Thus
E = SymGradu = w
0
e e +e
1
2
(v
0
+gradw) (3.11)
+
1
2
(v
0
+gradw) e+Symgradv, X C
.
A comparison of (3.9) and (3.11) gives (3.10). Equations (3.10) and (3.11)
give
_
_
trE = tr
E+,
tr
E = tr (Symgradv) = divv,
trE = w
0
+divv.
(3.12)
36
Substitution from (3.9)
2
and (3.9)
3
into (3.8) yields
_
_
= ( + 2) w
0
+divv,
t = 2 =(v
0
+gradw) ,
T = (w
0
+divv)
1+
h
(gradv) + (gradv)
T
i
.
(3.13)
From (6.18) and (3.9)
2
we have
DivT
= (+divt) e +t+div
T.
Therefore the balance of linear momentum (2.46)
1
in the absence of body
forces can be written as
_
0
+divt = 0, X C
,
t+div
T = 0, X C
.
(3.14)
We now substitute from (3.13) into (3.14) to obtain the following eld equa-
tions for the determination of displacements v and w.
_
_
( + 2)w
00
+ ( +) divv+ 4
R
w = 0, X C
,
div
2Symgradv +
1divv
+( +) gradw+v= 0, X C
,
(3.15)
or
_
_
( + 2)w
00
+ ( +) divv+ 4
R
w = 0, X C
,
v+ ( +) gradw+ 4
R
v+( +) graddivv = 0, X C
.
The condition of zero surface tractions on the mantle can be written as
TN = 0, X D[0, L] ,
t N = 0, X D[0, L] .
(3.16)
37
In terms of the displacement eld, it becomes
(w
0
+divv)
1N
+((gradv) + (gradv)
T
)N = 0, X D[0, L] ,
(v
0
+gradw) N = 0, X D[0, L] .
(3.17)
In the semi-inverse method used to solve the problem, the loads on the end
faces are not specied but are determined by the solution. However, they must
satisfy the global equilibrium equations:
Z
C
TndS = 0,
Z
C
x TndS = 0.
(3.18)
We now consider the equilibrium of the portion B
z
of the cylinder C that is
the image in the present conguration of the portion
B
Z
=
n
X =

X,

Z
X D ,
Z [0, Z]
o
of the cylinder C
in the reference conguration between the bases D(0) and D(Z).

Global equilibrium equations for B
Z
are
Z
B
Z
T
NdS
= 0,
Z
B
Z
(X+u) T
NdS
= 0.
In the linear theory, these become
Z
B
Z
T
NdS
= 0,
Z
B
Z
XT
NdS
= 0.
(3.19)
38
Let
_
_
f (Z)
Z
D(Z)
T
edS
,
m(Z)
Z
D(Z)
XT
edS
,
(3.20)
be the resultant force and the resultant moment with respect to

O, Z
of forces
acting on the section D
Z
.
We set
f (Z) = f
0
= f
p
0
+f
a
0
e, Z [0, L] ,
m(Z) = m
0
+Zm
1
= m
p
(Z) +m
a
e, Z [0, L] ,
(3.21)
where
m
p
(Z) = m
p
0
+Zm
1
,
m
1
= f
p
0
,
m
a
= m
a
0
,
(3.22)
and
f
p
0
=
Z
D(0)
tdS
,
f
a
0
=
Z
D(0)
dS
,
m
p
0
=
Z
D(0)
dS,
m
a
0
=
Z
D(0)
XtdS
e.
(3.23)
In order to prove (3.21), we recall that
N = e, X D{0} ,
N = e, X D{L} .
(3.24)
From (3.9)
2
, we get
T
e = t +e, X D.
39
Since B
Z
is in equilibrium and the mantle is traction free, therefore
Z
D(Z)
T
edS
=
Z
D(0)
T
edS
,
and Z [0, L]
f (Z) =
Z
D(Z)
T
edS
=
Z
D(0)
T
edS
=
Z
D(0)
tdS
+
Z
D(0)
edS
,
or
f
0
= f
p
0
+f
a
0
e.
In the same way (3.19)
2
gives
Z
D(0)
XT
edS
+
Z
D(Z)
X+Ze
edS
= 0,
which can be written as
Z
D(Z)
XT
edS
=
Z
D(0)
XT
edS
Ze
Z
D(Z)
T
edS
.
Thus
m(Z) =
Z
D(0)
X(t +e) dS
Ze
Z
D(Z)
tdS
,
=
Z
D(0)
XtdS
Z
D(0)
dS
Zf
p
0
, Z [0, L] ,
that is (3.21)
2
.
Quantities f and m equal, respectively, the resultant force and the resultant
moment that are applied to the basis. From the St.-Venant principle they repre-
sent
6
equivalence classes of possible distribution of loads on the basis.
40
Substitution for stresses in terms of displacement gradient (recall the consti-
tutive relations (3.13)) (3.21), (3.22) and (3.23) gives the following relationships
between f
p
0
, f
a
0
, m
p
(z) , m
a
and the displacement eld v, w.
f
p
0
=
Z
D(Z)
(v
0
+gradw) dS
=
Z
D(0)
(v
0
+gradw) dS
,
f
a
0
=
Z
D(Z)
[( + 2) w
0
+divv] dS
,
=
Z
D(0)
[( + 2) w
0
+divv] dS
,
m
p
(Z) =
Z
D(Z)
(( + 2) w
0
+divv)
dS
,
=
Z
D(0)
(( + 2) w
0
+divv) dS
Z
Z
D(0)
(v
0
+gradw) dS
,
m
a
=
Z
D(Z)
(v
0
+gradw)
dS
,
=
Z
D(0)
(v
0
+gradw)
dS
.
(3.25)
For the St.-Venants problem, we list below all equations in terms of displace-
ments:
4
R
w + ( +) divv
0
+ ( + 2)w
00
= 0, X C
,
div
2Symgradv+
1divv
+( +) gradw
0
+v
00
= 0, X C
2Symgradv+
1divv
N = (w
0
)N, X D[0, L] ,
(gradw) N = v
0
N, X D[0, L] ;
f
p
0
=
Z
D(Z)
(v
0
+gradw) dS
,
f
a
0
=
Z
D(Z)
[( + 2) w
0
+divv] dS
,
(3.26)
41
m
p
0
+Zm
1
=
Z
D(Z)
(( + 2) w
0
+divv)
dS
,
m
a
0
=
Z
D(Z)
(v
0
+gradw)
dS
.
Solution of St.-Venants problem by a semi-inverse method
We seek a solution of equations (3.26) of the form
w(X) =
M
X
i=0
1
i!
w
i
Z
i
, X C
,
v (X) =
M
X
i=0
1
i!
v
i
Z
i
, X C
.
(3.27)
With this hypothesis equations (3.26), in terms of the unknown functions v
i
and w
i
, become
4
R
w
i
+ ( +) divv
i+1
+ ( + 2)w
i+2
= 0,
X D,
div
2Symgradv
i
+
1divv
i
+( +) gradw
i+1
+v
i+2
= 0,
X D;
2Symgradv
i
+
1divv
i
N = (w
i+1
)N,
X D,
gradw
i
N = v
i+1
N,
X D;
Z
D(Z)
(v
i+1
+gradw
i
) dS
= 0, i > 0,
Z
D(Z)
[( + 2) w
i+1
+divv
i
] dS
= 0, i > 0,
Z
D(Z)
(( + 2) w
i+1
+divv
i
)
dS
= 0, i > 1,
Z
D(Z)
(v
i+1
+gradw
i
)
dS
= 0, i > 0.
(3.28)
Thus the three dimensional problem has been reduced to a sequence of two
dimensional problems.
42
In terms of the boundary-value problems A, B, C and D given in the Appendix,
the boundary-value problems (3.28) can be identied as follows.
For i = M, the boundary-value (b-v) problem dened by equations (3.28)
1,4
is problem A and it has the solution w
M
= w
0
M
. The b-v problem dened
by equations (3.28)
2,3
is problem B whose solution is v
M
= v
0
M
+
M
,
where
M
is a constant. Here and below, quantities with superscript zero signify
constants.
Similarly for i = M 1, equations (3.28)
1,4
, (3.28)
2,3
, (3.28)
5,8
and (3.28)
6,7
dene respectively problems A, B, C and D of the Appendix and their solutions
are
gradw
M1
=
h
v
0
M
+
M
X
i
+ 2
M
f
M
;
divv
M1
=
w
0
M
( +)
;
M
= 0, v
M
= v
0
M
and w
M1
= w
0
M1
v
0
M

X;
w
M
= 0 and v
M1
= v
0
M1
+
M1
.
Following the same procedure once more, we obtain
gradw
M2
=
h
v
0
M1
+
M1
X
i
+ 2
M1
f
M1
;
divv
M2
=

w
0
M1
v
0
M

X
( +)
;
M1
= 0, v
M1
= v
0
M1
and w
M2
= w
0
M2
v
0
M1

X;
w
M1
= 0, v
M2
= v
0
M2
+
M2
and v
M
= 0.
43
We can apply the same reasoning until i = 2 and thus get
v
4+i
= 0, i = 0, ..., M 4;
w
3+i
= 0, i = 0, ..., M 3;
v
3
= v
0
3
,
v
2
= v
0
2
+
2
,
w
2
= w
0
2
v
0
3

X.
For i = 1, equations (3.28)
1,4
, (3.28)
2,3
, (3.28)
5,8
and (3.28)
6,7
dene respec-
tively problems A, B, C and D
of the Appendix and their solutions are

gradw
1
=
h
v
0
2
+
2
X
i
+ 2
2
f
2
,
divv
1
=

w
0
2
v
0
3

X
( +)
,
2
= 0, v
2
= v
0
2
and w
1
= w
0
1
v
0
2

X,
w
0
2
= v
0
3

b. (3.29)
For i = 0, equations (3.28)
2,3
form problem B of the Appendix, and have the
solution
divv
0
=

w
0
1
v
0
2

X
( +)
.
Using (6.38) and (3.29) we get
v
1
= v
0
1
+
0
1
v
0
3

X+
n
sym
h
X
io
v
0
3
, X D;
v
0
= v
0
0
+
0
0

Xw
0
1
X+
n
sym
h
X
io
v
0
2
, X D.
Since the divergence of v
1
is not zero we cannot use the solution of problem
A. Equations (3.28)
1,4
for i = 0 give
44
4
R
w
0
+ 2v
0
3

b

X
= 0, X D;
gradw
0
N = v
0
1
N+
h
(
0
1
)

X
i
N
v
0
3

X N
+
n
sym
h
X
io
(v
0
3
) N, X D.
(3.30)
A solution of equations (3.26) of the form (3.27) is
4
v
4+i
= 0, i = 0, ..., M 4;
v
3
= v
0
3
,
v
2
= v
0
2
,
v
1
= v
0
1
+ (
0
1
)

X
v
0
3

X+
n
sym
h
X
io
(v
0
3
) ,
v
0
= v
0
0
+ (
0
0
)

Xw
0
1
X+
n
sym
h
X
io
(v
0
2
) ,
w
3+i
= 0, i = 0, ..., M 3;
w
2
= v
0
3

b

X
,
w
1
= w
0
1
v
0
2

X,
w
0
is a solution of (3.30).
(3.31)
Generalized Axisymmetric Problems
The only shape for the cross-section that is possible in this case is a circle
with one circular hole. In terms of the cylindrical coordinates with the origin at
the centroid of the cross-section, we can write
D
n
X = rr : r [a, b] , [0, 2]
o
.
4
It can be seen that v
0
3
, v
0
2
,
0
1
and w
0
1
are proportional to the loads on the basis.
45
For a generalized axisymmetric problem, the displacement eld does not de-
pend upon . However the circumferential or the tangential displacement need
not vanish. The solution (3.31) becomes
v
(2+i)r
= 0, i = 0, ..., M 4;
v
(2+i)
= 0, i = 0, ..., M 4;
v
1r
= 0,
v
1
=
0
1
r,
v
0r
= w
0
1
r,
v
0
=
0
0
r,
w
2+i
= 0, i = 0, ..., M 3;
w
1
= w
0
1
.
The function w
0
satises
_
_
4
R
w
0
= 0, X D;
gradw
0
N =
h
(
0
1
)

X
i
N, X D.
(3.32)
Equations (3.32) have the solution
5
:
gradw
0
= v
0
1
+
0
1
+ 2
0
1
f

,
where
divf = 0, X D;
rotf = 1, X D;
f N =0, X D.
For a circular cross-section,
5
See the problem A in the Appendix.
46
4
R
w
0
= 0, X D;
gradw
0
N = 0, X D;
and therefore
w
0
= w
0
0
, X D.
Summarizing the preceding results, the solution of the generalized axisym-
metric St.-Venant problem is
v
i
= 0, i = 2, ..., M;
v
1r
= 0,
v
1
=
0
1
r,
v
0r
= w
0
1
r,
v
0
=
0
0
r,
w
i
= 0, i = 2, ..., M;
w
1
= w
0
1
,
w
0
= w
0
0
,
where it is possible to prove that
0
1
and w
0
1
are proportional respectively to the
axial resultant moment and the axial resultant force on the basis.
Uniform pressure in the reference conguration
Formulation of the problem
We now study the case when the state of stress in the reference conguration
is that of uniform pressure,
T
0
= p1.
47
Since the body is in equilibrium in the reference conguration, therefore
DivT
0
= Grad (p) = 0, X C
;
T
0
N = pN =
t
0
, X C
.
When no additional loads are applied on the mantle of the cylindrical body,
the displacement u is governed by
Div (HT
0
+CE) = 0, X C
;
(HT
0
+CE) N = 0, X D[0, L] .
(3.33)
We use the decompositions
T
=
_
_
_
_
T t
t
_
_
_
_
, E =
_
_
_
_
E

_
_
_
_
, H =
_
_
_
_
gradv v
0
gradw
_
_
_
_
,
T
0
= p
_
_
_
_
1 0
0 1
_
_
_
_
, HT
0
= p
_
_
_
_
gradv v
0
gradw w
0
_
_
_
_
, e =
_
_
_
_
0
1
_
_
_
_
,
T
e = t +e,
and note that the rst Piola-Kirchho stress tensor is not symmetric because of
the prestress. Since
CE = (w
0
+divv)
_
_
_
_
1 0
0 1
_
_
_
_
+ 2
_
_
_
_
symgradv
w
0
_
_
_
_
,
=
_
_
_
_
(w
0
+divv)
1 + 2symgradv (v
0
+gradw)
(v
0
+gradw) (w
0
+divv) + 2w
0
_
_
_
_
,
48
the components of the rst Piola-Kirchho stress tensor are
T = p
1 p gradv+(w
0
+divv)
1 + 2symgradv,
t = pv
0
+(v
0
+gradw) ,
t = p gradw +(v
0
+gradw) ,
= p pw
0
+(w
0
+divv) + 2w
0
.
(3.34)
Thus
HT
0
+CE =
_
_
_
_
T+p
1 t
t +p
_
_
_
_
,
(HT
0
+CE) e = t + ( +p) e. (3.35)
Substitution from (3.34) and (3.35) into (3.33) yields
div
p gradv+(w
0
+divv)
1 + 2symgradv
+ [pv
0
+(v
0
+gradw)]
0
= 0, X C
;
div {p gradw +(v
0
+gradw)} + [pw
0
+(w
0
+divv) + 2w
0
]
0
= 0, X C
p gradv+(w
0
+divv)
1 + 2symgradv

N = 0, X D[0, L] ;
[p gradw +(v
0
+gradw)]

N =0, X D[0, L] .
Global balance of forces and moments
Subtracting the resultant of forces and moments of forces acting in the present
conguration from those in the reference conguration, we obtain
Z
B
Z
(HT
0
+CE) NdS
= 0,
Z
B
Z
(X+u) (HT
0
+CE) NdS
= 0.
(3.36)
49
With
f (Z) =
Z
D(Z)
(HT
0
+CE) edS
,
m(Z) =
Z
D(Z)
X(HT
0
+CE) edS
,
we can show that Z [0, L]
f (Z) = f
0
= f
p
0
+f
a
0
e,
m(Z) = m
0
+Zm
1
,
= m
p
(Z) +m
a
0
e,
= m
p
0
+m
a
0
eZ f
p
0
,
(3.37)
where
f
p
0
=
R
D(0)
tdS
,
f
a
0
=
R
D(0)
( +p) dS
,
m
p
0
=
R
D(0)
( +p) dS
,
m
a
0
=
h
R
D(0)

XtdS
i
e.
(3.38)
In order to prove (3.37), we recall (3.33)
2
and note that
Z
B
Z
(HT
0
+CE) NdS
=
Z
D(0)
(HT
0
+CE) edS
+
Z
D(Z)
(HT
0
+CE) edS
= 0.
Now using (3.35),
f (Z) =
Z
D(Z)
T
edS
=
Z
D(0)
(HT
0
+CE) edS
=
Z
D(0)
tdS
+
Z
D(0)
( +p) edS
,
= f
p
0
+f
a
0
e = f
0
, Z [0, L] .
50
Neglecting terms in u of order greater than 1 and considering (3.33)
2
, we
conclude from (3.36)
2
that
Z
B
Z
(X+u) (HT
0
+CE) NdS
=
Z
D(0)
X(HT
0
+CE) edS
+
Z
D(Z)
X+Ze
(HT
0
+CE) edS
= 0.
Thus
m(Z) =
Z
D(0)
X(HT
0
+CE) edS
Ze
Z
D(Z)
(HT
0
+CE) edS
,
m(Z) =
Z
D(0)
X(t + ( +p) e) dS
Zf
p
0
,
= m
a
0
e +m
p
0
Zf
p
0
, Z [0, L] .
Substituting in (3.38) for stresses in terms of displacement eld (3.34) yields
f
p
0
=
Z
D(Z)
pv
0
+(v
0
+gradw) dS
=
Z
D(0)
pv
0
+(v
0
+gradw) dS
,
f
a
0
=
Z
D(Z)
[( + 2 p) w
0
+divv] dS
,
=
Z
D(0)
[( + 2 p) w
0
+divv] dS
,
m
p
(Z) =
Z
D(Z)
[( + 2 p) w
0
+divv] dS
,
=
Z
D(0)
[( + 2 p) w
0
+divv] dS
Z
Z
D(0)
[pv
0
+(v
0
+gradw)] dS
,
m
a
=
Z
D(Z)
[pv
0
+(v
0
+gradw)] dS
,
=
Z
D(0)
[pv
0
+(v
0
+gradw)] dS
.
(3.39)
51
The nal equations are
( p) 4
R
w + ( +) divv+ ( + 2 p)w
00
= 0, X C
;
div
2Symgradv pgradv +
1divv
+( +) gradw+ ( p) v
00
= 0,X C
2Symgradv pgradv +
1divv

N = w
0
N, X D[0, L] ;
( p) (gradw)

N = v
0

N, X D[0, L] ;
f
p
0
=
Z
D(Z)
[( p) v
0
+gradw] dS
,
f
a
0
=
Z
D(Z)
[( + 2 p) w
0
+divv] dS
,
m
p
0
+Zm
1
=
Z
D(Z)
[( + 2 p) w
0
+divv]
dS
,
m
a
0
=
Z
D(Z)
[( p) v
0
+gradw] dS
.
(3.40)
With the hypothesis (3.27) on displacements, equations (3.40) yield
( p) 4
R
w
i
+ ( +) divv
i+1
+ ( + 2 p)w
i+2
= 0,
X D;
div
2Symgradv
i
pgradv
i
+
1divv
i
+( +) gradw
i+1
+ ( p) v
i+2
= 0,
X D;
2Symgradv
i
pgradv
i
+
1divv
i

N = (w
i+1
)
N,
X D;
( p) gradw
i

N = v
i+1

N,
X D;
Z
D(Z)
[( p) v
i+1
+gradw
i
] dS
= 0, i > 0;
Z
D(Z)
[( + 2 p) w
i+1
+divv
i
] dS
= 0, i > 0;
Z
D(Z)
[( + 2 p) w
i+1
+divv
i
]
dS
= 0, i > 1;
Z
D(Z)
[( p) v
i+1
+gradw
i
] dS
= 0, i > 0.
(3.41)
A solution of the system (3.41) is not yet available.
52
Solution of St.-Venants Problem by Signorinis Method
We apply Signorinis method to nd a solution of St-Venants problem when
the stress strain relation is quadratic in displacement gradients. However, instead
of solving the problem in complete generality, we assume that the rst-order de-
formations are pure torsional. Furthermore, we adopt hypothesis (3.27) on the
second order displacement eld.
As noted earlier, Signorinis method transforms a non-linear problem of n-th
order, into n linear problems.
The innitesimal twist per unit length, , is identied as the small parameter.
That is,

0
1
= ,
and from (2.83) we have
u(, X) =
2
X
n=1
n
u
n
(X) = u
1
(X) +u
2
(X)
2
,
where
u
1
(X) = Z
+e. (3.42)
The function is the warping function and is the solution of the problem
(3.30). We have eliminated the rigid body displacement by adopting the Green
and Adkins condition (2.89).
From (2.84) we have
H = H
1
+
2
H
2
,
53
where
H
1
= Z +
e +e grad,
E
1
=
1
2
nh
+grad
i
e +e
h
+grad
io
.
Thus
I
E
1
= 0.
We write (2.69) as
T
= T
(1)
+
2
T
(2)
,
where
T
(1)
= 2E
1
+I
E
1
I,
T
(2)
=

T
(2)
+I
h
1
2
I
H
1
H
T
1
+

4
2
I
(E
1
)
2
i
H
T
1
2
+
6
E
1
.
The displacement eld u
2
is a solution of the boundary value problem (2.87)
and (2.88) with n = 2.
In order to simplify the notation, we set
u
2
(X) = v (X) +w(X) e,
and from (3.26) obtain the following equations for the determination of v and w.
4
R
w + ( +) divv+ ( + 2)w
00
= 2 ( +) Z
2
, X C
,
div
2Symgradv +
1divv
+( +) gradw+v
=
2
( +) grad
h
(grad)
i
4
2
h
+grad
i
6
4
div
nh
+grad
i
+grad
io
, X C
,
54
2Symgradv+
1divv
N = (w
0
)N
+
_
grad
i
+
1
_
_
1
2
4
2
+grad
2
+
(grad) +
2
( +) Z
2
_
_
_
_
N X D[0, L] ,
(gradw) N = v
0
NZ
2
X N
, X D[0, L] .
(3.43)
We solve this linear problem by following the method outlined in the next
chapter for analyzing a linear Almansi-Michell problem. In particular, we assume
that v and w are given by (3.27). Since body forces present in (3.43)
1,2
are of
order
2
, we can use the solution (4.17) by setting m = 2 and obtain
w
i
= 0, i = 6, ..., M;
w
5
= v
0
6

b

X
,
w
4
= v
0
5

b

X
,
v
i
= 0, i = 7, ..., M;
v
i
= v
0
i
, i = 5, 6;
v
4
= v
0
4
+ (
0
4
)

X
v
0
6

X+
n
sym
h
X
io
(v
0
6
) ,
v
3
= v
0
3
+ (
0
3
)

Xw
0
5
X+
n
sym
h
X
io
(v
0
5
) , X D.
The other part of the displacement eld can be found by solving the following
system of equations:
4
R
w
0
+ ( +) divv
1
+ ( + 2)w
2
= 0,
4
R
w
1
+ ( +) divv
2
+ ( + 2)w
3
= 2 ( +)
2
,
4
R
w
2
+ ( +) divv
3
+ ( + 2)w
4
= 0,
X D;
55
div
2Symgradv
0
+
1divv
0
+( +) gradw
1
+v
2
=
2
( +) grad
h
(grad)
i
4
2
h
+grad
i
6
4
div
nh
+grad
i
+grad
io
,
div
2Symgradv
1
+
1divv
1
+( +) gradw
2
+v
3
= 0,
div
2Symgradv
2
+
1divv
2
+( +) gradw
3
+v
4
= 0,
X D;
2Symgradv
0
+
1divv
0
N = (w
1
)N
+
2
grad
i
N
1
2
4
2
+grad
(grad)
N,
2Symgradv
1
+
1divv
2
N = (w
3
)N,
2Symgradv
2
+
1divv
3
N = (w
4
)N+
2
( +) N,
X D;
(gradw
0
) N = v
1
N,
(gradw
1
) N = v
2
N
2
X N
,
(gradw
2
) N = v
3
N,
X D.
Some remarks should be considered in the application here of the principle of
Saint-Venant [25].
A complete solution of this problem is given in [18] where the global balance
equations are used to show that two of the constants vanish.
Here we note the usefulness of the Almansi-Michells solution to solve 2
nd
order St.-Venants problem.
56
Chapter 4
The Almansi-Michells problem
The dierence between the Almansi-Michell and the St.-Venant problems is that
in the former the mantle of the cylinder is subjected to surface tractions and in the
latter it is traction free. Here, we study a particular case of the Almansi-Michell
problem for which loads can be expressed as a polynomial in the axial co-ordinate
Z. That is,
T
N c (X) p(X) +a (X) e

m
X
j=0
Z
j
j!
h
p
j
+ea
j
X
i
,
m
X
j=0
Z
j
j!
c
j
, X D[0, L] .
(4.1)
As for the St.-Venant problem, we adopt the polynomial hypothesis with re-
spect to the axial variable Z for the displacement eld:
_
_
w(X) =
M
X
i=0
1
i!
w
i
Z
i
, X C
,
v (X) =
M
X
i=0
1
i!
v
i
Z
i
, X C
.
(4.2)
We note that m and M are arbitrary integers; however, generally M >> m.
In terms of the displacement eld (v, w), the boundary conditions become
2Symgradv
i
+
1divv
i
N = (w
i+1
)N+p
i
,
X D,
gradw
i
N = v
i+1
N+a
i
,
X D.
(4.3)
57
Recalling the denitions (3.20) of f (Z) and m(Z), we rst show that
f (Z) =
m+1
X
i=0
f
i
Z
i
i!
,
m(Z) =
m+2
X
i=0
m
i
Z
i
i!
,
(4.4)
where
f
0
= f
p
0
+f
a
0
e = q +ne,
f
p
0
= q =
Z
D(0)
tdS
,
f
a
0
= n =
Z
D(0)
dS
,
f
i
= f
p
i
+ef
a
i
=
Z
D
c
i1
dl
,
f
p
i
=
Z
D
p
i1
dl
,
f
a
i
=
Z
D
a
i1
dl
;
(4.5)
and
m
0
= m
p
0
+m
a
0
e = F
0
+Te,
m
p
0
=
Z
D(0)
dS
,
m
a
0
=
Z
D(0)
XtdS
e,
m
p
i
= i
f
p
i1
+
Z
D
a
i1
dl
,
m
a
i
=
Z
D
Xp
i1
dl
e,
m
p
m+2
= (m+ 1) f
p
m+1
,
m
a
m+2
= 0.
(4.6)
Here dl
is the element of arc length of the boundary of the cross-section in

the reference conguration.
58
Substitution from (4.1) into the global equilibrium of force (3.19)
1
yields
Z
D(0)
T
edS
+
Z
D(Z)
T
edS
+
Z
D[0,Z]
c (X) dS
= 0. (4.7)
Because of (3.9)
2
we have
Z
D(0)
T
edS
=
Z
D(0)
tdS
+e
Z
D(0)
dS
. (4.8)
Because of (3.20)
1
, the second integral on the left hand side of (4.7) equals
f (Z). On the mantle dS
= dl
d where is a parameter along the axial position

in the reference conguration. From (4.1) we obtain
Z
D[0,Z]
c (X) dS
=
m
X
j=0
Z
D[0,Z]
j
j!
c
j
dS
, (4.9)
=
m
X
j=0
Z
D
c
j
dl
Z
Z
0
j
j!
d,
=
m
X
j=0
Z
D
c
j
Z
j+1
(j + 1)!
dl
=
m+1
X
i=1
Z
i
i!
Z
D
c
i1
dl
.
Equations (4.7) and (3.20)
1
give
f (Z) =
Z
D(Z)
T
edS
=
Z
D(0)
T
edS
m+1
X
i=1
Z
i
i!
Z
D
c
i1
dl
. (4.10)
Thus, because of (3.9)
2
and (4.5)
2,3,4
, (4.10) is equivalent to (4.4)
1
.
Using (4.1) in the global equilibrium of momentum (3.19)
2
, we obtain
Z
D(0)
XT
edS
+
Z
D(Z)
X+e
edS
+
Z
D[0,Z]
X+e
cdS
= 0.
(4.11)
Because of (3.9)
2
,
Z
D(0)
XT
edS
=
Z
D(0)
dS
+
_
_
_
Z
D(0)
XtdS
_
_
_
e. (4.12)
59
Recalling (4.1) and (3.20)
2
, we get
Z
D(Z)
X+e
edS
=
Z
D(Z)
XT
edS
+Z
(
e
m
X
j=0
Z
j
j!
f
p
j
)
, (4.13)
=
Z
D(Z)
XT
edS
+
m+1
X
i=1
Z
i
(i 1)!
f
p
i1
,
= m(Z) +
(
m+1
X
i=1
Z
i
i!
i f
p
i1
)
,
and
Z
D[0,Z]
X+e
cdS
=
m
X
j=0
Z
D[0,Z]
j
j!
X+e
c
j
dS
, (4.14)
=
m
X
j=0
Z
D[0,Z]
j
j!
h
Xc
j
+
p
j
i
dS
,
=
m
X
j=0
Z
D
Xc
j
Z
j+1
(j + 1)!
+ (p
j
)
Z
j+2
j! (j + 2)
dl
,
=
m+1
X
i=1
Z
D
Xc
i1
Z
i
i!
dl
+
m+2
X
i=2
Z
D
(p
i2
)
Z
i
(i 2)!i
(i 1)
(i 1)
dl
,
=
m+1
X
i=2
Z
i
i!
Z
D
h
Xc
i1
+ (p
i2
) (i 1)
i
dl
+Z
Z
D
Xc
0
dl
+
Z
D
(p
m
)
Z
m+2
(m+ 2)!
(m+ 1) dl
.
Substitution from (4.12), (4.13) and (4.14) into (4.11) gives
0 =
Z
D(0)
dS
_
_
_
Z
D(0)
XtdS
_
_
_
e
+m(Z) +
m+1
X
i=1
Z
i
i!
i f
p
i1
+Z
Z
D
Xc
0
dl
60
+
m+1
X
i=2
Z
i
i!
Z
D
h
Xc
i1
dl
+ (p
i2
) (i 1)
i
dl
+
Z
D
(p
m
)
Z
m+2
(m+ 2)!
(m+ 1) dl
.
Thus
m(Z) =
Z
D(0)
dS
+
_
_
_
Z
D(0)
XtdS
_
_
_
e
(
m+1
X
i=1
Z
i
i!
i f
p
i1
)
Z
Z
D
Xc
0
dl
m+1
X
i=2
Z
i
i!
Z
D
h
Xc
i1
dl
+ (p
i2
) (i 1)
i
dl
Z
D
(p
m
)
Z
m+2
(m+ 2)!
(m+ 1) dl
,
=
Z
D(0)
dS
+
_
_
_
Z
D(0)
XtdS
_
_
_
e
Z
f
p
0
+
Z
D
Xc
0
dl
m+1
X
i=2
Z
i
i!
i f
p
i1
+
Z
D
h
Xc
i1
dl
+ (p
i2
) (i 1)
i
dl
Z
D
(p
m
)
Z
m+2
(m+ 2)!
(m+ 1) dl
.
Using (4.5)
5
and the denition (4.1) of c in terms of p and a we arrive at
m(Z) =
Z
D(0)
dS
+
_
_
_
Z
D(0)
XtdS
_
_
_
e
Z f
p
0
Z
Z
D
Xp
0
dl
+eZ
Z
D
a
0
dl
m+1
X
i=2
Z
i
i!
i f
p
i1
+
Z
D
h
Xp
i1
(a
i1
)
i
dl
f
p
i1
(i 1)
f
p
m+1
Z
m+2
(m+ 2)!
(m+ 1) ,
61
=
Z
D(0)
dS
+
_
_
_
Z
D(0)
XtdS
_
_
_
e
Z

f
p
0
+
Z
D
Xp
0
dl
e
Z
D
a
0
dl
m+1
X
i=2
Z
i
i!
f
p
i1
+
Z
D
h
Xp
i1
(a
i1
)
i
dl
f
p
m+1
Z
m+2
(m+ 2)!
(m+ 1) ,
=
Z
D(0)
dS
+
_
_
_
Z
D(0)
XtdS
_
_
_
e
m+1
X
i=1
Z
i
i!
f
p
i1
Z
D
h
Xp
i1
+e
(a
i1
)
i
dl
+
Z
m+2
(m+ 2)!
(m+ 1) f
p
m+1
.
Using (4.6)
2,3,4,5,6
, we arrive at (4.4)
2
.
We note that f
0
and m
0
equal the resultant force and the resultant moment
acting on the basis and f
1
, m
1
, f
2
, m
2
... are kinds of resultant forces and moments
acting on the mantle of the cylinder.
Substitution from the constitutive relations (3.13) into (3.20) gives
f (Z) =
Z
D(Z)
(v
0
+gradw) dS
+e
Z
D(Z)
[( + 2) w
0
+divv] dS
,
m(Z) =
Z
D(Z)
(( + 2) w
0
+divv)
dS
+e
Z
D(Z)
(v
0
+gradw) dS
.
(4.15)
We now substitute the series (3.27) and (4.4) into (4.15) and equate like powers
of Z on both sides to arrive at
Z
D(Z)
(v
i+1
+gradw
i
) dS
=
_
_
f
p
i
, i = 0, ..., m+ 1,
0, i > m+ 1;
62
Z
D(Z)
[( + 2) w
i+1
+divv
i
] dS
=
_
_
f
a
i
, i = 0, ..., m+ 1,
0, i > m+ 1;
Z
D(Z)
[( + 2) w
i+1
+divv
i
]
dS
=
_
_
m
p
i
, i = 0, ..., m+ 2,
0, i > m+ 2;
Z
D(Z)
(v
i+1
+gradw
i
)
dS
=
_
_
m
a
i
, i = 0, ..., m+ 1,
0, i > m+ 1.
(4.16)
Following the same procedure as that used to solve the St.-Venant problem,
we conclude the following X D.
w
i
= 0, i = m+ 4, ..., M,
w
m+3
= v
0
m+4

b

X
,
w
m+2
= v
0
m+3

b

X
,
v
i
= 0, i = m+ 5, ..., M;
v
i
= v
0
i
, i = m+ 3, m+ 4;
v
m+2
= v
0
m+2
+
0
m+2
v
0
m+4

X+
n
sym
h
X
io
v
0
m+4
,
v
m+1
= v
0
m+1
+
0
m+1

Xw
0
m+2
X+
n
sym
h
X
io
v
0
m+3
.
(4.17)
It is hard to evaluate the remaining terms; however, we can evaluate them for
a circular cylinder.
We consider the case of uniform tractions on the mantle of a cylinder; a
familiar case is that of pressure loading on its inner and outer surfaces.
63
Using formulae given in the part of the Appendix concerning cylindrical co-
ordinates, we conclude from (4.17) the following:
v
i
= 0, i = m+ 3, ..., M;
v
(m+2)r
= 0,
v
(m+2)
=
0
m+2
r,
v
(m+1)r
= w
0
m+2
r,
v
(m+1)
=
0
m+1
r,
w
i
= 0 i = m+ 3, ..., M;
w
m+2
= w
0
m+2
,
w
m+1
= w
0
m+1
.
(4.18)
In cylindrical coordinates equilibrium equations (3.28)
1,2
become
1
r
(rw
i,r
)
,r
+ ( +)
1
r
rv
(i+1)r
,r
+ ( + 2)w
i+2
= 0,
(2 +)
h
1
r
(rv
ir
)
,r
i
,r
+ ( +) w
(i+1),r
+v
(i+2)r
= 0,
h
1
r
(rv
i
)
,r
i
,r
+v
(i+2)
= 0, r [R
i
, R
e
] .
(4.19)
Boundary conditions (4.3) take the form
(2 +) v
ir,r
+
v
ir
r
= w
i+1
+p
e
ir
at r = R
e
,
(2 +) v
ir,r
+
v
ir
r
= w
i+1
p
i
ir
at r = R
i
,
h
r
1
r
v
i
,r
i
= p
e
i
at r = R
e
,
h
r
1
r
v
i
,r
i
= p
i
i
at r = R
i
,
w
i,r
= v
(i+1)r
+a
e
i
at r = R
e
,
w
i,r
= v
(i+1)r
a
i
i
at r = R
i
.
(4.20)
64
It is possible to extract from (4.19)
1
and (4.20)
5,6
the following equations for
the unknown eld w
m
.
1
r
(rw
m,r
)
,r
+ 2w
0
m+2
= 0, r [R
i
, R
e
] ,
w
m,r
= w
0
m+2
R
e
+a
e
m
, r = R
e
,
w
m,r
= w
0
m+2
R
i
a
i
m
, r = R
i
.
A simple integration gives
w
m
= w
0
m
R
e
R
i
(a
i
m
R
e
+a
e
m
R
i
)
[R
2
e
R
2
i
]
log(r)
1
2
w
0
m+2
r
2
,
and
w
0
m+2
=
(a
i
m
R
i
+a
e
m
R
e
)
[R
2
e
R
2
i
] (1 +)
.
2
and (4.20)
1,2
the unknown eld v
mr
.
h
1
r
(rv
mr
)
,r
i
,r
= 0 r [R
i
, R
e
] ,
(2 +) v
mr,r
+
v
mr
r
= w
m+1
+p
e
mr
, r = R
e
,
(2 +) v
mr,r
+
vmr
r
= w
m+1
p
i
mr
, r = R
i
.
v
mr
=
"
(R
i
R
e
)
2
(p
e
mr
+p
i
mr
)
2[R
2
e
R
2
i
]
#
1
r
+
p
e
mr
R
2
e
+p
i
mr
R
2
i
w
0
m+1
[R
2
e
R
2
i
]
2 [R
2
e
R
2
i
] ( +)
r.
3
and (4.20)
3,4
the unknown eld v
m
.
1
r
(rv
m
)
,r
,r
+
0
m+2
r = 0, r [R
i
, R
e
] ,
65
h
r
1
r
v
m
,r
i
= p
e
m
, r = R
e
,
h
r
1
r
v
m
,r
i
= p
i
m
, r = R
i
.
v
m
= v
m0
r +
p
i
m
R
4
e
R
2
i
+p
e
m
R
2
e
R
4
i
2[R
4
e
R
4
i
]
1
r

0
m+2
8
r
3
,
and
0
m+2
= 4
p
e
m
R
2
e
+p
i
m
R
2
i
[R
4
e
R
4
i
]
.
Summarizing the preceding results we have, for a constant load on the mantle
(m = 0), the following displacement eld:
v
i
= 0, i = 3, ..., M;
v
2r
= 0,
v
2
=
0
2
r,
v
1r
= w
0
2
r,
v
1
=
0
1
r,
v
0r
=

(R
i
R
e
)
2
(
p
e
0r
+p
i
0r
)
2
[
R
2
e
R
2
i
]
1
r
+
p
e
0r
R
2
e
+p
i
0r
R
2
i
w
0
1
[
R
2
e
R
2
i
]
2
[
R
2
e
R
2
i
]
(+)
r,
v
0
= v
0
0
r +
p
i
0
R
4
e
R
2
i
+p
e
0
R
2
e
R
4
i
2
[
R
4
e
R
4
i
]
1
r

0
2
8
r
3
,
w
i
= 0 i = 3, ..., M;
w
2
=
(
a
i
0
R
i
+a
e
0
R
e)
[
R
2
e
R
2
i
]
(1+)
,
w
1
= w
0
1
,
w
0
= w
0
0
R
e
R
i(
a
i
0
R
e
+a
e
0
R
i)
[
(R
e
)
2
(R
i
)
2
]
log(r)
1
2
w
0
2
r
2
,
66
where
w
0
2
=
(a
i
0
R
i
+a
e
0
R
e
)
[R
2
e
R
2
i
] (1 +)
,
0
2
= 4
p
e
0
R
2
e
+p
i
0
R
2
i
[R
4
e
R
4
i
]
.
It is possible to get a solution for other values of m since v
mr
, v
m
, w
m
are
governed by linear ordinary dierential equations.
If m = 1 then (4.18) and the calculation involved for the case m = 0 give
v
i
= 0, i = 4, ..., M;
v
3r
= 0,
v
3
=
0
3
r,
v
2r
= w
0
3
r,
v
2
=
0
2
r,
v
1r
=

(R
i
Re)
2
(p
e
1r
+p
i
1r
)
2[R
2
e
R
2
i
]
1
r
+
p
e
1r
R
2
e
+p
i
1r
R
2
i
w
0
2
[R
2
e
R
2
i
]
2[R
2
e
R
2
i
](+)
r,
v
1
= v
0
0
r +
p
i
1
R
4
e
R
2
i
+p
e
1
R
2
e
R
4
i
2
[
R
4
e
R
4
i
]
1
r

0
3
8
r
3
,
w
i
= 0 i = 4, ..., M;
w
3
=
(
a
i
1
R
i
+a
e
1
R
e)
[
R
2
e
R
2
i
]
(1+)
,
w
2
= w
0
2
,
w
1
= w
0
1
R
e
R
i(
a
i
1
R
e
+a
e
1
R
i)
[
(R
e
)
2
(R
i
)
2
]
log(r)
1
2
w
0
3
r
2
,
where
w
0
3
=
(a
i
1
R
i
+a
e
1
R
e
)
[R
2
e
R
2
i
] (1 +)
,
0
3
= 4
p
e
1
R
2
e
+p
i
1
R
2
i
[R
4
e
R
4
i
]
.
67
Thus,
(rv
1r
)
,r
= r
p
e
1r
R
2
e
+p
i
1r
R
2
i
w
0
2
[R
2
e
R
2
i
]
( +) [R
2
e
R
2
i
]
.
Dierential equation for w
0
is
1
r
(rw
0,r
)
,r
+ ( +)
1
r
(rv
1r
)
,r
+ ( + 2)w
2
= 0,
that is
1
r
(rw
0,r
)
,r
+
p
e
1r
R
2
e
+p
i
1r
R
2
i
[R
2
e
R
2
i
]
+ 2w
0
2
= 0.
Boundary conditions for w
0
are
w
0,r
= v
1r
+a
e
0
, at r = R
e
,
w
0,r
= v
1r
a
i
0
, at r = R
i
.
w
0
= w
0
0
1
2
w
0
2
r
2
1
4
p
e
1r
R
2
e
+p
i
1r
R
2
i
[R
2
e
R
2
i
]
r
2
R
e
R
i
2 (a
i
0
R
e
+a
e
0
R
i
) + (p
e
1r
+p
i
1r
) R
e
R
i
2[R
2
e
R
2
i
]
log(r),
where
w
0
2
=
(p
e
1r
R
2
e
+p
i
1r
R
2
i
) + 2 ( +) (a
i
0
R
i
+a
e
0
R
e
)
(3 + 2) [R
2
e
R
2
i
]
.
It is possible to give analytical solutions even for v
0r
, v
0
. However expressions
are too large to include in the thesis.
A Mathematica le is attached at the end of the chapter to nd the complete
solution for the case of m = 1. In this le computations of this section are done
by setting m = 1 at the beginning and dening everything in order to make
Mathematica do the calculations.
We now give a simple example.
68
An example
We choose the following parameters in the ISU.
R
e
= 0.5, R
i
= 0.2;
= 200 10
9
, = 77.2 10
9
;
0
1
= 10
3
, w
0
1
= 10
1
;
p
e
1r
= 10
5
, p
e
1
= 10
5
, a
e
1
= 10
5
;
p
i
1r
= 10
5
, p
i
1
= 10
5
, a
i
1
= 10
5
;
p
e
0r
= 10
5
, p
e
0
= 10
5
, a
e
0
= 10
5
;
p
i
0r
= 10
5
, p
i
0
= 10
5
, a
i
0
= 10
5
.
Thus,
v
i
= 0, i = 4, ..., M;
v
3r
= 0,
v
3
= 1.9 10
6
r,
v
2r
= 1.1 10
6
r,
v
2
= 1.9 10
6
r,
v
1r
= [6.2 10
8
]
1
r
+ [1.6 10
6
] r,
v
1
= [3.1 10
8
]
1
r
+ [10
3
] r [3.1 10
6
] r
3
,
v
0r
= [6.7 10
8
]
1
r
[3.6 10
2
] r [3.4 10
7
] r
3
+ [1.7 10
7
] r log(r),
v
0
= [3.1 10
8
]
1
r
+ [3.1 10
6
] r
3
,
69
w
i
= 0 i = 4, ..., M;
w
3
= 3.2 10
6
,
w
2
= 3.7 10
6
,
w
1
= 10
1
+ [1.6 10
6
] r
2
4.3 10
7
log(r),
w
0
= 1.4 10
6
r
2
4.9 10
7
log(r).
70
71
72
73
74
75
76
77
Chapter 5
Conclusions
We have used a semi-inverse method and the Signorini method to analyze three di-
mensional deformations of a prismatic body made of a homogeneous and isotropic
elastic material. The solution of the problem is reduced to that of solving a set
of two-dimensional problems over the cross-section of the prismatic body. For a
stress free reference conguration of the body, an explicit solution of these two
dimensional problems, except for the 0th order problem, is given. For a simple
cross-section, the 0th order problem may be solved analytically; otherwise its ap-
proximate solution can be found numerically. An explicit solution is given for a
generalized axisymmetric problem in which all three components of displacement
are independent of the angular position. Equations governing three dimensional
deformations of a homogeneous prismatic body made of an isotropic material and
subjected to a hydrostatic pressure in the reference conguration are also derived.
By assuming that the three components of displacement can be expressed as a
polynomial in the axial coordinate, the solution of the three dimensional problem
is reduced to that of solving a set of two dimensional problems dened over the
cross-section of the body. These equations are dicult to solve analytically.
78
By assuming that surface tractions on the mantle of a cylindrical prismatic
body and the three components of displacement can be expressed as a polynomial
in the axial coordinate, three dimensional elastostatics equations are reduced to
a set of two-dimensional elastostatics equations. These two dimensional problems
are solved analytically for generalized axisymmetric deformations of a circular
cylindrical body. Numerical results are given for a hollow cylindrical body with a
uniform pressure applied on its mantle and also when the pressure is a polynomial
of degree one in the axial coordinate.
79
Chapter 6
Appendix
1
Decomposition of a 2
nd
order tensor
Let A and B be two vector spaces. We dene the vector space A B as the
space of the linear combination a b such that a A , b B and
(a b) c = (b c) a c B;
Let us recall the denition given in (3.3)
U = W V,
and the linear properties of the tensor product. We have the following decompo-
sition
U U = (WW) (WV) (V W) (V V) (6.1)
of the vector space UU that is the space of the linear operators LIN of U in U:
LIN = U U.
1
Most of the results of the rst three paragraphs can be found in [1], [20], [32] and [36].
80
LIN can be seen as the space of second order tensors. Every second order ten-
sor can be decomposed into the sum of its symmetric part and its skew symmetric
part. Thus
LIN SY M SKW, (6.2)
where SY M is the space of symmetric tensors and SKW is the space of skew
symmetric tensors. That is SY M is a linear combination of
Sym(u u
0
) u U, u
0
U,
and SKW a linear combination of
Skw(u u
0
) u U, u
0
U.
These spaces can also be written as
SY M = U U,
SKW = U U.
For the subspace V V of U U, we have
lin V V = sym skw,
such that
sym = V V,
skw = V V.
The space W V is the space of all symmetric tensors
Sym(w v) w W, v V.
81
Finally we have the following decomposition for a 2
nd
order tensor:
SY M = (WW) (W V) sym,
SKW = (V W) skw.
The star of Hodge
It is possible to give the denition of the star of Hodge in an arbitrary vector
space; however we will need it only in the two-dimensional subspace V: for the
sake of simplicity we will give its denition only in this space.
The cross product of two vectors (u, v) belonging to the three dimensional
space U is another vector belonging to the same space. Let us suppose that the
two vectors belong to V; then, their cross product is an element of one-dimensional
vector space W and can be characterized as a real number. Let us give the
following denition:
2u v = , u, v V, R. (6.3)
Besides we can consider the skew symmetric tensor u v belonging to the
subspace skw. We note that it is possible to have a bijective mapping between
skw and the space R of the real numbers. So it is possible to give the following
denition for the star of Hodge in V:
(u v) , u, v V, R. (6.4)
If we are in an orthonormal system we can use the index notation, then the
82
star of Hodge has the following matrix representation:
[] =
_
_
_
_
0 1
1 0
_
_
_
_
. (6.5)
In index notation
[]
ij
=
ij
,
where
ij
is the Levi Civita skew symmetric tensor in two dimensions. In fact we
have
(u v)
ij
= (u
i
v
j
u
j
v
i
) , u, v V. (6.6)
The Levi Civita tensor is useful to represent the cross product:
u v =
ij
u
i
v
j
, u, v V. (6.7)
The matrix representation of the scalar obtained by the application of the star
of Hodge (6.5) to a generic skew symmetric tensor (for example (6.6)) is the scalar
product of the two tensors:
[ (u v)] = []
ij
(u
i
v
j
u
j
v
i
) , u, v V. (6.8)
If we add a symmetric tensor in the second factor of (6.8) we do not change
the result. Because of (6.7)
[ (u v)] = []
ij
[(u
i
v
j
u
j
v
i
) + (u
i
v
j
+u
j
v
i
)] ,
= 2 []
ij
u
i
v
j
= 2
ij
u
i
v
j
= 2u v, u, v V, (6.9)
we have veried the compatibility of (6.3) and (6.4).
83
It is possible to apply the star of Hodge even to a scalar and we have:
_
_
= 2, R,
(u v) = 2 (u v) , u, v V.
Thus
(u v) = 2 (u v) = 2 (u v) . (6.10)
It is possible to demonstrate these in an orthonormal system via the index
notation and with the formula
[]
ij
[]
ij
=
_
_
_
_
0 1
1 0
_
_
_
_
:
_
_
_
_
0 1
1 0
_
_
_
_
= 1 + 1 = 2.
We note that
[]
ij
[]
jh
=
_
_
_
_
0 1
1 0
_
_
_
_
_
_
_
_
0 1
1 0
_
_
_
_
=
_
_
_
_
1 0
0 1
_
_
_
_
=
ih
. (6.11)
We can apply the star of Hodge to a vector and it is possible to demonstrate
that the vector is rotated in the anticlockwise direction. In an orthonormal system,
v =
_
_
_
_
0 1
1 0
_
_
_
_
_
_
_
_
v
1
v
2
_
_
_
_
=
_
_
_
_
v
2
v
1
_
_
_
_
, v V.
Thus
v = v, v V.
An useful denition is the following
v (v) = v, v V, R. (6.12)
84
The useful formula
u v = (u) v, u, v V, (6.13)
is easy to prove in an orthonormal system.
u v =
ij
u
i
v
j
=
ji
u
i
v
j
= (u)
j
v
j
= (u) v, u, v V.
Furthermore,
() v = (v) , v V, R, (6.14)
in an orthonormal system is trivial. Besides we have
u (u) = () u, u V, R, (6.15)
and because of (6.13), (6.9) and (6.10),
[(u) v] = (u v) =
1
2
[ (u v)] = (u v) . (6.16)
Dierential operator
We rst dene a dierential operator in an absolute way, and later we special-
ize the denition to Cartesian and cylindrical coordinates.
Let be a eld that can be either a scalar belonging to R, or a vector
belonging to U, or a tensor belonging to U U. Let Z be the space of . The
gradient of evaluated at a point X =

X, Z
of the cylinder C
is an operator
Grad Z U = Z W Z V,
85
that can be decomposed as follows:
Grad =
0
e +grad,
where
0

Z
,
and
grad Z V
is the gradient on the cross-section of the cylinder. The gradient operator in the
cross-section is dened as
grad () = ()
grad = ()
X
1
e
1
+
X
2
e
2
.
The last expression means that we apply the derivative on the term appearing
on the left side of the last parenthesis. This denition is correct even if we use
non-Cartesian coordinates; it is illustrated by means of an example at the end of
this section.
For = , a scalar,
Grad =
0
e +grad,
where in Cartesian coordinates
grad () =

X
1
e
1
+

X
2
e
2
.
Note that Grad is the 3-dimensional gradient operator but grad is the gradient
operator in a plane containing the cross-section of the cylinder.
86
If = u = v +ew is a vector, then
Gradu U U = U W U V = (WW) (V W) U V,
and
Gradu = u
0
e +gradu =w
0
e e +v
0
e +gradu,
where
gradu U V = WV V V,
or
gradu = e (gradw) +gradv,
and
w
0
=
w
Z
.
Thus
Gradu = w
0
e e +v
0
e +e (gradw) +gradv. (6.17)
If = L = e e +u e +e v +

L is a tensor eld, then
GradL U U U,
= U U W U U V,
= U WW U V W U WV U V V,
= WWW V WW WV W V V W
WWV V WV WV V V V V,
87
and
GradL = L
0
e +gradL =
0
e e e +u
0
e e +e v
0
e +

L
0
e
+grad (e e) +grad (u e +e v) +grad
L,
=
0
e e e +u
0
e e +e v
0
e +

L e
+e e
grad +u e
grad +e v
grad +grad
L.
The trace of this eld is given by
2
tr [GradL] =
0
e +u
0
+tr [gradv] e +tr
grad
.
Another important operator is the divergence operator; it can be derived from
the gradient operator because
Div = tr (Grad) ,
for any eld . If = u = v +ew is a vector, then
Divu = w
0
+divv.
If = L = e e +u e +e v +

L is a tensor, then
DivL = (
0
+divv) e +u
0
+div
L. (6.18)
Another useful operator is the rot. It is possible to dene this operator in a
general way, however, for the sake of simplicity we analyze it only for a cylindrical
2
The trace of a third order tensor of this kind is, roughly speaking, a vector where the
last tensor product of the tensor is substituted by a scalar product. In rectangular Cartesian
coordinates:
[tr(L)]
i
= L
ihh
.
88
body. The rot operator on the cross-section is dened as
rot = skwgrad.
If = v is a vector, then
rotv = skwgradv.
In Cartesian coordinates, we have
rotv = []
ij
1
2
(v
i,j
v
j,i
) =
1
2
ij
(
j
v
i
+
i
v
j
) =
ij
i
v
j
=
1
v
2
2
v
1
.
If =

L is a tensor,
3
then
rot
L = skwgrad
L.
In Cartesian coordinates
rot
i
= []
jh
1
2
L
ij,h

L
ih,j
=
jh
1
2

L
ij,h
+
jh
1
2

L
ih,j
, (6.19)
=
1
2
jh
L
ij
+
j
L
ih
=
jh
L
ih
=
1
L
i2
L
i1
.
Thus, in direct notation
rot
b = rot
L
T
b
, b V.
3
We have not dened the star of Hodge of a third order tensor. However in Cartesian
coordinates
(L)
i
= []
jh
L
ijh
.
Furthermore
(skwL)
ijh
=
1
2
(L
ijh
L
ihj
) .
89
Proposition 3 The following useful formulae hold:
rot
= (grad) , R;
rot () = [grad ()] , R.
(6.20)
Proof. Following the denition of the operator rot, we have
rot
b = rot (b) = skwgrad (b) ,

= skw[b grad ()] = [b grad ()] ,
= b grad () = grad () b = (grad) b, b V, R,
and
[rot ()] b = rot (() b) = skwgrad (() b) ,
= skwgrad ((b)) = skw
h
(b)
grad
i
,
= skw[(b) grad ()] = [(b) grad ()] ,
= (b) grad () = b [grad ()] , b V, R.
Results (6.20) can be proved in an easier way using the Cartesian coordinates.
For example,
rot
i
=
jh
j
(
ih
) =
ji
j
= []
ij

j
= (grad)
i
,
[rot ()]
i
=
jh
j
([]
ih
) =
jh
hi
j
=
ji
j
=
i
= [grad]
i
.
We now review some more useful formulae. R and u U,
_
_
Grad (u) = Gradu +uGrad,
Div (gradu)
T
= Grad (Divu) .
(6.21)
90
These can be proved by noting that
Grad (u) =

u
Grad
= uGrad +
Grad
= uGrad +Gradu,
Div (Gradu)
T
= Div (Gradu) =tr
h
Gradu
Grad
i
,
= Grad [tr (Gradu)] = Grad [Divu] .
R, a, b, c U, v V, L sym, W skw , we have
Div
= Grad, rotgradv = 0, tr (rotL) = 0, (6.22)

a (b c) = b(a c) c (a b) .
In Cartesian coordinates, the validity of formulae (6.22) can be easily established.
Div
i
=
j
(
ij
) =
i
= [Grad]
i
,
rotgradv =
ij
i
(gradv)
hj
=
ij
ij
v
h
= 0,
[tr (rotL)]
ii
=
ihk
h
L
ik
,
hij
hlm
=
il
jm
im
jl
,
[a (b c)]
i
=
ijh
a
j
(b c)
h
=
ijh
a
j
hlm
b
l
c
m
= a
j
b
l
c
m
(
hij
hlm
) ,
= a
j
b
l
c
m
(
il
jm
im
jl
) = a
m
b
i
c
m
a
l
b
l
c
i
= [b(a c) c (a b)]
i
.
All of the formulae illustrated above for vectors in a three dimensional space
U are also valid in a two dimensional space V. The following two relations are
also quite useful.
div
=
i
(
ij
X
j
) =
ij
ij
= 0, (6.23)
rot
=
ij
i
(
jh
X
h
) = (
ih
)
ih
= 2. (6.24)
Because of (6.11) the Laplacian operator,
R
, is dened as
R
u Div (Gradu) , u U. (6.25)
91
An Integral formula
Let the path from the origin O to a general point

X of the cross-section be
denoted by

O,

X
=

. The integral I of a function f

on the path

can
be written as
I =
Z
ds, (6.26)
where
l
=

R
X

X
=
1
X

X,
R =

R
X

X,
X =

R =

.
The integration in (6.26) can also be performed from 0 to

X. That is,
I =
Z

X
0
f
X

X
!

X
1
X
d
R. (6.27)
Solution of four boundary value problems
Problem A
The following Neumann problem
_
_
4
R
w = 0, X D,
gradw N =
h
v
0
+
X
i
N, X D,
(6.28)
92
has a unique solution to within an additive constant. A solution for the gradient
of w is
gradw =
h
v
0
+
X
i
+ 2f, (6.29)
where
divf = 0, rotf = 1,
X D,
f N =0,
X D.
(6.30)
Here N is an outward unit normal to D, and v
0
and are constants.
Because of (6.23)
1
,
div
n
h
v
0
+
X
i
+ 2f
o
= div
+ 2divf = 0 + 0 = 0.
Using (6.24), we get X D
rot (gradw) = rot
n
h
v
0
+
X
i
+ 2f
o
=
n
rot
+ 2
o
= 0.
Thus (6.29) satises (6.28) provided that (6.30) holds.
Problem B
The following boundary value problem
_
_
div
2Symgradv+
divv v
0
= 0, X D,
2Symgradv+
divv v
0
N = w
0
N, X D,
(6.31)
has a unique solution to within a rigid motion. Here , , v
0
and w
0
are constants
and

1 is the two dimensional identity tensor. A solution of (6.31) satises
2Symgradv+
divv v
0
= w
0
1. (6.32)
93
Taking the trace of both sides of (6.32), we obtain
divv =

w
0
v
0
( +)
,
where we have implicitly assumed that + 6= 0.
From (6.32) we can evaluate the gradient of the vector eld v:
2Symgradv = w
0
1
_
_
w
0
v
0
( +)
v
0

X
_
_
,
=
w
0
+ v
0
1 +
1
_
_
w
0
v
0
( +)
_
_
,
=
w
0
v
0
1

+
w
0
v
0
1.
Recalling the denition

2 ( +)
of the Poisson ratio and dening
w w
0
v
0

X, (6.33)
we get
Symgradv = w
1.
From the theorem of completion we know that there exists a unique skew-
symmetric part of the gradient of v and it is denoted by the Hodge star of a
scalar function
. Thus
gradv = w
1. (6.34)
94
The compatibility condition
0 = rot (gradv) = rot
,
gives a form for the function . From (6.20) we have
0 = grad () (gradw) ,
and now using (6.33) we arrive at
grad () =
v
0
.
Hence
=
0
+
v
0

X. (6.35)
The application of the Hodge operator to both sides of (6.35) gives
=
0
+
h
v
0

X
i
=
0
v
0
, (6.36)
where we have used (6.16). Substitution from (6.36) into (6.34) gives
gradv =
0
v
0
1. (6.37)
Substituting in (6.37) for w from (6.33), and using (6.27) to integrate (6.37),
we get
v
= v
0
+
Z
|
X|
0
(
v
0
X

"
w
0
v
0

X

R
X
#
1
)

X
1
X
d
R,
= v
0
+
v
0
1
2
+
v
0

X
1
2
w
0

X,
= v
0
+

X+
1
2
v
0

X+
v
0

X
i
w
0
X.
95
Because of (6.16)
v
0

X+
v
0

X =
h
v
0

X
i
+
h
v
0
X
i

X,
=
h

X
i
v
0
+
h
X
i
v
0
= 2
n
sym
h

X
io
v
0
.
Thus
v
= v
0
+

X+
n
sym
h
X
io
v
0
w
0
X,
X D. (6.38)
Lemma 4 It is important to note that if
v
0
= 0, w
0
= 0,
then the solution of the boundary value problem (6.31) is
v
= v
0
+

X,
or equivalently
v
= v
0
+
,
which is a rigid body motion.
Problem C
For
Z
D(Z)
(v+gradw) dS
= 0, (6.39)
Z
D(Z)
(v+gradw) dS
= 0, (6.40)
96
where X D,
v = v
0
+
,
gradw =
h
v
0
+
X
i
+ 2f,
divf = 0, rotf = 1,
and X D,
f N =0,
nd .
Solution. Substitution for v and gradw into (6.39) and (6.40) gives
Z
D(Z)
v
0
+
h
v
0
+
X
i
+ 2f
dS
= 2
Z
D(Z)
fdS
= 0,
2
Z
D(Z)
fdS
= 0.
Since conditions on f do not guarantee that
Z
D(Z)
fdS
always vanishes,
therefore we conclude that
= 0.
Problem D
Find and v
0
such that
97
_
_
Z
D(Z)
[( + 2) w +divv] dS
= 0,
Z
D(Z)
[( + 2) w +divv]
dS
= 0,
divv =
w
( +)
,
w =

w
0
v
0
.
(6.41)
Solution. We substitute in the rst integral the given expressions for w and
divv and obtain
Z
D(Z)
( + 2)
w
0
v
0
+

( +)
w
0
v
0
dS
= A
D
+ 2

2
( +)
w
0
v
0
= 0,
(6.42)
where A
D
is the area of cross-section. For the integral (6.41)
2
we have
_
_
Z
D(Z)
( + 2)
w
0
v
0
+

( +)
w
0
v
0
dS
=

+ 2

2
( +)
A
D
w
0
Z
D(Z)
v
0
dS
= 0.
(6.43)
Thus
w
0
= v
0
b. (6.44)
We note that (6.44) holds even if (6.41)
2
does not (problem D
). Substituting
(6.44) into (6.43) we get
n
A
D
h
b
io
v
0
=
_
_
Z
D(Z)
h

X
i
dS
_
v
0
.
Since the matrices in braces are not equal to each other, therefore
v
0
= 0, w
0
= 0.
98
Cylindrical coordinates
With the origin of the cylindrical and the Cartesian coordinates coincident
with each other, the in-plane cylindrical (r, ) and Cartesian coordinates (X
1
, X
2
)
of the same point are related to each other by
X
1
= r cos ,
X
2
= r sin,
or
r =
q
(X
1
)
2
+ (X
2
)
2
,
= arctan
X
2
X
1
.
Thus
r
X
1
= cos ,
r
X
2
= sin;
X
1
=
X
2
(X
1
)
2
1+
X
2
X
1
2
=
X
2
r
2
=
sin
r
,
X
2
=
1
X
1
1+
X
2
X
1
2
=
X
1
r
2
=
cos
r
.
The in-plane position vector of a point can be written as
X = X
1
e
1
+X
2
e
2
= rr,
where
r =cos e
1
+ sin e
2
,
and the unit vector normal to it
= sine
1
+ cos e
2
,
99
when combined with e
3
make a right handed orthonormal set. Conversely:
e
1
=cos r sin
,
e
2
=sin r+cos
.
Whereas a constant vector v
0
in a Cartesian co-ordinate system has constant
components, its components in a cylindrical co-ordinate system are not constants.
For example,
v
0
= v
0
x
e
1
+v
0
y
e
2
= v
0
x
cos r sin
+v
0
y
sinr+cos
,
=

v
0
x
cos +v
0
y
sin
r +
v
0
x
sin +v
0
y
cos
,
= v
0
r
r +v
0
.
Thus
v
0
r
= v
0
x
cos +v
0
y
sin ,
v
0
= v
0
x
sin +v
0
y
cos ,
and values of v
0
r
and v
0
vary with the angle . We note that r and

depend upon
and not upon r. Furthermore,
r
r
= 0,

r
= 0,
r
=

,

= r;
(r r)
=

r +r
r r;
= r r +
= r
r.
The nabla operator, 5, is dened as
() 5 = ()
X
1
e
1
+
X
2
e
2
, (6.45)
100
where

x
means the derivative with respect to x of the quantity preceding it; in
the same way
x
means the derivative with respect to x of the quantity following
it. In polar coordinates we have
X
1
=
r
X
1
r
+

X
1
= (cos )
r
sin
r

X
2
=
r
X
2
r
+

X
2
= (sin)
r
+
cos
r

,
and
5 = e
1
X
1
+e
2
X
2
=
h
cos r sin
(cos )
r
sin
r

+
h
sinr+cos
(sin)
r
+
cos
r

,
= r
r
+
1
r
.
From the denition (6.45) we have the following forms for the grad and the
div operators
_
_
grad 5 =
r
r +

1
r
,
div tr [5] ,
where is a general scalar, vector or tensor eld. More specically if = w is a
scalar, then
gradw = w
,r
r +
1
r
w
,
,
and the divergence of a scalar is not dened.
101
For = v, a vector,
gradv =

v
r
r +v
r
r +

1
r
= v
r,r
(r r) +v
,r
+
1
r
v
r,
+
1
r
v
r
+
1
r
v
,
1
r
v
,
= v
r,r
(r r)+
1
r
v
r,
1
r
v
+v
,r
1
r
v
r
+
1
r
v
,
,
where
v
r,r
=
v
r
r
, etc.
In matrix form
gradv =
_
_
_
_
v
r,r
1
r
(v
r,
v
)
v
,r
1
r
(v
r
+v
,
)
_
_
_
_
,
and
divv = tr (gradv) = v
r,r
+
1
r
v
r
+
1
r
v
,
=
1
r
(rv
r
)
,r
+
1
r
v
,
.
When = T is a tensor eld, we usually do not need its gradient; however
we do need its divergence.
divT =
h
T
rr
(r r) +T
r
+T
r
+T
r
r +

1
r
,
=
h
T
rr,r
r +T
r,r
i
+
1
r
h
T
r,
r +T
,
i
+
1
r
T
rr
r +
1
r
T
r
+
1
r
T
r

1
r
T
r,
=r
T
rr,r
+
1
r
(T
r,
+T
rr
T
T
r,r
+
1
r
(T
,
+T
r
+T
r
)
.
In matrix form
divT = div
_
_
_
_
T
rr
T
r
T
r
T
_
_
_
_
=
_
_
_
_
T
rr,r
+
1
r
(T
r,
+T
rr
T
)
T
r,r
+
1
r
(T
,
+T
r
+T
r
)
_
_
_
_
.
102
The matrix representation for the star of Hodge operator is
=
_
_
_
_
0 1
1 0
_
_
_
_
.
We can develop the following representations:
X =
_
_
_
_
r
0
_
_
_
_
,
X =
_
_
_
_
0 1
1 0
_
_
_
_
_
_
_
_
r
0
_
_
_
_
=
_
_
_
_
0
r
_
_
_
_
,
=
_
_
_
_
0
0
_
_
_
_
,
()

X =
_
_
_
_
0
0
_
_
_
_
_
_
_
_
r
0
_
_
_
_
=
_
_
_
_
0
r
_
_
_
_
,
v =
_
_
_
_
0 1
1 0
_
_
_
_
_
_
_
_
v
r
v
_
_
_
_
=
_
_
_
_
v
v
r
_
_
_
_
,
=
_
_
_
_
r
0
_
_
_
_
0 r
=
_
_
_
_
0 r
2
0 0
_
_
_
_
,
Sym
h
X
i
=
_
_
_
_
0
1
2
r
2
1
2
r
2
0
_
_
_
_
,
n
Sym
h
X
io
(v) =
_
_
_
_
0
1
2
r
2
1
2
r
2
0
_
_
_
_
_
_
_
_
v
v
r
_
_
_
_
=
r
2
2
_
_
_
_
v
r
v
_
_
_
_
.
103
We consider the class of problems for which the displacement eld is a general
function of the radial co-ordinate but a polynomial in the axial co-ordinate Z.
The tangential or the circumferential component of the displacement need not
vanish.
We assume that the shape of the cross-section, loads and components of the
displacement eld satisfy following conditions.
1. The external boundary D
0
is a circle of radius R
e
and the cross- section
has at most one circular hole ( = 1) of radius R
i
. That is
D
n
X = rr : r [R
i
, R
e
] , [0, 2]
o
.
2. The end faces of the cylindrical body are subjected to only extensional and
torsional loads: bending and exure loads are zero. Thus
f
p
0
= 0, m
p
0
= 0.
The loads on the mantle (4.1) are
p
i
=p
ir
(r)r +p
i
(r)
,
a
i
= a
i
(r) .
Hence
p
ir
(r) =
_
_
p
e
ir
r = R
e
,
p
i
ir
r = R
i
;
104
p
i
(r) =
_
_
p
ext
i
r = R
e
,
p
int
i
r = R
i
;
a
i
(r) =
_
_
a
ext
i
r = R
e
,
a
int
i
r = R
i
.
3. The displacement eld has the form
v
i
=v
ir
(r)r +v
i
(r)
,
w
i
= w
i
(r) .
In cylindrical coordinates, and recalling that displacements and stresses are
independent of the angular position , dierential operators introduced above have
the following simple forms.
gradw = w
,r
r,
gradv = v
r,r
(r r)
1
r
v
+v
,r
+
1
r
v
r
,
divv =
1
r
(rv
r
)
,r
,
divT =r
h
1
r
(rT
rr
)
,r
1
r
T
i
+
T
r,r
+
1
r
T
r
+
1
r
T
r
.
In matrix form the gradient of a vector is
gradv =
_
_
_
_
v
r,r

1
r
v
v
,r
1
r
v
r
_
_
_
_
,
and its symmetric part can be written as
Symgradv =
_
_
_
_
v
r,r
1
2
r
1
r
v
,r
1
2
r
1
r
v
,r
1
r
v
r
_
_
_
_
.
105
Equivalently,
Symgradv = v
r,r
(r r) +
1
2
h
r
r
i
"
r
1
r
v
,r
#
+
1
r
v
r
,
and
[Symgradv] r =rv
r,r
+
1
2
h
r
1
r
v
,r
i
.
Furthermore
div [Symgradv] = r
1
r
(rv
r,r
)
,r
1
r
1
r
v
r
_
_
1
2
"
r
1
r
v
,r
#
,r
+
2
r
1
2
r
1
r
v
,r
_
_
,
= r
1
r
(rv
r
)
,r
,r
+
1
2
1
r
(rv
)
,r
,r
,
and
grad [divv] =

1
r
(rv
r
)
,r
,r
r = {div [Symgradv]}
r
,
4
R
w = div [gradw] =
1
r
(rw
,r
)
,r
.
The system of dierential equations
4
R
w + ( +) divv + ( + 2) w = 0,
div
2Symgradv+
1divv
+( +) gradw + v = 0,
X D,
becomes in cylindrical coordinates
1
r
(rw
,r
)
,r
+ ( +)
1
r
(rv
r
)
,r
+ ( + 2) w = 0,
(2 +)
h
1
r
(rv
r
)
,r
i
,r
+ ( +) w
,r
+ v
r
= 0,
h
1
r
(rv
)
,r
i
,r
+ v
= 0, r [R
i
, R
e
] ,
and the boundary conditions
2Symgradv+
1divv
N = wN+p,
106
gradw N = v N+a,
become
(2 +) v
r,r
+
v
r
r
= w +p
r
,
h
r
1
r
v
,r
i
= p
,
w
,r
= v
r
+a, at r = R
i
and r = R
e
.
107
Bibliography
[1] J. Abram. Tensor Calculus through Dierential Geometry. London Butter
Worths, London, 1965.
[2] E. Almansi, "Sopra la deformazione dei cilindri sollecitati lateralmente". Atti
Real Accad. naz. Lincei Rend, Cl sci s, mat e natur. Ser. 5 10 (I), 333-338,
1901a.
[3] E. Almansi, "Sopra la deformazione dei cilindri sollecitati lateralmente". Atti
Real Accad. naz. Lincei Rend, Cl sci s, mat e natur. Ser. 5 10 (II), 400-408,
1901b.
[4] R. C. Batra. "Non Classical Boundary Conditions". Arch. Rational Mech.
Anal. 48(3): 163-& 1972.
[5] R. C. Batra. "Saint-Venants principle for a helical spring". J Appl Mech.
45(2): 297-301, 1978.
[6] R. C. Batra. "Saint-Venants principle in linear elasticity with microstruc-
ture". J Elasticity 13(2): 165-173, 1983.
[7] R. C. Batra, J. S. Yang. "Saint-Venants principle in linear piezoelectricity".
J Elasticity 38(2): 209-218, FEB 1995.
108
[8] R. C. Batra, J. S. Yang. "Saint-Venants principle for linar elastic porous
materials". J Elasticity 39(3): 265-271, JUN 1995.
[9] R. C. Batra, "Saint-Venants principle for helical piezoelectric body". J Elas-
ticity 43(1): 69-79, APR 1996.
[10] C. I. Bors, "Almansi-Michell problem for orthotropic beams". Atti Real Ac-
cad. naz. Lincei 48(6): 602, Roma, 1970.
[11] C. I. Bors, "Almansi-Michell problem for anisotropic beams". Atti Real Ac-
cad. naz. Lincei 50(3): 299, Roma, 1971.
[12] C. I. Bors, "Almansi-Michell problem for an elastic orthotropic cylinder".
Atti Real Accad. naz. Lincei 54(3): 441-446, Roma, 1973.
[13] B. H. Bransden, C. J. Joachain. Physics of atoms and molecules. Longman
group limited, 1983.
[14] A. Clebsch. Thorie de llasticit des corps solides (Traduite par MM:
Barr de Saint-Venant et Flamant, avec des Notes tundes de M. Barr de
Saint-Venant). Dunod, Paris 1883 [Reprinted by Johnson Reprint Corpora-
tion, New York 1996].
[15] S.C. Cowin, J. W. Nunziato. Linear Elastic Materials with voids. Journal
of Elasticity 13: 125-147, 1983.
[16] F. dellIsola, R. C. Batra. Saint-Venants Problem for Porous Linear Elastic
Materials. Journal of Elasticity 47: 73-81, 1997.
[17] F. dellIsola, L. Rosa. Saint-Venants Problem in linear piezoelectricity.
Mathematics and Control in Smart Structures, V. Varadhan, ed., SPIE, Vol.
2715, 399-409, 1996.
109
[18] F. dellIsola, G. C. Ruta, R. C. Batra. A Second-Order Solution of Saint-
Venants Problem for an Elastic Pretwisted Bar Using Signorinis Perturba-
tion Method. Journal of Elasticity 49: 113-127, 1998.
[19] F. dellIsola, G. C. Ruta, R. C. Batra. Generalized Poynting Eect in Pre-
deformed Prismatic Bars. Journal of Elasticity 50: 181-196, 1998.
[20] A. Di Carlo. Il problema di Saint-Venant. Manoscritto relativo ad un ciclo
di seminari per il corso di Dottorato di Ricerca in Ingegneria delle Strutture,
Roma 1993.
[21] A. E. Green, J. E. Adkins. Large Elastic Deformations and Nonlinear Con-
tinuum Mechanics. Claderon Press, Oxford (1960).
[22] M. Gurtin, An Introduction to Continuum Mechanics. Academic Press,
Inc. New York 1970.
[23] M. E. Gurtin. The linear theory of elasticity. Handbuch der Physik VIa/2
(ed. C. A. Truesdell) Springer-Verlag, New York, 1972.
[24] C. O. Horgan, J. K. Knowles. "Recent developments concerning Saint-
Venants principle". Adv Appl Mech 23: 179-269, 1983.
[25] C. O. Horgan, J. K. Knowles. "The eect of non linearity on a principle of
Saint-Venant type". J Elasticity 11(3): 271-291, 1981.
[26] C. O. Horgan, L. E. Payne. "Saint-Venants principle in linear isotropic elas-
ticity for incompressible or nearly incompressible materials". J Elasticity
46(1): 43-52, JAN 1997.
[27] D. Iesan. Saint-Venants Problem for inhomogeneous and anisotropic elastic
bodies. Journal of Elasticity 6: 277-294, 1976.
110
[28] D. Iesan. On Saint-Venants Problem for elastic dielectrics. Journal of Elas-
ticity 21: 101-115, 1989.
[29] D. Iesan. Saint-Venants Problem. Springer-Verlag, New York, NY 1987.
[30] D. Iesan and L. Nappa. Saint-Venants Problem for microstretch elastic
solids. Int. J. Engng. Sci. 32: 229-236, 1994.
[31] J. K. Knowles, E. Sternberg. "On Saint-Venants principle and torsion of
solids of revolution". Arch. Ration. Mech. Anal. 22(2): 100-&, 1966.
[32] A. Lichnerowicz. Elements of Tensor Calculus. Spottiswoode Ballantyne &
Co Ltd, London & Colchester, 1962.
[33] J. W. Nunziato, S.C. Cowin. A nonlinear theory of elastic materials with
voids. Arch. Rational Mech. Anal. 72: 175-201, 1979.
[34] R. S. Rivlin. The solution of problems in second order elasticity theory. J.
Rational Mechs. Analysis 2: 53-81, (1953).
[35] A. Romano, "Lezioni di sica matematica". Anno accademico 1982-83.
[36] G. C. Ruta. Espansioni formali per il problema di Saint-Venant. Universit
degli studi di Roma "La Sapienza" .Facolt di Ingegneria. Dottorato di ricerca
in meccanica teorica e applicata VIII ciclo. Tesi di Dottorato, Febbraio 1996.
[37] R. A. Toupin, "Saint- Venant and a matter of principle". T New York Acad
sci 28(2): 221-&, 1965.
[38] C. A. Truesdell, W. Noll, The nonlinear eld theories of mechanic. In Hand-
buch der Physik (S. Fl
..
ugge, ed.), vol. III/3, Springer-Verlag, Berlin, 1965.
111
[39] A. J. C. Barr de Saint-Venant, Mmoire sur la exion des prismes. Mm.
Savants Etrangers 14, 223, 1855.
[40] A. J. C. Barr de Saint-Venant, Mmoire sur la torsion des prismes. Journal
Math. Liouville (2), 1, 89, 1856.
[41] I. S. Sokolniko, Mathematical theory of elasticity. McGraw-Hill, NewYork
1946.
[42] C. J. Thompson. Mathematical Statistical Mechanics. Princeton University
Press. B.
[43] C. C. Wang, C. A. Truesdell. Introduction to Rational Elasticity.Leyden.
Noordho, 1973.
112
VITA
Luca Placidi was born in Rome on 21 April 1976. In 1995 he took his high school
degree in Rome at "Liceo Scientico Talete". In 2001 he earned his master degree
in Physics from "Universit degli studi di Napoli Federico II" in Naple. In 2000
he began his graduate studies in Engineering Mechanics at Virginia Polytechnic
Institute and State University, also known as Virginia Tech.
113

Solution of St.-Venant's and

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Solution of St.-Venant's and

Uploaded by

Copyright:

Available Formats

SOLUTION OF ST.

to the present conguration C . Let X be a generic point in the reference

and x its position in the present conguration C . The nite

to C is described by the following vector valued function:

its corresponding position in C.

be the innitesimal element of area at X; N is the outward

be the innitesimal volume at X, and dc be the corresponding

into the intermediate conguration C

in the reference conguration is

the mass density does not depend upon the material

= T in linear elasticity. When

are such that they do not

, then they are called dead loads.

= then the problem is said to be a traction-value problem; if C

(X) then the load is called dead

on F is unknown. Even for dead loads we have the

equals the Cauchy stress tensor T

) in Taylor series around the point (X, 0, 0), we get

) (X, 0)k) , (2.79)

depends analytically upon the

= (D[0, L]) (D{0}) (D{L}) .

( = 1, ..., ) clockwise; these are shown

in the reference conguration between the bases D(0) and D(Z).

of the Appendix and their solutions are

N c (X) p(X) +a (X) e

is the element of arc length of the boundary of the cross-section in

d where is a parameter along the axial position

b = rot (b) = skwgrad (b) ,

= Grad, rotgradv = 0, tr (rotL) = 0, (6.22)

vary with the angle . We note that r and

You might also like