Vous êtes sur la page 1sur 34

CHAPTER 1

REVIEW OF
CONTINUUM MECHANICS

This chapter is aimed at reviewing some concepts of continuum mechanics and introducing the no-
tation that will be used subsequently.

1.1 Notation and useful properties of tensors

This section briefly reviews some notation and properties of tensors used in these notes.

The inner product of two first-order tensors (vectors) a and b is a scalar:

a · b = ai b i = a1 b 1 + a2 b 2 + a3 b 3 (1.1)

Note that a · b = b · a.
The inner product of a second-order tensor T and a vector a produces a vector given by:
⎧ ⎫
⎨T11 a1 + T12 a2 + T13 a3 ⎪
⎪ ⎬
b=T·a ⇒ bi = Tij aj = T21 a1 + T22 a2 + T23 a3 (1.2)
⎪ ⎪
T31 a1 + T32 a2 + T33 a3
⎩ ⎭

In general, the above operation is non-commutative. Indeed, we have the following:

b = a · T = TT · a ⇒
⎧ ⎫
⎨a1 T11 + a2 T21 + a3 T31 ⎪
⎪ ⎬ (1.3)
bi = aj Tji = Tji aj = a1 T12 + a2 T22 + a3 T32
⎪ ⎪
a1 T13 + a2 T23 + a3 T33
⎩ ⎭

Therefore, it is commutative only when T is a symmetric tensor, i.e., TT = T.


The inner product of two second-order tensors S and T produces a second-order tensor P defined as:

P= S·T ⇒ Pij = Sik Tkj (1.4)

Moreover,
T
(S · T) = TT ST (1.5)
The inverse of a second-order tensor is defined as

A · A−1 = A−1 · A = I (1.6)


Lecture Notes. 3
By Lorenzo Dozio Copyright ⃝
c 2014 Lorenzo Dozio
4 REVIEW OF CONTINUUM MECHANICS

The inverse operation has the following properties


( −1 )−1
A =A
−1 −1
= α−1 A
( )
αA (1.7)
−1 −1 −1
(A · B) =B ·A

The double inner product of two second-order tensor produces a scalar:

p=S:T ⇒ p = Sij Tij = S11 T11 + S12 T12 + S13 T13


+ S21 T21 + S22 T22 + S23 T23 (1.8)
+ S31 T31 + S32 T32 + S33 T33

The above operation is also known as contraction or contractive product. Note that for the double inner product holds

S:T=T:S (1.9)

The double inner product of a fourth-order tensor S and a second-order tensor T produces a second-order tensor as
follows:
P=S:T ⇒ Pij = Sijkl Tkl (1.10)
The outer product of two vectors a and b is a second-order tensor T defined as
⎡ ⎤
a1 b 1 a1 b 2 a1 b 3
T=a⊗b ⇒ Tij = ai bj = ⎣a2 b1 a2 b2 a2 b 3 ⎦ (1.11)
⎢ ⎥

a3 b 1 a3 b 2 a3 b 3

Any second-order tensor can be uniquely decomposed into a symmetric and skew-symmetric tensor as follows:

A = sym(A) + skew(A) (1.12)

where
1( 1
A + AT
)
sym(A) = ⇒ sym(A)ij = (Aij + Aji ) (1.13)
2 2
1( 1
A − AT
)
skew(A) = ⇒ skew(A)ij = (Aij − Aji ) (1.14)
2 2
Let B a symmetric tensor, i.e., B = BT . Since the double inner product of a symmetric tensor and a skew-symmetric
tensor is zero, we have
A : B = (sym(A) + skew(A)) : B = sym(A) : B (1.15)

1.2 Deformation map

Consider a body (continuum) in an initial state, which occupies the domain (volume) V0 . This is called the initial
configuration. Unless we specify otherwise, the initial configuration will be used as the reference configuration of
the body, i.e. the configuration to which various equations are referred. In many (but not all) situations, the initial
configuration is an undeformed configuration.
The domain (volume) occupied by the body in the current configuration will be denoted by V . The current configu-
ration will also be called the deformed configuration.
The position of a point in the reference configuration is given by X. The variables X do not change with time and are
called Lagrangian or material coordinates. For simplicity, they will be referred here to a rectangular Cartesian coordinate
system, i.e., ⎧ ⎫
⎨X1 ⎪
⎪ ⎬
X = X2 (1.16)
⎩ ⎪

X3

MOTION 5

φ(X, t)

X3

X
x3
x
X2

X1
x2

x1

Figure 1.1 Deformation mapping.

The position of a point in the current configuration is given by x. The variables x are called Eulerian or spatial coordi-
nates. In rectangular coordinates ⎧ ⎫
⎪x1 ⎬
⎨ ⎪
x = x2 (1.17)

⎩ ⎪
x3

The response of a continuum can be described by assuming X and time t as independent variables (Lagrangian formula-
tion) or by assuming x and time t as independent variables (Eulerian formulation). An Eulerian description is typically
used in fluid mechanics, while Lagrangian formulations are prevalent in solid mechanics. We adopt here a Lagrangian
description.
In a Lagrangian formulation, the deformation of the body is described with respect to a reference configuration as
follows
x = φ(X, t) (xi = φi (X, t)) (1.18)
where the vectorial function φ is a map from the initial configuration to the current configuration (see Figure 1.1). It is
assumed that the mapping φ(X, t) is continuously differentiable and is one-to-one. This ensures that compatibility is
satisfied (there are no gaps or overlaps in the deformed body) and that φ is invertible. Therefore, we can write

X = φ−1 (x, t) (1.19)

1.3 Motion

If we take the coordinate system describing the deformed configuration to be identical to the coordinate system describing
the reference configuration, we can define the displacement s(X, t) of a point of the body at initial position X as the
difference between its current and reference configuration

s(X, t) = x − X (si = xi − Xi )
(1.20)
= φ(X, t) − X

When X is held constant then the derivative is called Lagrangian time derivative or total time derivative. The material
velocity of a generic point is given by
∂x ∂s(X, t)
v(X, t) = = (1.21)
∂t ∂t
6 REVIEW OF CONTINUUM MECHANICS

The material acceleration is the rate of change of the velocity, or the total time derivative of the velocity expressed by

∂v(X, t) ∂vi (X, t)


a(X, t) = or ai (X, t) = (1.22)
∂t ∂t
It is known that when the velocity is expressed in Eulerian coordinates, its total time derivative in indicial notation is
given by:
Dvi (x, t) ∂vi (x, t) ∂vi (x, t)
= + vj (x, t) (1.23)
Dt ∂t ∂xj
Therefore, we have
Dv(X, t) ∂v(X, t)
a(X, t) = = (1.24)
Dt ∂t

s(X, t)
X3 , x3

X
x

X2 , x2

X1 , x1

Figure 1.2 Deformation mapping and displacement.

1.4 Deformation gradient

The deformation of the body can be characterized by the deformation gradient tensor F, which is defined as the jacobian
matrix of the mapping φ
∂φ ∂x
F= ≡ (1.25)
∂X ∂X
In indicial notation we have:
Fij = φi/j = xi/j (1.26)
In matrix notation: ⎡ ∂x ∂x1 ∂x1
⎡ ⎤ ⎤
F11 F12 F13 ∂X
1
∂X2 ∂X3
⎢ ∂x21 ∂x2 ∂x2 ⎥
F = ⎣F21 F22 F23 ⎦ = (1.27)
⎢ ⎥
⎣ ∂X1 ∂X2 ∂X3 ⎦
∂x3 ∂x3 ∂x3
F31 F32 F33 ∂X1 ∂X2 ∂X3

According to Eq. (1.18), we have

x + dx = φ(X + dX) ⇒ dx = φ(X + dX) − φ(X)


∂φ (1.28)
= · dX
∂X
It follows that
dx = F · dX (dxi = Fij dXj ) (1.29)
and
dX = F−1 · dx (dXi = Fij−1 dxj ) (1.30)
GREEN-LAGRANGE STRAIN TENSOR 7

Using Eq. (1.20) in Eq. (1.25), we can write


∂s
F= I+ (1.31)
∂X
or, in indicial notation,
Fij = δij + si/j (1.32)
where δij is the Kronecker delta (δij = 1 if i = j and δij = 0 if i ̸= j). Since

∂s ∂s ∂x
= · (1.33)
∂X ∂x ∂X
we have that the displacement gradient tensor in the initial configuration is related to the displacement gradient in the
current configuration by
∂s ∂s
= ·F (1.34)
∂X ∂x
The determinant of F is denoted by J = J(X) and is called the jacobian determinant or the determinant of the
deformation gradient. Recall that the volume in the current and reference configuration are related by

dV = JdV0 (1.35)

Note that it is always J > 0 to ensure that material volume elements remain positive.

1.5 Green-Lagrange strain tensor

A deformation measure must vanish for any rigid body motion and in particular for rigid body rotation. If a strain
measure fails to meet this requirement, it will predict nonzero strains, and in turn nonzero stresses, in rigid body rotation.
The deformation gradient F is not a good candidate as a strain measure since it contains the rigid body mode of rotation
in addition to stretch. This can be seen by considering a rigid body motion which consists of a translation xT and a
rotation about the origin
x(X, t) = xT (t) + R · X (1.36)
where R is the orthonormal rotation matrix (RT · R = I). From Eq. (1.25) it follows that

F=R (1.37)

Then, a general measure of deformation independent of both translation and rotation is required. This is satisfied
by the Green-Lagrange (GL) strain tensor which measures the difference of the square of the length of an infinitesimal
segment in the current (deformed) configuration and in the reference configuration:

|dx|2 − |dX|2 = (F · dX) · (F · dX) − dX · dX


= dX · FT · (F · dX) − dX · dX
( )
( T )
= dX · F · F · dX − dX · dX (1.38)
= dX · FT · F − I dX
( )

= 2 dX · E · dX

Thus, the Green-Lagrange tensor is defined by

1( T )
E= F ·F−I (1.39)
2

or, in indicial notation,


1( T ) 1
Eij = F Fkj − δij = (Fki Fkj − δij ) (1.40)
2 ik 2
8 REVIEW OF CONTINUUM MECHANICS

Using Eq. (1.31) the GL strain tensor can be expressed in terms of displacements as follows
0 1 2T 1 2T 3
1 ∂s ∂s ∂s ∂s
E= + + · (1.41)
2 ∂X ∂X ∂X ∂X

In indicial notation,
1( )
Eij =si/j + sj/i + sk/i sk/j (1.42)
2
where indices repeated twice in a term are summed. In explicit form
01 22 1 22 1 22 3
∂s1 1 ∂s1 ∂s2 ∂s3
E11 = + + +
∂X1 2 ∂X1 ∂X1 ∂X1
01 22 1 22 1 22 3
∂s2 1 ∂s1 ∂s2 ∂s3
E22 = + + +
∂X2 2 ∂X2 ∂X2 ∂X2
01 22 1 22 1 22 3
∂s3 1 ∂s1 ∂s2 ∂s3
E33 = + + +
∂X3 2 ∂X3 ∂X3 ∂X3
1 2 4 5
1 ∂s1 ∂s2 1 ∂s1 ∂s1 ∂s2 ∂s2 ∂s3 ∂s3
E12 = + + + +
2 ∂X2 ∂X1 2 ∂X1 ∂X2 ∂X1 ∂X2 ∂X1 ∂X2
1 2 4 5
1 ∂s1 ∂s3 1 ∂s1 ∂s1 ∂s2 ∂s2 ∂s3 ∂s3
E13 = + + + +
2 ∂X3 ∂X1 2 ∂X1 ∂X3 ∂X1 ∂X3 ∂X1 ∂X3
1 2 4 5
1 ∂s2 ∂s3 1 ∂s1 ∂s1 ∂s2 ∂s2 ∂s3 ∂s3
E23 = + + + +
2 ∂X3 ∂X2 2 ∂X2 ∂X3 ∂X2 ∂X3 ∂X2 ∂X3
where the terms in square brackets are the non-linear part of the strains. Note that the G-L strain tensor is symmetric
(Eij = Eji ) and it vanishes in rigid body motion.

——————————————–

EXAMPLE 1.1

Consider a body which is rotated by a finite (large) angle θ about the axis X3 . Let the right-handed sense of rotation correspond
to positive values of θ. A generic point of the body at a distance R from the origin in initial configuration has coordinates

X1 = R cos φ; X2 = R sin φ; X3
!
2 2
where φ is the angle with respect to X1 axis and R = X1 + X2 . In the current configuration, the position of the point is

x1 = R cos (φ + θ) ; x2 = R sin (φ + θ) ; x3 = X3

Using angle transformation formulas we can write

x1 = R (cos φ cos θ − sin φ sin θ) = X1 cos θ − X2 sin θ


x2 = R (sin φ cos θ + cos φ sin θ) = X2 cos θ + X1 sin θ

In matrix form, ⎧ ⎫ ⎡ ⎤⎧ ⎫
⎨ x1 ⎪
⎪ ⎬ cos θ − sin θ 0 ⎪⎨X1 ⎪⎬
x2 = ⎣ sin θ cos θ 0⎦ X2 (1.43)
⎢ ⎥

⎩ ⎪ ⎪
⎩ ⎪
x3 0 0 1 X3
⎭ ⎭

According to Eq. (1.20), the motion is expressed as


⎧ ⎫ ⎡ ⎤⎧ ⎫
⎨s1 ⎪
⎪ ⎬ cos θ − 1 − sin θ 0 ⎪ ⎨X1 ⎪

s2 = ⎣ sin θ cos θ − 1 0⎦ X2
⎢ ⎥

⎩ ⎪ ⎪
⎩ ⎪
s3 0 0 0 X3
⎭ ⎭
NANSON’S FORMULA 9

The in-plane (1, 2) components of the GL strain tensor are given by


1/
(cos θ − 1)2 + sin2 θ = 0
0
E11 = cos θ − 1 +
2
1/ 2
sin θ + (cos θ − 1)2 = 0
0
E22 = cos θ − 1 +
2
1 1
E12 = (− sin θ + sin θ) + [− (cos θ − 1) sin θ + sin θ (cos θ − 1)] = 0
2 2

——————————————–

1.6 Nanson’s formula

As we will see later, we need to transform quantities that are defined with respect to areas in a deformed (current)
configuration to those relative to areas in a reference (initial) configuration, and vice versa.
By denoting dA an area of a region in the deformed configuration with outward normal n, the infinitesimal volume
element in the current configuration can be written as

dV = dx · n dA (1.44)

In a similar way, the infinitesimal volume element in the reference configuration can be written as

dV0 = dX · n0 dA0 (1.45)

where n0 is the outward normal of the area dA0 . Using Eq. (1.35) and (1.29), one obtains

dV = JdV0
= JdX · n0 dA0
= J F−1 · dx · n0 dA0
( )
(1.46)
= J dx · F−T · n0 dA0
( )

= Jdx · F−T · n0 dA0


( )

By equating (1.44) and (1.46) we can write

n dA = JF−T · n0 dA0 (1.47)

which is known as Nanson’s formula.

1.7 Conservation of mass

Mass conservation requires that the mass, during deformation, remains the same, i.e.,
6 6
ρ(x, t)dV = ρ0 dV0 = const. (1.48)
V V0

where ρ0 is the density of the body in the reference configuration. Using Eq. (1.35) yields
6 6 6
ρ(x, t)dV = ρ(X, t)JdV0 = ρ0 dV0 (1.49)
V V0 V0

which must be valid for any V0 . Therefore, the conservation of mass for Lagrangian descriptions is written as follows

ρ0 (X) = ρ(X, t)J (1.50)


10 REVIEW OF CONTINUUM MECHANICS

1.8 Conservation of linear momentum

The linear momentum of a body in the current (deformed) configuration is given by


6
p(t) = ρ(x, t)v(x, t)dV (1.51)
V

where v is the velocity. If the body is subjected to body forces ρb(x, t) and surface tractions t(x, t), the total force is
given by 6 6
f (t) = ρ(x, t)b(x, t)dV + t(x, t)dA (1.52)
V A

The momentum conservation principle states that the time derivative of the linear momentum equals the net force:

D
6 6 6
ρv dV = ρb dV + t dA (1.53)
Dt V V A

Using what derived in the previous sections, we can write the time variation of the linear momentum in the current
configuration in terms of quantities related to the reference configuration as follows:

D D
6 6
ρ(x, t)v(x, t)dV = ρ(X, t)v(X, t)JdV0
Dt V Dt V0
D
6
= ρ0 (X)v(X, t)dV0
Dt V0
(1.54)
Dv(X, t)
6
= ρ0 dV0
Dt
6V0
= ρ0 a(X, t)dV0
V0

where a = a(X, t) is the material acceleration of point X.


Similarly, using the Cauchy’s fundamental relation,

t = σT · n (ti = σji nj ) (1.55)

where ⎡ ⎤
σ11 σ12 σ13
σ = ⎣σ21 σ22 σ23 ⎦ (1.56)
⎢ ⎥

σ31 σ32 σ33


is the Cauchy stress tensor, defined in the current configuration (also called true stress tensor since it is the physical stress
of the true-current configuration), we can write
6 6
t(x, t)dA = σ T · n dA
A A
6
σ T · JF−T · n0 dA0
( )
=
6A0 (1.57)
( −1 )T
= JF · σ · n0 dA0
A
6 0
= PT · n0 dA0
A0

In the above we have introduced the nominal stress tensor defined as


−1
P = JF−1 · σ (Pij = JFik σkj ) (1.58)
CONSERVATION OF LINEAR MOMENTUM 11

which is similar to the Cauchy stress expect that it is expressed in terms of the area and normal of the reference (unde-
formed) surface. Note also that we can define

t0 = PT · n0 (t0i = PijT n0j = Pji n0j ) (1.59)

which is the counterpart in the reference configuration of the Cauchy’s relation. Finally, since the term related to body
forces can be written as 6 6 6
ρb dV = ρbJ dV0 = ρ0 b dV0 (1.60)
V V0 V0

the conservation of linear momentum for the Lagrangian description is expressed as follows
6 6 6
ρ0 a dV0 = ρ0 b dV0 + PT · n0 dA0 (1.61)
V0 V0 A0

In indicial notation, 6 6 6
ρ0 ai dV0 = ρ0 bi dV0 + Pji n0j dA0 (i = 1, 2, 3) (1.62)
V0 V0 A0

Using the divergence theorem on the last term in Eq. (1.62) yields
6 6 6
ρ0 ai dV0 = ρ0 bi dV0 + Pji/j dV0 (i = 1, 2, 3) (1.63)
V0 V0 V0

Since the above relation holds for an arbitrary domain, we can write

ρ0 ai = ρ0 bi + Pji/j (i, j = 1, 2, 3) (1.64)

or, in tensor notation,


ρ0 a = ρ0 b + ∇0 · P (1.65)
where ∇0 is the gradient operator in reference coordinates, i.e.,
∂ ⎪
⎧ ⎫

⎪ ⎪
∂X1 ⎪

⎪ ⎪

⎪ ⎪


⎪ ⎪


⎪ ⎪
⎨ ∂ ⎪ ⎬
∇0 = (1.66)

⎪ ∂X2 ⎪


⎪ ⎪


⎪ ⎪

⎪ ∂ ⎪

⎪ ⎪


⎩ ⎪

∂X3
The above are the equations of motion (dynamic equilibrium equations) in the Lagrangian formulation.
Note that the Eulerian formulation with respect the current volume V and current surface A can be obtained directly
from Eq. (1.61) using what derived above. From Eq. (1.35) and the definition of the nominal stress tensor we have
6 6 6
ρ0 a J −1 dV = ρ0 b J −1 dV + Jσ T · F−T · n0 dA0 (1.67)
V V A

Note that now the integrals are calculated with respect to the current (deformed) values. Using Eq. (1.50) and Nanson’s
formula yields 6 6 6
ρa dV = ρb dV + σ T · n dA (1.68)
V V A
After applying the divergence theorem we obtain the well known equations of motion in the Eulerian description:

ρa = ρb + ∇ · σ (1.69)

or, in indicial notation,


ρai = ρbi + σji/j (i, j = 1, 2, 3) (1.70)
12 REVIEW OF CONTINUUM MECHANICS

In Eq. (1.69) the operator ∇ indicates the gradient vector in current coordinates
∂ ⎪
⎧ ⎫

⎪ ⎪
∂x1 ⎪

⎪ ⎪

⎪ ⎪


⎪ ⎪


⎪ ⎪
⎨ ∂ ⎪ ⎬
∇= (1.71)

⎪ ∂x2 ⎪⎪

⎪ ⎪


⎪ ⎪



⎪ ⎪


⎪ ⎪

⎩ ⎭
∂x3
Comparing Eq. (1.69) with the Lagrangian form (1.65), we can see that they are quite similar: the Cauchy stress is
replaced by the nominal stress, the density is replaced by the initial density and the spatial derivatives are calculated with
respect to the reference configuration X instead of the current configuration x.

1.9 Second Piola-Kirchhoff tensor

It is useful to introduce another measure of stress, the so-called second Piola-Kirchhoff (PKII) stress tensor. The second
Piola-Kirchhoff (PKII) stress tensor is defined as follows

S = JF−1 · σ · F−T (1.72)

and relates forces in the reference configuration with areas in the reference configuration. The nominal stress and the
PKII stress tensors are related by

P = S · FT T
(Pij = Sik Fkj = Sik Fjk ) (1.73)
−T −1
S = P · F−T (Sij = Pik Fkj = Pik Fjk ) (1.74)
Therefore, the linear momentum equation (1.65) can be written in terms of PKII tensor as

ρ0 a = ρ0 b0 + ∇0 · S · FT
( )
(1.75)

1.10 Conservation of angular momentum

As already known, the Cauchy stress is a symmetric tensor, i.e., σ = σ T . Since from Eq. (1.58)

σ = J −1 F · P (1.76)

we have )T
J −1 F · P = J −1 F · P
(
(1.77)
Multiplying both sides by J and performing the transpose gives

F · P = PT · FT (Fik Pkj = Pki Fjk ) (1.78)

It is seen from above that the nominal stress tensor is not symmetric.
Substituting Eq. (1.73) into Eq. (1.77) yields
)T
J −1 F · S · FT = J −1 F · S · FT
(
(1.79)

which gives
F · S · FT = F · ST · FT (1.80)
So the conservation of angular momentum in Lagrangian coordinates requires the PKII stress tensor to be symmetric

S = ST (1.81)
PRINCIPLE OF VIRTUAL WORK 13

1.11 Principle of virtual work

Conservation of linear momentum with respect to the reference configuration is expressed in Eq. (1.64). This holds
for a body having prescribed displacement (Dirichlet) boundary conditions on its boundary AD 0 and prescribed traction
(Neumann) boundary conditions on its boundary AN 0 . The whole boundary A 0 is defined as A0 = AD N
0 ∪ A0 .
After multiplying each term of Eq. (1.64) by a variation δsi of the reference configuration displacement (i.e. a virtual
displacement) which is arbitrary except at the kinematic boundary condition locations where it vanishes

S0 = δsi |δsi ∈ C0 (X), δsi = 0 on AD


7 8
δsi (X) ∈ S0 , 0 (1.82)

and integrating over the reference domain V0 yields


6 6 6
δsi ρ0 ai dV0 = δsi ρ0 bi dV0 + δsi Pji/j dV0 (1.83)
V0 V0 V0

The last integral can be further expressed as follows


6 6 9 :
δsi Pji/j dV0 = (δsi Pji )/j − δsi/j Pji dV0
V0
6V0 6
= δsi Pji n0j dA0 − δsi/j Pji dV0 (1.84)
A0 V0
6 6
= δsi t̄0i dA0 − δsi/j Pji dV0
AN
0 V0

where t0i are the prescribed tractions on AN0 . Therefore, we can write
6 6 6
δsi/j Pji dV0 = δsi ρ0 bi dV0 + δsi t0i dA0
V0 V0 AN
0
6 (1.85)
− δsi ρ0 ai dV0
V0

The above can be considered the weak form for the linear momentum equation. It can be also written in tensor notation
as 1 2T
∂s
6 6 6 6
δ : P dV0 = ρ0 δs · b dV0 + δs · t0 dA0 − ρ0 δs · a dV0 (1.86)
V0 ∂X V0 AN
0 V0

Using Eq. (1.73), the weak form can be expressed in terms of the second Piola-Kirchhoff tensor as follows
6 6 6 6
δsi/j Sjk Fik dV0 = δsi ρ0 bi dV0 + δsi t0i dA0 − δsi ρ0 ai dV0
V0 V0 AN
0 V0

Since
6 6
δsi/j Sjk Fik dV0 = δFij Sjk Fik dV0
V0 V0
6 6
= δFij Fik Sjk dV0 = δFjiT Fik Sjk dV0
V0 V0
6
( T )
= δF · F : S dV0
V
6 04 5
1( T ) 1( T
δF · F + FT · δF + δF · F − FT · δF : S dV0
)
=
2 2
6V0
; ( T ) ( T )<
= sym δF · F + skew δF · F : S dV0
V
6 0 6
sym δFT · F : S dV0 =
( )
= δE : S dV0
V0 V0
14 REVIEW OF CONTINUUM MECHANICS

we can write 6 6 6 6
δEij Sij dV0 = δsi ρ0 bi dV0 + δsi t0i dA0 − δsi ρ0 ai dV0 (1.87)
V0 V0 AN
0 V0

which is the principle of virtual work written with respect to the reference configuration (Lagrangian formulation). In
tensor notation it is given by
6 6 6 6
δE : S dV0 = ρ0 δs · b dV0 + δs · t0 dA0 − ρ0 δs · a dV0 (1.88)
V0 V0 AN
0 V0

The principle of virtual work can be expressed as

δW (s, s̈, t) = δWint − δWext = 0 (1.89)


where 6
δWint = δE : S dV0 (1.90)
V0
is the internal virtual work, and 6
δWext = δs · fext dV0 (1.91)
V0
is the external virtual work, where fext is the sum of the external body and surface loads applied on the body, including
inertial forces in dynamic analysis.

1.12 Infinitesimal strain tensor

The principle of virtual work derived so far applies to both conservative and non-conservative systems where no en-
ergy functional exists. It is completely general as no particular material was taken into account and large displace-
ments/rotations and large strains are allowed. When both strains and displacements/rotations are assumed to maintain
a small amplitude, and the material is considered to be linear, important simplifications can be made, which are shown
below.
Assume that all components of the displacement gradient ∂s/∂X are small:
∂si
≪1 (i, j = 1, 2, 3) (1.92)
∂Xj
From Eq. (1.34) we have
∂s ∂s
= · F−1
∂x ∂X
= (F − I) · F−1
= I − F−1
and, by using the binomial expansion,

I − F−1 = I − [I + (F − I)]−1
9 :
2 3
= I − I − (F − I) + (F − I) − (F − I) + . . .

we obtain that each component of the displacement gradient in the current configuration may be expressed as follows
∂si ∂si ∂si ∂sk ∂si ∂sk ∂sl
= − + − ...
∂xj ∂Xj ∂Xk ∂Xj ∂Xk ∂Xl ∂Xj
Thus, using assumption in Eq. (1.92), we can write
1 2
∂s ∂s ∂si ∂si
≈ ≈
∂x ∂X ∂xj ∂Xj
INFINITESIMAL STRAIN TENSOR IN RIGID-BODY MOTION 15

This implies that, in case of small displacement gradients, we can assume that there is no distinction to be made between
the undeformed (initial) and deformed (current) configurations (x ≈ X). That is, the Lagrangian description and the
Eulerian description are approximately the same. Accordingly, the GL strain tensor simplifies as follows
0 1 2T 3 0 1 2T 3
1 ∂s ∂s 1 ∂s ∂s
E≈ + = + =ε (1.93)
2 ∂X ∂X 2 ∂x ∂x

where ε is the well known infinitesimal strain tensor. In explicit form, its components are given by
1 2
∂s1 1 ∂s1 ∂s2
ε11 = ε12 = +
∂x1 2 ∂x2 ∂x1
1 2
∂s2 1 ∂s1 ∂s3
ε22 = ε13 = +
∂x2 2 ∂x3 ∂x1
1 2
∂s3 1 ∂s2 ∂s3
ε33 = ε23 = +
∂x3 2 ∂x3 ∂x2

Furthermore, it follows that


1(
F + FT − I
)
ε=
2
Clearly, ε is symmetric. Recall that the diagonal term of the infinitesimal strain tensor εii is relative elongation along
axis xi and off-diagonal term εij is half decrease of the angle between line elements initially parallel to axes xi and xj .
Note that the displacement gradient may be decomposed into a symmetric and a skew-symmetric part as follows
0 1 2T 3 0 1 2T 3
∂s 1 ∂s ∂s 1 ∂s ∂s
= + + −
∂x 2 ∂x ∂x 2 ∂x ∂x
=ε+ω (si/j = εij + ωij )

where ω is the skew-symmetric infinitesimal rotation tensor


⎡ ⎤
0 −θ3 θ2
ω = ⎣ θ3 0 −θ1 ⎦
⎢ ⎥

−θ2 θ1 0

and
1( ) 1( ) 1( )
θ1 = s3/2 − s2/3 , θ2 = s1/3 − s3/1 , θ3 = s2/1 − s1/2
2 2 2

1.13 Infinitesimal strain tensor in rigid-body motion

It is important to point out that the infinitesimal strain tensor ε is not an exact measure of deformation because it is
not zero in the case of a finite rigid body motion. However, when the rotations are small, it provides an excellent
approximation to such a measure. This is illustrated in the following example.

——————————————–

EXAMPLE 1.2

Consider again the example of a body rigidly rotating about the X3 axis by a finite angle θ. The displacement components in the
(1, 2) plane are given by 1 2 3 41 2
s1 cos θ − 1 − sin θ X1
=
s2 sin θ cos θ − 1 X2
16 REVIEW OF CONTINUUM MECHANICS

According to Eq. (1.93), the in-plane (1, 2) components of the infinitesimal strain tensor are given by

ε11 = cos θ − 1
ε22 = cos θ − 1 (1.94)
1
ε12 = (− sin θ + sin θ) = 0
2
It is clear that, if θ is large, the extensional strains ε11 and ε22 do not vanish. Therefore, the linear strain tensor is not suitable for
large rotation problems, i.e., in geometrically nonlinear problems. Instead, if θ is small, cos θ ≈ 1, and we have

ε11 ≈ 0
ε22 ≈ 0

as expected by a correct strain measure.

——————————————–

A question arises from the previous example: how large the rotations can be before a nonlinear analysis is required? The
magnitude of the strains predicted in Eq. (1.94) are an indication of the error due to the small strain assumption. Indeed,
by expanding cos θ in a Taylor’s series we can write
θ2 θ2
ε11 = cos θ − 1 = 1 − + ···− 1 ≈ −
2 2
This shows that the error in the linear strain is second order in the rotation. If the strains of interest are of order 10−2
and 1% error is acceptable (it almost always is), then the rotations can be of order 10−2 , since the error due to the small
strain assumption is of order 10−4 . If the strains of interest are smaller, the acceptable rotations are smaller: for strains
of order 10−4 , the rotations should be of order 10−3 for 1% error.

1.14 Linear elasticity

Up to now, we have seen that the relationship between strain and motion does not depend upon stress. Further, the
relationship between stress and the applied force does not depend upon strain. As such, we must introduce a link
between equations of kinematics and equations of equilibrium to complete the theory. The gap between the twos is filled
by a constitutive model, which is the mathematical law describing how strain and stress are related each other for a given
material. It is known that there are models of elasticity, plasticity, viscoelasticity, and many others. In addition to the
assumption of small displacements and displacement gradients, here we shall consider only elastic materials. An elastic
material will return to its initial configuration upon unloading and the configuration adopted by a stressed elastic material
does not depend upon the history of loading. Elastic materials can be characterized by a linear or a nonlinear behavior.
We will focused on the class of elastic materials with linear behavior.

1.14.1 Generalized Hooke’s law


Path independence of the stress state can be guaranteed by assuming the existence of a strain energy function W given
by
6 t2
W (ε(t)) = σ : ε̇ dt (1.95)
t1
where ε̇ is the derivative with respect to time of the small strain tensor. By construction, we have

Ẇ = σ : ε̇ (1.96)

Therefore, path independence is assured since


6 t2 6 t2
σ : ε̇ dt = Ẇ dt = W (t2 ) − W (t1 ) (1.97)
t1 t1
LINEAR ELASTICITY 17

Using the chain rule of differentiation yields


∂W
Ẇ = ε̇ij (1.98)
∂εij
Comparing the above with Eq. (1.96) we have found the following important result
∂W
σij = (1.99)
∂εij

which shows that the components of the stress tensor are the derivatives of the strain energy function with respect to the
components of strain. If the elastic material is linear, it follows that

σ=C:ε (σij = Cijkl εkl ) (1.100)

where the 81 components of the fourth-order tensor C (called stiffness tensor) are given by

∂σij ∂2W
Cijkl = = (1.101)
∂εkl ∂εij ∂εkl
The strain energy function is quadratic. In components, we may write
1
W = εij Cijkl εkl (1.102)
2
Owing to the well-known symmetry of the infinitesimal strain and Cauchy’s stress tensors (εij = εji , σij = σji ), only
six independent components of stress are related to six independent components of strain. As a result, there are at most
36 distinct elastic coefficients since
Cijkl = Cjikl (1.103)
Furthermore, smoothness of the strain energy function (W is a C 1 function of ε) implies that

∂2W ∂2W
= (1.104)
∂εij ∂εkl ∂εkl ∂εij

Hence, the elastic coefficients also possess the following symmetry:

Cijkl = Cklij (1.105)

which reduces the number of independent components of the elastic tensor C to 21. For the sake of convenience, we can
introduce the following strain and stress vectors
⎧ ⎫ ⎧ ⎫


⎪ ε 11 ⎪

⎪ ⎪σ11 ⎪

⎪ ⎪

⎪ ⎪ ⎪ ⎪



⎪ ε22 ⎪⎪





⎪ σ22 ⎪⎪



⎨ε ⎪ ⎬ ⎪
⎨σ ⎪ ⎬
33 33
ϵ= , σ= (1.106)
⎪2ε23 ⎪
⎪ ⎪ ⎪σ23 ⎪
⎪ ⎪

⎪ ⎪
⎪ ⎪
⎪ ⎪

⎪2ε13 ⎪ ⎪σ13 ⎪

⎪ ⎪ ⎪
⎪ ⎪
⎪ ⎪
⎪ ⎪ ⎪


⎩2ε ⎭ ⎪ ⎩σ ⎪
⎪ ⎭
12 12

Therefore, the elastic coefficients can be collected into a symmetric 6 × 6 elastic or stiffness matrix as follows
⎡ ⎤
C11 C12 C13 C14 C15 C16
⎢ ⎥
⎢C12 C22 C23 C24 C25 C26 ⎥
⎢ ⎥
⎢C
⎢ 13 C23 C33 C34 C35 C36 ⎥

C=⎢ ⎥ (1.107)
⎢C14 C24 C34 C44 C45 C46 ⎥
⎢ ⎥
⎢C
⎣ 15 C25 C35 C45 C55 C56 ⎦

C16 C26 C36 C46 C56 C66
18 REVIEW OF CONTINUUM MECHANICS

The constitutive model for a linear elastic material can then be written in matrix form as
σ = Cϵ (1.108)
It is also assumed that the above relation is invertible. Thus, the components of strain are related to the components of
stress by
ϵ = C−1 σ = Sσ (1.109)
where S is the material compliance matrix.

1.14.2 Orthotropic materials


What presented above refers to the general case of a solid body made of anisotropic material. Further reduction in
the number of independent stiffness (or compliance) parameters comes from the so-called material symmetry. When
three mutually orthogonal planes of material symmetry exist, the number of elastic coefficients is reduced to 9 and such
materials are called orthotropic. The stress-strain relations for an orthotropic material take the form
⎡ ⎤
C11 C12 C13 0 0 0
⎢ ⎥
⎢C12 C22 C23 0 0 0 ⎥
⎢ ⎥
⎢C
⎢ 13 C23 C33 0 0 0 ⎥
σ=⎢ ⎥ϵ (1.110)

⎢ 0 0 0 C 44 0 0 ⎥
⎢ ⎥
⎢ 0 0 0 0 C55 0 ⎥
⎣ ⎦
0 0 0 0 0 C66
The inverse relation is ⎡ ⎤
S11 S12 S13 0 0 0
⎢ ⎥
⎢S12 S22 S23 0 0 0 ⎥
⎢ ⎥
⎢S S23 S33 0 0 0 ⎥
⎢ 13
ϵ=⎢ ⎥σ (1.111)

⎢ 0 0 0 S44 0 0 ⎥
⎢ ⎥
⎢ 0 0 0 0 S55 0 ⎥
⎣ ⎦
0 0 0 0 0 S66
and the elastic and compliance parameters are related by
2 2
S22 S33 − S23 S33 S11 − S13 1
C11 = C22 = C44 =
S S S44
S13 S23 − S12 S33 S12 S13 − S23 S11 1
C12 = C23 = C55 =
S S S55
2
S12 S23 − S13 S22 S11 S22 − S12 1
C13 = C33 = C66 =
S S S66

and
2 2 2
S = S11 S22 S33 − S11 S23 − S22 S13 − S33 S12 + 2S12 S13 S23
Most often, instead of the abstract quantities Cij or Sij , the material properties are expressed in terms of engineering
constants such as Young’s modulus, shear modulus and Poisson’s ratio, since they are typically measured in a laboratory
using simple uniaxial tension tests and pure shear tests. Accordingly, the material compliance matrix for an orthotropic
material is expressed as ⎡ ⎤
1
E1 − νE212 − νE313 0 0 0
⎢ ν12 1
− νE323

⎢− E E2 0 0 0 ⎥
⎢ 1 ⎥
⎢− ν13 − ν23 1
0 0 0 ⎥
⎢ E1 E2 E3
S=⎢ (1.112)

⎢ 0 1 ⎥
⎢ 0 0 G23 0 0 ⎥

⎢ 0 1
⎣ 0 0 0 G31 0 ⎥⎦
1
0 0 0 0 0 G12
LINEAR ELASTICITY 19

where E1 , E2 , E3 are Young’s moduli in 1,2, and 3 material directions, respectively, νij is Poisson’s ratio and G23 , G31 , G12
are shear moduli in 2-3, 3-2 and 1-2 planes, respectively. Note that, for symmetry reasons, the following reciprocal rela-
tions hold
ν21 ν12
=
E2 E1
ν31 ν13
=
E3 E1
ν32 ν23
=
E3 E2
By inversion of the compliance matrix in Eq. (1.112) the elastic coefficients can be expressed in terms of material
engineering constants as follows
1 − ν23 ν32 1 − ν31 ν13
C11 = C22 = C44 = G23
∆ E2 E3 ∆ E1 E3
ν21 + ν23 ν31 ν32 + ν12 ν31
C12 = C23 = C55 = G31
∆ E2 E3 ∆ E1 E3
ν31 + ν21 ν32 1 − ν12 ν21
C13 = C33 = C66 = G12
∆ E2 E3 ∆ E1 E2

where
1 − ν12 ν21 − ν23 ν32 − ν31 ν13 − 2ν21 ν32 ν13
∆=
E1 E2 E3

1.14.3 Isotropic materials


When there exist no preferred directions in the material, i.e., the material has infinite number of planes of material sym-
metry, the material is called isotropic and the number of independent elastic coefficients reduces to 2. The corresponding
compliance matrix is written as
⎡ ⎤
1 −ν −ν 0 0 0
⎢ ⎥
⎢−ν 1 −ν 0 0 0 ⎥
⎢ ⎥
1 ⎢−ν −ν 1
⎢ 0 0 0 ⎥
S= ⎢ (1.113)

E⎢ 0

⎢ 0 0 2(1 + ν) 0 0 ⎥

⎢ 0 0 0 0 2(1 + ν) 0 ⎥
⎣ ⎦
0 0 0 0 0 2(1 + ν)
Consequently, the material stiffness matrix is expressed as
⎡ ⎤
1−ν ν ν 0 0 0
⎢ ⎥
⎢ ν
⎢ 1 − ν ν 0 0 0 ⎥

E ⎢ ν
⎢ ν 1−ν 0 0 0 ⎥
C=

(1 + ν)(1 − 2ν) ⎢ 0 1−2ν
⎢ ⎥
⎢ 0 0 2 0 0 ⎥

⎢ 0 1−2ν
⎣ 0 0 0 2 0 ⎥

1−2ν
0 0 0 0 0 2
Alternatively, the stress-strain relations for isotropic materials can be written in more compact form as
σij = λεkk δij + 2µεij (1.114)
where λ and µ are called Lamé constants, which are related to the material engineering constants by the following
equations

λ=
(1 + ν)(1 − 2ν)
µ=G
20 REVIEW OF CONTINUUM MECHANICS

1.14.4 Reduced constitutive equations


1.14.4.1 Plane stress. When analyzing solids that are relatively flat and thin, simplifying assumptions are made about
the stress state to facilitate the solution of practical problems. One such assumption is that the stresses in a thin, flat body,
that are normal to the plane of flatness, are negligible compared to the other stresses. This simplification is commonly
referred to as the plane-stress assumption. For a state of plane stress in an anisotropic solid, with respect to the x1 − x2
plane, the stress field is approximated such that
σ33 = σ23 = σ13 = 0 (1.115)
For this special case, the general matrix constitutive strain-stress equation uncouples directly into
⎧ ⎫ ⎡ ⎤⎧ ⎫
⎨ ε11 ⎪
⎪ ⎬ S11 S12 S16 ⎪ ⎨σ11 ⎪⎬
ε22 = ⎣S12 S22 S26 ⎦ σ22 (1.116)
⎢ ⎥
⎪ ⎪ ⎪
⎩ ⎪
2ε12 S16 S26 S66 σ12
⎩ ⎭ ⎭

and ⎧ ⎫ ⎡ ⎤⎧ ⎫
⎨ ε33 ⎪
⎪ ⎬ S13 S23 S36 ⎪ ⎨σ11 ⎪

2ε23 = ⎣S14 S24 S46 ⎦ σ22
⎢ ⎥
⎪ ⎪ ⎪
⎩ ⎪
2ε13 S15 S25 S56 σ12
⎩ ⎭ ⎭

In the case of orthotropic materials we have


1 ν21
S11 = S12 = −
E1 E2
1 ν12
S22 = S21 =−
E2 E1
1
S16 = S26 = 0 S66 =
G12
The strain-stress relation in Eq. (1.116) with S16 = S26 = 0 is inverted to obtain the stress-strain relations for orthotropic
materials as ⎧ ⎫ ⎡ ⎤⎧ ⎫
⎨σ11 ⎪
⎪ ⎬ Q11 Q12 0 ⎪ ⎨ ε11 ⎪⎬
σ22 = ⎣Q12 Q22 0 ⎦ ε22 (1.117)
⎢ ⎥

⎩ ⎪ ⎪ ⎪
σ12 0 0 Q66 2ε12
⎭ ⎩ ⎭

where Qij are the reduced stiffness coefficients given by


E1 ν21 E1
Q11 = Q12 =
1 − ν12 ν21 1 − ν12 ν21
E2
Q22 = Q66 = G12
1 − ν12 ν21
1.14.4.2 Axial stress. Another case of practical interest involves simplifying assumptions when slender bodies are
considered, i.e., solids with one dimension far greater than the others. If x1 is this prevalent dimension, the assumption
is that
σ22 = σ33 = σ23 = 0 (1.118)
Such simplification is denoted as axial-stress assumption and yields the following decoupled constitutive equations
⎧ ⎫ ⎡ ⎤⎧ ⎫
⎨ ε11 ⎪
⎪ ⎬ S11 S15 S16 ⎪ ⎨σ11 ⎪⎬
2ε13 = ⎣S15 S55 S56 ⎦ σ13 (1.119)
⎢ ⎥
⎪ ⎪ ⎪
⎩ ⎪
2ε12 S16 S56 S66 σ12
⎩ ⎭ ⎭

and ⎧ ⎫ ⎡ ⎤⎧ ⎫
⎨ ε22 ⎪
⎪ ⎬ S12 S25 S26 ⎪ ⎨σ11 ⎪

ε33 = ⎣S13 S35 S36 ⎦ σ13
⎢ ⎥
⎪ ⎪ ⎩ ⎪

2ε23 S14 S45 S46 σ12
⎩ ⎭ ⎭
PRINCIPLE OF VIRTUAL WORK IN LINEAR ELASTICITY: 3-D FORMULATION 21

In the case of orthotropic materials we have


1
S11 = S15 = 0
E1
1
S16 = 0 S55 =
G13
1
S56 = 0 S66 =
G12
Therefore, the corresponding stress-strain relations are
⎧ ⎫ ⎡ ⎤⎧ ⎫
⎨σ11 ⎪
⎪ ⎬ Q11 0 0 ⎨ ε11 ⎪
⎪ ⎬
σ13 = ⎣ 0 Q55 0 ⎦ 2ε13 (1.120)
⎢ ⎥

⎩ ⎪ ⎪ ⎪
σ12 0 0 Q66 2ε12
⎭ ⎩ ⎭

where the reduced stiffness coefficients are given by

Q11 = E1 Q55 = G13 Q66 = G12 (1.121)

1.15 Principle of virtual work in linear elasticity: 3-D formulation

Since for small displacement gradients, as seen above, the initial configuration is approximately equal to the deformed
current configuration, the PKII stress tensor can be approximated by the Cauchy’s stress tensor. Therefore, the principle
of virtual work in the pure linear case is written as follows:
6 6 6 6
δεij σij dV0 = δsi bi dV0 + δsi ti dA0 − δsi ρ0 ai dV0 (1.122)
V0 V0 AN
0 V0

or, in tensor notation,


6 6 6 6
δε : σ dV0 = δs · b dV0 + δs · t dA0 − δs · ρ0 a dV0 (1.123)
V0 V0 AN
0 V0

where b(fi , i = 1, 2, 3) is the vector of generic external forces per unit volume.
By assuming a linearly elastic material, we may write
6 6 6 6
δεij Cijkl εkl dV0 = δsi bi dV0 + δsi ti dA0 − δsi ρ0 ai dV0 (1.124)
V0 V0 AN
0 V0

or, in tensor notation,


6 6 6 6
δε : C : ε dV0 = δs · b dV0 + δs · t dA0 − δs · ρ0 a dV0 (1.125)
V0 V0 AN
0 V0

The principle of virtual work for the linear case can be also put into a convenient (more familiar) matrix form.
Therefore, the principle of virtual work becomes
6 6 6 6
δϵT Cϵ dV0 = δsT b dV0 + δsT t dA0 − δsT ρ0 s̈ dV0 (1.126)
V0 V0 AN
0 V0

The principle described in the above Eq. (1.126) will be used thoroughly these notes as the basis to derive exact
and approximate dynamic models of structural systems. First, we will see that the principle of virtual work yields the
correct exact boundary-value problem in almost a routine fashion. Since work quantities of mechanical components are
well defined, there is no room for sign errors, provided the various steps involved are carried out correctly. Second,
the principle of virtual work can be seen as a variational principle (this view is supported by the way we have derived
22 REVIEW OF CONTINUUM MECHANICS

it). Consequently, direct variational methods can be employed from Eq. (1.126) to obtain approximate models of the
dynamic behaviour. Since the principle of virtual work is here selected as the main tool to find solutions of structural
dynamic problems, the general 3-D formulation presented in Eq. (1.126) will be specialised in the following sections to
one- and two-dimensional engineering models of structural members.
It is worth noticing that in many textbooks on structural dynamics the equations of motion are derived on the basis of
the Hamilton’s principle. It can be shown that Hamilton’s principle can be either postulated as a first principle or derived
from the principle of virtual work. Therefore, for our purposes, they can be considered to be equivalent.

1.16 Strength of materials theories

The formulation of the principle of virtual work presented in Eq. (1.126) applies to linear dynamics of 3-D flexible bodies.
This approach is typically referred to as theory of elasticity, since no simplifications of the actual deformation field are
introduced. There are two basic ways in which the theory of elasticity can be simplified in order to obtain a so-called
strength of materials theory, which is the classical engineering approach for some structural elements or components.
Resulting models are easier to solve and, under proper assumptions, are accurate and reliable.
The first strategy in simplifying the 3-D formulation involves an approximation regarding the displacements of the
body under consideration. Generally speaking, this approach relies on properly assuming in advance a prescribed dis-
placement field which governs the kinematics of the deformable body. Such displacement field is typically called kine-
matic theory. The second way is to introduce an approximation regarding the stress field within the body. The most
common examples are the plane stress and the axial stress models mentioned above. In dynamic problems, the dis-
placement approximation is typically preferred and many different kinematic theories of structural members have been
historically developed. It is worth noting that, in some cases, assumptions on the displacement field can be compatible
with assumptions related to the stress field. This is typically the case of classical theories of beams and plates, as shown
later.
In this section, a presentation of the principle of virtual work for one-dimensional continuous systems (beam-like
structures) is provided.

1.16.1 Timoshenko beam theory


The most common displacement approximation for the extension and bending motion of beam-like structures is the so-
called Timoshenko beam theory, which is presented in the following. Note that, provided the assumption on negligible
shear deformation of the beam cross-section, the well-known Euler-Bernoulli theory can be directly derived from the
Timoshenko theory as shown later.
Timoshenko theory for axial-flexural (extension/bending) response of beams relies on the following assumed kine-
matic field: ⎧
⎨s1 = u(x, y, z, t) = u0 (x, t) + zθy (x, t) − yθz (x, t)

s2 = v(x, y, z, t) = v0 (x, t) (1.127)

s3 = w(x, y, z, t) = w0 (x, t)

In the above Eq. (1.127), the axis x is taken as the reference beam axis in the spanwise direction and (y, z) are the
orthogonal axes defining the arbitrary cross-section of the beam (see Fig. 1.3). (u0 , v0 , w0 ) are the rigid displacements
of the reference axis in the x, y, and z directions, respectively, and (θy , θz ) are the (small) rigid rotations about the y and
z axis, respectively.
It is worth noticing that in Eq. (1.127) there is no twisting of the cross-section due to a possible applied torque
(torsional rotation is explicitly omitted in the assumed kinematic field). It is known that, in order to have bending without
twisting, all the transverse distributed and concentrated loads must act through the shear center of the cross section1 .
Having assumed that, it is thus reasonable to take the loci of shear centers, called elastic axis, as the reference beam axis
x.
According to the assumed kinematics, the internal virtual work reduces to
6
δWint = (δεx σx + δγxy τxy + δγxz τxz ) dV (1.128)
V

1 The shear center may be defined as the point of the cross-section where lateral loads acting through it do not cause any twisting.
STRENGTH OF MATERIALS THEORIES 23

z, w0
θz
w
v
y, v0
u θy

x, u0
A

Figure 1.3 Reference system for Timoshenko beam model.

where the strains are given by ⎧


⎨εx = u0/x + zθy/x − yθz/x

γxy = v0/x − θz (1.129)

γxz = w0/x + θy

Let’s assume the following reduced constitutive law


⎧ ⎫ ⎡ ⎤⎧ ⎫
⎨ σx ⎪
⎪ ⎬ E 0 0 ⎨ εx ⎪
⎪ ⎬
τxy = ⎣0 Gy 0 ⎦ γxy (1.130)
⎢ ⎥

⎩ ⎪ ⎪
⎩ ⎪
τxz 0 0 Gz γxz
⎭ ⎭

Note that the material properties can vary in the spanwise direction x and with the cross-sectional coordinates, i.e.,
E = E(x, y, z), G = G(x, y, z). After substituting Eqs. (1.129) and (1.130) into Eq. (1.128), the internal virtual work
reads
6 6
δWint = δu0/x (EAu0/x + ESy θy/x − ESz θz/x ) dx + δθy/x (ESy u0/x + EJy θy/x − EJyz θz/x ) dx
ℓ ℓ
6 6
+ δθz/x (−ESz u0/x − EJyz θy/x + EJz θz/x ) dx + δ(v0/x − θz )GAy (v0/x − θz ) dx
6ℓ ℓ

+ δ(w0/x + θy )GAz (w0/x + θy ) dx


where the following cross-sectional rigidities are introduced:


6
EA(x) = E dA (axial rigidity)
6A
EJy (x) = Ez 2 dA (bending rigidity about y axis)
A
6
EJz (x) = Ey 2 dA (bending rigidity about z axis)
A
6
GAy (x) = Gy dA (shear rigidity along y axis)
6A
GAz (x) = Gz dA (shear rigidity along z axis)
6A
ESy (x) = Ez dA (axial-bending rigidity)
6A
ESz (x) = Ey dA (axial-bending rigidity)
6A
EJyz (x) = Eyz dA (bending-bending rigidity)
A
24 REVIEW OF CONTINUUM MECHANICS

The internal virtual work can be also written in compact matrix form as follows
⎧ ⎫T ⎡ ⎤⎧ ⎫

⎪ δu0/x ⎪
⎪ EA ESy −ESz 0 0 ⎪
⎪ u0/x ⎪ ⎪
⎪ ⎪ ⎪ ⎪
δθy/x ES EJy −EJyz 0 0 ⎥⎪ θ
⎪ ⎪ ⎢ ⎥⎪ ⎪
6 ⎪ y y
⎪ ⎪
/x

⎨ ⎪
⎬ ⎢⎢ ⎪ ⎪
⎥⎨ ⎬
δθz/x ⎢−ESz
⎢ −EJyz EJz 0 0 ⎥⎥⎪ θ z/x dx (1.131)
ℓ⎪ ⎪ ⎪
δ(v − θ ) ⎣ 0 0 0 GAy 0 ⎦⎪⎪ v0/x − θz ⎪
⎪ ⎪ ⎢ ⎥⎪ ⎪
0 z
⎪ ⎪ ⎪ ⎪
/x

⎪ ⎪
⎪ ⎪
⎪ ⎪ ⎪ ⎪
δ(w0/x + θy ) 0 0 0 0 GAz w0/x + θy
⎩ ⎭ ⎩ ⎭

According to the assumed displacement field in Eq. (1.127), the external virtual work due to inertia is given by
6 6
in
δWext = − δu0 (mü0 + Sy θ̈y − Sz θ̈z ) dx − δθy (Sy ü0 + Iy θ̈y − Iyz θ̈z ) dx
6ℓ ℓ
6 6 (1.132)
− δθz (−Sz ü0 − Iyz θ̈y + Iz θ̈z ) dx − δv0 mv̈0 dx − δw0 mẅ0 dx
ℓ ℓ ℓ

where
6
m = m(x) = ρ0 dA (mass per unit length)
6A
Iy = Iy (x) = ρ0 z 2 dA (mass moment of inertia per unit length)
A
6
Iz = Iz (x) = ρ0 y 2 dA (mass moment of inertia per unit length)
A
6
Iyz = Iyz (x) = ρ0 yz dA (mass product of inertia per unit length)
A

and
6
Sy = Sy (x) = ρ0 z dA
6A
Sz = Sz (x) = ρ0 y dA
A

Note that the material density can vary along x and over the cross-section, i.e., ρ0 = ρ0 (x, y, z). Recall that the above
formulation does not take into account any bending/twisting coupling. Since the inertial forces act through the mass
center of the beam cross-section, it follows that the model developed thus far is valid if and only if the loci of mass
centers (mass axis) is coincident with the loci of shear centers. In this case, since x coincides with the axis of centers of
mass, we have 6 6
Sy = ρ0 z dA = 0 Sz = ρ0 y dA = 0 (1.133)
A A
and the inertial virtual work Eq. (1.132) can be written as
6 6
in
δWext =− δu0 mü0 dx − δθy (Iy θ̈y − Iyz θ̈z ) dx
6ℓ ℓ
6 6 (1.134)
− δθz (−Iyz θ̈y + Iz θ̈z ) dx − δv0 mv̈0 dx − δw0 mẅ0 dx
ℓ ℓ ℓ

and, in matrix form, as


⎧ ⎫T ⎡ ⎤⎧ ⎫
δu0 ⎪ m 0 0 0 0 ü ⎪
⎪ 0⎪

⎪ ⎪ ⎪

⎪ ⎪ ⎪
⎨ δv0 ⎪ ⎢0 m 0 0 0 ⎥⎪ v̈ ⎪
⎪ ⎪ ⎢ ⎥⎪ ⎪
⎥⎨ 0⎬
6 ⎪
⎪ ⎪

⎪ ⎪

− δw0 ⎢0
⎢ 0 m 0 0 ⎥ ⎥ ⎪ 0 ⎪ dx
ẅ (1.135)
ℓ⎪ ⎪
⎪ δθy ⎪ ⎣0 0 0 Iy −Iyz ⎦ ⎪
⎪ θ̈y ⎪

⎪ ⎪ ⎢ ⎥⎪ ⎪
⎪ ⎪
⎪ ⎪ ⎪

⎪ ⎪ ⎩ ⎪

δθz 0 0 0 −Iyz Iz θ̈z
⎩ ⎭ ⎭
STRENGTH OF MATERIALS THEORIES 25

z, w0
v0/x
w
v
y, v0
u −w0/x

x, u0
A

Figure 1.4 Reference system for Euler-Bernoulli beam model.

The external virtual work due to applied distributed loads is expressed by


6 6 6
load
δWext = δu0 qx dx + δv0 qy dx + δw0 qz dx
ℓ ℓ
6 6 ℓ (1.136)
+ δθy (my + za qx ) dx + δθz (mz − ya qx ) dx
ℓ ℓ

where qx is the distributed axial load per unit length passing through a point of coordinates (ya , za ) over the beam cross-
section, qy and qz are distributed transverse loads per unit length acting through the shear center2 , and my and mz are
distributed flexural moments per unit length about y and z axis, respectively.

1.16.2 Euler-Bernoulli beam theory


In the Euler-Bernoulli beam theory, shear deformation is considered to be negligible. Assuming γxy = γxz = 0 leads to

θy = −w0/x
(1.137)
θz = v0/x

Therefore, the displacement field is described by



⎨u(x, y, z, t) = u0 (x, t) − zw0/x (x, t) − yv0/x (x, t)

v(x, y, z, t) = v0 (x, t) (1.138)

w(x, y, z, t) = w0 (x, t)

and the out-of-plane normal stress εx is given by

εx = u0/x − zw0/xx − yv0/xx (1.139)

In this case, the internal virtual work reduces to


6
δWint = δu0/x (EAu0/x − ESy w0/xx + ESz v0/xx ) dx

6
+ δw0/xx (−ESy u0/x + EJy w0/xx + EJyz v0/xx ) dx (1.140)
6ℓ
+ δv0/xx (ESz u0/x + EJyz w0/xx + EJz v0/xx ) dx

2 This is a requirement to avoid coupling between bending and twisting.


26 REVIEW OF CONTINUUM MECHANICS

or, in matrix form,


⎧ ⎫T ⎡ ⎤⎧ ⎫
⎨ δu0/x ⎪
6 ⎪ ⎬ EA −ESy ESz ⎪ ⎨ u0/x ⎪⎬
δw0/xx ⎣−ESy EJy EJyz ⎦ w0/xx dx (1.141)
⎢ ⎥
ℓ⎪ ⎪ ⎪ ⎪
δv0/xx ESz EJyz EJz v0/xx
⎩ ⎭ ⎩ ⎭

According to the assumed displacement field in Eq. (1.138) and the coincidence of the reference axis with the mass
axis of the beam, the external virtual work due to inertia becomes
6 6 6
in
δWext = − δu0 mü0 dx − δv0 mv̈0 dx − δw0 mẅ0 dx
6ℓ ℓ
6ℓ (1.142)
− δw0/x (Iy ẅ0/x + Iyz v̈0/x ) dx − δv0/x (Iyz ẅ0/x + Iz v̈0/x ) dx
ℓ ℓ

or, in matrix form,


⎧ ⎫T ⎡ ⎤⎧ ⎫

⎪ δu0 ⎪⎪ m 0 0 0 0 ⎪
⎪ ü0 ⎪⎪
⎪ ⎪ ⎪ ⎪
⎨ δv0 ⎪ 0 m 0 0 0 ⎥⎪ v̈0 ⎪

⎪ ⎪
⎪ ⎢ ⎥⎪ ⎪
6 ⎪ ⎬ ⎢ ⎪
⎨ ⎪

⎢ ⎥
− δw0 ⎢0
⎢ 0 m 0 0 ⎥⎥⎪ ẅ 0 dx (1.143)
ℓ⎪ ⎪ ⎪
⎪δw0/x ⎪⎪ ⎣0 0 0 Iy Iyz ⎦ ⎪
⎪ẅ0/x ⎪

⎪ ⎪ ⎢ ⎥⎪⎪ ⎪


⎪ ⎪
⎪ ⎪ ⎪

δv0/x 0 0 0 Iyz Iz v̈0/x
⎩ ⎭ ⎩ ⎭

The external virtual work due to applied distributed loads is expressed by


6 6 6
load
δWext = δu0 qx dx + δv0 qy dx + δw0 qz dx
ℓ ℓ
6 6ℓ (1.144)
− δw0/x (my + za qx ) dx + δv0/x (mz − ya qx ) dx
ℓ ℓ

1.16.3 Decoupled dynamics


In previous sections, we have written the principle of virtual work for dynamic response of Timoshenko and Euler-
Bernoulli beam models having elastic axis coincident with mass axis and applied external transverse loads acting on this
reference axis. In this case, bending (flexural) behavior is decoupled from torsional response. However, it is clear from
off-diagonal terms in the cross-sectional stiffness matrix and inertial matrix (see Eqs. (1.131),(1.135) for Timoshenko
theory and (1.141),(1.143) for Euler-Bernoulli theory) that there is a coupling between axial (longitudinal) and bending
behavior and between bending effects in the two orthogonal planes xy and xz.
The longitudinal dynamic response can be decoupled from flexural dynamic behavior if
6 6
ESy = Ez dA = 0 ESz = Ey dA = 0 (1.145)
A A

This conditions is satisfied if the x reference axis coincides with the centroidal axis.
A decoupling of bending behavior between two orthogonal planes is obtained if
6 6
EJyz = Eyz dA = 0 Iyz = ρ0 yz dA = 0 (1.146)
A A

If we assume that such conditions are satisfied, we can write the PVW for longitudinal and flexural dynamics of beams
separately.
Note that, in the following, a beam undergoing decoupled longitudinal vibration will be called rod, i.e., a rod is a
beam subject only to extensional motion.

1.16.4 The rod problem


Let us consider a rod of length ℓ subjected to a distributed longitudinal load of magnitude qx = qx (x, t) per unit length
and two axial concentrated loads N0 = N0 (t) and Nℓ = Nℓ (t) at the left (x = 0) and right (x = ℓ) end points of the rod,
respectively (see Figure 1.5).
STRENGTH OF MATERIALS THEORIES 27
replacements
N0 (t) Nℓ (t)

EA(x), m(x), ℓ qx (x, t)

Figure 1.5 The rod problem.

The internal virtual work due to the longitudinal deformation of the rod is given by
6 ℓ
δWint = δu/x EA(x)u/x dx (1.147)
0

where u = u(x, t) is the longitudinal displacement of the reference axis whose origin coincides with the left end of the
rod. The external virtual work is the sum of the work due to the loads qx , N0 and Nℓ and the work due to the longitudinal
inertia of the rod 6 ℓ 6 ℓ
δWext = δu [qx − N0 δ(x) + Nℓ δ(x − ℓ)] dx − δu m(x)ü dx (1.148)
0 0
Note that the concentrated loads can be treated in the same way as distributed loadings by introducing the Dirac delta
function δ(x).
The equation of the principle of virtual work δWint = δWext is then expressed in a concise way as follows
6 ℓ 6 ℓ 6 ℓ
δu/x EAu/x dx = δu qx dx − δu(0)N0 + δu(ℓ)Nℓ − δu mü dx (1.149)
0 0 0

where we have used the property


6 +∞
f (x)δ(x − x0 ) dx = f (x0 )
−∞
It is worth noticing that the boundary forcing terms N0 and Nℓ should not be strictly considered as derived only by
external applied loads. Indeed, they represent the resulting axial forces on the ends of the rod. For example, consider the
same problem introduced thus far but now the rod is connected at x = ℓ to a mass and a grounded spring. The internal
virtual work remains the same, while the external virtual work reads
6 ℓ 6 ℓ 6 ℓ
δWext = δu qx dx − δu N0 δ(x) dx − δu M ü δ(x − ℓ) dx
0 0 0
6 ℓ 6 ℓ (1.150)
− δu Ku δ(x − ℓ) dx − δu mü dx
0 0

where M and K are the mass and spring values, respectively. The external virtual work can be also written as follows
6 ℓ 6 ℓ
δWext = δu qx dx − δu(0)N0 + δu(ℓ) [−M ü(ℓ) − Ku(ℓ)] − δu mü dx (1.151)
0 0

where you can see that the (reaction) force Nℓ is the sum of elastic and inertia forces due to the connected spring-mass
system.

1.16.5 The beam problem


Let us now consider the problem of flexural motion of a prismatic beam of length ℓ. Suppose that the beam is subjected in
the xz plane to a distributed transverse load qz = qz (x, t) per unit length, a distributed bending moment my = my (x, t)
per unit length, two boundary concentrated transverse loads FL and FR and two boundary concentrated bending loads
ML and MR (see Figure 1.6).
Let first assume that the beam is slender, so that shear deformations can be considered to be negligible. Therefore,
we can write the principle of virtual work of the beam according to Euler-Bernoulli theory. The internal virtual work is
expressed as follows
6 ℓ
δWint = δw/xx EJy (x)w/xx dx (1.152)
0
28 REVIEW OF CONTINUUM MECHANICS

replacements
z

qz (x, t)

ML (t) FR (t)
MR (t)
x
my (x, t)
FL (t)
EJy (x), m(x), ℓ

Figure 1.6 The beam problem.

where w = w(x, t) is the transverse deflection of the elastic axis. The external virtual work may be written as
6 ℓ
δWext = δw [qz − FL δ(x) + FR δ(x − ℓ)] dx
0
6 ℓ
− δw/x [my + ML δ(x) − MR δ(x − ℓ)] dx (1.153)
0
6 ℓ 6 ℓ
− δw m(x)ẅ dx − δw/x Iy (x)ẅ/x dx
0 0

Therefore, the principle of virtual work for flexural dynamics in the xz plane of an Euler-Bernoulli beam having
coincidence of elastic axis and mass axis is given by
6 ℓ 6 ℓ 6 ℓ
δw/xx EJy w/xx dx = δw qz dx − δw/x my dx − δw(0) FL + δw(ℓ) FR
0 0 0
6 ℓ 6 ℓ
(1.154)
− δw/x (0) ML + δw/x (ℓ) MR − δw mẅ dx − δw/x Iy ẅ/x dx
0 0

It is worth noting that often the contribution of the rotary inertia (last term in Eq. (1.154)) is neglected3 .
As discussed before for the longitudinal case, the boundary forces and moments can represent the reaction resultants
arising from what is actually connected to the ends of the beam. Consider for example that the ends of the beam carry
masses and are supported by springs. In this case, the external virtual work is given by
6 ℓ 6 ℓ
δWext = δw qz dx −
δw/x my dx − δw(0) [KlL w(0) + ML ẅ(0)]
0 0
; <
+ δw(ℓ) [−KlR w(ℓ) − MR ẅ(ℓ)] − δw/x (0) KrL w/x (0) + IL ẅ/x (0)
6 ℓ
; <
+ δw/x (ℓ) −KrR w/x (ℓ) − IR ẅ/x (ℓ) − δw m(x)ẅ dx
0

where KlL and KlR are the linear springs, KrL and KrR are the rotational springs, ML and MR and the end masses,
and IL and IR are the end mass moments of inertia.

Let us consider again the case presented before, but now we want to write the principle of virtual work for planar bending
motion including the effects of shear deformation. The internal and external virtual work are given, respectively, by
6 ℓ 6 ℓ 6 ℓ
δWint = δθy/x EJy (x)θy/x dx + δw/x GAz (x)(w/x + θy ) dx + δθy GAz (x)(w/x + θy ) dx (1.155)
0 0 0

3 Inthe beam-related literature, the Euler-Bernoulli beam model is typically presented without the contribution of the rotary inertia. The model which
includes the inertia due to the axial displacement of the beam is often referred to as Rayleigh’s theory.
STRENGTH OF MATERIALS THEORIES 29

T0 (t) GJ(x), Ip (x), ℓ

mx (x, t)

Tℓ (t) x

Figure 1.7 The bar problem.

and
6 ℓ 6 ℓ
δWext = δw [qz − FL δ(x) + FR δ(x − ℓ)] dx + δθy [my + ML δ(x) − MR δ(x − ℓ)] dx
0 0
6 ℓ 6 ℓ
(1.156)
− δw m(x)ẅ dx − δθy Iy (x)θ̈y dx
0 0

Therefore, the principle of virtual work for planar flexural dynamics of a Timoshenko beam having coincidence of
elastic axis and mass axis is given by
6 ℓ 6 ℓ 6 ℓ
δθy/x EJy θy/x dx + δw/x GAz (w/x + θy ) dx + δθy GAz (w/x + θy ) dx =
0 0 0
6 ℓ 6 ℓ
δw qz dx + δθy my dx − δw(0) FL + δw(ℓ) FR + δθy (0) ML − δθy (ℓ) MR (1.157)
0 0
6 ℓ 6 ℓ
− δw mẅ dx − δθy Iy θ̈y dx
0 0

1.16.6 Torsion of bars


The motion that is excluded from the beam theories presented thus far is the twisting of the beam cross-section due to
torsional loadings. This type of behavior is briefly reviewed in this section. For the sake of clarity, first the case of
uncoupled torsional dynamics is treated. A beam that is loaded only in torsion will be referred here as a bar. Note also
that only the beam model based on the assumption of uniform torsion will be discussed. Uniform torsion, known also as
Saint-Venant’s theory of torsion, basically assumes that the bar ends are loaded only by two equal and opposite twisting
moments and that they are free to warp without any form of restraint. Clearly, the actual situation is not as simple as
the restricted cases listed above. Think for instance to the common case of a bar fixed at one end and loaded by a non-
uniform distributed torque along its length. However, the effects of varying twisting moments and of warping constraint
are relatively small for long bars. We will restrict our analysis to such cases where the approximation of uniform torsion
is still satisfactory.
Let us consider a bar of length ℓ and cross-section A, subjected to an external distributed torque of magnitude mx =
mx (x, t) per unit length and two concentrated torques TL = TL (t) and TR = TR (t) at the end points (see Fig. 1.7).
According to the Saint-Venant’s torsion theory of bars with free warping, the assumed displacement field is given by

⎨s1 = u = ψ(y, z)θ/x

s2 = v = −zθ (1.158)

s3 = w = yθ

30 REVIEW OF CONTINUUM MECHANICS

where θ = θ(x, t) is the angular displacement (twist) of the cross section of the beam and ψ(y, z) is a warping function
which describes the warping out of the cross section after twisting. Since in the Saint-Venant’s theory of torsion, out-
of-plane warping is allowed to occur without restraint, we assume that εx = 0. The non-null deformation and stress
components reduce to
( )
γxy = ψ/y − z θ/x
( ) (1.159)
γxz = ψ/z + y θ/x

and
( )
τxy = G ψ/y − z θ/x
( ) (1.160)
τxz = G ψ/z + y θ/x

respectively.
The internal virtual work due to torsional deformation is given by
6
δWint = (δγxy τxy + γxz τxz ) dV
Vbar
6 ℓ 6 9( (1.161)
)2 ( )2 :
= δθ/x G ψ/y − z + ψ/z + y dA θ/x dx
0 A

The external virtual work is the sum of the work due to the torques mx , TL and TR and the work due to the torsional
inertia
6 ℓ
δWext = δθ [mx − TL δ(x) + TR δ(x − ℓ)] dx
0
6 ℓ 6 6 ℓ 6 (1.162)
2
( 2 2
)
− δθ/x ρ ψ dA θ̈/x dx − δθ ρ y + z dA θ̈ dx
0 A 0 A

Therefore, the principle of virtual work for torsional dynamics of a bar according to Saint-Venant’s theory reads
6 ℓ 6 ℓ
δθ/x GJ(x)θ/x dx = δθ mx dx − δθ(0) TL + δθ(ℓ) TR
0 0
6 ℓ 6 ℓ
(1.163)
− δθ/x Iψ (x)θ̈/x dx − δθ Ip (x)θ̈ dx
0 0

where 6 9( )2 ( )2 :
GJ = GJ(x) = G ψ/y − z + ψ/z + y dA (1.164)
A

is the torsional rigidity of the bar, 6


ρ y 2 + z 2 dA
( )
Ip = Ip (x) = (1.165)
A

is the polar mass moment of inertia per unit length, and


6
Iψ = Iψ (x) = ρ ψ 2 dA (1.166)
A

is the mass moment of inertia due to warping of the cross-section.


In many models, the axial inertia term due to the warping of the cross section is neglected. In this case, the PVW
simplifies as follows
6 ℓ 6 ℓ 6 ℓ
δθ/x GJθ/x dx = δθ mx dx − δθ(0) TL + δθ(ℓ) TR − δθ Ip θ̈ dx (1.167)
0 0 0
STRENGTH OF MATERIALS THEORIES 31

What observed before about forcing boundary terms in the longitudinal and flexural dynamics of rods and beams
applies as well to torsional behavior. For example, if the bar is connected to torsional grounded springs at both ends, the
external virtual work is given by
6 ℓ 6 ℓ 6 ℓ
δWext = δθ mx dx − δθ KL θ δ(x) dx − δθ KR θ δ(x − ℓ) dx
0 0 0
6 ℓ 6 6 ℓ 6
ρψ 2 dA θ̈/x dx − ρ y 2 + z 2 dA θ̈ dx
( )
− δθ/x δθ
0 A 0 A

or, using the property of Dirac delta function and neglecting axial inertia,
6 ℓ 6 ℓ
δWext = δθ mt dx − δθ(0) KL θ(0) − δθ(ℓ) KR θ(ℓ) − δθ Ip θ̈ dx
0 0

1.16.7 Coupled flexural-torsional dynamics


Let us consider now a beam of length ℓ with elastic axis (taken as the beam reference axis x) different from the mass
axis. The beam is subjected to a resultant distributed transverse load qz = qz (x, t) per unit length and a distributed
torque mx = mx (x, t) per unit length. It is evident that, since the load due to the beam inertia does not pass through the
shear center axis, the beam will undergo both flexural and torsional deflections. According to Euler-Bernoulli theory and
Saint-Venant’s assumptions for torsion, the internal virtual work is the sum of flexural and torsional deformation:
6 ℓ 6 ℓ
δWint = δw/xx EJy (x)w/xx dx + δθ/x GJ(x)θ/x dx (1.168)
0 0

where w = w(x, t) is the transverse deflection of the elastic axis and θ = θ(x, t) is the twist about elastic axis. The
external virtual work is due to external distributed load and inertia of the beam
6 ℓ 6 ℓ 6 ℓ 6 ℓ
δWext = δw qz dx + δθ mx dx − δwG mẅG dx − δθ Ip θ̈ dx (1.169)
0 0 0 0

where wG is the transverse deflection of the mass axis. If the distance between the elastic axis and mass axis is denoted
by e, we have for small rotations that wG = w0 + eθ. Therefore, the principle of virtual work reads
6 ℓ 6 ℓ 6 ℓ 6 ℓ
δw/xx EJy w/xx dx + δθ/x GJθ/x dx = δw qz dx + δθ mx dx
0 0 0 0
6 ℓ 6 ℓ 6 ℓ
(1.170)
− δw m(ẅ + eθ̈) dx − δθ me(ẅ + eθ̈) dx − δθ Ip θ̈ dx
0 0 0

where it is clear that there is a coupling between flexural and torsional dynamics.

1.16.8 Thin plates


In this section, the principle of virtual work for thin plate problems is presented. Thin plates are historically referred
to flat structural components characterized by one dimension, the thickness, that is very much less than the other two
dimensions. In such a view, the thinness property is referred to a pure geometrical property of the structure. Accordingly,
the mechanical behaviour of thin plates can be represented assuming a plane-stress state (see section 1.14.4.1). Indeed,
models based on the so-called thin-plate assumptions give acceptable accuracy for small thickness-to-length ratios.
The above definition of thin plates should be properly extended when the dynamic response is of interest. Broadly
speaking, we can say that a plate can be considered to be thin if the characteristic wavelength of the vibration field is
much greater than its thickness dimension. Accordingly, a plate with small thickness-to-length ratio but vibrating at
relatively high frequency cannot be considered a thin plate and more advanced models than those based on the thin-plate
theory must be used to accurately represent its mechanical behaviour.
In the following, plates made of perfectly bonded layers of orthotropic materials (laminated composite plates) are
considered. Equations for isotropic plates will be derived as a special case. The plate occupies the domain A in the
32 REVIEW OF CONTINUUM MECHANICS

z
y

x
h

Figure 1.8 The plate problem.

(x, y) plane corresponding to the plate middle surface. The plate midplane is taken as the reference surface. The plate
consists of Nℓ layers, which are assumed to be homogeneous and made of orthotropic material. The thickness of the
plate and the k-th layer are denoted by h and hk , respectively. The k-th layer is located between the points z = zk and
z = zk+1 in the thickness direction. Displacements of the plate reference surface along the x, y and z directions are
indicated by u0 = u0 (x, y, t), v0 = v0 (x, y, t) and w0 = w0 (x, y, t), respectively. The plate is assumed to be subjected
to a distributed transverse load q = q(x, y, t) per unit area.

z
plate top h
z= 2
layer k + 1

layer k

(x, y)
plane
layer k − 1

z = − h2
plate bottom

Figure 1.9 A laminated plate.

Analogously to what presented for beams, approximations regarding the actual deformation are introduced in order
to avoid the complexity involved in the theory of elasticity. A strength of materials theory is assumed based on the so-
called Kirchhoff hypothesis which can be viewed as the extension of the Euler-Bernoulli beam theory to two-dimensional
problems. It is noted that the prescribed kinematic behavior is compatible with the plane stress assumptions.
According to Kirchhoff’s theory, the displacement of a generic point is referred to the displacement of the plate surface
by the following assumed kinematic field ⎧
⎨s1 = u = u0 − zw0/x

s2 = v = v0 − zw0/y (1.171)

s3 = w = w0

where, just as is the case of Euler-Bernoulli beam approximation, the rotations are the same as the bending slopes. The
transverse displacement w is considered to be independent of thickness coordinate z. This is is consistent with the
hypothesis of neglecting thickness stretching of the plate due to its thinness. In view of the assumed displacements, the
non-null strains are
⎧ ⎫ ⎧ ⎫ ⎧ ⎫ ⎧ ⎫ ⎧ ⎫
0 1
⎨ εx ⎪
⎪ ⎬ ⎪ ⎨ εx ⎪⎬ ⎨ εx ⎪
⎪ ⎬ ⎨ u0/x
⎪ ⎪
⎬ ⎨ −w0/xx ⎪
⎪ ⎬
ϵ = εy = ε0y + z ε1y = v0/y + z −w0/yy (1.172)
⎩ ⎪
⎪ ⎭ ⎪ ⎩ 0 ⎪ ⎪
⎩ 1 ⎪ ⎪ ⎪ ⎪ ⎪
γxy γxy γxy u0/y + v0/x −2w0/xy
⎭ ⎭ ⎩ ⎭ ⎩ ⎭
STRENGTH OF MATERIALS THEORIES 33

where ϵ0 = (ε0x , ε0y , γxy


0
) are the membrane strains and ϵ1 = (ε1x , ε1y , γxy
1
) are the flexural strains or curvatures. The
corresponding components of the stress tensor are function of the strains through the following relation:

⎧ ⎫(k) ⎡ ⎤(k) ⎧ ⎫
⎨ σx ⎪
⎪ ⎬ Q̃11 Q̃12 Q̃16 ⎨ εx ⎪
⎪ ⎬
σ (k) = σy = ⎣Q̃12 Q̃22 Q̃26 ⎦ εy (1.173)
⎢ ⎥

⎩ ⎪ ⎪
⎩ ⎪
τxy Q̃16 Q̃26 Q̃66 γxy
⎭ ⎭

where
(k) (k) (k) (k) (k)
Q̃11 = Q11 c4k + 2(Q12 + 2Q66 )c2k s2k + Q22 s4k
(k) (k) (k) (k)
Q̃12 = (Q11 + Q22 − 4Q66 )c2k s2k + Q12 (c4k + s4k )
(k) (k) (k) (k) (k)
Q̃22 = Q11 s4k + 2(Q12 + 2Q66 )c2k s2k + Q22 c4k
(k) (k) (k) (k) (k) (k) (k)
Q̃16 = (Q11 − Q12 − 2Q66 )sk c3k + (Q12 − Q22 + 2Q66 )s3k ck
(k) (k) (k) (k) (k) (k) (k)
Q̃26 = (Q11 − Q12 − 2Q66 )s3k ck + (Q12 − Q22 + 2Q66 )sk c3k
(k) (k) (k) (k) (k) (k)
Q̃66 = (Q11 + Q22 − 2Q12 − 2Q66 )s2k c2k + Q66 (s4k + c4k )

and sk = sin θ(k) , ck = cos θ(k) . Qij are the reduced stiffness coefficients previously introduced within the plane-stress
assumption. The internal virtual work is given by
6 66 6 h/2
δWint = δϵT σ dV0 = δϵT σ dz dy dx
V0 A −h/2
Nℓ 6
66 = zk+1
(1.174)
T (k)
= δ (ϵ0 + zϵ1 ) σ dz dy dx
A k=1 zk

In more compact notation,


66 66
δWint = δϵT0 N dy dx + δϵT1 M dy dx (1.175)
A A

where
⎧ ⎫ ⎧ ⎫(k)
⎨ Nx ⎪
⎪ ⎬ = ⎨ σx ⎪
Nℓ 6 zk+1 ⎪ ⎬ =Nℓ 6 zk+1
N = Ny = σy dz = σ (k) dz (1.176)
⎪ ⎭ k=1 zk
⎪ ⎪
⎩ ⎪ k=1 zk
Nxy τxy
⎩ ⎭

and
⎧ ⎫ ⎧ ⎫(k)
⎨ Mx ⎪
⎪ ⎬ = ⎨ σx ⎪
Nℓ 6 zk+1 ⎪ ⎬ =Nℓ 6 zk+1
M = My = σy zdz = σ (k) z dz (1.177)
⎪ ⎭ k=1 zk
⎪ ⎩ ⎪

k=1 z k
Mxy τxy
⎩ ⎭

are the membrane, bending and twisting stress resultants. Using Eq. (1.173) yields

Nℓ 6
= zk+1
N= Q̃(k) (ϵ0 + zϵ1 ) dz = Aϵ0 + Bϵ1 (1.178)
k=1 zk

Nℓ 6
= zk+1
M= Q̃(k) z (ϵ0 + zϵ1 ) dz = Bϵ0 + Dϵ1 (1.179)
k=1 zk
34 REVIEW OF CONTINUUM MECHANICS

where
Nℓ
(k)
=
Aij = Q̃ij (zk+1 − zk )
k=1
N

1= (k) ( 2
− zk2
)
Bij = Q̃ij zk+1 (1.180)
2
k=1
N

1= (k) ( 3
− zk3
)
Dij = Q̃ij zk+1
3
k=1

are the stiffness coefficients arising from the piecewise integration over the thickness of the thin plate. Therefore, the
internal virtual work can be expressed in a compact form as follows
66 66
δWint = δϵT0 (Aϵ0 + Bϵ1 ) dy dx + δϵT1 (Bϵ0 + Dϵ1 ) dy dx (1.181)
A A

The external virtual work is given by the summation of the work done by external loads and the work done by inertial
forces
load in
δWext = δWext + δWext (1.182)
The expression of the virtual work due to inertia follows from the assumed displacement field
66 = Nℓ 6 zk+1 9
in
δu0 − zδw0/x ρ(k) ü0 − z ẅ0/x
( ) ( )
δWext =−
A k=1 zk
:
+ δv0 − zδw0/y ρ(k) v̈0 − z ẅ0/y + δw0 ρ(k) ẅ0 dzdydx
( ) ( )

After performing the calculations, we obtain


66
in
δWext =− [δu0 mü0 + δv0 mv̈0 + δw0 mẅ0
A
(1.183)
− δu0 I1 ẅ0/x − δw0/x I1 ü0 + δw0/x I2 ẅ0/x
<
−δv0 I1 ẅ0/y − δw0/y I1 v̈0 + δw0/y I2 ẅ0/y dy dx

where
Nℓ
=
m= ρ(k) (zk+1 − zk )
k=1
N

1=
ρ(k) zk+1
( 2
− zk2
)
I1 = (1.184)
2
k=1
N

1=
ρ(k) zk+1
( 3
− zk3
)
I2 =
3
k=1

The laminated plate is assumed to be loaded by a transverse distributed load q(x, y, t) and subjected to specified
stress components (σ̄nn , σ̄ns , σ̄nz ) on the portion Γσ of the boundary, where n and s denote the normal and tangential
directions on the boundary, respectively. The corresponding virtual work is given by
66 6 6 h/2
load
δWext = δwq dydx + [δun σ̄nn + δus σ̄ns + δwσ̄nz ] dzds
A Γσ −h/2

where δun and δus are the virtual displacements along the normal and tangential directions, respectively, on the boundary
Γ. Using the assumed kinematic field in Eq. (1.171) yields
66 6 6 h/2
load
; <
δWext = δw0 q dydx + δu0n σ̄nn − δw0/n σ̄nn z + δu0s σ̄ns − δw0/s σ̄ns z + δw0 σ̄nz dzds
A Γσ −h/2
STRENGTH OF MATERIALS THEORIES 35

where δu0n and δu0s are the virtual displacements of the mid-surface. Introducing the following stress resultants
6 h/2 6 h/2
N̄nn = σ̄nn dz N̄ns = σ̄ns dz
−h/2 −h/2

6 h/2 6 h/2
M̄nn = σ̄nn z dz M̄ns = σ̄ns z dz
−h/2 −h/2

and
6 h/2
Q̄n = σ̄nz dz
−h/2

the virtual work due to external loads can be written as


66 6
load
; <
δWext = δw0 q dydx + δu0n N̄nn − δw0/n M̄nn + δu0s N̄ns − δw0/s M̄ns + δw0 Q̄n ds (1.185)
A Γσ

In the case of plates made of a single isotropic layer (isotropic plates) with density ρ we have B = 0 and I1 = 0. In
addition, A16 = A26 = D16 = D26 = 0. Therefore, the internal and inertial virtual work reduce to
66
( T
δϵ0 Aϵ0 + δϵT1 Dϵ1 dy dx
)
δWint =
6 6A
; ( ) ( )
= δu0/x A11 u0/x + A12 v0/y + δv0/y A12 u0/x + A22 v0/y
A
( ) ( )<
+δu0/y A66 u0/y + v0/x + δv0/x A66 u0/y + v0/x dy dx
66
; ( ) ( ) <
+ δw0/xx D11 w0/xx + D12 w0/yy + δw0/yy D12 w0/xx + D22 w0/yy + δw0/xy 4D66 w0/xy dy dx
A

and
66
in
δWext =− [δu0 ρhü0 + δv0 ρhv̈0 + δw0 ρhẅ0 ] dy dx
A
ρh3 ρh3
66 4 5
− δw0/x ẅ0/x + δw0/y ẅ0/y dy dx
A 3 3

It is clear that, in the case of isotropic thin plates, the in-plane motion is decoupled to the out-of-plane motion. Note also
that, if we are interested in the dynamic analysis at low frequencies and the plate is very thin, the contribution due to
rotary inertia can be neglected without affecting significantly the accuracy of the results.

1.16.9 Thick plates


When the thickness dimension is not so small with respect to the side dimensions and/or we are interested in the vibration
response within a frequency range where the wavelength is not so large with respect to the thickness, the thin-plate
assumptions are no more adequate to represent with acceptable accuracy the mechanical response of the plate. In those
cases the plate is considered to be thick and the designer should rely on more advanced models.
The simplest theory of thick plates is due to Mindlin and is based on the following assumed kinematic field

⎨s1 = u = u0 + zφx

s2 = v = v0 + zφy (1.186)

s3 = w = w0

[...] TODO [...]


36 REVIEW OF CONTINUUM MECHANICS

1.16.10 Shells
[...] TODO [...]

REFERENCES

1. P. Chadwick (1999). Continuum Mechanics, Dover Publications.


2. L. E. Malvern (1969). Introduction to the Mechanics of a Continuous Medium, Prentice-Hall.
3. K. D. Hjelmstad (2005). Fundamentals of Structural Mechanics, Springer.
4. J. N. Reddy (2008). An Introduction to Continuum Mechanics, Cambridge University Press.
5. J. Bonet, R. D. Wood (1997). Nonlinear Continuum Mechanics in Finite Element Analysis, Cambridge University Press.
6. O. A. Bauchau, J. I. Craig (2009). Structural Analysis with Applications to Aerospace Structures, Springer.
7. S. Timoshenko, J. N. Goodier (1970). Theory of Elasticity, McGraw-Hill.
8. B. K. Donaldson (1993). Analysis of Aircraft Structures, McGraw-Hill.

Vous aimerez peut-être aussi