Lecture 1: Constant Strain Triangle

The triangular elements with different numbers of nodes are used for solving two dimensional
solid members. The linear triangular element was the first type of element developed for the
finite element analysis of 2D solids. However, it is observed that the linear triangular element is
less accurate compared to linear quadrilateral elements. But the triangular element is still a very
useful element for its adaptivity to complex geometry. These are used if the geometry of the 2D
model is complex in nature. Constant strain triangle (CST) is the simplest element to develop
mathematically. In CST, strain inside the element has no variation (Ref. module 3, lecture 2) and
hence element size should be small enough to obtain accurate results. As indicated earlier, the
displacement is expressed in two orthogonal directions in case of 2D solid elements. Thus the
displacement field can be written as
u
d     (5.1.1)
v
Here, u and v are the displacements parallel to x and y directions respectively.

5.1.1 Element Stiffness Matrix for CST

A typical triangular element assumed to represent a subdomain of a plane body under plane
stress/strain condition is represented in Fig. 5.1.1. The displacement (u, v) of any point P is
represented in terms of nodal displacements
u = N1u1 + N 2u2 + N 3u3
(5.1.2)
v = N1v1 + N 2 v2 + N 3v3
Where, N1, N2, N3 are the shape functions as described in module 3, lecture 2.

Fig. 5.1.1 Linear triangular element for plane stress/strain

The strain-displacement relationship for two dimensional plane stress/strain problem can be
simplified in the following form from three dimensional cases (eq.1.3.9 to1.3.14).

u 1  u   v  
2 2

x      
x 2  x   x  

v 1  u   v  
2 2

y      
y 2  y   y  
(5.1.3)
 
v u  u u v v 
 xy     
x y  x y x y 

In case of small amplitude of displacement, one can ignore the nonlinear term of the above
equation and will reach the following expression.
∂u
εx =
∂x
∂v
εy = (5.1.4)
∂y
∂v ∂u
γ= +
∂x ∂y
xy

Hence the element strain components can be represented as,

 ∂u ∂N ∂N ∂N
ε x = =1 u1 + 2 u2 + 3 u3
 ∂x ∂x ∂x ∂x
 ∂v ∂N1 ∂N 2 ∂N
ε=
ε y == v1 + v2 + 3 v3
 ∂y ∂y ∂y ∂y
 ∂u ∂v ∂N1 ∂N ∂N ∂N ∂N ∂N
γ xy = + = u1 + 2 u2 + 3 u3 + 1 v1 + 2 v2 + 3 v3
 ∂y ∂x ∂y ∂y ∂y ∂x ∂x ∂x
Or,

 ∂N ∂N 2 ∂N 3   u1 
 1 0 0 0  u 
 ε x   ∂x ∂x ∂x  2
   ∂N1 ∂N 2 ∂N 3  u3 
=ε =εy   0 0 0   (5.1.5)
γ   ∂y ∂y ∂y   v1 
 xy   ∂N ∂N 2 ∂N 3 ∂N1 ∂N 2 ∂N 3   v2 
 
1

 ∂y ∂y ∂y ∂x ∂x ∂x   v3 

Or, ε = [ B ] {d } (5.1.6)
In the above equation [B] is called as strain displacement relationship matrix. The shape
functions for the 3 node triangular element in Cartesian coordinate is represented as,

 

1 
           


  x 2 y 3 x 3 y 2 y 2 y 3 x x 3 x 2 y 


 N  2A 
 1    1  

 
N 2     x 3 y1  x1y3    y3  y1  x   x1  x 3  y  
     2A  

   

 3
 N   1 
   

  x1y 2  x 2 y1    y1  y 2  x   x 2  x1  y  
 2A
 

Or,
 1
 


    x   y  

 
1 1 1

 N1   2A 
     

  1
 N 2     2  2 x   2 y   (5.1.7)

 
 
 2A 


 3
 N  
  1 


  3  3 x  3 y

 2A
 

Where,
 1   x 2 y 3  x 3 y 2 ,  2   x 3 y1  x1y3 ,  3   x1y 2  x 2 y1 ,
1   y 2  y3 , 2   y3  y1 , 3   y1  y 2 , (5.1.8)
1   x 3  x 2 ,  2   x 2  x1 ,  3   x 2  x1 ,
Hence the required partial derivatives of shape functions are,
∂N1 β1 ∂N 2 β 2 ∂N 3 β3
= , = , = ,
∂x 2 A ∂x 2 A ∂x 2 A
∂N1 γ 1 ∂N 2 γ 2 ∂N 3 γ 3
= , = , = ,
∂y 2 A ∂x 2A ∂x 2 A
Hence the value of [B] becomes:
 ∂N ∂N 2 ∂N 3 
 1 0 0 0 
 ∂x ∂x ∂x 
 ∂N1 ∂N 2 ∂N 3 
[ B] =  0 0 0 
 ∂y ∂y ∂y 
 ∂N1 ∂N 2 ∂N 3 ∂N1 ∂N 2 ∂N 3 
 
 ∂y ∂y ∂y ∂x ∂x ∂x 
 β1 β2 β3 0 0 0
1 
Or, [ B] =  0 0 0 γ1 γ 2 γ 3  (5.1.9)
2A
 γ 1 γ2 γ3 β1 β 2 β3 
 According to Variational principle described in module 2, lecture 1, the stiffness matrix is
represented as,
=[k ] ∫∫∫ [ B ] [ D ][ B ] d Ω
T
(5.1.10)
Ω

Since, [B] and [D] are constant matrices; the above expression can be expressed as

=[k ] [=
B ] [ D ][ B ] ∫∫∫ d V [ B ] [ D ][ B ]V
T T
(5.1.11)
V

For a constant thickness (t), the volume of the element will become A.t . Hence the above
equation becomes,
[ k ] = [ B ] [ D ][ B ] At
T
(5.1.12)
For plane stress condition, [D] matrix will become:
 
1 µ 0 
E  
[ D] = 2 
µ 1 0  (5.1.13)
1− µ 
1− µ 
0 0 
 2 
Therefore, for a plane stress problem, the element stiffness matrix becomes,
 β1 0 γ1 
β 0 γ 2   
 2 1 µ 0   β1 β 2 β3 0 0 0
 β3 γ3   
0   0 0 γ 3 
Et 0
[k ] =   µ 1 0 γ1 γ 2 (5.1.14)
4 A (1 − µ 2 )  0 γ1 β1  
1 − µ   γ 1 γ 2 γ3 β1 β 2 β3 
0 γ2 β2   0 0 
  2 
 0 γ3 β3 
Or,
 2 (1 + µ ) β γ 
 β1 + Cγ 1 β1β 2 + Cγ 1γ 2 β1β3 + Cγ 1γ 3 µβ1γ 2 + C β 2γ 1 µβ1γ 3 + C β3γ 1 
2
1 1
 2 

β + Cγ
2 2
β 2 β3 + Cγ 2γ 3 µβ 2γ 1 + C β1γ 2
(1 + µ ) β γ 
µβ 2γ 3 + C β 3γ 2  (5.1.15)
 2 2 2 2
 2 
Et
[k ] =  (1 + µ ) β γ 
4 A (1 − µ ) 
2
β32 + Cγ 32 µβ3γ 1 + C β1γ 3 µβ 3γ 2 + C β 2γ 3 3 3 
 2 
 γ + Cβ
2
1 1
2
γ 1γ 2 + C β1β 2 γ 1γ 3 + C β1β3 
 Sym. γ 22 + C β 22 γ 2γ 3 + C β 2 β 3 
 
 γ 32 + C β32 

Where, C =
(1 − µ )
2
Similarly for plane strain condition, [D] matrix is equal to,
  
(1 − µ ) µ 0 
E  
= [ D]  µ (1 − µ ) 0  (5.1.16)
(1 + µ )(1 − 2µ )  1 − 2µ 
 0 0 
 2 
Hence the element stiffness matrix will become:
 M β12 + γ 12 M β1β 2 + γ 1γ 2 M β1β3 + γ 1γ 3 ( µ + 1) β1γ 1 µβ1γ 2 + β 2γ 1 µβ1γ 3 + β3γ 1 
 
 M β 22 + γ 22 M β 2 β 3 + γ 2γ 3 µβ 2γ 1 + β1γ 2 ( µ + 1) β 2γ 2 µβ 2γ 3 + β3γ 2 
 M β32 + γ 32 µβ3γ 1 + β1γ 3 µβ3γ 2 + C β 2γ 3 ( µ + 1) β3γ 3 
(5.1.17)
Et
[k ] =  
2 A (1 + µ )  M γ 12 + β12 M γ 1γ 2 + β1β 2 M γ 1γ 3 + β1β3 
 Sym. M γ 22 + β 22 M γ 2γ 3 + β 2 β 3 
 
 M γ 32 + β32 

Where M= (1 − µ )

5.1.2 Nodal Load Vector for CST

From the principle of virtual work,

   d    u  F d    u  F d

T T T
(5.1.18)
  

Where, FΓ, and FΩ are the surface and body forces respectively. Using the relationship between
stress-stain and strain displacement, one can derive the following expressions:
D ][ B ]{d } , δ {ε } [ B ]δ=
{σ } [ = {d } and δ {u} [ N ]δ {d }
(5.1.19)
Hence eq. (5.1.18) can be rewritten as,
 NS  F d   dT  N T F d (5.1.20)
  d B DBdd    d 
T T T T

 
  

Or,
  B  D  Bdd    N  F d    N  F d
S T (5.1.21)
T T

  
s
Here, [N ] is the shape function along the boundary where forces are prescribed. Eq.(5.1.21) is
equivalent to  k d  F , and thus, the nodal load vector becomes

F    N  F d    N  F d

T T
S
 

  (5.1.22)
For a constant thickness of the triangular element eq.(5.1.22) can be rewritten as

F  t   N  F ds  t   N  F dA

T T
S
  (5.1.23)
S A

For the a three node triangular two dimensional element, one can represent F and F as,
 Fx  Fx 
F     and F    
Fy  Fy 
Fx   0 
For example, in case of gravity load on CST element, F   
    
Fy  g
For this case, the shape functions in terms of area coordinates are:
 L1 L 2 L3 0 0 0
 N     (5.1.24)
0 0 0 L1 L 2 L3 
As a result, the force vector on the element considering only gravity load, will become,
L 0  0   0 
1
    0 
L 0  0   
2
  
L 0  0   0 
  dA  t  
0 
F  t  3
 dA  gt   dA (5.1.25)
0 L   L g   L 
A
 1
 g A
 
1 A
 
1

0 L  L g L 

     
2 2 2

0 L 
 3 L g 3 L 3

The integration in terms of area coordinate is given by,

p! q! r !
∫A L1 L2 L3dA = ( p + q + r + 2 )! 2 A
p q r
(5.1.26)

Thus, the nodal load vector will finally become


 0  
  0  

 
  
  
 



0 
  0  0
 
  
 
 
 



0 
  0   
    0

 1!0!0!  
   gAt

 
 

(5.1.27)
  
 
  
0
 
2A gAt
F  gt  (1  0  0  2)!       

 
 

3 
 3 1


 0!1!0! 
 
 gAt  
 

 2A    1

 (0  1  0  2)!   
 3  
 


 
 
 
 

1


 0!0!1!    gAt 

 2A 
 
 

 (0  0  1  2)! 

  
  3 
Lecture 2: Linear Strain Triangle

5.2.1 Element Stiffness Matrix for LST

In case of CST, it is observed that the strain within the element remains constant. Though, these
elements are able to provide enough information about displacement pattern of the element, but
it is unable to provide adequate information about stress inside an element. This limitation will
be significant enough in regions of high strain gradients. The use of a higher order triangular
element called Linear Strain Triangle (LST) significantly improves the results at these areas as
the strin inside the element is varying. The LST element has six nodes (Fig. 5.2.1) and hence,
twelve degrees of freedom. Thus the displacement function can be chosen as follows.
u =α 0 + α1 x + α 2 y + α 3 x 2 + α 4 xy + α 5 y 2
(5.2.1)
v =a6 + a7 x + a8 y + a9 x 2 + a10 xy + a11 y 2

Fig. 5.2.1 Linear strain triangle element

Therefore, the element strain matrix is obtained as

u
x   1  2 3 x   4 y
x
v (5.2.2)
y    8  10 x  211y
y
v u
 xy    ( 2   7 )  ( 4  2 9 )x  (2 5  10 )y
x y
In the area coordinate system as discussed in module 3, lecture 3 we can write the shape function
for the six node triangular element as
N1  L1 2L1 1 N 2  L 2 2L 2 1 N 3  L3 2L3 1
(5.2.3)
N 4  4L1L 2 N 5  4L 2 L3 N 6  4L3L1
The displacement (u,v) of any point within the element can be represented in terms of their nodal
displacements with the use of interpolation function.
6
u = ∑ N i ui
i =1
6
(5.2.4)
v = ∑ N i vi
i =1

Using eq.(5.2.4) we can rewrite eq.(5.2.2) as,

 u1 
u 
 2
u3 
 
 ∂N ∂N 2 ∂N 3 ∂N 4 ∂N 5 ∂N 6  u4 
 1 0 0 0 0 0 0  u 
 ∂x ∂x ∂x ∂x ∂x ∂x  5
 ∂N1 ∂N 2 ∂N 3 ∂N 4 ∂N 5 ∂N 6  u6 
ε = 0 0 0 0 0 0  
 ∂y ∂y ∂y ∂y ∂y ∂y   v1 
 ∂N1 ∂N 2 ∂N 3 ∂N 4 ∂N 5 ∂N 6 ∂N1 ∂N 2 ∂N 3 ∂N 4 ∂N 5 ∂N 6   v2 
  
 ∂y ∂y ∂y ∂y ∂y ∂y ∂x ∂x ∂x ∂x ∂x ∂x   v3 
v 
 4
 v5 
v 
 6
Or,
ε = [ B ] {d } (5.2.5)
Where,
 ∂N ∂N 2 ∂N 3 ∂N 4 ∂N 5 ∂N 6 
 1 0 0 0 0 0 0 
 ∂x ∂x ∂x ∂x ∂x ∂x 
 ∂N1 ∂N 2 ∂N 3 ∂N 4 ∂N 5 ∂N 6  (5.2.6)
[ B] =  0 0 0 0 0 0 
 ∂y ∂y ∂y ∂y ∂y ∂y 
 ∂N1 ∂N 2 ∂N 3 ∂N 4 ∂N 5 ∂N 6 ∂N1 ∂N 2 ∂N 3 ∂N 4 ∂N 5 ∂N 6 
 
 ∂y ∂y ∂y ∂y ∂y ∂y ∂x ∂x ∂x ∂x ∂x ∂x 

Using Chain rule,

∂N1 ∂N1 ∂L1 ∂N1 ∂L2 ∂N1 ∂L3
= . + . + .
∂x ∂L1 ∂x ∂L2 ∂x ∂L3 ∂x
As discussed in module 3, lecture 1, we can write the above expression as,
 ∂N1 b1 ∂N1 b2 ∂N1 b3 ∂N1
= . + . + .
∂x 2 A ∂L1 2 A ∂L2 2 A ∂L3
∂N1 b1
= . ( 4 L1 − 1)
∂x 2 A
Similarly we can evaluate expressions for other terms and can be written as,
∂N1 b1 ∂N 2 b2 ∂N 3 b3
= . ( 4 L1 − 1) = . ( 4 L2 − 1) = . ( 4 L3 − 1)
∂x 2 A ∂x 2 A ∂x 2 A
∂N 4 ∂N 5 ∂N 6
=4 ( L2b1 + L1b2 ) =4 ( L3b2 + L2b3 ) = 4 ( L1b3 + L3b1 )
∂x ∂x ∂x
And,
∂N1 a1 ∂N 2 a2 ∂N 3 a3
= . ( 4 L1 − 1) = . ( 4 L2 − 1) = . ( 4 L3 − 1)
∂y 2 A ∂y 2 A ∂y 2 A
∂N 4 ∂N 5 ∂N 6
=4 ( L2 a1 + L1a2 ) =4 ( L3 a2 + L2 a3 ) =4 ( L1a3 + L3 a1 )
∂y ∂y ∂y
Where,
a1 =
x2 − x3 a2 =
x3 − x1 a3 =
x1 − x2
b1 =
y2 − y3 b2 =
y3 − y1 b3 =
y1 − y2

The stiffness matrix of the element is represented by,

=[k ] ∫∫∫ [ B ] [ D ][ B ] d Ω
T
(5.2.7)
Ω

The, [D] matrix is the constitutive matrix which will be taken according to plane stress or plane
strain condition. The nodal strain and stress vectors are given by,
{ε n } = {ε x1 ε x 2 ε x3 ε y1 ε y 2 ε y 3 γ xy1 γ xy 2 γ xy 3}
T

(5.2.9a)
{σ n } = {σ x1 σ x 2 σ x3 σ y1 σ y 2 σ y 3 τ xy1 τ xy 2 τ xy 3}
T

(5.2.9b)

 [ Bn1 ] [ 0] 
{ε n } =  [ 0] [ Bn 2 ] {d } (5.2.10)
[ Bn 2 ] [ Bn1 ] 
Referring to section 3.3.1, using proper values of area coordinates in [B] matrix, one can find
 3b1 −b2 −b3 4b2 0 4b3 
1 
[ Bn1 ] = −b1 3b2 −b3 4b1 4b3 0  (5.2.11a)
2A
 −b1 −b 3b3 0 4b2 4b1 
And,
 3a1 −a2 −a3 4a2 0 4a3 
[ Bn 2 ] = −a1 3a2 −a3 4a1 4a3 0 
1 
(5.2.11b)
2A
 −a1 −a 3a3 0 4a2 4a1 
Thus, the element stiffness can be evaluated by putting the values from eq. (5.2.11) in eq. (5.2.7).

5.2.2 Nodal Load Vector for LST

Similar to 3-node triangular element, the load will be lumped at each node which can be
computed using the earlier expression,

F    N  F d    N  F d

T T
S
  (5.2.12)
 

And for element with constant thickness,

F  t   N  F ds  t   N  F dA
T T
S
 
(5.2.13)
S A

5.2.3 Numerical Example using CST

Determine the displacements at the nodes for the following 2D solid continuum considering a
constant thickness of 25 mm, Poisson’s ratio, µ as 0.25 and modulus of elasticity E as 2 x 105
N/mm2. The continuum is discritized with two CST plane stress elements.

Fig. 5.2.2 Geometry and discretization of the continuum

The element 1 is connected with node 1, 3 and 4 and let assume its Cartesian coordinates are (x1,
y1), (x3, y3) and (x4, y4) respectively. If we consider nodes 1, 3 and 4 are similar to node 1, 2 and
3 in eq.(5.1.9) then the [B] can be written as
  β1 β 2 β3 0 0 0 
[ B ] =  0 0 0 γ 1 γ 2 γ 3 
1 
2A
 γ 1 γ 2 γ 3 β1 β 2 β3 
By introducing values of β & γ discussed in previous lecture note, we can get value of [B] as
 0 1 −1 0 0 0 
1 
=[ B] 0 0 0 −3 0 3 
1500 
 −3 0 3 0 1 −1
For plain stress problem, putting the values of E and µ one can find the following values.
 
1 µ 0  16 4 0 
E   4 ×104  
= [ D] = µ 1
1 − µ 2 
0   4 16 0 
3
1− µ   0 0 6 
0 0 
 2 
Therefore the stiffness matrix for the element 1 will be
[ k ]1 = tA [ B ] [ D ][ B ]
T

Putting values of t, A, [B] & [D]we will get,

750 0 -750 0 -250 -250 

0 222.2222 -222.2222 -166.6667 0 166.6667 

-750 -222.2222 972.2222 166.6667 250 -416.6667 
[ k ]1 =4 ×104 × 0 -166.6667 166.6667 2000 0 -2000

 
-250 0 250 0 83.3333 -83.3333 
 
 250 166.6667 -416.6667 -2000 -83.3333 2083.3333
Similarly element 2 is connected with nodes 1, 2 and 3 and global coordinates of these nodes are
(x1, y1), (x2, y2) and (x3, y3) respectively. For this element, by proceeding in a similar manner to
element 1 we can calculate [B] matrix as,
 −1 1 0 0 0 0 
1 
= [ B]  0 0 0 0 −3 3
1500
 0 −3 3 −1 1 0 
Hence, the elemental stiffness matrix becomes,
  222.2222 -222.2222 0 0 166.6667 -166.6667 
-222.2222 972.2222 -750 250 -416.6667 166.6667 

0 -750 750 -250 250 0 
[ k ]2 =4 ×10 × 
4

0 250 -250 83.3333 -83.3333 0 
166.6667 -416.6667 250 -83.3333 2083.3333 -2000 
 
-166.6667 166.6667 0 0 -2000 2000 

By assembling the stiffness matrices into global stiffness matrix [K],

972.2222 -222.2222 0 -750 0 166.6667 -416.6667 -250 
-222.2222 972.2222 -750 0 250 -416.6667 166.6667 0 
 
0 -750 972.2222 -222.2222 -416.6667 250 0 166.6667 
 
4  -750 0 -222.2222 972.2222 166.6667 0 250 -416.6667 
[ K ] =×
4 10 ×
0 250 -416.6667 166.6667 2083.3333 -83.3333 0 -2000 
 
166.6667 -416.6667 250 0 -83.3333 2083.3333 -2000 0 
-416.6667 166.6667 0 250 0 -2000 2083.3333 -83.3333 
 
 250 0 166.6667 -416.6667 -2000 0 -83.3333 2083.3333

Now, applying equation [ F ] = [ K ]{d } , the following expression can be written.

 Fu1  972.2222 -222.2222 0 -750 0 166.6667 -416.6667 -250   u1 
F  -222.2222 972.2222 -750 0 250 -416.6667 166.6667 0  u 
 u2    2
 Fu 3  0 -750 972.2222 -222.2222 -416.6667 250 0 166.6667  u3 
    
 u4 
F 4  -750 0 -222.2222 972.2222 166.6667 0 250 -416.6667  u4 
  =× ×
  v1 
4 10
0
 Fv1  
250 -416.6667 166.6667 2083.3333 -83.3333 0 -2000
 
 Fv 2  166.6667 -416.6667 250 0 -83.3333 2083.3333 -2000 0   v2 
  -416.6667  
 Fv 3  166.6667 0 250 0 -2000 2083.3333 -83.3333   v3 
 
F   250 2083.3333  v4 
 v4  0 166.6667 -416.6667 -2000 0 -83.3333

Putting boundary conditions u=

1 v=
1 u=
2 u=
4 v=
4 0 and adopting elimination technique for
applying boundary condition we get expression,
 0  972.2222 250 0  v2 
    u 
25000  =4 × 10 ×  250
4
2083.3333 2000  3
 0     
   0 2000 2083.3333   v3 
Solving the above expression, the unknown nodal displacements may be obtained as follows.
v2 25.96 ×10−5 mm, u3 =
= −10.02 ×10−5 mm and
= v3 96.92 ×10−5 mm.
Lecture 3: Rectangular Elements

The rectangular elements are widely used for solving two dimensional continuums. The main
advantage of this type of element is the easy formulation and easy development of computer
code. The element stiffness of such elements is derived here using the concept of isoparametric
formulation.

5.3.1 Computation of Element Stiffness

In case of a four node rectangular element, the geometry and displacement filed can be expressed
in terms of their nodal values with the help of interpolation function. As the formulation will be
isoparametric, the interpolation function will become same for expressing both the variables.
Thus, coordinates and displacements at any point inside the element (Fig. 5.3.1) can be expressed
as
x  N1 x1  N 2 x2  N 3 x3  N 4 x4
(5.3.1)
y  N1 y1  N 2 y2  N 3 y3  N 4 y4
And
u  N1u1  N 2u2  N 3u3  N 4u4
(5.3.2)
v  N1v1  N 2 v2  N 3v3  N 4 v4

The above equations can be written in matrix form as


 x1 


 


 y1


x  


 2 
 


   N1
x 0 N2 0 N3 0 N4 0  y2 

    
   (5.3.3)

 
 x 0
 N1 0 N2 0 N3 0  
N 4   x3 

 

 y3 
 

 



x4 


 y4 


 

And
 
 u1 


 


 v1


 


 u 2
  

u
   N1 0 N2 0 N3 0 N4 0  v2 
      (5.3.4)

v 
  0
 N1 0 N2 0 N3 0 N 4  u3 

 
 v3 


 


 u 
4

 
v4 

 


The shape function four node rectangular element is derived and shown in module 3, lecture 4.
However the shape functions are reproduced here for easy reference for the derivation of the
stiffness matrix.

 1 x1 h 

 

 4 

 N1  
 


 
  1  x 1  h  
   
N2   

 
Ni      4
 (5.3.5)

 N3 
 
 1  x1  h

 
 
 


N4  
 4 



 1 x1  h

 


 4 


Fig. 5.3.1 Four node rectangular element

The strain-displacement relationship for two dimensional plane stress/strain problem can be
simplified in the following form from three dimensional cases (eq.1.3.9 to1.3.14).

u 1  u   v  
2 2

x      
x 2  x   x  

v 1  u   v  
2 2

y      
y 2  y   y  
(5.3.6)
 
v u  u u v v 
 xy     
x y  x y x y 
In case of small amplitude of displacement, one can ignore the nonlinear term of the above
equation and will reach the following expression.
∂u
εx =
∂x
∂v
εy = (5.3.7)
∂y
∂v ∂u
γ= +
∂x ∂y
xy

Using the shape function the above expression can be written as

 u1 
 
 ∂N1 ∂N 2 ∂N 3 ∂N 4   v1 
 0 0 0 0  u 
εx   ∂x ∂x ∂x ∂x  2
   ∂N1 ∂N 2 ∂N 3 ∂N 4   v2 
εy  = 0 0 0 0   [ B ]{d } (5.3.8)
   ∂y ∂y ∂y ∂y  u3 
γ xy   ∂N1 ∂N1 ∂N 2 ∂N 2 ∂N 3 ∂N 3 ∂N 4 ∂N 4   v3 
  
 ∂y ∂x ∂y ∂x ∂y ∂x ∂y ∂x  u4 
v 
 4
Here, [B] is known as strain displacement relationship matrix. The derivatives of the shape
functions are calculated using the chain rule.
∂N i ∂N i ∂x ∂N i ∂y
= . + .
∂ξ ∂x ∂ξ ∂y ∂ξ
(5.3.9)
∂N i ∂N i ∂x ∂N i ∂y
= . + .
∂η ∂x ∂η ∂y ∂η
Here, i is referred to number of nodes in an element and will be 4 in this case. Converting above
expression in matrix form
  N i   x y   N i   N i 
       
 x   x x   x   x 
       J   (5.3.10)
 N i   x y   N i   N i 
      
 h   h h   y   y 
The matrix [J] is referred to Jacobian matrix which is discussed in Lecture 7, module 3. Using
eq. (5.3.1) one can write
x N1 N N N
 x1  2 x2  3 x3  4 x4 (5.3.11)
x x x x x
Putting the values of the nodal coordinates and shape functions of the four node element in the
above equation the following relations will be obtained.
x 11 h 11 h 11  h 11  h a
 0  a  a  0 
x 4 4 4 4 2
(5.3.12a)
Similarly,
y 11 h 11 h 11  h 11  h
 0  0  b  b  0
x 4 4 4 4
(5.3.12b)
x 11 x  11  x  11  x  11 x 
 0  a  a  0  0 (5.3.12c)
h 4 4 4 4
y 11 x  11  x  11  x  11 x  b
 0  0  b  b 
h 4 4 4 4 2
(5.3.12d)
Substituting above values in Jacobian matrix the following relations will be obtained.
a  2 
 0  0
   
 J    2  and  J    a 
1
(5.3.13)
 b  2
0  0 
 2   b 
Thus, eq.(5.3.10) can be written as
 N i   N   N   N 
 i   2 0   i   2 . i 
 x 
    J 1  x    a   x   a x 
    (5.3.14)
 N i   N i   2   N i   2 N i 
     0    . 
 y   h   b   h   b h 
After derivation of the shape functions expressed in eq.(5.3.5), the following values will be
obtained.
 N1 2 N1 1 h N 2 1 h N 3 1  h N 4 1  h
 .  ;  ;  ; 
x a x 2a x 2a x 2a x 2a (5.3.15)
N1 2 N1 1 x N 2 1  x N 3 1  x  N 4 1 x 
 .  ;  ;  ; 
y b h 2b y 2b y 2b y 2b
So, the strain displacement relationship matrix, [B] will become as follows.
 (1 − η ) (1 − η ) (1 + η ) (1 + η ) 
− 0 0 0 − 0 
 2a 2a 2a 2a 
 (1 − ξ ) (1 + ξ ) (1 + ξ ) (1 − ξ )  (5.3.16)
[ B] =
 0 − 0 − 0 0 
 2b 2b 2b 2b 
 (1 − ξ ) (1 − η ) (1 + ξ ) (1 − η ) (1 + ξ ) (1 + η ) (1 − ξ ) (1 + η ) 
− − − − 
 2b 2a 2b 2a 2b 2a 2b 2a 

The element stiffness matrix will become

[k ] t =
∫∫ [ B ] [ D ][ B ] dx dy t ∫∫ [ B ] [ D ][ B ] J dξ dη
T T
= (5.3.17)
It is seen that the above is expressed in terms of ξ and η and hence can be numerically integrated
by the Gauss Quadrature rule. The stiffness matrix for each element can be found which needs
to be globally assembled for getting the global stiffness matrix to obtain the solution. The
stiffness matrix of higher order rectangular element can be derived in a similar fashion. For
example, in case of eight node rectangle element, the size of [B] matrix will become 16 × 3
which was 8 × 3 for four node element. Thus the size of element stiffness for eight node element
will become 16 × 16.

5.3.2 Computation of Nodal Loads

If a distributed load acts on a side of a four node rectangular element, the nodal load vector can
be calculated the similar procedure as discussed in case of triangular element. If an element as
shown below is subjected to a linearly varying intensities of load at its one side, then the
magnitude of this at any point on the side can be expressed by its interpolation function as
follows.
1 − η 1 + η  qx 2 
qx =    (5.3.18)
 2 2   qx 3 
Here, qx2 and qx3 are the force intensities per unit length at nodes 2 and 3 respectively. The load
at nodes can be calculated from the following expression.
{Fx }
= ∫ {N } q d Γ
S
2 x 1 (5.3.19)
As ξ=1 along the side 2-3, the interpolation function will become
  (1 − ξ )(1 − η ) 
   0 
 4   
 (1 + ξ )(1 − η )   (1 − η ) 

 
  2 
=N 2S ={ } 4
   (5.3.20)
 (1 + ξ )(1 + η )   (1 + η ) 
 4   2 
   
 (1 − ξ )(1 + η )   0 

 4 


If the element thickness is t, then dΓ1 =t.dl. Thus the eq.(5.3.19) can be replaced as
 0 

 (1 − η ) 
+1
  1 −η 1 + η  qx 2 
{Fx } = t ∫  2    dl (5.3.21)
−1  (1 + η )  
2 2   qx 3 
 2 
 
 0 

Fig. 5.3.2 Varying load on a four node element

After integrating the above expression, the nodal load vector along x direction will become as
follows.
 0 
 2q + q 
{Fx } =  x 2 x3 
t
(5.3.22)
3  q x 2 + 2q x 3 

 0 

Lecture 4: Numerical Evaluation of Element Stiffness

Derivation of element stiffness for a four node rectangle element has been demonstrated in last
lecture. The stiffness matrix of each element can be calculated easily by developing a suitable
computer algorithm. To help students for developing their own computer code, a numerical
example has been solved and demonstrated here.

5.4.1 Numerical Example

Calculate the stiffness matrix for the given four node rectangular element by the Gauss
Quadrature integration rule using one point and two point formula assuming plane stress
formulation. Consider, the thickness of element = 20 cm, E=2 × 103 kN/cm2 and µ =0.

Fig. 5.4.1 Element Dimension

5.4.2 Evaluation of Stiffness using One Point Gauss Quadrature

For the calculation of stiffness matrix, first, 1×1 Gauss Quadrature integration procedure has
been carried out. Thus, the natural coordinate of the sampling point will become 0,0 and weight
will become 2.0 which is shown in the figure below.

Fig. 5.4.2 Natural coordinates for one point Gauss Quadrature

For a four node quadrilateral element, the shape functions and their derivatives are as follows.

(1 − ξ )(1 − η ) (1 + ξ )(1 − η ) (1 + ξ )(1 + η ) (1 − ξ )(1 + η )

N1 = , N2 = and N 3 = and N 4 =
4 4 4 4
∂N1 − (1 − ξ ) ∂N 2 (1 − ξ ) ∂N 3 (1 + ξ ) ∂N 4 − (1 + ξ )
= ; = ; = ; =
∂ξ 4 ∂ξ 4 ∂ξ 4 ∂ξ 4
∂N1 − (1 − η ) ∂N 2 − (1 + η ) ∂N 3 (1 + η ) ∂N 4 (1 − η )
= ; = ; = ; =
∂η 4 ∂η 4 ∂η 4 ∂η 4

The Jacobian matrix can be found from the following relations.

 ∂N1 ∂N 2 ∂N 3 ∂N 4   x1 y1  x y1 
 −(1 − ξ ) +(1 − ξ ) +(1 + ξ ) −(1 + ξ )   1 
 ∂ξ  
∂ξ ∂ξ ∂ξ   x 2 y2   4 4 4 4   x2 y2 

|J| =   
 ∂N1 ∂N 2 ∂N 3 ∂N 4   x3 y3   −(1 − η ) −(1 + η ) +(1 + η ) +(1 − η )   x y3 

 ∂η    3
 ∂η ∂η ∂η   x4 y4   4 4 4 4   x y4 
 4

Considering the sampling point, (ξ=0 and η=0 ), the value of the Jacobian, [J] is

−1 +1 1 − 1  0 0 
 

[J ] =  4 4 4 4  70 0 
−1 −1 1 1  70 50
 
4 4 4 4   0 50

1 
 0
35 0  35  J11* J12* 
Thus, [J ] =   =
and [ J ]−1
 =   * * 
; and |𝐽| = 875
 0 25 0 1   J 21 J 22 
 25 
Now, the strain vector for the element will become
 ∂u 
 ∂ξ 
 
 J 11* J 12* 0 0   ∂u 
 *   ∂η 
[ε ] =  0 0 J 21 *
J 22  ×  ∂v 
 J 21

* *
J 22 J 11* J 12*   
 ∂ξ 
 ∂v 
 ∂η 
 
  u1 
 ∂N 1 ∂N 2 ∂N 3 ∂N 4  
 ∂ξ 0 0 0 0  v1
∂ξ ∂ξ ∂ξ  
  u 
0   ∂N 1 ∂N 2 ∂N 3 ∂N 4
 J 11* 0  
2
J 12* 0 0 0 0
 *   ∂η ∂η ∂η ∂η  2 
v
[ε ] =  0 0 *
J 21 J 22 × ∂N 1 ∂N 2 ∂N 3 ∂N 4  u 3 
 J 21
* *
J 22 J 11* J 12*   0 0 0 0  
  ∂ξ ∂ξ ∂ξ ∂ξ   v3 
 ∂N 1 ∂N 2 ∂N 3 ∂N 4   
 0 0 0 0 u4
 ∂η ∂η ∂η ∂η   
 v 4 

 J 11* J 12* 0 0 
 * 
[ε ] =  0  × [B']{d } = [ B]{d }
*
0 J 21 J 22
 J 21
* * * 
 J 22 J 11* J 12 

 1 1 1 1 
− 4 0 4 0 4
0 −
4
0
1 
 1 1 1 1   35 0 0 0
− 0 − 0  1
0 0
[B ] =
/  4 4 4 4  and [ B] =  0 0 0 [B']
 0 −1 0 1 0 − 
1 1  25 
 0
0 0 
4 4 4 4 1 1
  
0 −
1
0 −
1
0
1
0
1
 25 35 
 4 4 4 4

7.143 0 7.143 0 7.143 0 7.143 0 

 
[B]   0 10 0 10 0 10 0 10 
 
 10 7.143 10 7.143 10 7.143 10  7.143
 

For plane stress condition

 
1 µ 0  1 0 0 
0  = 2 × 10 0 1 0 
E 
[C ] = µ 1 3

1− µ 2  1− µ 
0 0  0 0 1 / 2
 2 
 1 0 0 
[C ][ B] = 2 × 10 0 1 0  ×
3

0 0 1 / 2

7.143 0 7.143 0 7.143 0 7.143 0 

 
 0 10 0 10 0 10 0 10 
 
 10 7.143 10 7.143 10 7.143 10  7.143
 
Assume the values of gauss weight, w = 2, the stiffness matrix [k] at this sampling point is
[k ] = tw [ B]T [C ] [ B] | J | , Where t is thickness of the element. Thus,
 −7.143 0 −10 
 0 −10 −7.143 

 7.143 0 −10 
  1 0 0 
−10 7.143  
10 × 0 1 0 
 0
[ k ] = 20 × 2 × 875 × 2 ×10 × 
3

7.143 0 10
  0 0 1/ 2 
 0 10 7.143  
 −7.143 0 4 
 
 0 10 −7.143 
 −7.143 0 7.143 0 7.143 0 −7.143 0 

× 0 −10 0 −10 0 10 0 10 
 −10 −7.143 −10 7.143 10 7.143 10 −7.143

 7.071 2.5 −0.714 −2.5 −7.071 −2.5 0.714 2.5 

 2.5 8.785 2.5 5.214 −2.5 −8.785 2.5 −5.214 

 −0.714 2.5 7.071 −2.5 0.714 −2.5 −7.071 2.5 
 
 −2.5 5.214 −2.5 8.785 2.5 −5.214 2.5 −8.785 
= 10 ×
3
 −7.071 −2.5 0.714 2.5 7.071 2.5 −0.714 −2.5 
 
 −2.5 −8.785 −2.5 −5.214 2.5 8.785 2.5 5.214 
 0.714 −2.5 −7.071 2.5 −0.714 2.5 7.071 −2.5 
 
 2.5 −5.214 2.5 −8.785 −2.5 5.214 −2.5 8.785 

5.4.3 Evaluation of Stiffness using Two Point Gauss Quadrature

In this case, 2×2 Gauss Quadrature integration procedure has been carried out to the calculate the
stiffness matrix of the same element for a comparison. The natural coordinate of the sampling
point is shown in the figure below.

Fig. 5.4.3 Natural Coordinates for Two Points Gauss Quadrature

The natural co-ordinates of the sampling points for 2×2 Gauss Quadrature integration are
1 +0.57735 +0.57735
2 -0.57735 +0.57735
3 -0.57735 -0.57735
4 +0.57735 -0.57735

For a four node quadrilateral element, the shape functions and their derivatives are as follows.

(1 − ξ )(1 − η ) (1 + ξ )(1 − η ) (1 + ξ )(1 + η ) (1 − ξ )(1 + η )

N1 = , N2 = , N3 = and N 4 =
4 4 4 4
∂N1 − (1 − ξ ) ∂N 2 (1 − ξ ) ∂N 3 (1 + ξ ) ∂N 4 − (1 + ξ )
= ; = ; = ; =
∂ξ 4 ∂ξ 4 ∂ξ 4 ∂ξ 4
∂N1 − (1 − η ) ∂N 2 − (1 + η ) ∂N 3 (1 + η ) ∂N 4 (1 − η )
= ; = ; = ; =
∂η 4 ∂η 4 ∂η 4 ∂η 4
The Jacobian matrix will be
 ∂N1 ∂N 2 ∂N 3 ∂N 4   x1 y1  x y1 
 −(1 − ξ ) +(1 − ξ ) +(1 + ξ ) −(1 + ξ )   1 
 ∂ξ  
∂ξ ∂ξ ∂ξ   x 2 y2   4 4 4 4   x2 y2 
| J | =   
 ∂N1 ∂N 2 ∂N 3 ∂N 4   x3 y3   −(1 − η ) −(1 + η ) +(1 + η ) +(1 − η )   x y3 

 ∂η    3
 ∂η ∂η ∂η   x4 y4   4 4 4 4   x y4 
 4
(a) At sampling point 1, (ξ=0.57735, η=0.57735)
The value of the Jacobian, [J] at sampling point 1 will become

 − (1 − 0.57735) + (1 − 0.57735) + (1 + 0.57735) − (1 + 0.57735)   0 0 

 4 4 4 4  70 0 
[J ] = 
− (1 − 0.57735) − (1 + 0.57735) + (1 + 0.57735) + (1 − 0.57735)  70 50
 
 4 4 4 4   0 50

1 
 0
35 0  35  J11* J12* 
Thus, [J ] =  0 25 =
and [ J ]−1
=   * * 
; Thus |𝐽| = 875
  0 1   J 21 J 22 
 25 
 ∂u 
 ∂ξ 
 
 J 11* J 12* 0 0   ∂u 
 *   ∂η 
[ε ] =  0 0 *
J 21 J 22  ×  ∂v 
 J 21

* *
J 22 J 11* J 12*   
 ∂ξ 
 ∂v 
 ∂η 
 
 u1 
 ∂N 1 ∂N 2 ∂N 3 ∂N 4  
 ∂ξ 0 0 0 0  v1
∂ξ ∂ξ ∂ξ  
  u 
0   ∂N 1 ∂N 2 ∂N 3 ∂N 4
 J 11* 0  
2
J 12* 0 0 0 0
 *   ∂η ∂η ∂η ∂η  v2 
[ε ] =  0 0 J *
J 22 ×
21
∂N 1 ∂N 2 ∂N 3 ∂N 4  u 3 
 J 21
*
J *
J *
J 12*   0 0 0 0  
 22 11
 ∂ξ ∂ξ ∂ξ ∂ξ   v3 
 ∂N 1 ∂N 2 ∂N 3 ∂N 4   
 0 0 0 0 u4
 ∂η ∂η ∂η ∂η   
 v 4 
 J 11* J 12* 0 0 
 * 
[ε ] =  0  × [B']{d } = [ B]{d }
*
0 J 21 J 22
 J 21

* *
J 22 J *
11 J 12* 
  1 − 0.57735 1 − 0.57735 1 + 0.57735 1 + 0.57735 
− 4
0
4
0
4
0 −
4
0 
 
 − 1 − 0.57735 0 −
1 + 0.57735
0
1 + 0.57735
0
1 − 0.57735
0 
 4 4 4 4 
′
[B ] =  
 1 − 0.57735 1 − 0.57735 1 + 0.57735 1 + 0.57735 
0 − 0 0 0 −
 4 4 4 4 
 1 − 0.57735 1 + 0.57735 1 + 0.57735 1 − 0.57735 
 0 − 0 − 0 0 
 4 4 4 4 

 −0.1057 0 0.1057 0 0.3943 0 −0.3943 0 

 −0.1057 0 −0.3943 0 0.3943 0 0.1057 0 
[ B '] = 
 0 −0.1057 0 0.1057 0 0.3943 0 −0.3943
 
 0 −0.1057 0 −0.3943 0 0.3943 0 0.1057 

Thus,
1 
 35 0 0 0  −0.1057 −0.3943
 0 0.1057 0 0.3943 0 0 
  
1   −0.1057 0 −0.3943 0 0.3943 0 0.1057 0 
[ B]  0 0 0 ×
 25   0 −0.1057 0 0.1057 0 0.3943 0 −0.3943
   
0 1 1
0  0 −0.1057 0 −0.3943 0 0.3943 0 0.1057 
 25 35 

 − 0.003 0 0.003 0 0.0113 0 − 0.0113 0 


[B] =  0 − 0.0042 0 − 0.0158 0 0.0158 0 0.0042 
− 0.0042 − 0.003 − 0.0158 0.003 0.0158 0.0113 0.0042 − 0.0113

For plane stress condition

 
1 µ 0  1 0 0 
0  = 2 × 10 0 1 0 
E 
[C ] = µ 1 3

1− µ 2  1− µ 
0 0  0 0 1 / 2
 2 
 1 0 0 
[C][B] = 2 × 10 0 1 0  ×
3

0 0 1 / 2

 − 0.003 0 0.003 0 0.0113 0 − 0.0113 0 


 0 − 0.0042 0 − 0.0158 0 0.0158 0 0.0042 
− 0.0042 − 0.003 − 0.0158 0.003 0.0158 0.0113 0.0042 − 0.0113

The values of gauss weights are wi=wj=1.0. Therefore, the stiffness matrix [k] at this sampling
point is [k ] = twi w j [ B ]ij [C ]ij [ B ]ij | J | , where t is thickness of the element. Thus at sampling point
T

1,
 −0.003 0 −0.0042 
 0 −0.0042 −0.003 

 0.003 0 −0.0158
  1 0 0 
−0.0158 0.003  
× 0 1 0  ×
0
[ k1 ]= 20 ×1×1× 875 × 2 ×103 × 
0.0113 0 0.0158  
   0 0 1/ 2 
 0 0.0158 0.0113  
 −0.0113 0 0.0042 
 
 0 0.0042 −0.0113
 −0.003 0 0.003 0 0.0113 0 −0.0113 0 
 0 −0.0042 0 −0.0158 0 0.0158 0 0.0042 

 −0.0042 −0.003 −0.0158 0.003 0.0158 0.0113 0.0042 −0.0113
 0.0632 0.0223 0.0848 −0.0223 −0.2358 −0.0834 0.0878 0.0834 
 0.0785 0.0834 0.2174 −0.0834 −0.2929 −0.0223 −0.0030 

 0.4672 −0.0834 −0.3162 −0.3109 −0.2358 0.3109 
 
 0.8866 0.0834 −0.8111 0.0223 −0.2929 
[ k=
1 ] 10 ×
4
 0.8795 0.3109 −0.3275 −0.3109 
 
 sym 1.0927 0.0834 0.0113 
 0.4755 −0.0834 
 
 0.2847 

(b) At sampling point 2, (ξ=-0.57735, η=0.57735)

The value of the Jacobian, [J] at sampling point 2 can be calculated in a similar way and finally
the strain-displacement relationship matrix and then the stiffness matrix [k2] can be evaluated
and is shown below.

 −0.0113 −0.0042 
0
 0 −0.0113
−0.0042

 0.0113 −0.0158
0
 
0 −0.0158
0.0113 
[ k2 ]= 20 ×1×1× 875 × 2 ×103  ×
0.003 0.0158 
0
 
 0 0.003 
0.0158
 −0.003 0.0042 
0
 
 0 −0.0030 
0.0042
 −0.0113 0 0.0113 0 0.0030 0 −0.0030 0 
 0 −0.0042 0 −0.0158 0 0.0158 0 0.0042 

 −0.0021 −0.0056 −0.0079 0.0056 0.0079 0.0015 0.0021 −0.0015

0.4756 0.0833 −0.3276 −0.0833 −0.2357 −0.0223 0.0878 0.0223 

 0.2847 0.3110 0.0112 −0.3110 −0.2929 −0.0833 −0.0030 

 0.8797 −0.3110 −0.3164 −0.0833 −0.2357 0.0833 
 
 1.0930 0.3110 −0.8113 0.0833 −0.2929 
[ k=
2 ] 10 ×
4
 0.4673 0.0833 0.0848 −0.0833
 
 sym 0.8868 0.0223 0.2174 
 0.0632 −0.0223
 
 0.0785 
(c) At sampling point 3, (ξ=-0.57735, η=-0.57735)
The value of the strain-displacement relationship matrix and then the stiffness matrix [k3] can be
evaluated and is shown below.

 −0.0113 0 −0.0158
 0 −0.0158 −0.0113

 0.0113 0 −0.0042 
 
0 −0.0042 0.0113 
[ k3 ]= 20 ×1×1× 875 × 2 ×103  ×
0.0030 0 0.0042 
 
 0 0.0042 0.0030 
 −0.0030 0 0.0158 
 
 0 0.0158 −0.0030 
 −0.0030 0 0.0113 0 0.0030 0 −0.0030 0 
 0 −0.0158 0 −0.0042 0 0.0042 0 0.0158 

 −0.0079 −0.0056 −0.0021 0.0056 0.0021 0.0015 0.0079 −0.0015

0.8797 0.3110 −0.3276 −0.3110 −0.2357 −0.0833 −0.3164 0.0833 

 1.0930 0.0833 0.0112 −0.0833 −0.2929 −0.3110 −0.8113

 0.4756 −0.0833 0.0878 −0.0223 −0.2357 0.0223 
 
 0.2847 0.0833 −0.0030 0.3110 −0.2929 
[ k=
3 ] 10 ×
4
 0.0632 0.0223 0.0848 −0.0223
 
 sym 0.0785 0.0833 0.2174 
 0.4673 −0.0833
 
 0.8868 

(d) At sampling point 4, (ξ=0.57735, η=-0.57735)

The value of the strain-displacement relationship matrix and then the stiffness matrix [k3] can be
evaluated and is shown below.
  −0.0030 0 −0.0158
 0 −0.0158 −0.0030 

 0.0030 0 −0.0042 
 
0 −0.0042 0.0030 
[ k4 ]= 20 ×1×1× 875 × 2 ×103  ×
0.0113 0 0.0042 
 
 0 0.0042 0.0113 
 −0.0113 0 0.0158 
 
 0 0.0158 −0.0113
 −0.0030 0 0.0030 0 0.0113 0 −0.0113 0 
 0 −0.0158 0 −0.0042 0 0.0042 0 0.0158 

 −0.0079 −0.0015 −0.0021 0.0015 0.0021 0.0056 0.0079 −0.0056 

0.4673 0.0833 0.0848 −0.0833 −0.2357 −0.3110 −0.3164 0.3110 

 0.8868 0.0223 0.2174 −0.0223 −0.2929 −0.0833 −0.8113

 0.0632 −0.0223 0.0878 −0.0833 −0.2357 0.0833 
 
 0.0785 0.0223 −0.0030 0.0833 −0.2929 
[ k=
4 ] 10 ×
4
 0.4756 0.0833 −0.3276 −0.0833
 
 sym 0.2847 0.3110 0.0112 
 0.8797 −0.3110 
 
 1.0930 

The stiffness matrix of the element can be computed as the sum of the values at the four
sampling points: [k ] = [k1 ] + [k2 ] + [k3 ] + [k4 ] . Thus, the final value of the stiffness matrix will
become

1.8857 0.5000 −0.4857 −0.5000 −0.9429 −0.5000 −0.4571 0.3110 

 2.3429 0.5000 0.4571 −0.5000 −1.1714 −0.5000 −1.6286 

 1.8857 −0.5000 −0.4571 −0.5000 −0.9429 0.5000 
 
2.3429 −0.5000 −1.6286 0.5000 −1.1714 
k ] 104 × 
[=  1.8857 0.5000 −0.4857 −0.5000 
 
 sym 2.3429 0.5000 0.4571 
 1.8857 −0.5000 
 
 2.3429 
Lecture 5: Computation of Stresses, Geometric Nonlinearity and Static Condensation

5.5.1 Computation of Stresses

After solving the static equation of {F} = [K]{d}, the nodal displacement {d} can be obtained in
global coordinate system. The element nodal displacement d  can then be calculated from the
nodal connectivity of the element. Using strain-displacement relation and then stress-strain
relation, the stress at the element level are derived.
σ
ε
D
= DB  d       (5.5.1)
Here,  is the stress at the Gauss point of the element as the sampling points for the integration
has been considered as Gauss points. Here, [D] is the constitutive matrix, [B] is the strain
displacement matrix of the element. As a result these stresses at Gauss points need to extrapolate
to the corresponding nodes of the element. It is well established that 2 × 2 Gauss integration
points are the optimal sampling points for two dimensional isoparametric elements. The ‘local
stress smoothing’ is a technique that can be used to extrapolate stresses computed at Gauss
points to nodal points. The stresses are computed at four Gauss points (I, II, III and IV) of an 8
node element as shown in Fig. 5.5.1. For example, at point III, r = s =1 an d ξ = η = 1 3.
Therefore the factor of proportionality is 3 ; i.e.,
r = ξ 3 and s = η 3 (5.5.2)
Stresses at any point P in the element are found by the usual shape function as
′ σ i for i = 1,2,3,4
σ P = ∑ N di (3.5.3)
In the above equation, σ P is σ x , σ y and τ xy at point P. N di
′ are the bilinear shape functions
written in terms of r and s rather than ξ and η as

′ = (1 ± r )(1 ± s )
1
N di (5.5.4)
4
′ are evaluated at r and s coordinates of point P. Let the point P coincides with the corner 1.
N di
To calculate stress σ x1 at corner 1 from σ x values at the four Gauss points, substitution of r and
s into the shape functions will give
σ xI =1.8666σ xI − 0.500σ xII + 0.134σ xIII − 0.500σ xIV (5.5.5)
Fig. 5.5.1 Natural coordinate systems used in extrapolation of stresses from Gauss points

The resultant extrapolation matrix thus obtained may be written as

σ 1  (
 1+ 3 2 ) − 0.5 (1 −3 2 ) − 0.5 
σ 
 2

 − 0.5 (
1+ 3 2 ) − 0.5 (
1− 3 2 

)
σ 3  (
 1− 3 2

) − 0.5 (
1+ 3 2 ) − 0.5 

σI 
 
σ 4  =  − 0.5 (
1− 3 2 ) − 0.5 (
1+ 3 2  ) σ 
 II 
σ 5 
 
( )
 1+ 3 4

( )
1+ 3 4 (
1− 3 4 ) ( )
1 − 3 4


σ III 
 (5.5.6)

σ 6  ( )
 1− 3 4 ( )
1+ 3 4 (
1+ 3 4 ) ( )
1 − 3 4 σ IV 
σ 
 7
(
 )
 1− 3 4 ( )
1− 3 4 (
1+ 3 4 ) ( )
1 + 3 4

σ 8  ( )
 1 + 3 4 ( )
1− 3 4 (
1− 3 4 ) ( )
1 + 3 4

Here, σ 1 , σ 2 …….. σ 8 are the smoothened nodal values and σ I … σ IV are the stresses at the
Gauss points. Smoothened nodal stress values for four node rectangular element can be also be
evaluated in a similar fashion. The relation between the stresses at Gauss points and nodal point
for four nodel element will be
  1 
1  3 
1
1 
3

 2 2 2 2 

1   3  I 
   1 3 1
2    2 1  1  
2  II 
    2 2
 III 
(5.5.7)
3   3 1 3 1
  1   1    

4   2 2 2 2  IV 
 3 
 1 1 
3

1
1 
 2 2 2 2 

The stress at particular node joining with more than one element will have different magnitude as
calculated from adjacent elements (Fig. 5.5.2(a)). The stress resultants are then modified by
finding the average of resultants of all elements meeting at a common node. A typical stress
distribution for adjacent elements is shown in Fig. 5.5.2(b) after stress smoothening.
 Fig. 5.5.2 Stress smoothening at common node

5.5.2 Geometric Nonlinearity

As discussed earlier, nonlinear analysis is mainly of two types: (i) Geometric nonlinearity and
(ii) Material nonlinearity. For geometric nonlinearity consideration, the relation between strain
and displacement is of utmost importance in the finite element formulation for stress analysis
problems. In case of plane stress/strain problem, the nonlinear term of the strain expression are
dropped for the sake of simplicity in the analysis. However, for large displacement problems, the
nonlinear strain term plays a vital role to obtain accurate response. The generalized strain-
displacement relations for the two-dimensional plane stress/strain problems are rewritten here to
derive the nonlinear solution.

∂u 1  ∂u   ∂v  
2 2

εx =+   +  
∂x 2  ∂x   ∂x 
∂v 1  ∂u   ∂v  
2 2

εy = +   +   
∂y 2  ∂y   ∂y  
(5.5.8)
∂v ∂u  ∂u ∂u ∂v ∂v 
γ xy = + +  +
∂x ∂y  ∂x ∂y ∂x ∂y 

The displacements at any point inside the node are expressed in terms of their nodla
displacements. Thus,
n n
=u N u [ N ]{u } =
∑= and v ∑
= N v [ N ]{v }
i =1
i i i i
i =1
i i i i (5.5.9)
Therefore,
 ∂u ∂ [ N i ]
= = {ui } [ B1 ]{ui }
∂x ∂x
∂v
= [ B1 ]{vi }
∂x
∂u
= [ B2 ]{ui }
∂y
∂v
= [ B2 ]{vi }
∂y
(5.5.10)
Here, [B1] and [B2] are the derivative of the shape function [Ni] with respect to x and y
respectively. The vectors {ui} and {vi} represent the nodal displacements vectors in x and y
directions respectively. The vector of strains at any point inside an element, {ε } may be
expressed in terms of nodal displacement as
{ε } = [ B ] {d }
(5.5.11)
where [B] is the strain displacement matrix. {d} is the nodal displacement vector and may be
expressed as
{u }
{d } =  i 
{vi } 
(5.5.12)
The matrix [B] may be expressed with two components as
[=
B ] [ Bl ] + [ Bnl ]
(5.5.13)
where, [Bl] and [Bnl] are the linear and nonlinear part of the strain-displacement matrix
respectively and are expressed as follows:
 [B1 ] [0] 
[Bl ] =  [0] [B2 ]
[B2 ] [B1 ]
(5.5.14)
and
 1 T 
 2 {u } [B1 ] [B1 ] 2 {v } [B1 ] [B1 ] 
T 1 T T

 
[Bnl ] =  1 {u }T [B2 ]T [B2 ] 1 {v }T [B2 ]T [B2 ]
2 T 2 
 {u } [B2 ]T [B1 ] {v }T [B2 ]T [B1 ] 
 
(5.5.15)
5.5.2.1 Steps to include effect of geometrical nonlinearity
The nonlinear geometric effect of the structure at a particular instant of time can be obtained by
performing the following steps.
1. Calculation of displacement {d}1 considering linear part of strain matrix [Bl].
2. Evaluation of nonlinear part of the strain matrix [Bnl] (eq5.5.15) adopting {d}1 from
previous step.
3. Evaluation of total strain matrix [B] = [Bl] + [Bnl].
4. Calculation of displacement {d}2 considering both linear and nonlinear part of strain
matrix [B].
5. Repetition of steps 2 to 4 with {d}2, from which modified displacement, {d}3 are
obtained.
6. Step 5 is carried out until the displacements for two consecutive iteration converge
i.e.,
{d } j − {d } j −1
<ε
{d } j
Where ε is any pre-assigned small value and j is the number of iterations.

5.5.3 Static Condensation

The higher order Lagrangian elements (i.e., nine node, sixteen node rectangular element) contain
number of internal nodes. This is necessary sometimes for the completeness of the desired
polynomial used in displacement function for derivation of interpolation function. These internal
nodes are not connected to the adjoining elements in the assemblage (Fig. 5.5.3).
 Fig. 5.5.3 Internal nodes of Nine node elements

Thus, the displacements of these nodes are not required to formulate overall equilibrium
equations of the structure. This limits the usefulness of these elements. A technique known as
“static condensation” can be used to suppress the degrees of freedom associated with the internal
nodes in the final computation. The technique of static condensation is explained below. The
equilibrium equation for a system are expressed in the finite element form as
F  =  K d  (5.5.16)
Where, {F}, [K] and {d} are the load vector, stiffness matrix and displacement vector for the
entire structure. The above equation can be rearranged by separating the relevant terms
corresponding to internal and external nodes of the elements.
Fi   K ii   K ie di 
 =    (5.5.17)
Fe   K ei   K ee  d e 
    
Here, {di} and {de} are the displacement vectors corresponding to internal and external nodes
respectively. Similarly, {Fi} and {Fe} are force vectors corresponding to internal and external
nodes. Now, the above expression can be written in the following form separately.

Fi    Kii di    Kie de  (5.5.18)

Fe    K ei di    K ee de  (5.5.19)

The stiffness matrix and nodal load vector corresponds to the internal nodes can be separated out.
For this, eq.(5.5.18) can be rewritten as
1 1
di    Kii  Fi    Kii   Kie de  (5.5.20)
Substituting the value of {di} obtained from the above equation in eq.(5.5.19), the following
expression will be obtained.
 Fe    K ei  Kii  Fi    Kii   Kie de    K ee de 
1 1
(5.5.21)

Here, the equations are reduced to a form involving only the external nodes of the elements. The
above reduced substructure equations are assembled to achieve the overall equations involving
only the boundary unknowns. Thus the above equation can be rewritten as

F   K
e ei  Kii  Fi    K ee    K ei  Kii   Kie de  de 
1 1
(5.5.22)

Or
Fc  =  K c de  (5.5.23)

Fc  = Fe  -  K ei  Kii  Fi  and  K c  =  K ee  -  K ei  Kii   Kie de  . Here, [Kc]
-1 -1
Where,
is called condensed or reduced stiffness matrix and {Fc} is the condensed or effective nodal load
vector corresponding to external nodes of the elements. In this process, the size of the matrix for
inversion will be comparatively small. The unknown displacements of the exterior nodes, {de}
can be obtained by inverting the matrix [Kc] in eq.(5.5.23). Once, the values of {de} are obtained,
the displacements of internal nodes {di} can be found from eq.(5.5.20).
Lecture 6: Axisymmetric Element

5.6.1 Introduction
Many three-dimensional problems show symmetry about an axis of rotation. If the problem
geometry is symmetric about an axis and the loading and boundary conditions are symmetric
about the same axis, the problem is said to be axisymmetric. Such three-dimensional problems
can be solved using two-dimensional finite elements. The axisymmetric problem are most
conveniently defined by polar coordinate system with coordinates (r, θ, z) as shown in Fig. 5.6.1.
Thus, for axisymmetric analysis, following conditions are to be satisfied.
1. The domain should have an axis of symmetry and is considered as z axis.
2. The loadings on the domain has to be symmetric about the axis of revolution, thus they
are independent of circumferential coordinate θ.
3. The boundary condition and material properties are symmetric about the same axis and
will be independent of circumferential coordinate.

Fig. 5.6.1 Cylindrical coordinates

Axisymmetric solids are of total symmetry about the axis of revolution (i.e., z-axis), the field
variables, such as the stress and deformation is independent of rotational angle θ. Therefore, the
field variables can be defined as a function of (r, z) and hence the problem becomes a two
dimensional problem similar to those of plane stress/strain problems. Axisymmetric problems
includes, circular cylinder loaded with uniform external or internal pressure, circular water tank,
pressure vessels, chimney, boiler, circular footing resting on θsoil
• P( r,z, ) mass, etc.

5.6.2 Relation between Strain and Displacement

An axisymmetric problem is readily described in cylindrical polar coordinate system: r, z and θ.
Here, θ measures the angle between the plane containing the point and the axis of the coordinate
system. At θ = 0, the radial and axial coordinates coincide with the global Cartesian X and Y
coordinates. Fig. 5.6.2 shows a cylindrical coordinate system and the definition of the position
ˆ zˆ and ˆ be unit vectors in the radial, axial, and circumferential directions at a point
vectors. Let r,
in the cylindrical coordinate system.

Fig. 5.6.2 Cylindrical Coordinate System

If the loading consists of radial and axial components that are independent of θ and the material
is either isotropic or orthotropic and the material properties are independent of θ, the
displacement at any point will only have radial (𝑢𝑟 ) and axial (𝑢𝑧 ) components. The only stress
components that will be nonzero are 𝜎𝑟𝑟 , 𝜎𝑧𝑧 , 𝜎𝜃𝜃 𝑎𝑛𝑑 𝜏𝑟𝑧 .

(a) Element in r-z plane (b) Element in r-θ plane

 Fig. 5.6.3 Deformation of the axisymmetric element
A differential element of the body in the r-z plane is shown in Fig. 5.6.3(a). The element
undergoes deformation in the radial direction. Therefore, it initiates increase in circumference
and associated circumferential strain. Let denote the radial displacement as u, the circumferential
displacement as v, and the axial displacement as w. Dashed line represents the deformed
positions of the body in Fig. 5.6.3(b). The radial strain can be calculated from the above diagram
as
1  u  u
=
εr u +- u =
× dr  (5.6.1)
dr  r  r
Since the rz plane is effectively the same as a rectangular coordinate system, the axial strain will
become
1  w  w
=
εz ×dzw+- w =  (5.6.2)
dz  z  z

Considering the original arc length versus the deformed arc length, the differential element
undergoes an expansion in the circumferential direction. Before deformation, let the arc length is
assumed as ds = rdθ. After deformation, the arc length will become ds = (r+u) dθ. Thus, the
tangential strain will be
r +u  d  - rd u
ε =  (5.6.3)
rd  r
Similarly, the shear strain will be

u w
 rz  
z r (5.6.4)
 r  0 and  z  0

Thus, there are four strain components present in this case and is given by
 ∂u  ∂ 
 ∂r   ∂r 0
εr     
   ∂w  0 ∂
εz   ∂z   ∂z   u 
{ε } =
=   =     (5.6.5)
 εθ   u  1 0  
 w
γ rz   r  r 
 ∂u ∂w  ∂ ∂
 +   
 ∂z ∂r   ∂z ∂r 
5.6.3 Relation between Stress and Strain
The stress strain relation for axisymmetric case can be derived from the three dimensional
constitutive relations. We know the stress-strain relation for a three-dimensional solid is

1−𝜇 𝜇 𝜇 0 0 0
⎡ ⎤
𝜎𝑥 ⎢ 𝜇 1−𝜇 𝜇 0 0 0 ⎥ ϵ𝑥
⎧𝜎𝑦 ⎫ ⎢ ⎥ ⎧ϵ ⎫
𝑦
⎪𝜎 ⎪ ⎢ 𝜇 𝜇 1−𝜇 0 0 0 ⎥ ⎪ϵ ⎪
𝐸
𝑧
= (1+𝜇)(1−2𝜇) ⎢ ⎥ 𝑧
𝜏 0 ⎥ ⎨𝜈𝑥𝑦 ⎬
1−2𝜇
⎨ ⎬𝑥𝑦 ⎢ 0 0 0 2
0
⎪𝜏𝑦𝑧 ⎪ ⎢ 1−2𝜇 ⎥ ⎪𝜈𝑦𝑧 ⎪
⎩𝜏𝑧𝑥 ⎭ ⎢ 0 0 0 0 2
0 ⎥ ⎩𝜈𝑧𝑥 ⎭
⎢ 1−2𝜇 ⎥
⎣ 0 0 0 0 0 2 ⎦
(5.6.6)

The stresses acting on a differential volume of an axisymmetric solid under axisymmetric

loading is shown in Fig. 5.6.4.

Fig. 5.6.4 Stresses acting on a differential volume

Now, comparing the stress-strain components present in the axisymmetric case, the stress-strain
relation can be expressed from the above expression as follows
 𝜎𝑟 1−𝜇 𝜇 𝜇 0 𝜀
⎧ ⎫ ⎡ ⎤⎧ 𝑟 ⎫
⎪𝜎𝑧 ⎪ ⎢ 𝜇 1−𝜇 𝜇 0 ⎥ ⎪𝜀𝑧 ⎪
𝐸
= ⎢ ⎥ (5.6.7)
⎨𝜎𝜃 ⎬ (1+𝜇)(1−2𝜇) ⎢ 𝜇 𝜇 1−𝜇 0 ⎥ ⎨𝜀𝜃 ⎬
⎪ ⎪ ⎢ 1−2𝜇 ⎥
⎪ ⎪
⎩𝜏𝑟𝑧 ⎭ ⎣ 0 0 0 2 ⎦
⎩𝜈𝑟𝑧 ⎭

Thus, the constitutive matrix [D] for the axisymmetric elastic solid will be
1−𝜇 𝜇 𝜇 0
⎡ ⎤
⎢ 𝜇 1−𝜇 𝜇 0 ⎥
𝐸
[D] = (1+𝜇)(1−2𝜇) ⎢ ⎥ (5.6.8)
⎢ 𝜇 𝜇 1−𝜇 0 ⎥
⎢ 1−2𝜇 ⎥
⎣ 0 0 0 2 ⎦

5.6.4 Axisymmetric Shell Element

A cylindrical liquid storage container like structures (Fig. 5.6.5) may be idealized using
axisymmetric shell element for the finite element analysis. It may be noted that the liquid in the
container may be idealized with two dimensional axisymmetric elements. Let us consider the
radius, height and, thickness of the circular tank are R, H and h respectively.

Fig. 5.6.5 Thin wall cylindrical container

The strain energy of the axisymmetric shell element (Fig. 5.6.6) including the effect of both
stretching and bending are expressed as
H
U = ε ∫+( N
N y εθy θ+ M χy y2πRdy)
1
(5.6.9)
20
Here, Ny and Nθ are the membrane force resultants and My is the bending moment resultant. The
shell is assumed to be linearly elastic, homogeneous and isotropic. Thus the force and moment
resultants can be expressed in terms of the mid-surface change in curvature χy as follows.

Fig 5.6.6 Axisymmetric plate element

Here, the strain-displacement relation is given by

 {σ} = [ D ] {ε} (5.6.10)

In which,
 
N y  ε y  1 µ 0 
     Eh 
 {σ } =  Nθ  ,  {ε } = ε θ  and  [ D ] =µ 1 0  (5.6.11)
M  χ  1− µ 2 2 
 y  y 0 0 h 
 12 
The generalized strain vector can be expressed in terms of the displacement vectors as follows.
{ε } = [ B ] {d } (5.6.12)
Where,
  ∂
 0 ∂y 
 
u   1 
 {d } =   and  [ B ] =  0 (5.6.13)
v  
R

 ∂ 
2

 − ∂y 2 0 
 
Here, u and v are the displacement components in two perpendicular directions. With the use of
stress and strain vectors, the potential energy expression are written in terms of displacement
vectors as

∫ ({d } [ B ] [ D ][ B ]dy
{d } )
H
1
U = πR× 2
T T
(5.6.14)
2 0

Thus, the element stiffness are derived as

H
 [ k ] = 2πR ∫ [ B ] [ D ][ B ] dy
T
(5.6.15)
0

Similarly, neglecting the rotary inertia, the kinetic energy can be expressed as

{d})
∫ ({d} [ N ] m [ N ]dy
H
1 T
T = πR× 2
T
(5.6.16)
2 0

{}
Where, m denotes the mass of the shell element per unit area and   d represents the velocity
vector. Thus, the element mass matrix is given by
Le

 [ M ] = 2πRm ∫ [ N ] [ N ] dy
T
(5.6.17)
0

Lecture 7: Finite Element Formulation of Axisymmetric Element

Finite element formulation for the axisymmetric problem will be similar to that of the two
dimensional solid elements. As the field variables, such as the stress and strain is independent of
rotational angle θ, circumferential displacement will not appear. Thus, the displacement field
variables are expressed as
n
u r,z  =  N i r,z ui
i=1
n
(5.7.1)
wr,z  =  N i r,z  wi
i=1
Here, ui and wi represent radial and axial displacements respectively at nodes. Ni (r, z) are the
shape functions. As the geometry and field variables are independent of rotational angle θ, the
interpolation function Ni (r, z) can be expressed similar to 2-dimensional problems by replacing
the x and y terms with r and z terms respectively.
5.7.1 Stiffness Matrix of a Triangular Element
Fig. 5.7.1 shows the cylindrical coordinates of a three node triangular element. Hence the
analysis of the axisymmetric element can be approached in a similar way as the CST element.
Thus the field variables of such an element can be expressed as
u =α 0 + α1r + α 2 z
(5.7.2)
w =α 3 + α 4 r + α 5 z
Or,
{d } = [φ ]{α } (5.7.3)
Where,
1 r z 0 0 0 
u  ,
 [φ ] =  {α } = {α 0 α1 α 2 α 3 α 4 α 5 }
T
{d } =   and
 w 0 0 0 1 r z 
Using end conditions,
 u1  1 ri zi 0 0 0  α 0 
 u  1 r zj 0 0 0  α1 
 2  j

 u3  1 rk zj 0 0 0  α 2 
 =   (5.7.4)
 w1  0 0 0 1 ri zi  α 3 
 w2  0 0 0 1 rj z j  α 4 
    
 w3  0 0 0 1 rk zk  α 5 
Or,
{d } = [ A]{α }
(5.7.5)
[ A] {d }
⇒ {α } =
−1

{}
Here d are the nodal displacement vectors.
 Fig. 5.7.1 Axisymmetric three node triangle in cylindrical coordinates

Putting above values in eq.(5.7.3), the following relations will be obtained.

][ A] {d } [ N ]{d }
{d } [φ=
−1
= (5.7.6)
Or,
 r1 
r 
 2
 u   Ni Nj Nk 0 0 0   r3 
{=
d} =    
N k   z1 
(5.7.7)
 w  0 0 0 Ni Nj
 z2 
 
 z3 
Using a similar approach as in case of CST elements, the three shape functions [ N1 , N 2 , N 3 ] can
be assumed as,

1
N1 (=
r, z ) ( r2 z3 − r3 z2 ) + ( z2 − z3 ) r + ( r3 − r2 ) z 
2A 
1
N 2 (=
r, z ) ( r3 z1 − r1 z3 ) + ( z3 − z1 ) r + ( r1 − r3 ) z 
2A 
1
N 3 (=
r, z ) ( r1 z2 − r2 z1 ) + ( z1 − z2 ) r + ( r2 − r1 ) z 
2A 
 Or,
1
N i ( r , z=
) (α i + r βi + zγ i )
2A
N j ( r , z=
)
1
2A
(α j + r β j + zγ j ) (5.7.8)

1
N k ( r , z=
) (α k + r β k + zγ k )
2A

Where,
αi =
rj zk − rk z j αj =
rk zi − ri zk αk =
ri z j − rj zi
βi =
z j − zk βj =
zk − zi βk =
zi − z j
γi =
rk − rj γi =
ri − rk γi =
rj − ri (5.7.9)

2 A=
1
2
( ri z j + rj zk + rk zi − ri zk − rj zi − rk z j )
Putting the value of {u,w} in eq. (5.7.7) from eq. (5.6.5),
 ∂N i ∂N j ∂N k 
 0 0 0 r 
 ∂r ∂r ∂r 1
 
 Ni Nj Nk   r2 
 r 0 0 0   r3 
{ε } =
r r
  [ B ]{d } (5.7.10)
 ∂N i ∂N j ∂N k   z1 
 0 0 0
∂z ∂z ∂z   z2 
  
 ∂N i ∂N j ∂N k ∂N i ∂N j ∂N k   z3 
 ∂z ∂z ∂z ∂r ∂r ∂r 
Thus, the strain displacement matrix can be expressed as,
 βi βj βk 0 0 0
 Nj 
N Nk
1  i 0 0 0
[ B] =  r r r  (5.7.11)
2A 
0 0 0 γi γj γk 
 
 γ i γj γk βi βj β k 

ri + rj + rk
Where, r = . Thus the stiffness matrix will become
3
=[k ] ∫∫∫ [ B ] [ D ][ B ] d Ω
T

2π
= [k ] ∫ ∫ [ B ] [ D ][ B ] r dθ dA 2π ∫ ∫ [ B ] [ D ][ B ] r dr dz
∫=
T T
Or, (5.7.12)
0
Since, the term [B] is dependent of ‘r’ terms; the term [ B ] [ D ][ B ] cannot be taken out of
T

integration. Yet, a reasonably accurate solution can be obtained by evaluating the [B] (denoted as
[B]) matrix at the centroid.
Hence, [ k ] = 2π r [ B ] [ D ][ B ] ∫ ∫ dr dz
T

Or,
[ k ]  [ B ] [ D ][ B ] 2π rA
T
(5.7.13)

5.7.2 Stiffness Matrix of a Quadrilateral Element

 50

The strain-displacement relation for axisymmetric problem derived earlier (eq.(5.6.5)) can be
rewritten as
 ∂u 
 ∂u   ∂r 
 ∂r   
  1 0 0 0 0   ∂u 
εr 
ε   ∂w  0 0 0 1 0   ∂z 
 ∂z  
{ε } =
= z 
  =  
 
1   ∂w  (5.7.14)
 εθ   u  0 0 0 0 
r   ∂r 
γ rz   r    
 ∂u ∂w  0 1 1 0 0   ∂w 
 +   ∂z 
 ∂z ∂r  u 
 

Applying chain rule of differentiation equation we get,

 ∂u   ∂u 
 ∂r   ∂ξ 
   J*  
*
0   ∂u 
 ∂u   11
J12 0 0

 ∂z   J*21 J*22 0 0 0   ∂η 
    
 ∂w  = 0 0 *
J11 *
J12 0   ∂w  (5.7.15)
 ∂rξ  0 J*21 J*22

0  ∂ 
  
0
 
 ∂w   0 0 0 0 1   ∂w 
 ∂z   ∂η 
 u   
   u 
Hence, the strain components are calculated as
 ∂u 
 ∂ξ 
 
0 0 0 0   J11 0   ∂u 
* *
1 J12 0 0
ε r    * 
ε   0 0 0 1 0   J 21 J*22 0 0 0   ∂η 
 z  
 = 1  0 0   ∂w 
* *
0 J11 J12
ε θ   0 0 0 0 
r   0 0 J*21 J*22

0   ∂ξ 
γ rz    
0 1 1 0 0   0 0 0 0 1   ∂w 
 ∂η 
 
 u 
Or,
 51


 u 

 


  

 
 
 J11

* *
J12 0 0 0  
   u 
  r  
 

  0 
  0    
*
  0 J J*22 
 z   
 
21
   
1  
 w (5.7.16)
  
    
  0 0 0 0
r   
   
 
 rz 
   J* J*22 * * 
0  w 
 
 21 J11 J12
 


  

 


 
 u 

 u u w w 
With the use of interpolation function and nodal displacements,  , , ,  can be expressed
 x h x h 
for a four node quadrilateral element as

 u 
   N1 N 2 N 3 N 4  u1 


   0   

x 
0 0 0

 
  x x x x  u2 

    

 u   N1 N 2 N 3 N 4   u3 
 

 
  0 0 0 0   

  h   h h h h  u4 

      

 (5.7.17)
 w  N1 N 2 N 3 N 4  
 w1 

   0 0 0 0   

 x   x x x x   w2 


 
 
 

  w   N1 N 2 N 3 
N 4   

 
  0 0 0 0 

w3

  h   h h h 
h   
    
 
 w4 


Putting eq. (5.7.17) in eq. (5.7.16) we get,

 N1 N 2 N 3 N 4 
 0 0 0 0   u1 

      

  u2 
 J11
* *   N N 2 N 3 N 4  

0   u3 

 r    J 0 0 0  1
 12 0 0 0 
   0 0 J*21 J*22 0       

  z     u4 
   1   0 N1 N 2 N 3 N 4   
 

  
    0 0 0 0 0 0 0 
 w1 

   r      
   


 rz    J* J* J* J* 0   w 2 

 21 22 11 12  0 N1 N 2 N 3 N 4  
 

 0 0 0  
 w 
3
       
 
w 4 

 N 4  
 
 N1 N2 N3 N4 N1 N2 N3
(5.7.18)
Thus, the strain displacement relationship matrix [B] becomes
 52

 N1 N 2 N 3 N 4 
 0 0 0 0 
     
 
 * * 
0 0  N1 N 2 N 3 N 4 
 J11 J12 0 0 0 0 0 
0
 0 J 21 J*22 0  
*    

 B   
1   0 N1 N 2 N 3 N 4  (5.7.19)
0 0 0 0 0 0 0
      
 * r 
 J 21 J*22 J11

* *
J12 0  N1 N 2 N 3 N 4 
 0 0 0 0
    

 N
 1 N2 N3 N4 N1 N2 N3 N 4 
For a four node quadrilateral element,
(1ξ− )1( η− )          and     
N1 = ⇒
∂N1
=
−
( 1η− ) ∂N1 1( ξ− )
=−
4ξ 4 ∂ η 4 ∂
(1ξ+ 1) ( η− )          and   
N2 = ⇒
∂N 2 ( 1η− )
=
∂N1 1( ξ+ )
=− (5.7.20)
4ξ 4 ∂ η 4 ∂

=N3
(1ξ+ 1) ( η+ )
⇒=
∂N 2 ( 1η+ )
         and   
∂N1 1( ξ+ )
=
4ξ 4 ∂ η 4 ∂
(1ξ− )1( η+ )          and   
N4 = ⇒
∂N 2
=
−
( 1η+ ) ∂N1 1( ξ− )
=
4ξ 4 ∂ η 4 ∂
Thus, the [B] matrix will become
 J11
* *
J12 0 0 0

0 0 J 21 J*22 0
*

 B   1  
0 0 0 0
r 
 *
 J 21 J 22 J11 J12 0
* * *
 
 1 1 1   1   
   0 0 0 0 
 4 4 4 4 
 
 1 1   1   1 
   0 0 0 0 
 4 4 4 4 
 
 1 1 1   1   
 0 0 0 0  
 4 4 4 4 
 
 1 1   1   1 
 0 0 0 0  
 4 4 4 4 
 
 
 11 1  1 1  1   11   11 1  1 1  1   11  
 4 4 4 4 4 4 4 4 

(5.7.21)
The stiffness matrix for the axisymmetric element finally can be found from the following
expression after numerical integration.
+1+1
=[k ] ∫ [ B=
] [ D ][ B ] d Ω ∫ ∫ [ B] [ D][πr.B].2.dξdηJ
T T
(5.7.22)
Ω −1−1
 53

Lecture 8: Finite Element Formulation for 3 Dimensional Elements

5.8.1 Introduction
Solid elements can easily be formulated by the extension of the procedure followed for two
dimensional solid elements. A domain in 3D can be discritized using tetrahedral or hexahedral
elements. For example, the eight node solid brick element is analogous to the four node rectangular
element. Regardless of the possible curvature of edges or number of nodes, the solid element can be
mapped into the space of natural co-ordinates, i.e the ξ = ±1,η = ±1, ζ = ±1 just like a plane element.

Fig. 5.8.1Eight node brick element

For three dimensional cases, each node has three degrees of freedom having u, v, and w as
displacement field in three perpendicular directions (X, Y and Z). In this case, one additional
dimension increases the computational expense manifolds.

5.8.2 Strain Displacement Relation

The strain vector for three dimensional cases can be written in the following form
 54

 ∂u 
 ∂x 
 
 ∂u 
 ∂y 
 
 ∂u 
 ∂z 
 ε x  1 0 0 0 0 0 0 0 0   ∂v 
 ε  0  
 y  0 0 0 1 0 0 0 0   ∂x 
 ε z  0 0 0 0 0 0 0 0 1   ∂v 
 =  
γ xy  0 1 0 1 0 0 0 0 0   ∂y 
γ yz  0 0 0 0 0 1 0 1 0   ∂v 
    
γ zx  0 0 1 0 0 0 1 0 0   ∂z 
 ∂w 
 ∂x 
 
 ∂w 
 ∂y 
 
 ∂w  (5.8.1)
 ∂z 

The following relation exists between the derivative operators in the global co-ordinates and the
natural co-ordinate system by the use of chain rule of partial differentiation.
 ∂   ∂x ∂y ∂z   ∂ 
 ∂ξ   ∂ξ
   ∂ξ ∂ξ   ∂x 
 ∂   ∂x ∂y ∂z   ∂ 
 =   (5.8.2)
 ∂η   ∂η ∂η ∂η   ∂y 
 ∂   ∂x ∂y ∂z   ∂ 

 ∂ζ   ∂ζ ∂ζ ∂ζ   ∂z 
Where the Jacobian Matrix will be
 ∂x ∂y ∂z 
 ∂ξ
 ∂ξ ∂ξ 
[J ] =  ∂x ∂y ∂z 
(5.8.3)
∂η ∂η ∂η 
 ∂x ∂y ∂z 
 
 ∂ζ ∂ζ ∂ζ 
For an isoparamatric element the coordinates at a point inside the element can be expressed by its
nodal coordinate.
n n n
=x i i
=i 1 =i 1
∑=
N x ; y ∑ N=
y and z ∑ N z i i
=i 1
i i (5.8.4)
 55

Substituting the above equations into the Jacobian matrix for an eight node brick element, we get

 ∂N i ∂N i ∂N i 
 ∂ξ xi ∂ξ 
yi zi
 ∂ξ
 ∂N i x ∂N i ∂N i 
8
[J ] = ∑ yi zi (5.8.5)
i =1
 ∂η i ∂η ∂η 
 ∂N ∂N i ∂N i 
 i xi yi zi 
 ∂ζ ∂ζ ∂ζ 
u 
 
The strain displacement relation is given by {ε } = [B ]{d }, where, {d } =  v  .
 w
 
The displacements in the x, y and z direction are u, v, and w respectively. Let consider the inverse of
Jacobian matrix as
 J 11* J 13* 
J 12*
 * * 
[J ]−1=  J 21 *
J 22
J 23  (5.8.6)
 J 31
* ** 
 J 32
J 33 
Thus, the relation between two coordinate systems can be rewritten as
−1
∂  ∂x ∂y ∂z   ∂   ∂ 
 ∂x   ∂ξ ∂ξ ∂ξ   ∂ξ   ∂ξ 
       J11
* *
J12 *
J13  
∂  ∂x ∂y ∂z   ∂   * *  ∂ 
=  =   ∂η   J 21
*
J 22 J 23   (5.8.7)
 ∂y   ∂η ∂η ∂η    *   ∂η 
 J 31
* *
J 32 J 33 
∂  ∂x ∂y ∂z   ∂    ∂ 
   ∂ζ ∂ζ ∂ζ   ∂ζ   
 ∂z       ∂ζ 

Thus, one can write the following relations

∂u ∂u ∂u ∂u 8
 ∂N i ∂N i ∂N i  8
= J11* + J12* + J13* = ∑  J11* + J12* + J13*  ui = ∑ ai ui
∂x
= ∂ξ ∂η ∂ζ i 1 =
 ∂ξ ∂η ∂ζ  i 1

∂u * ∂u * ∂u * ∂u
8
 * ∂N i * ∂N i * ∂N i 
8
= J 21 + J 22 + J 23 = ∑  J 21 + J 22 + J 23  i ∑ bi ui
u =
∂y
= ∂ξ ∂η ∂ζ i 1 =
 ∂ξ ∂η ∂ζ  i 1

∂u * ∂u * ∂u * ∂u
8
 * ∂N i * ∂N i * ∂N i 
8
= J 31 + J 32 + J 33 = ∑  J 31 + J 32 + J 33  i ∑ ci ui
u =
∂z
= ∂ξ ∂η ∂ζ i 1 =
 ∂ξ ∂η ∂ζ  i 1

Similarly,
∂v 8 ∂v 8 ∂v 8 ∂w 8 ∂w 8 ∂w 8
= ∑ = ai vi ; ∑ i i ∂z ∑
= b v ; = c v ; ∑ i i ∂y ∑
=
a w ; b w= and ∑ ci wi
∂x i 1 = ∂y i 1 = ∂x i 1 = ∂z i 1
i i i i
= i 1 = i 1 =

Using above relations, the strain vector can be written as

 56

 ∂u 
 ∂x 
 
 ∂v 
εx   ∂y   ai 0 0
     0 
ε y   ∂w  0 bi
u 
 ε z   ∂z  8 0 ci   i 
{ε } =
0
=   ∂u =  ∑   v  (5.8.8)
γ xy   + ∂v  i =1  bi ai 0 i 
bi   i 
w
γ yz   ∂y ∂x  0 ci
     
γ zx   ∂v + ∂w   ci 0 ai 
 ∂z ∂y 
 
 ∂u + ∂w 
 ∂z ∂x 

Now, the strain displacement relationship matrix [B] can be identified from the above equation by
comparing it to {ε } = [B ]{d }.

5.8.3 Element Stiffness Matrix

The element stiffness matrix can be generated similar to two dimensional case using the following
relations.
+1 +1 +1

[ k ] = ∫ ∫ ∫ [ B ] T [ D ][ B ]dξ dη dζ J (5.8.9)
−1 −1 −1

The size of the constitutive matrix [D] for solid element will be 6 × 6 and is already discussed in
module 1, lectures 3. For eight node brick element, the size of stiffness matrix will become 24 × 24
as number of nodes in one element is 8 and the degrees of freedom at each node is 3. It is well
established that 2 × 2 × 2 Gauss integration points are the optimal sampling points for eight node
isoparametric brick elements.

5.8.4 Element Load Vector

The forces on an element can be generated due to its self weight or externally applied force which
may be concentrated or distributed in nature. The distributed load may be uniform or non-uniform.
All these types of loads are to redistributed to the nodes using finite element formulation.

5.8.4.1 Gravity load

The load vector due to body forces in general is given by
={Q} ∫[N ]
T
{ X }d Ω (5.8.10)
Ω
 57

where {X } is the body forces per unit volume. The nodal load vector at any node i may be expressed
as
={Qi } ∫[N ] i
T
{ X }d Ω (5.8.11)
Ω

In case of gravity load, the force will act in the global negative Z direction. Therefore,
Ni 0 0   0 
   
[ N i ] =  0 N i 0  and { X } =  0  (5.8.12)
 0 0 N i   
− ρ g 
Here, the mass density of the material is ρ and the acceleration due to gravity is g. Thus, eq.(5.8.11)
will become
 0 
= {Qi } ∫  0  d Ω (5.8.13)
Ω 
− Ni ρ g 
For isoparametric element the, the above expression will become
 0 

+1 +1 +1 
{Qi } = ∫ ∫ ∫  0  J dξdηdζ (5.8.14)
−1 −1 −1
− N ρ 
 i 
Using Gauss Quadrature integration rule, the above expression may be evaluated as
 0 
n n n
 
{ Qi } = ∑∑∑ wi w j wk J ( ξi ,η j ,ζ k )  0  (5.8.15)
=i 1 =j 1 =
k 1 − N ρ g 
 i ( ξi ,η j ,ζ k )
Where, n is the number of nodes in an element. For eight node linear brick element the value of n
will be 8 and the integration order suggested is 2×2×2. Similarly, for twenty node quadratic brick
element, the value of n will be 20 and the integration order suggested is 3×3×3.

5.8.4.2 Surface pressure

Let assume a uniform surface pressure of intensity q is acting normal to the element face. The load
vector due to surface pressure is given by
{Q} = ∫ ∫ [ N s ]T { p}dA (5.8.16)

The nodal load at any node i may be expressed as

{Qi } = ∫ ∫ [ N i s ]T { p}dA (5.8.17)

In case of surface load, the value of [ N is ] in the above equation will become
 58

 N is 0 0 
 
[ N is ] =  0 N i
s
0  (5.8.18)
0
 0 N is 

Here N is is the interpolation function for the node i. For example, the value of N is can be obtained
by substituting ξ =1 in N is for face 1. Thus, the surface pressure is expressed as,
 ql3 
 
{ p} = qm3  (5.8.19)
 qn 
 3
Where, l3 , m3 , n3 are the direction cosines. Thus, eq.(5.8.17) can be expressed using eq.(5.8.19) in the
following form.
 N is ql3 
{Qi } = ∫−1 ∫−1  N is qm3 dA
+1 +1
(5.8.20)
 N s qn 
 i 3
The value of dA is can be evaluated considering the cross product of vectors along the natural
coordinates parallel to the loaded faces of the element. Thus,
dA = e2 × e3 dηdζ (5.8.21)

5.8.5 Stress Computation

Using the relation of {F} = [K]{d}, the unknown nodal displacement vector {d} are calculated in
global coordinate system. Once the nodal displacements are obtained, the strain components as each
node can be computed using strain-displacement relations for each element. Similarly element stress
can be calculated using stress-strain relation. These stresses at Gauss points are extrapolated to the
corresponding nodes of the element to find the nodal stresses. In general, for three dimensional state
of stress there are at least three planes, called principal planes. The corresponding stress vector is
perpendicular to the plane and where there are no normal shear stresses. These three stresses which
are normal to these principal planes are called principal stresses. The principal stresses σ1, σ2, and σ3
are computed from the roots of the cubic equation represented by the determinant of the flowing.

σ x −σ τ xy τ xz
τ xy σ y −σ τ yz =
0 (5.8.22)
τ xz τ yz σ z −σ
The characteristic equation has three real roots σi, due to the symmetry of the stress tensor. The
principal stresses are arranged so that σ1 > σ2 > σ3. The maximum shear stress can be computed
from the following relations.
 59

 σ1 − σ 2 σ2 −σ3 σ 3 − σ1 
τ max = largest of  , and  (5.8.23)
 2 2 2 
These three shear stress components will occur on planes oriented at 450 from the principal planes.
The distortion energy theory suggests that the total strain energy can be divided into two
components. They are (i) volumetric strain energy and (ii) distortion or shear strain energy. It is
anticipated that yield develops if the distortion component exceeds that at the yield point for a simple
tensile test. From the concept of distortion energy theory, the equivalent stress which is historically
known as Von Mises stress are defined as
 (σ 1 − σ 2 ) 2 + (σ 2 − σ 3 ) 2 + (σ 3 − σ 1 ) 2 
σe =   (5.8.24)
 2 
 
The Von Mises stresses offer a measure of the shear or distortional stress in the material. In general,
this type of stress tends to cause yielding in metals.
 60

Worked out Examples

Example 5.1 Calculation of nodal loads on a triangular element

A CST element as shown in Fig. 5.I gets axial loading of (Fx1) 10 kN/m in X direction and (Fy1) 20
kN/m in Y direction. Compute the nodal loads in the element.

Fig. 5.I Distributed loading on a triangular element

From the above figure, the length of sides 1, 2 and 3 are calculated and will be 10, 8 and 6 cm
respectively. First, let consider side 1:
l1

{F }1 = ∫ [ N ] {FΓ } ds
T

Or,
0 0 0 0 0
L 0  L 0  L 
 2  2  2
l1
L 0   Fx1  l1
 L3 0  10  l1
 L3 
{F }=1 ∫  3    ds= ∫0  0    ds= 10 × ∫  ds
0 
0 0   Fy1   0  0  0
0
0 L2  0 L2  0
     
 0 L3   0 L3  0
p p !q !
Putting, ∫ L
1 L2 q ds =
( p + q + 1)!
l , we will get,
 61

0  0   0 
1  1  0.5
     
l1 1  0.1 1  0.5
{F }1 =10 × ×   =10 × ×   = 
2 0  2 0   0 
0  0   0 
     
0  0   0 
Similarly for side 2,
 L1 0  L1 0 0   0 
0 0  0 0  0   0 
     
l2
L 0   Fx 2  l2
 L3 0  0  l2  0   0 
{F }2 = ∫ 3    ds = ∫0  0    ds = 20 × ×   =   kN
0 
0 L1   Fy 2   L1  20  2 1  0.8
0 0 0 0 0   0 
       
 0 L3   0 L3  1  0.8
Since no force is acting on side 3,
0 
0 
 
0 
{F }3 =  
0 
0 
 
0 
Hence, the nodal load vector in all the nodes in x and y directions will become,
0 
0.5
 
0.5
{F } =   kN.
0.8
0 
 
0.8