FINITE ELEMENT ANALYSIS

IN STRUCTURES

Chapter 5
PLANE STRESS
AND PLANE STRAIN

Zahit Mecitoğlu

© 2008 İstanbul Technical University

2-D PROBLEMS
„ Many problems in theory of elasticity are two
dimensional in nature.

„ Hence they can be modeled as plane stress or

plane strain.
PLANE STRESS
„ A thin plate subjected to in-plane loading acting in its
own plane, the state of stress and deformation within
the plate is called plane stress.

z
PLANE STRAIN
If a long body is subjected to transverse loading and its cross
section and loading do not vary significantly in the longitudinal
direction, a small thickness in the loaded area can be treated
as subjected to plane strain.
y
x

z p
z
Strain-Displacement Relationships
„ The displacement vector δ has two components for two
dimensional problems.
⎧u ⎫
{δ} = ⎨ ⎬
⎩v ⎭
 The strain-displacement relationships are given as follows.
⎡∂ ⎤
⎢ 0⎥
⎧ ε x ⎫ ⎢ ∂x ⎥ ⎧ ε x0 ⎫
⎪ ⎪ ⎢ ∂ ⎥ ⎧u ⎫ ⎪ ⎪ or {ε} = [∂ ]{δ} + {ε 0 }
⎨ εy ⎬ = ⎢ 0 ⎥ ⎨ ⎬ + ⎨ ε y0 ⎬
⎪γ ⎪ ⎢ ∂y ⎩v ⎭ ⎪ ⎪
⎩ xy ⎭ ∂ ∂ ⎥ γ
⎩ xy 0 ⎭
⎢ ⎥
⎢⎣ ∂y ∂x ⎥⎦
Strain-Displacement Relationships
In the case of thermal strains, the initial strain vector can
be written for the plane strain ad plane stress cases as
follows

⎧ ε x0 ⎫ ⎧1 ⎫ ⎧ ε x0 ⎫ ⎧1 ⎫
⎪ ⎪ ⎪ ⎪
{ε 0 } = ⎨ ε y 0 ⎬ = α Δ T ⎨ 1 ⎬ {ε 0 } = ⎪⎨ ε y 0 ⎪⎬ = (1 + ν )α Δ T ⎪⎨ 1 ⎪⎬
⎪γ ⎪ ⎪0 ⎪ ⎪γ ⎪
⎩ ⎭
⎪0 ⎪
⎩ ⎭
⎩ xy 0 ⎭ ⎩ ⎭ xy 0
Stress-Strain Relationships
„ Assuming an isotropic material, we shall write
relationships between stresses and strains for both plane
stress and plane strain.
(i) Plane Stress
σ z = τ zx = τ yz = 0 γ yz = γ zx = 0
 Writing strains in terms of stresses, we have:
⎧ εx ⎫ ⎡ 1 −ν 0 ⎤⎧σ x ⎫ ⎧ ε x0 ⎫
⎪ ⎪ 1⎢ ⎥⎪σ ⎪ + ⎪ ε ⎪
ε
⎨ y⎬ = ⎢ − ν 1 0 ⎥ ⎨ y ⎬ ⎨ y0 ⎬
⎪γ ⎪ E ⎢ 0 ( + ν )⎥ ⎪τ ⎪ ⎪ γ ⎪
⎩ xy ⎭ ⎣ 0 2 1 ⎦ ⎩ xy ⎭ ⎩ xy 0 ⎭
ν
εz = − (σ x + σ y ) + αΔT
E
Stress-Strain Relationships
„ Solving for stresses in terms of the strains, we find:
⎧σ x ⎫ ⎡ ⎤⎛ ⎧ ⎫ ⎧ ⎞
⎪ ⎪ E ⎢ 1 ν 0 ⎥⎜ ⎪ ε x ⎪ ⎪ ε x 0 ⎫⎪ ⎟
σ
⎨ y⎬ = ⎢ν 1 0 ⎥⎜ ⎨ ε y ⎬ − ⎨ ε y 0 ⎬ ⎟
⎪τ ⎪ (
1 − ν 2
) ⎢ 1 − ν ⎥⎜⎜ ⎪ ⎪ ⎪
γ γ ⎪ ⎟⎟
⎩ xy ⎭ ⎢ ⎥
2 ⎦⎝ ⎩ ⎭ ⎩
0 0 xy 0 ⎭ ⎠
⎣
xy

 We can write the constitutive relations as the matrix expressions:

{ε} = [C]{σ} + {ε 0 } {σ} = [E]({ε} − {ε 0 })
where ⎡ ⎤
⎡ 1 −ν 0 ⎤ ⎢ 1 ν 0 ⎥
1⎢
[C] = ⎢− ν 1 0 ⎥⎥ and [E ] = E
⎢ν 1 0 ⎥
E (
1− ν2 ) ⎢ 1− ν⎥
⎢⎣ 0 0 2(1 + ν )⎥⎦ ⎢
⎣
0 0
2 ⎦
⎥
Stress-Strain Relationships
(ii) Plane Strain
„ The case of plane strain is based on the assumptions that:
∂w
w = 0 and =0 at every cross section.
∂z
τ yz = τ zx = 0 ε z = γ zx = γ yz = 0
 When strains are written in terms of stresses, we get:
⎧ εx ⎫ ⎡1 − ν − ν 0⎤ ⎧ σ x ⎫ ⎧ ε x 0 ⎫
⎪ ⎪ 1+ ν ⎢ ⎥⎪σ ⎪ + ⎪ ε ⎪
ε
⎨ y⎬ = ⎢ − ν 1 − ν 0 ⎥ ⎨ y ⎬ ⎨ y0 ⎬
⎪γ ⎪ E ⎪τ ⎪ ⎪ γ ⎪
⎩ xy ⎭ ⎢
⎣ 0 0 2 ⎥
⎦ ⎩ xy ⎭ ⎩ xy 0 ⎭
Stress-Strain Relationships
 Asbefore we can solve for the stresses in terms of the
corresponding strains to obtain:

⎡ ⎤⎛ ⎧ ⎫ ⎧
⎧σ x ⎫ ⎢1 − ν ν 0 ⎥ εx ε x0 ⎫⎞
⎪ ⎪ E ⎜ ⎪ ⎪ ⎪ ⎪ ⎟
⎨σ y ⎬ = ⎢ ν 1− ν 0 ⎥⎜ ⎨ ε y ⎬ − ⎨ ε y 0 ⎬ ⎟
⎪τ ⎪ (1 + ν )(1 − 2ν ) ⎢ 0 1 − 2ν ⎥⎜⎜ ⎪ ⎪ ⎪
γ γ ⎪ ⎟⎟
⎩ xy ⎭ ⎢ ⎥
2 ⎦⎝ ⎩ ⎭ ⎩
0 xy 0 ⎭ ⎠
⎣
xy

σ z = ν (σ x + σ y ) − EαΔT
Stress-Strain Relationships
„ The strain-stress operator [C] is found as
⎡1 − ν − ν 0 ⎤
1+ ν ⎢ ⎥
[C] = − ν 1 − ν 0
E ⎢ ⎥
⎢⎣ 0 0 2⎥⎦

 In addition, the stress-strain operator [E] is seen from

above equation to be:
⎡ ⎤
⎢1 − ν ν 0 ⎥
[E ] = E
⎢ ν 1− ν 0 ⎥
(1 + ν )(1 − 2ν ) ⎢ 1 − 2ν ⎥
⎢ 0 0 ⎥
⎣ 2 ⎦
Stress-Strain Relationships
Orthogonally Anisotropic (Orthotropic) Material:
„ If a material is orthotropic in the x and y directions, the
operators can be written for the plane stress case as
follows:
⎡ 1 ν xy ⎤
⎢ − 0 ⎥
⎢ Ex Ey ⎥ ⎡ Ex ν xy E x 0 ⎤
⎢ ν
[C] = ⎢− xy 1
0 ⎥
⎥ [E ] = ⎢⎢ν yx E y Ey 0
⎥
⎥
⎢ Ex
⎢
Ey
1 ⎥
⎥ ⎢ 0
⎣ 0 (1 − ν xy ν yx )G xy ⎥⎦
⎢ 0 0
G xy ⎥⎦
⎣
Stress-Strain Relationships
„ Here the symbol νxy denotes the strain in the x direction
due to strain in the y direction. If we use for an
approximation shear modulus as
1 1 + ν yx 1 + ν xy
≈ +
G xy Ex Ey

we need only three independent constants.

 From the reciprocal theorem, we have
E y ν xy
=
E x ν yx
Stress-Strain Relationships
„ The case of plane strain in an orthotropic material has the
following strain-stress and stress-strain operators:
⎡ d b ⎤
⎢ − 0 ⎥
⎢ Ex Ey ⎥ ⎡aE x bE x 0 ⎤
[C] = ⎢⎢− c a
0 ⎥
⎥
[E ] = 1 ⎢⎢cE y dE y 0
⎥
⎥
⎢ Ex Ey ⎥ ad − bc
⎢ 0 1 ⎥
⎢ 0
⎣ 0 (ad − bc )G xy ⎥⎦
⎢ 0
⎣ G xy ⎥⎦

 where a = 1 − ν yz ν zy b = ν xy + ν xz ν zy
c = ν yx + ν yz ν zx d = 1 − ν xz ν zx

 From the reciprocal theorem we find Ey b

=
Ex c
STRESS TRANSFORMATION σy
σy’
τyx τy’x’ σx’
τx’y’
τxy

σx σx

τxy

σx’
τyx τx’y’
τy’x’ σy’
y’ y
σy
x’
θ
σx’ σy’
τx’y’ x
τxy
σx τy’x’
σx

τxy

τyx τyx

σy σy

 Figure shows stresses in the directions of axes x’ and y’, which are inclined at the
angle θ with axes x and y.
STRESS TRANSFORMATION
„ We are interested with the inclined stresses.

„ The stresses in the inclined axes can be obtained as

follows:
σx + σy σx − σy
σ′x = + cos 2θ + τ xy sin 2θ
2 2
σx + σy σx − σy
σ′y = − cos 2θ − τ xy sin 2θ
2 2
σx − σy
τ′xy = − sin 2θ + τ xy cos 2θ
2
STRESS TRANSFORMATION
„ It is possible to find the directions of principal stresses by
differentiating the first equation with respect to θ and setting the
results equal to zero.
dσ′x 2τ xy
=0 tan 2θ P =
dθ σx − σy

„ On the other hand, if we differentiate the third equation with respect

to θ and setting the results equal to zero, we obtain an angle for
the maximum shearing stress τ’max .

dτ x′y′ σx − σy
=0 tan 2θ S = −
dθ 2τ xy
θS = θP ± π/4
STRESS TRANSFORMATION
„ Principle normal stresses and the accompanying shear
stress can be obtained by substitution.

σx + σy ⎛ σx − σy ⎞
2

σ P1 = + ⎜⎜ ⎟⎟ + τ 2xy = σ′max
2 ⎝ 2 ⎠

σx + σy ⎛ σx − σy ⎞
2

σ P2 = − ⎜⎜ ⎟⎟ + τ 2xy = σ′min
2 ⎝ 2 ⎠
τ′xy = 0

„ Thus, the shearing stress is zero when the normal stresses

have principal values (maximum and minimum)
STRESS TRANSFORMATION
 Thefollowing expression can be found for maximum
shear stress:

⎛ σx − σy ⎞
2

τ′max = ⎜⎜ ⎟⎟ + τ 2xy
⎝ 2 ⎠

 The accompanying normal stresses as:

σx + σy
σ x ′ = σ y′ =
2
TRIANGULAR ELEMENTS
 Finite
elements with a triangular shape prove to be quite
versatile for the purpose of discretizing any two
dimensional continuum.

 Oneof the earliest and best-known elements is the

constant strain triangle described by the Turner et al.

 Inthe present section stiffnesses and equivalent nodal

loads will be developed in detail for this triangle. In
addition, a brief description will be given for the linear
strain triangles.
 TRIANGULAR ELEMENTS
y

Q2i
Q2i-1

node i

y x

x
CONSTANT STRAIN TRIANGLES (CST)
 Figure shows a constant strain triangle of thickness t,
having the following generic displacements (translations)
in the x-y plane:
{δ } = ⎣u v⎦ T

q6
(x3, y3)
3 q5

y v
q4
u (x2, y2)
(x, y) q3
q2 e 2
1 q1
(x1, y1)
x
CST ELEMENT
 Itsthree corners, points 1, 2, and 3, serve as nodes that
are numbered in a counterclockwise sequence.
 At each node there are two nodal translations in the
directions of x and y, with the former preceding the latter.
We denote the element displacement vector as,

{q} = ⎣q1 q2 q3 q4 q5 q6 ⎦
T
CST ELEMENT
A triangular element of area A appears in the Figure and any
point P(x,y) on the triangle may be located by dividing it into
subtriangles having areas A1, A2, and A3. Dimensionless area
coordinates for the triangle are defined as
A1 A2 A3
η=0 ξ= η= ς=
y
3 A A A
A1 η=1
A2

A1 + A 2 + A 3 = A
(x, y)
2
A3
1
ξ=0

ξ=1 x
ξ +η + ς = 1
This relation shows that ξ, η,
and ζ are interdependent.
CST ELEMENT
3

 Figure indicates that ξ = 1

N1
y

N1 = ξ = 1 2 at point 1 and ξ = 0 along

1 edge 2-3.
x
x  Also indicated is a linear
N2
3
variation of ξ from point 1
y N2 = η = 1
to the opposite edge, and
2
similarly for η and ζ.
1
x
x

N3 = 1-ξ−η = 1
3
N3
y

1
x
CST ELEMENT
 Theindependent shape functions are conveniently
represented by the pair ξ, η, as follows
N1 = ξ N2 = η N3 = 1 − ξ −η

 Thedisplacements inside the element are now written

using the shape functions and the nodal values of the
unknown displacement field.
u = N 1q 1 + N 2 q 3 + N 3 q 5

v = N 1q 2 + N 2 q 4 + N 3 q 6
CST ELEMENT
 The relations can be expressed in a matrix form
⎧q1 ⎫
⎪q ⎪
⎪ 2⎪
⎧u ⎫ ⎡ N 1 0 N2 0 N3 0 ⎤ ⎪⎪q 3 ⎪⎪
{δ } = ⎨ ⎬ = ⎢ ⎨ ⎬
⎩v ⎭ ⎣ 0 N1 0 N2 0 N 3 ⎥⎦ ⎪q 4 ⎪
⎪q 5 ⎪
⎪ ⎪
⎪⎩q 6 ⎪⎭
where [N] is the shape function matrix, given by
[N] = ⎡⎢
N1 0 N2 0 N3 0⎤
⎣0 N1 0 N2 0 N 3 ⎥⎦

 The displacement function can be written in a compact form

{δ } = [N]{q}
Isoparametric Formulation
A finite element is said to be isoparametric if the same
interpolation formulas define both the geometric and the
displacement shape functions.
 Such elements satisfy both geometric and displacement
compatibility conditions.
 If the geometric interpolation functions are of lower than
the displacement shape functions, the element is called
subparametric.
 If the reverse is true, the element is referred to as
superparametric.
Isoparametric Formulation
 Because isoparametric elements are usually curved, they
tend to be more suitable than subparametric elements for
modeling geometric boundary conditions.

 Inan isoparametric formulation, the generic displacements

are expressed in terms of local coordinates but the strain-
displacement relations are required differentiations with
respect to the global coordinates.

 Also,
it becomes necessary to employ numerical integration
whenever explicit integrations are impossible.
CST ELEMENT
 Forthe triangular element, the coordinates x, y can also be
represented in terms of nodal coordinates using the same
shape functions, as follows
x = N1 x 1 + N 2 x 2 + N 3 x 3
y = N 1 y1 + N 2 y 2 + N 3 y 3

 In matrix form we may write

⎧ x1 ⎫
⎪y ⎪
⎪ 1⎪
⎧x ⎫ ⎡ N1 0 N2 0 N3 0 ⎤ ⎪⎪x 2 ⎪⎪
⎨ ⎬=⎢ ⎨ ⎬
⎩y⎭ ⎣ 0 N1 0 N2 0 N 3 ⎥⎦ ⎪ y 2 ⎪
⎪x 3 ⎪
⎪ ⎪
⎪⎩ y 3 ⎪⎭
CST ELEMENT
 Using the notation, xij = xi – xj and yij = yi – yj we can write
x = x 13ξ + x 23η + x 3
y = y13ξ + y 23η + y 3

equation relates x and y coordinates to the ξ and η

 This
coordinates.
CST ELEMENT
the chain rule for partial derivatives of any function φ,
 Using
we have ∂ϕ ∂ϕ ∂x ∂ϕ ∂y
= +
∂ξ ∂x ∂ξ ∂y ∂ξ
∂ϕ ∂ϕ ∂x ∂ϕ ∂y
= +
∂η ∂x ∂η ∂y ∂η
which can be written in matrix notation as
⎧ ∂ϕ ⎫ ⎡ ∂x ∂y ⎤ ⎧ ∂ϕ ⎫
⎪⎪ ∂ξ ⎪⎪ ⎢ ∂ξ ∂ξ ⎥ ⎪⎪ ∂x ⎪⎪
⎨ ∂ϕ ⎬ = ⎢ ∂x ⎥
∂y ⎥ ⎨ ∂ϕ ⎬
⎪ ⎪ ⎢ ⎪ ⎪
⎪⎩ ∂η ⎪⎭ ⎢⎣ ∂η ∂η ⎥⎦ ⎪⎩ ∂y ⎪⎭
CST ELEMENT
 The (2 x 2) square matrix in the above equation is denoted
as the Jacobian of the transformation, [J]:
⎡ ∂x ∂y ⎤
⎢ ∂ξ ∂ξ ⎥ ⎡ x 13 y13 ⎤
[J ] = ⎢ ∂x ⎥=
∂y ⎥ ⎢⎣ x 23
⎢ y 23 ⎥⎦
⎢⎣ ∂η ∂η ⎥⎦
Also, from the above equation,
⎧ ∂ϕ ⎫ ⎧ ∂ϕ ⎫ 1 ⎡ y 23 − y13 ⎤
[Γ] = [J]
−1
=
⎪⎪ ∂x ⎪⎪ ⎡ Γ11 Γ12 ⎤ ⎪⎪ ∂ξ ⎪⎪ det[J ] ⎢⎣− x 23 x 13 ⎥⎦
⎨ ∂ϕ ⎬ = ⎢ ⎨ ⎬
⎪ ⎪ ⎣Γ21 Γ22 ⎥⎦ ⎪ ∂ϕ ⎪
⎪⎩ ∂y ⎪⎭ ⎪⎩ ∂η ⎪⎭ det[J] = x13y23 – x23y13
CST ELEMENT
 From the knowledge of the area of the triangle, it can be
seen that the magnitude of det [J] is twice the area of the
triangle.
 If the points 1, 2, and 3 are ordered in a counterclockwise
manner, det [J] is positive sign. We have
A = det[J ]
1
2

Most computer codes use a counterclockwise order for the

nodes and use det [J] for evaluating the area.
 CST - Strain-Displacement Relations
⎧ ∂u ⎫
⎪ ⎪ ⎧∂u ∂x ⎫ ⎡ Γ11 Γ12 0 0 ⎤ ⎧∂u ∂ξ ⎫
⎧εx ⎫ ⎪ ∂x ⎪ ⎡1 0 0 0⎤ ⎪ ⎡1 0 0 0⎤ ⎢ ⎥ ⎪∂u
∂u ∂y ⎪⎪ ⎢ ⎥ ⎢Γ21 Γ22 ∂η ⎪⎪
{ε } = ⎪⎨ ε y ⎪⎬ = ⎪⎨ ∂v ⎪⎬ = ⎢⎢0 0 0 1⎥⎥ ⎪⎨ ⎥ ⎪⎨
0 0
⎬ = 0 0 0 1 ⎬
⎪γ ⎪ ⎪ ∂y ⎪ ⎢0 1 1 0⎥ ⎪∂v ∂x ⎪ ⎢ ⎥⎢ 0 0 Γ11 Γ12 ⎥ ⎪∂v ∂ξ ⎪
⎩ xy ⎭ ⎪ ∂u ∂v ⎪ ⎣ ⎦ ⎪∂v ⎢⎣0 1 1 0⎥⎦ ⎢ ⎥
⎩ ∂y ⎪⎭ ⎣0 Γ21 Γ22 ⎦ ⎪⎩∂v ∂η ⎪⎭
⎪ ∂y + ∂x ⎪
0
⎩ ⎭
⎧ q1 ⎫
⎡∂N1 ∂ξ 0 ∂N 2 ∂ξ 0 ∂N 3 ∂ξ 0 ⎤ ⎪⎪q 2 ⎪⎪
⎡ Γ11 Γ12 0 0 ⎤⎢ ⎥ ⎪q ⎪
∂ ∂η ∂ ∂η ∂ ∂η ⎥ ⎪⎨ 3 ⎪⎬
= ⎢⎢ 0 ⎥ N 0 N 0 N 0
0 Γ21 Γ22 ⎥ ⎢ 1 2 3

⎢ 0 ∂N1 ∂ξ 0 ∂N 2 ∂ξ 0 ∂N 3 ∂ξ ⎥ ⎪q 4 ⎪
⎢⎣Γ21 Γ22 Γ11 Γ12 ⎥⎦ ⎢ ⎥
⎣ 0 ∂N1 ∂η 0 ∂N 2 ∂η 0 ∂N 3 ∂η ⎦ ⎪q 5 ⎪
⎪ ⎪
⎪⎩q 6 ⎪⎭
CST - Strain-Displacement Relations
⎧ q1 ⎫
⎡1 0 0 0 − 1 0 ⎤ ⎪⎪q 2 ⎪⎪
⎡ Γ11 Γ12 0 ⎤⎢
0 − 1 0 ⎥⎥ ⎪⎪q 3 ⎪⎪
0
{ε } = ⎢⎢ 0 0 Γ22 ⎥⎥ ⎢
0 0 1
Γ21 ⎨ ⎬
⎢0 1 0 0 0 − 1⎥ ⎪q 4 ⎪
⎢⎣Γ21 Γ22 Γ11 Γ12 ⎥⎦ ⎢ ⎥
⎣0 0 0 1 0 − 1⎦ ⎪q 5 ⎪
⎪ ⎪
⎪⎩q 6 ⎪⎭
⎧ q1 ⎫
⎪q ⎪
⎡ Γ11 0 Γ12 0 − (Γ11 + Γ12 ) 0 ⎤⎪ ⎪
2

⎪⎪q 3 ⎪⎪
= ⎢⎢ 0 Γ21 0 Γ22 0 ⎥
− (Γ21 + Γ22 )⎥ ⎨ ⎬
⎪ q4 ⎪
⎢⎣Γ21 Γ11 Γ22 Γ12 − (Γ21 + Γ22 ) − (Γ11 + Γ12 )⎥⎦
⎪q 5 ⎪
⎪ ⎪
⎪⎩q 6 ⎪⎭
 CST - Strain-Displacement Relations
⎧q ⎫ 1
⎪q ⎪
⎡ y 23 0 y 31 0 y12 0 ⎤⎪ 2 ⎪
1 ⎢ ⎥ ⎪⎪q 3 ⎪⎪
{ε } = 0 x 32 0 x13 0 x 21 ⎥ ⎨ ⎬
det[J ] ⎢
⎪ q4 ⎪
⎢⎣ x 32 y 23 x13 y 31 x 21 y12 ⎥⎦
⎪q 5 ⎪
⎪ ⎪
⎪⎩q 6 ⎪⎭
 The above equation can be written in matrix form as {ε } = [B]{q}
where [B] is a (3 x 6) element strain-displacement matrix relating
the three strains to the six nodal displacements and given by
⎡ y 23 0 y 31 0 y12 0 ⎤
1 ⎢ ⎥
[B] = 0 x 0 x 0 x
det[J ] ⎢
32 13 21 ⎥

⎢⎣ x 32 y 23 x 13 y 31 x 21 y12 ⎥⎦
It may be noted that all the elements of the [B] matrix are constants
expressed in terms of the nodal coordinates.
CST - Stiffness Matrix
 The element stiffness matrix given by te

[k ] = ∫ [B] [E][B]dV = [B] [E][B] ∫ dV

T T Ae

Ve Ve

∫ dV = t A e e
Ve

[k ] = t e A e [B] [E][B]
T
CST - Body Forces
fb4
y
fb6 2 fb3

b(x,y) fb5
3
fb2

x fb1
1
 The vector of body forces
⎧b x ⎫
{f b } = ∫ [N] {b}dV {b} = ⎨ ⎬
T
and dV = tedA
Ve ⎩b y ⎭
 CST - Body Forces
⎡ N1 0⎤ ⎡ N1b x ⎤
⎢0 N1 ⎥ ⎢N b ⎥
⎢ ⎥ ⎢ 1 y⎥
⎢N2 0 ⎥ ⎧bx ⎫ ⎢ N 2bx ⎥
{f b } = t e ∫ ⎢ ⎥ ⎨ b ⎬ dA = t e ∫ ⎢ N b ⎥dA
Ae ⎢
0 N2 ⎥ ⎩ y ⎭ Ae ⎢ 2 y⎥

⎢ N3 0⎥ ⎢ N 3b x ⎥
⎢ ⎥ ⎢ ⎥
⎣0 N3 ⎦ ⎢⎣ N 3b y ⎥⎦

 The force components, bx and by, can be expressed in terms

of ξ and η using the isoparametric formulation.
 CST - Body Forces
of polynomial terms in the area coordinates ξ and η
 Integrals
can be obtained as follows:
a!b!c!
∫ η ζ dA =
ξ a b c
2A e
Ae
(a + b + c + 2)!
a!b!
∫ η dA =
ξ a b
2A e
 If c = 0, it becomes Ae
(a + b + 2)!

If the body force is constant, we need the integrations of ∫ N i dA .

For N1 = ξ, we calculate the integral as follows: A e

1!0! 1
∫ dA =
ξ 2A e = A e
1

Ae
(1 + 0 + 2)! 3
CST - Body Forces
1
 Similarly, ∫A 2N dA = ∫A 3 . 3 A e
N dA =
e e

 Hence, the body force vector is given as,

⎧bx ⎫
⎪b ⎪
⎪ y⎪
Aete ⎪⎪bx ⎪⎪
{ fb } = ⎨b ⎬
3 ⎪ y⎪
⎪b ⎪
⎪ x⎪
⎩⎪by ⎭⎪
CST - Surface Tractions
 The vector of surface tractions
Φ ( x, y )
y Φ(x,y) y

x x

fs4

2 fs3
y fs6

fs5
3
fs2

fs1
1
x
CST - Surface Tractions
A traction force is a distributed load acting on the surface of
the body.
 We can integrate the traction force to obtain a distributed
load on the midline of the surface. Φ = Φ t
e
 Consider an edge 2 – 3 acted on by traction , shown in the
above figure. We have
⎡ N1 0 ⎤ ⎡ N1Φ x ⎤
{fs } = ∫ [ N ] {Φ} dl
T
⎢ 0 ⎢ ⎥
N1 ⎥ N
⎢ 1 y⎥Φ
⎢ ⎥
l2−3 ⎢ N2 0 ⎥ ⎧⎪ Φ x ⎫⎪ ⎢N Φ ⎥
{ s} ∫ ⎢
f = ⎥ ⎨Φ ⎬ dl = te ∫ ⎢ N Φ ⎥dl
2 x

⎧⎪ Φ x ⎫⎪ l2 − 3 ⎢
0 N 2 ⎥ ⎪⎩ y ⎪⎭
l2 − 3 ⎢ 2 y ⎥
{Φ} = ⎨Φ ⎬ ⎢ N3
⎢
0 ⎥
⎥
⎢
N
⎢ 3 x⎥Φ
⎥
⎪⎩ y ⎪⎭ ⎣ 0 N3 ⎦ ⎢⎣ N 3Φ y ⎥⎦
 CST - Surface Tractions
 The force components, , can be expressed in terms of ξ and
η using the isoparametric formulation.
of polynomial terms in the area coordinates ξ and
 Integrals
η can be obtained as follows:
a !b !c !
∫ η ζ dl =
ξ a b c
li- j
li- j
( a + b + c + 1)!

 Ifthe traction force is constant, we need the integrations of ∫ Ni dl

. We note here that N1 = 0 along the edge 2 - 3 and li-j

∫ N1dl = 0
l2−3
CST - Surface Tractions
 On the other hand, for N2 = η and li-j = l2-3 we calculate the
integral as follows:
0!1!0! 1
∫ ξ η ς dl = l2 − 3 = l2 − 3
0 1 0

l2 − 3
( 0 + 1 + 0 + 1)! 2
 Similarly,
0!0!1! 1
∫ N3dl = ∫ ς dl
( 0 + 0 + 1 + 1)!
l2−3 = l2−3
2
l 2 −3 l 2 −3

 Hence, the vector of surface forces is given as,

l2 − 3
{fs } =
T
⎢0 0 Φ x Φy Φx Φ y ⎥⎦
2 ⎣
CST - Thermal Forces

 We may designate the element temperature load as

{fT } = ∫ [B ] [E]{ε0 } dV
T

Ve
where {ε0} is initial strain vector due to the temperature
change, ΔT. From the theory of elasticity, {ε0} can be
represented by
⎧ε ⎫ ⎧1 ⎫ ⎧ε ⎫
⎪⎪ x0
⎪⎪ ⎧1 ⎫
⎪ ⎪ ⎪⎪ x0
⎪⎪
{ 0} ⎨ y0 ⎬
ε = ε = α Δ T ⎨1 ⎬ {ε0 } = ⎨ ε y 0 ⎬ = (1 +ν ) αΔT ⎪⎨1 ⎪⎬
⎪ ⎪ ⎪0 ⎪ ⎪ ⎪ ⎪0 ⎪
γ
⎪⎩ xy 0 ⎪⎭ ⎩ ⎭ γ
⎪⎩ xy 0 ⎪⎭ ⎩ ⎭
for plane stress for plane strain
CST - Thermal Forces
 Ifthe temperature change is constant on the element
the integration may be performed as

{fT } = ∫ [B] [E]{ε0 } dV = Aete [B] [E]{ε0 }

T T

Ve

The stresses in an element are then obtained by using

the previous equation in the form:

{σ} = [ E] ({ε} − {ε0 }) = [ E] ([ B ]{q} - {ε0 })

 Example
 Forthe two-dimensional loaded plate shown in the figure,
determine the displacements of nodes 1 and 2 and the
element stresses using plane stress conditions. Body force
may be neglected in comparison to the external forces.

500 N/mm

Thickness t = 5 mm
E = 70 GPa 200 mm

ν = 0.3 300 mm
40 kN
 Example
 SOLUTION Finite element model

y Global numbers

Q6 Q4 Local numbers
3
Q5
3
2 Q3 1 2
2
2

1
Q8 1 Q2
2 3
x 1

4 1
Q7 Q1
 Example
Connectivity Coordinates

Element Nodes Node x y

Number Number (mm) (mm)
1 2 3 Local 1 300 0
Numbers
2 300 200
1 1 2 4
Global
3 0 200
2 3 4 2 Numbers
4 0 0
 Example
 For the plane stress conditions, the material property matrix
is given by ⎡ ⎤
⎢1 ν 0 ⎥ ⎡7.69 2.31 0 ⎤
[E ] = 0 ⎥ = 10 4 ⎢⎢ 2.31 7.69 0 ⎥⎥
E
⎢ν 1
(1 −ν 2 ) ⎢ 1 −ν ⎥
⎢⎣ 0 0 2.69⎥⎦
⎢0 0 ⎥
⎣ 2 ⎦
 We found first Jacobian matrices for the elements:
⎡ x 13 y13 ⎤ ⎡300 0 ⎤
[ ]
J =⎢1
⎥ =⎢
y 23 ⎦ ⎣300 200⎥⎦
[ ]
det J1 =
300 0
300 200
= 6 x10 4 mm 2
⎣ x 23

− 300
[ ] ⎡x
J 2 = ⎢ 13
y13 ⎤ ⎡− 300
=
0 ⎤ [ ]
det J =
2 0
− 300 − 200
= 6 x10 4 mm 2
⎣ x 23 y 23 ⎥⎦ ⎢⎣− 300 − 200⎥⎦
 Example
 The strain-nodal displacement matrices [Be] can be
determined as
⎡2 0 0 0 −2 0 ⎤ ⎡− 2 0 0 0 2 0⎤
[ ]
B1 =
1 ⎢
0 − 3 0 3 0 0 ⎥ [ ]
B2 =
1 ⎢
6x10 2 ⎢
0 3 0 − 3 0 0 ⎥
⎥
6x10 2 ⎢ ⎥
⎢⎣ 3 − 2 − 3 0 0 2⎥⎦
⎢⎣− 3 2 3 0 0 − 2⎥⎦

 On performing the matrix multiplication [E][B]e , we get

⎡7.69 2.31 0 ⎤ ⎡ 2 0 0 0 −2 0 ⎤
[ ]
[E] B1 = 1 2 10 4 ⎢⎢ 2.31 7.69 0 ⎥⎥ ⎢⎢ 0 − 3 0 3 0 0 ⎥⎥
6x10
⎢⎣ 0 0 2.69⎥⎦ ⎢⎣− 3 2 3 0 0 − 2⎥⎦
⎡ 256 − 115.5 0 115.5 − 256 0 ⎤
= ⎢⎢ 77 − 384.5 0 384.5 − 77 0 ⎥⎥
⎢⎣− 134.5 89.7 134.5 0 0 − 89.7⎥⎦
 Example
 Similarly, [E][B2] is found, as follows
⎡− 256 115.5 0 − 115.5 256 0 ⎤
[ ]
[E] B 2 = ⎢⎢ − 77 384.5 0 − 384.5 77 0 ⎥⎥
⎢⎣134.5 − 89.7 − 134.5 0 0 89.7⎥⎦
 Example
 The stiffness matrices can be calculated as
⎡2 0 −3⎤
⎢0 −3 2 ⎥⎥
⎢ ⎡ 256 −115.5 0 115.5 −256 0 ⎤
⎢0 0 3 ⎥⎢
0 ⎥⎥
T
⎡ k1 ⎤ = A1t1 ⎡ B1 ⎤ [ E ] ⎡ B1 ⎤ = 30000 x5 ⎢ ⎥ ⎢ 77 −384.5 0 384.5 −77
⎣ ⎦ ⎣ ⎦ ⎣ ⎦ 6 x102 ⎢ 0 3 0⎥
⎢⎣ −134.5 89.7 134.5 0 0 −89.7 ⎥⎦
⎢ −2 0 0 ⎥
⎢ ⎥
⎢⎣ 0 0 −2 ⎥⎦
⎡ 229 −125 −101 57.8 −128.2 67.2 ⎤
⎢ −125 333 67.2 −288.4 57.8 −44.8⎥⎥
⎢
3 ⎢ −101 67.2 101 0 0 −67.2 ⎥
= 10 ⎢ ⎥
⎢ 57.8 −288.4 0 288.4 −57.8 0 ⎥
⎢ −128.2 57.8 0 −57.8 128.2 0 ⎥
⎢ ⎥
⎣⎢ 67.2 −44.8 −67.2 0 0 44.8 ⎦⎥
 Example
 Similarly, [k2] is found, as follows

⎡ 229 − 125 − 101 57.8 − 128.2 67.2 ⎤

⎢ − 125 333 67 . 2 − 288 .4 57 . 8 − 44 .8⎥
⎢ ⎥
⎢ − 101 − 67.2⎥
[ ] [ ] [ ]
k 2 = A 2 t 2 B 2 [E ] B 2 = 10 3 ⎢
T

57. 8 −
67.2
288 .4
101
0 288
0
. 4 −
0
57 .8 0
⎥
⎢ ⎥
⎢− 128.2 57.8 0 − 57.8 128.2 0 ⎥
⎢ ⎥
⎢⎣ 67.2 − 44.8 − 67.2 0 0 44.8 ⎥⎦
 Example
 In the above matrices, the global dof association is shown
on top. In the problem under consideration, Q2, Q5, Q6, Q7,
and Q8 are all zero.
 Using the elimination approach, it is now sufficient to
consider the stiffness associated with the degrees of
freedom Q1, Q3, Q4.
 The force vector due to the distributed load on the 3-1 edge
of element 2 can be calculated as:

{ }
fs2 =
l3−1
2 ⎣
⎢Φ x Φy 0 0 Φx
T
Φ y ⎥⎦ =
300
2
⎢⎣0 −500 0 0 0 −500 ⎥⎦
T
 Example
 The forces are neglected. A concentrated load of 40 kN is
applied in the direction of Q1.
 The set of equations is given by the matrix representation

⎡ 229 − 101 57.8⎤ ⎧ Q1 ⎫ ⎧ 40 ⎫

⎪ ⎪ ⎪ ⎪
10 3 ⎢⎢− 101 229 0 ⎥⎥ ⎨Q 3 ⎬ = 10 3 ⎨ 0 ⎬
⎢⎣ 57.8 0 333 ⎥⎦ ⎪⎩Q 4 ⎪⎭ ⎪− 75⎪
⎩ ⎭

 Solving for Q1, Q3 and Q4, we get

Q1 = 0.30326 mm, Q3 = 0.13361 mm, Q4 = -0.27749 mm

 Example
 For element 1, the element nodal displacement vector is
given by
{q1} = ⎣0.30326 0 0 0.13361 − 0.27749 0 0⎦T
 The element stresses σ1 are calculated from [E][B1]{q1} as
{σ }= ⎣45.742
1
− 83.398 − 22.839⎦ MPa
T

 Similarly
{q } = ⎣0
2
0 0 0 0.13361 − 0.27749⎦
T

 The element stresses σ1 are calculated from [E][B1]{q1} as

{σ } = ⎣34.258
2
10.277 − 24.903⎦ MPa
T
 LINEAR STRAIN TRIANGLE (LST)
 The linear strain triangles in the Figure have a constant
thickness t and the following generic displacements.
3 ζ=1
ξ=0 vi

y ζ=½ 5 i
ui
v η=½

{δ } = ⎣u
η=0 6
v⎦
T u
(x, y) η=1
q2
2
4
q1 ξ = ½ ζ=0
1
ξ=1

 Wewill approach this isoparametric element by first

examining its straight-sided parent.
LST ELEMENT
 Inaddition to the three corner nodes (numbered 1, 2, and 3),
there are also three midedge nodes (numbered 4, 5, and 6).
The linear strain triangle is shown in the previous figure with
its natural coordinates, and its isoparametric counterpart
element appears in this figure.
3 ζ=1
ξ=0 vi
5 i
y ζ=½ ui
η=½
6 v
u
η=0 (x, y) η=1
q2
2

q1
1 ξ=½ 4 ζ=0
ξ=1

x
LST ELEMENT
 Nodal displacements for either element are:
{δ } = ⎣q1 q2 ,..., q12 ⎦ = ⎣u 1
T
v1 ,..., v 6 ⎦
T

 Geometric interpolation functions for the parent element are

linear, as is the case for any triangle with straight sides.
Quadratic displacement shape functions may be written in
natural coordinates as:
N1 = (2ξ − 1)ξ N 4 = 4ξη
6
u = ∑ Niui
i =1 N 2 = (2η − 1)η N 5 = 4ηζ
6
v = ∑ N i vi N 3 = (2ζ − 1)ζ N 6 = 4ξζ
i =1
LST ELEMENT
 Forthe isoparametric triangle T6, we take the geometric
interpolation functions to be:
6 6
x = ∑ Ni xi y = ∑ Ni yi
i =1 i =1

 where Ni are the interpolation functions. Thus, the edges of the

element become quadratic curves, as indicated in the figure.
Because the natural coordinates are curvilinear, the Jacobian
matrix is required. Thus,
⎡ ∂x ∂y ⎤ ⎡ 6 ∂N i 6
∂N i ⎤
⎢ ∂ξ ∂ξ ⎥ ⎢ ∑ ∂ξ
xi ∑
i =1 ∂ξ
yi ⎥
[J ] = ⎢ ∂x ⎥ = ⎢ i =61
∂y ⎥ ⎢ ∂N i 6
∂N i ⎥
⎥
⎢
∂η ⎥⎦ ⎢⎣∑ ∑
xi yi
⎢⎣ ∂η i =1 ∂ξ i =1 ∂η ⎥⎦
 LST ELEMENT
 The inverse of the Jacobian [J] is denoted [Γ] and given by
1 ⎡ J 22 − J12 ⎤
[Γ] = [J ] −1
= det[J] = J11J22 – J12J21
det[J ] ⎢⎣− J 21 J11 ⎥⎦
⎧ ∂u ⎫
⎪ ⎪ ⎧∂u ∂x ⎫ ⎡ Γ11 Γ12 0 0 ⎤ ⎧∂u ∂ξ ⎫
⎧ ε x ⎫ ⎪ ∂x ⎪ ⎡1 0 0 0⎤ ⎪ ⎡ ⎤⎢
∂y ⎪⎪ ⎢ 0 ⎥⎥ ⎪⎪∂u ∂η ⎪⎪
1 0 0 0
{ε } = ⎪⎨ ε y ⎪⎬ = ⎪⎨ ∂v ⎪⎬ = ⎢⎢0 0 0 1⎥⎥ ⎪⎨
∂u
= ⎥ ⎢Γ21 Γ22 0
⎬ 0 0 0 1 ⎥⎢ 0 ⎨ ⎬
⎪γ ⎪ ⎪ ∂y ⎪ ⎢0 1 1 0⎥ ⎪∂v ∂x ⎪ ⎢ 0 Γ11 Γ12 ⎥ ⎪ ∂v ∂ξ ⎪
⎩ xy ⎭ ⎪ ∂u ∂v ⎪ ⎣ ⎦ ⎪∂v ⎢⎣0 1 1 0⎥⎦ ⎢ ⎥
+ ⎩ ∂y ⎭⎪ ⎣0 0 Γ21 Γ22 ⎦ ⎪⎩∂v ∂η ⎪⎭
⎪ ∂y ∂x ⎪
⎩ ⎭
⎧ q1 ⎫
⎡ ∂N ∂ξ 0 , ... , ∂N 6 ∂ξ 0 ⎤ ⎪⎪ q 2 ⎪⎪
⎡ Γ11 Γ12 0 ⎤⎢ 1
0 ⎥⎥ ⎪⎪ . ⎪⎪
0
∂N ∂η ∂N 6 ∂η
= ⎢⎢ 0 Γ22 ⎥⎥ ⎢ 1
0 , ... ,
0 Γ21 ⎨ ⎬
⎢ 0 ∂N1 ∂ξ , ... , 0 ∂N 6 ∂ξ ⎥ ⎪ . ⎪
⎢⎣Γ21 Γ22 Γ11 Γ12 ⎥⎦ ⎢ ⎥
⎣ 0 ∂N1 ∂η , ... , 0 ∂N 6 ∂η ⎦ ⎪ . ⎪
⎪ ⎪
⎪⎩q12 ⎪⎭
LST ELEMENT
 The above equation can be written in matrix form as
{ε } = [B]{q}
where [B] is a (3 x 12) element strain-displacement matrix
relating the three strains to the twelve nodal displacements
and given by

⎡∂N1 ∂ξ 0 , ... , ∂N 6 ∂ξ 0 ⎤
⎡ Γ11 Γ12 0 0 ⎤⎢ ⎥
∂N1 ∂η , ... , ∂N 6 ∂η
[B] = ⎢⎢ 0 ⎥ ⎢ 0 0 ⎥
0 Γ21 Γ22 ⎥
⎢ 0 ∂N1 ∂ξ , ... , 0 ∂N 6 ∂ξ ⎥
⎢⎣Γ21 Γ22 Γ11 Γ12 ⎥⎦ ⎢ ⎥
⎣ 0 ∂N1 ∂η , ... , 0 ∂N 6 ∂η ⎦
 LST ELEMENT
 The element stiffness matrix given by

[k ]12 x12 = ∫ [ B ]12 x 3 [ E ]3 x 3 [ B ]3 x12 dV = te ∫ [ B ]12 x 3 [ E ]3 x 3 [ B ]3 x12 dA

T T

Ve Ae
Integration in the Natural Coordinates
 In order to obtain any term of the stiffness matrix, the type of
integration to be performed as:
I = ∫∫ f (ξ ,η )dxdy

 However, this integral is more easily evaluated if it is first

transformed to the natural coordinates ξ and η.
 In addition the limits of each integration must be changed to
become –1 to 1; and the infinitesimal area dA = dxdy must be
replaced by an appropriate expression in terms of dξ and δη.
 For this purpose next figure shows an infinitesimal area dA in
the natural coordinates. Vector r locates a generic point in the
Cartesian coordinates x and y as follows:
Integration in the Natural Coordinates
 For this purpose following figure shows an infinitesimal area
dA in the natural coordinates. Vector r locates a generic point
in the Cartesian coordinates x and y as follows:
∂r
∂x dη
dη ξ ∂η
y ∂η

∂y ∂r
dη dξ
∂η dA ∂ξ
ξ
x ∂y
∂r dξ
r+ dη ∂ξ
∂η r ∂x
dξ

∂r
y
∂ξ
r = x + y = xi + yj
r+ dξ
j ∂ξ
x
k
i

z
Integration in the Natural Coordinates
 The rates of change of r with respect to ξ and η are:
∂r ∂x ∂y ∂r ∂x ∂y
= i+ j = i+ j
∂ξ ∂ξ ∂ξ ∂η ∂η ∂η
 When multiplied by dξ and dη, the derivatives in the above
equations form two adjacent sides of the infinitesimal
parallelogram of area dA in the figure. This area may be
determined from the following vector triple product:
⎛ ∂r ∂r ⎞
dA = ⎜⎜ dξ x dη ⎟⎟ ⋅ k
⎝ ∂ξ ∂η ⎠
Integration in the Natural Coordinates
 By substitutions we obtain
⎛ ∂x ∂y ∂x ∂y ⎞
dA = ⎜⎜ − ⎟⎟dξdη
⎝ ∂ξ ∂η ∂η ∂ξ ⎠
 The expression in the parentheses of above equation may be
written as a determinant. That is,
∂x ∂y
∂ξ ∂ξ
dA = dξ dη = J dξ dη
∂x ∂y
∂η ∂η
in which [J] is the Jacobian matrix and | J | is its determinant.
Thus, the new form of the integral becomes:
1 1
I = ∫ ∫ f (ξ ,η ) J dξdη
−1 −1
Numerical Integration
 Gaussian quadrature procedure will be applied for the
numerical integration. For triangles in natural coordinates the
numerical integration formula is [4]:
n
I = A ∑ Wj f (ξ ,η ) j
j=1

in which Wj is the weighting factor for the jth sampling point.

Integration points for n = 1, 3, and 4 appear in the next
figure, and their locations and weighting factors are given in
the following table.
Numerical Integration
3 3 3
b d

c
a a c
2 b 2
2
a

1 Linear 1 Quadratic 1 Cubic

QUADRILATERAL ELEMENTS
ISOPARAMETRIC QUADRILATERAL [ Q4 ]
 Intwo-dimensional problems the isoparametric quadrilaterals
serve a useful purpose. This element is the isoparametric
parent of the bilinear displacement rectangle, developed by
Melosh. The elements depicted in the next figure. The generic
displacements indicated in the figure are:
{δ } = ⎣u, v⎦ T

An x and a y translation are shown at each node. Thus, the

nodal displacement vector is:
{q} = ⎣q1 , q 2 ,..., q 8 ,⎦ T
= ⎣u 1 , v1 ,..., v 4 ⎦
T
Q4 ELEMENT
q8 η = y/b
q6

q7 q5

ξ = x/a
2b v
u q6
3
η
q5
q2 q3 η=1
q1 q4
2a q7 ξ
y 4 q8

η=½
v ξ=1
η=0
u
2
η = -½
ξ=½ q3
η = -1 ξ=0
q2 1 ξ = -1 ξ = -½ q4

q1
x
Q4 ELEMENT
 Now let us assume the following displacement functions
6
u = N1 u 1 + N 2 u 2 + N 3 u 3 + N 4 u 4 = ∑ N i u i
i =1
6
v = N1 u 1 + N 2 u 2 + N 3 u 3 + N 4 u 4 = ∑ N i v i
i =1

in which N1, N2, N3, and N4 are the interpolation functions

and they are given as follows:
N1 = 1
4 (1 − ξ )(1 − η ) N2 = (1 + ξ )(1 − η )
1
4

N 3 = 14 (1 + ξ )(1 + η ) N 4 = 14 (1 − ξ )(1 + η )
Q4 ELEMENT
 We may write in the matrix form:
⎧ q1 ⎫
⎧u ⎫ ⎡ N 1 0 N2 0 N3 0 N4 0 ⎤ ⎪⎪q 2 ⎪⎪
{δ } = ⎨ ⎬ = ⎢ ⎨ ⎬
⎩v ⎭ ⎣ 0 N1 0 N2 0 N3 0 N 4 ⎥⎦ ⎪ M ⎪
⎪⎩q 8 ⎪⎭
or
{δ }i = [N]i {q}i (i = 1,2,3,4)
For example, for node 1 we may write following equations:
{δ }1 = [N]1 {q}1 ⎧u 1 ⎫ ⎡ N 1
⎨ ⎬=⎢
⎩ v1 ⎭ ⎣ 0
0 ⎤ ⎧ q1 ⎫
N1 ⎥⎦ ⎩q 2 ⎭
⎡1 0⎤ ⎧ q1 ⎫
⎨ ⎬ = N1 ⎢
⎣ 0 1 ⎥ ⎨q ⎬
⎦⎩ 2 ⎭
Q4 ELEMENT
 The generic displacements {δ}i in above equation represent
the translations at any point due to the displacements {q}i at
node i. As a further efficiency of notation, we can write the
functions Ni as
N i = (1 + ξ 0 )(1 + η0 )
1
4
where
ξ0 = ξιξ η0 = ηιη (i = 1,2,3,4)
Q4 ELEMENT
 Inthe isoparametric formulation, we use the same shape
functions Ni to also express the coordinates of a point within
the element in terms of nodal coordinates. Thus,
x = N1 x 1 + N 2 x 2 + N 3 x 3 + N 4 x 4
y = N 1 y1 + N 2 y 2 + N 3 y 3 + N 4 y 4 ⎧ x1 ⎫
⎪y ⎪
⎪ 1⎪
In matrix form we may write ⎪x 2 ⎪
⎪ ⎪
⎧x ⎫ ⎡ N1 0 N2 0 N3 0 N4 0 ⎤⎪y 2 ⎪
⎨ ⎬=⎢ ⎨ ⎬
⎩y⎭ ⎣ 0 N1 0 N2 0 N3 0 N 4 ⎥⎦ ⎪ x 3 ⎪
⎪y3 ⎪
⎪ ⎪
⎪x 4 ⎪
⎪y ⎪
⎩ 4⎭
Q4 ELEMENT
 The Jacobian matrix, [J], is given as follows:
⎡ ∂x ∂y ⎤ ⎡ 4 ∂N i 4
∂N i ⎤
⎢ ∂ξ ∂ξ ⎥ ⎢ ∑ ∂ξ
xi ∑
i =1 ∂ξ
yi ⎥
[J] = ⎢ ∂x ⎥ = ⎢ i =41
∂y ⎥ ⎢ ∂N i 4
∂N i ⎥
⎥
⎢
⎢⎣ ∂η ⎢ ∑
∂η ⎥⎦ ⎣ i =1 ∂ξ
xi ∑
i =1 ∂η
yi
⎥⎦

⎡ x1 y1 ⎤
J12 ⎤ 1 ⎡− (1 − η ) (1 − η ) (1 + η ) − (1 + η )⎤ ⎢⎢ x 2 y 2 ⎥⎥
[J] = ⎡⎢
J11
⎥ = ⎢ = [D L ]{C N }
⎣J 21 J 22 ⎦ 4 ⎣ − (1 − ξ ) − (1 + ξ ) (1 + ξ ) (1 − ξ ) ⎥⎦ ⎢ x 3 y3 ⎥
⎢ ⎥
⎣x 4 y4 ⎦
Q4 ELEMENT
The inverse of the Jacobian [J] is denoted [Γ] and given by
1 ⎡ J 22 − J12 ⎤
[Γ] = [J] −1
=
det[J ] ⎢⎣− J 21 J11 ⎥⎦

To determine the derivatives of all the functions with respect to

x and y, we can apply the transformation equation repeatedly.
Hence,
⎧ ∂N i ⎫ ⎧ ∂N i ⎫
⎪⎪ ∂x ⎪⎪ ⎡ Γ11 Γ12 ⎤ ⎪⎪ ∂ξ ⎪⎪
[D G ] = ⎨ ∂N ⎬ = ⎢ ⎥ ⎨ ∂N ⎬ = [Γ ][D L ] i = 1,2,3,4
⎪ i ⎪ ⎣Γ21 Γ22 ⎦ ⎪ i ⎪
⎪⎩ ∂y ⎪⎭ ⎪⎩ ∂η ⎪⎭
Q4 ELEMENT
−1
 Altogether, we have [D G ] = [J ] [ D L ] = [ Γ ][ D L ]

 The matrix [DG] given by this expression consists of

derivatives of Ni with respect to global coordinates. That is,

⎡ ∂N1 ∂N 2 ∂N3 ∂N 4 ⎤
⎢ ∂x ∂x ∂x ∂x ⎥
⎢ ⎥
[ DG ] = ⎢ ⎥
⎢ ∂N ∂N 2 ∂N3 ∂N 4 ⎥
⎢ 1 ⎥
⎢⎣ ∂y ∂y ∂y ∂y ⎥⎦
Q4 ELEMENT
 Evaulating terms in [DG] yields:
D G11 =
1
[− (1 − η)J 22 + (1 − ξ)J12 ] D G 21 =
1
[(1 − η)J 21 − (1 − ξ)J11 ]
4J 4J

D G12 =
1
[(1 − η)J 22 + (1 + ξ)J12 ] D G 22 =
1
[− (1 − η)J 21 − (1 + ξ)J11 ]
4J 4J

D G13 =
1
[(1 + η)J 22 − (1 + ξ)J12 ] D G 23 =
1
[− (1 + η)J 21 + (1 + ξ)J11 ]
4J 4J

D G14 =
1
[− (1 + η)J 22 − (1 − ξ)J12 ] D G 24 =
1
[(1 + η)J 21 + (1 − ξ)J11 ]
4J 4J
Q4 ELEMENT
 By this approach we can find all the terms in [DG] numerically.
 Because of the appearance of the determinant of [J] in
denominator positions, we usually cannot integrate explicitly to
obtain stiffnesses and equivalent nodal loads. Instead, it
becomes necessary to use numerical integration.
 The strain-displacement relationships for the Q4 element can
be written as:
⎧ ∂u ⎫ ⎡ ∂ ⎤
⎪ ⎪ ⎢ 0⎥
⎧ ε x ⎫ ⎪ ∂x ⎪ ⎢ ∂x ⎥
{ε } = ⎪⎨ ε y ⎪⎬ = ⎪⎨ ∂v ⎪⎬ = ⎢⎢ 0 ∂ ⎥ ⎧u ⎫
⎥ ⎨ ⎬ = [∂ ]{δ } = [∂ ][N ]{q} {ε } = [B]{q}
⎪γ ⎪ ⎪ ∂y ⎪ ⎢ ∂y ⎩ v ⎭
⎩ xy ⎭ ⎪ ∂u ∂v ⎪ ∂ ∂⎥
+ ⎢ ⎥
⎪ ∂y ∂x ⎪ ⎢ ∂y ∂x ⎥⎦
⎩ ⎭ ⎣
Q4 ELEMENT
⎡∂ ⎤
⎢ 0⎥
⎢ ∂x ⎥
∂ ⎥ ⎡ N1 0 N2 0 N3 0 N4 0 ⎤
[B] = [∂ ][N] = ⎢⎢ 0
∂y ⎥ ⎢⎣ 0 N1 0 N2 0 N3 0 N 4 ⎥⎦
⎢∂ ∂⎥
⎢ ⎥
⎢⎣ ∂y ∂x ⎥⎦

 Similarly, the relations can be concisely expressed as:

{ε }i = [B]i {q}i (i = 1,2,3,4)

 Q4 ELEMENT
⎡∂ ⎤ ⎡ ∂N i ⎤
⎢ 0⎥ ⎢ 0 ⎥
⎢ ∂x ⎥ ⎢ ∂x ⎥
∂ ⎥ ⎡N i 0⎤ ⎢ ∂N i ⎥
[B]i = [∂ ][N]i = ⎢⎢ 0 = 0
∂y ⎥ ⎢⎣ 0 N i ⎥⎦ ⎢ ∂y ⎥
⎢∂ ∂⎥ ⎢ ∂N ∂N i ⎥
⎢ ⎥ ⎢ i ⎥
⎢⎣ ∂y ∂x ⎥⎦ ⎢⎣ ∂y ∂x ⎥⎦

 Referring
to the previous Eqs. we see that the submatrix [B]I
may also be written as follows:
⎡ D G1i 0 ⎤
[B]i = ⎢⎢ 0 D G 2i ⎥⎥
⎢⎣D G 2i D G1i ⎥⎦
Q4 ELEMENT
 Next, we express the stiffness matrix for Q4 element (with
constant thickness t) in natural coordinates. Thus,
1 1
[k ] = t e ∫ ∫ [B ξ ,η ]8 x 3 [E ]3x 3 [B(ξ ,η )]3x 8 J(ξ ,η )dξdη
( ) T

−1 −1

 Similarly,equivalent nodal loads due to body forces may be

written in natural coordinates as:
1 1
[f b ] = t e ∫ ∫ [N(ξ ,η )]8Tx 2 [b(ξ ,η )]2 x1 J(ξ ,η )dξdη
−1 −1
Q4 ELEMENT
 In addition, equivalent nodal loads due to initial strains in
natural coordinates are:
1 1
[f0 ] = te ∫ ∫ ⎡⎣B (ξ ,η )⎤⎦8 x3 [E]3 x3 ⎡⎣ε0 (ξ ,η )⎤⎦3 x1 J (ξ ,η ) dξ dη
T

−1 − 1
 If a temperature change ΔT varies bilinearly, it is defined as
4
ΔT = ∑ N i ΔTi
i =1

 Except in special cases, the integrals must be performed by

ZM2
numerical integration.
Slide 87

ZM2 . However, if the element is rectangular, direct explicit integration may be used. Also, line loadings with x or h constant may be handled by
explicit line integrations. Of course, if the body forces consist of point loads, no integration is required at all. Furthermore, note that the
determinant of the Jacobian matrix appears in the denominators of all the terms in matrix [B]. Therefore, the determinant [J] in Eq. (3.4-10)
cancels throughout; so the equivalent loads for initial strains may be integrated either explicitly or numerically.
ZAHIT; 3/7/2004
Q4 ELEMENT – Numerical Integration
 Forquadrilaterals in natural coordinates, the type of integration
to be performed as
1 1
I= ∫ ∫ f (ξ ,η ) J dξdη
−1 −1
 Two successive applications of Gaussian quadrature produce:

I = ∑∑ R j R k f (ξ j , ηk ) J (ξ j , ηk )
n n

k =1 j=1

 Rjand Rk are weighting factors for the point (ξj, ηk).

Integration points for n = 1,2,3 each way on a quadrilateral
are shown in the next figure.
Q4 ELEMENT – Numerical Integration
η η η 3
3 3
4 ξ
4 ξ 4 ξ

1 ±
3

2 2
2

1
1 1
n=2 n=3
n=1

Numerical Integration Constants for quadrilaterals

n ξj ηk Rj Rk

1 0 0 2 2

2 1 1 1 1
± ±
3 3
3 ±0.7745966 ±0.7745966 0.555555555 0.555555555
692 692 6 6

0.888888888 0.888888888
0.0 0.0
9 9
 Example
Derive numerically the stiffness term k12 for the isoparametric
Q4 element in the figure, using Gaussian integration with n =
2 each way. Assume that thickness t is constant .The
coordinates of nodes are given in the figure.
3 (14,14)
η
y
G3
ξ

(5,10)
4 G4

G2
G1 (20,5)
2
1
(7,1)
x
 Example
Solution:
The formula for numerical integration of terms in [k] is:

[k ] = t e ∑∑ R jR k [B(ξ j ,η k )]8Tx 3 [E]3x 3 [B(ξ j ,η k )]3x8 J(ξ j ,η k )

n n

k =1 j=1

In particular, for k12 with n = 2, Rj = Rk = 1 so that

n n
k 12 = t e ∑∑ [B]1,1 [E ]3x 3 [B]1, 2 J
T

k =1 j=1

In this formula the symbol [B]1,1 denotes the first column of

submatrix [B]1 and [B]1,2 is the second column.
 Example
Substituting these columns in above equation, we obtain
n n
k 12 = t e (E12 + E 33 )∑∑ D G11D G 21 J
k =1 j=1

To evaluate this equation, we first calculate Jacobian matrix as

⎡7 1⎤
1 ⎡− (1 − η ) (1 − η ) (1 + η ) − (1 + η )⎤ ⎢⎢20 5 ⎥⎥ 1 ⎡ 11 − 2η 4⎤
[J] = [D L ][C N ] = ⎢ = ⎢
4 ⎣ − (1 − ξ ) − (1 + ξ ) (1 + ξ ) (1 − ξ ) ⎥⎦ ⎢14 14⎥ 2 ⎣− 4 − 2ξ 9 ⎥⎦
⎢ ⎥
⎣ 5 10 ⎦

J = (115 − 18η + 8ξ )
1
Then the determinant of J is
4
 Example
Terms required from DG are:
D G11 =
1
[− (1 − η )J 22 + (1 − ξ )J12 ] = − 5 + 9η − 4ξ
4J 2(115 − 18η + 8ξ )
3(− 5 + 2η + 3ξ )
D G 21 =
1
[(1 − η )J 21 − (1 − ξ )J11 ] =
4J 2(115 − 18η + 8ξ )

Evaluating the product DG11DG21| J | at each of the four

integration points and summing the results can be obtained as:
k 12 = 0.1578t e (E12 + E 33 )
It can be finalized using numerical values of t, E12, and E33.
ISOPARAMETRIC QUADRILATERAL (Q8)
η = y/b
q8 q14
q13
q6
q5 isoparametric quadrilateral (Q8)
q7
q6
q15 q11 3
2b v η
ξ = x/a q14 q5
u
q16 q12
q12
q7 7 q13
ξ

q2 q3 4 q
q9 y 8
6
q1 q10 q4 q11
2a q15
v
u
8
q16 2
q9
rectangular parent q2 1 5 q4
q3

q10
q1
x
Q8 ELEMENT
 Inorder to understand this higher-order isoparametric element,
it is helpful to study first its rectangular parent. Nodal
displacements for either element consist of x and y
translations at each node.
{q} = ⎣q1 , q2 ,..., q16 ⎦T = ⎣u1 , v1 ,..., v8 ⎦T
 Now let us assume the following displacement functions
8 8
u = ∑ N i ui v = ∑ N i vi
i =1 i =1
Q8 ELEMENT
 The interpolation functions are given as follows
N i = (1 + ξ 0 )(1 + η0 )(ξ 0 + η0 − 1)
1
(i = 1,2,3,4)
4
N i = (1 − ξ 2 )(1 + η0 )
1
(i = 5,7)
2
N i = (1 + ξ 0 )(1 − η 2 )
1
(i = 6,8)
2
ξ 0 = ξ iξ η0 = ηiη
 This
rectangle is called a Serendipity element [1] because
nodes appear only on its edges.
Q8 ELEMENT
 For isoparametric (Q8) element, we take the geometric
interpolation functions to be the same as the displacement
shape functions. Physically, this means that the natural
coordinates ξ and η are curvilinear, and all sides of the
element become quadratic curves.
8 8
x = ∑ N i xi y = ∑ N i yi
i =1 i =1

 Theformulation of stiffnesses and equivalent nodal loads for

element Q8 is very similar to that for Q4 element given earlier.
Numerical integration also follows the same pattern as before,
even though the local coordinates are curved.
Q8 vs Q4 - COMPARISON
When higher-order elements
are available, it is not
always clear whether to
use them or to refine
network of lower-order
elements to achieve the
same accuracy.
Figure depicts a two-
dimensional continuum
discretized into Q4
elements and a smaller
number of Q8 elements.
Q8 vs Q4 - COMPARISON
In this case it seems clear that using Q8 elements instead of
four times as many Q4 elements has advantages.
 First, we see that the number of nodes is smaller when Q8
elements are used.
 Second, this type of element is able to model the curved
boundary with superior accuracy.
 Third, we can expect greater numerical accuracy in the results
because of quadratic displacement functions in place of linear
functions.
 Of course, the validity of these predictions can only be verified
by numerical experimentation with practical applications.