NOVEMBER 30, 2021

TWO-DIMENSIONAL ELEMENT

LT COL MD JAHIDUL ISLAM, PHD

DEPARTMENT OF CIVIL ENGINEERING
MIST, Mirpur Cantonment
CE 6109: Finite Element Methods I Chapter 5

5 TWO-DIMENSIONAL ELEMENTS
5.1 INTRODUCTION

One-dimensional elements are introduced in the previous chapters where truss and frame
elements were formulated. Constant and linear strain elements have been introduced and also
considerations like h – and p- convergence, normalized coordinates, isoparametric elements,
numerical integration and coordinate transforms were presented.
For the case of 2-dimensional problems, 2-D elements are needed. In this chapter, some
commonly used elements will be introduced. These include the triangular, quadrilateral and
axisymmetric elements. Modelling considerations, substructuring and transition elements will
also be elaborated.

5.2 CONSTANT STRAIN TRIANGULAR ELEMENT

Constant strain triangular (CST) is the simplest 2-D element.

To proceed to solve problems using such element, we must derive the

• Displacement interpolation function N,

• Strain shape function B,
• Element stiffness matrix K, and
• Nodal load vector P.

CST element has three nodes at the vertices of the triangle, which are numbered around the element in
the counter-clockwise direction.

Each node has two degree of freedom (can move in the x and y directions).

by Dr. Md. Jahidul Islam Page 1

CE 6109: Finite Element Methods I Chapter 5

The displacement interpolation functions (u and v) are assumed to be linear functions within the
element,

u ( x, y ) = 1 +  2 x +  3 y (5.1)
v ( x, y ) = 1 +  2 x + 3 y (5.2)

Where, α and β are constants. node has two degree of freedom (can move in the x and y directions).

Substituting the nodal coordinates and displacement give,

u1 = 1 +  2 x1 +  3 y1
u2 = 1 +  2 x2 +  3 y2
u3 = 1 +  2 x3 +  3 y3
(5.3)
v1 = 1 +  2 x1 + 3 y1
v2 = 1 +  2 x2 +  3 y2
v3 = 1 +  2 x3 + 3 y3

Solving these equations, we can find the coefficients α1, α2, … β3 in terms of nodal displacements and
coordinates.

Substituting the coefficients into the displacement Eq. (5.3) and rearranging in matrix form we obtain,

 u1 
v 
 1
u   N1 0 N2 0 N3 0  u2 
 =  
N 3   v2 
(5.4)
v   0 N1 0 N2 0
u3 
 
 v3 

The displacement shape functions (linear functions in x and y) are,

by Dr. Md. Jahidul Islam Page 2

CE 6109: Finite Element Methods I Chapter 5

N1 =
1
2A
( x2 y3 − x3 y2 ) + ( y2 − y3 ) x + ( x3 − x2 ) y
N2 =
1
2A
( x3 y1 − x1 y3 ) + ( y3 − y1 ) x + ( x1 − x3 ) y (5.5)
N3 =
1
2A
( x1 y2 − x2 y1 ) + ( y1 − y2 ) x + ( x2 − x1 ) y
where

1 x1 y1 
A = det 1 x2 y2  = area of the triangle
1
2
1 x3 y3 

Strains are found to be,

u v u v
x = = 2 , y = = 3 ,  xy = + = 3 + 2
x y y x

Which are constant, and hence, the element is named as “constant strain element” (CST).

In Matrix form:

  x    x 0   x 0 
   u  

ε =   y  =  0  y    =  0  y   N q = B q (5.6)
   y  x   v   y  x 
 xy     

The strain shape function,

 y23 0 y31 0 y12 0

1 
B= 0 x32 0 x13 0 x21  (5.7)
2A 
 x32 y23 x13 y31 x21 y12 

where, xij = xi − x j and yij = yi − y j (i, j = 1, 2,3)

The element stiffness matrix for the CST element,

K =  BT EBdV = tA ( BT EB ) (5.8)
V

Where, t is the thickness of the element.

K is a 6 x 6 symmetric matrix.

Load vector: (T for traction, b for body forces),

Tx  bx 
P = p +  NT bdV = p + pT + pb = p +  NT   hdS +  NT   hdA
V A Ty  A by 

by Dr. Md. Jahidul Islam Page 3

CE 6109: Finite Element Methods I Chapter 5

Nodal forces:

 F1x 
F 
 1y 
 F2 x 
p= 
 F2 y 
 F3 x 
 
 F3 y 

Forces distributed along the edge of element:

• T = force/unit A along the edge

• h = thickness of element
• S = along the edge

Tx  Tx 
 We =   u  v   hdS =  qT  NT   hdS
S
T  y T S  y

5.2.1 Formulation using Natural Coordinates

In previous section, the equations are sufficient to solve problems using CST within FEM. However,
for programming and numerical computation, a normalized form may be more convenient and hence
given in this section.

Area of triangle: 3

1 x y
2 A3 = 1 x1 y1  = x1 y2 − x2 y1 + xy12 + yx21
1 x2 y2 

Now, 1-- = A3/A = N3

by Dr. Md. Jahidul Islam Page 4

CE 6109: Finite Element Methods I Chapter 5

The CST element can be formularized using the natural coordinates (ξ, η).

The displacement shape function,

N1 =  , N 2 =  , N 3 = 1 −  −  (5.9)

Notice that
N1 + N 2 + N 3 = 1 (5.10)

The plot for shape function N1 is shown in the following figure.

N2 and N3 have similar features.

Two coordinate systems for the element:

 The global coordinates (x, y)

 The natural coordinates (ξ, η).

The relation between the two is given by,

x = N1 x1 + N 2 x2 + N3 x3
(5.11)
y = N1 y1 + N 2 y2 + N 3 y3

or

by Dr. Md. Jahidul Islam Page 5

CE 6109: Finite Element Methods I Chapter 5

x = x13 + x23 + x3
(5.12)
y = y13 + y23 + y3

where, xij = xi − x j and yij = yi − y j (i, j = 1, 2,3)

Displacement u or v on the element can be viewed as functions of (x, y) or (ξ, η).

Using the chain rule for derivatives we have,

 u   x y   u   u 
      
  x    x 
      = J  
 = (5.13)
 u   x y   u   u 
     
   y   y 

where, J is called the Jacobian matrix of the transformation.

From Eq. (5.12), we have,

x y13 
J =  13
y23 
(5.14)
 x23
1  y23 − y13  1  y23 − y13 
J −1 = − x =
x13  2 A  − x23 x13 
(5.15)
x13 y23 − x23 y13  23

From Eqs. (5.13-5.15), we obtain,

 u   u 
 x  1  y − y13     1  y23 − y13   u1 − u3 
 u  =  =
23
  
x13   u  2 A  − x23 x13  u2 − u3 
(5.16)
  2 A  − x23
 y    

Similarly,

 v 
 x  1  y − y13   v1 − v3 
 v  =
23
  
x13  v2 − v3 
(5.17)
  2 A  − x23
 y 

Using Eqs. (5.16 & 5.17), we obtain the strain-displacement matrix,

ε = Bq
Where,

by Dr. Md. Jahidul Islam Page 6

CE 6109: Finite Element Methods I Chapter 5

 y23 0 y31 0 y12 0

1 
B= 0 x32 0 x13 0 x21  (5.18)
2A 
 x32 y23 x13 y31 x21 y12 

Limitations of CST Element

 In bending problems, the mesh of CST elements will produce a model that is stiffer than the
actual problem.
 The CST model converges very slowly to the exact solution. This is partly due to the element
predicting only constant stress within each element, when for a bending problem, the stress
actually varies linearly through the depth of the beam.
 For a beam subjected to pure bending, the CST has a spurious or false shear stress and hence a
spurious shear strain in parts of the model that should not have any shear stress or shear strain.
This spurious shear strain absorbs energy; therefore, some of the energy that should go into
bending is lost. The CST is then too stiff in bending, and the resulting deformation is smaller than
actually should be.
 This phenomenon of excessive stiffness developing in one or more modes of deformation is
sometimes described as shear locking or parasitic shear.
 Furthermore, in problems where plane strain conditions exist (recall this means when z = 0) and
the Poisson’s ratio approaches 0.5, a mesh can actually lock, which means the mesh then cannot
deform at all.

Application of the CST Element

 Use in area where the strain gradient is small.

 Use in mesh transition areas (fine mesh to coarse mesh).
 Avoid using CST in stress concentration or other crucial areas in the structure, such as edges of
holes and corners.
 Recommended for quick and preliminary FE analysis of 2-D problems.

5.3 LINEAR STRAIN TRIANGULAR ELEMENT

 Linear strain triangular (LST) element has six nodes: three corner nodes and three mid-side nodes.

 Each node has two degree of freedom (can move in the x and y directions).

 This element is also known as quadratic triangular element.

by Dr. Md. Jahidul Islam Page 7

CE 6109: Finite Element Methods I Chapter 5

Figure 5.3-1 Relation between type of plane triangular element and polynomial coefficients based on
a Pascal triangle.

The displacement interpolation functions (u and v) are assumed to be quadratic functions within the
element,

u ( x, y ) = 1 +  2 x +  3 y +  4 x 2 +  5 xy +  6 y 2 (5.19)
v ( x, y ) = 1 +  2 x + 3 y +  4 x 2 + 5 xy +  6 y 2 (5.20)

Where, α and β are constants.

The strains are found to be,

 x =  2 + 2 4 x +  5 y
 y = 3 + 5 x + 2 6 y (5.21)
 xy = ( 3 +  2 ) + ( 5 + 2 4 ) x + ( 2 6 + 5 ) y

by Dr. Md. Jahidul Islam Page 8

CE 6109: Finite Element Methods I Chapter 5

Which are linear functions.

Thus comes the “linear strain triangle” (LST).

It provides better results than the CST.

In the natural coordinate system, the six shape functions for the LST element are,

N1 =  ( 2 − 1)
N 2 =  ( 2 − 1)
N 3 =  ( 2 − 1)
(5.22)
N 4 = 4
N 5 = 4
N 6 = 4

In which,  = 1 −  −  (5.23)

Each of these six shape functions represents a quadratic form on the element as shown in the figure.

Displacements can be written as,

6 6
u =  Ni ui , v =  Ni vi (5.24)
i =1 i =1

The element stiffness matrix is still given by

K =  BT EBdV (5.25)
V

But here, BT EB is quadratic in x and y.

Figure below shows a specific LST and its coordinates. The triangle is of base dimension b
and height h, with mid-side nodes.

by Dr. Md. Jahidul Islam Page 9

CE 6109: Finite Element Methods I Chapter 5

u ( x, y ) = 1 +  2 x +  3 y +  4 x 2 +  5 xy +  6 y 2 (5.19)
v ( x, y ) = 1 +  2 x + 3 y +  4 x 2 + 5 xy +  6 y 2 (5.20)

We can calculate the coefficients a1 through a6 by evaluating the displacement u at each of the six
known coordinates of each node as follows:

u1 = u ( 0, 0 ) = a1
u2 = u ( b, 0 ) = a1 + a2b + a4b 2
u3 = u ( 0, h ) = a1 + a3 h + a6 h 2
2 2
b h b h b bh h
u4 = u  ,  = a1 + a2 + a3 + a4   + a5 + a6  
2 2 2 2 2 4 2

5.4 BASIC CONCEPTS OF PLANE STRESS AND PLANE STRAIN

Depending on the structure and loading conditions, it is not always necessary to solve a problem using
a fully 3-D model.

With negligible loss in accuracy, it is often faster to solve using a 2-D model. Two-dimensional
(planar) elements are defined by three or more nodes in a two-dimensional plane (that is, x-y).

There are 3 distinct 2-D models that can be used, namely,

• Plane stress
• Plane strain, and
• Axisymmetric

5.4.1 Plane Stress

Plane stress is defined to be a state of stress in which the normal stress and the shear stresses directed
perpendicular to the plane are assumed to be zero. For instance, in Figures 5.4–1(a) and 5.4–1(b), the

by Dr. Md. Jahidul Islam Page 10

CE 6109: Finite Element Methods I Chapter 5

plates in the x-y plane shown subjected to surface tractions T (pressure acting on the surface edge or
face of a member in units of force/area) in the plane are under a state of plane stress; that is, the
normal stress and the shear stresses assumed to be zero.

 z =  yz =  zx = 0

Figure 5.4-1 Plane stress problems: (a) plate with hole; (b) plate with fillet.

Generally, members that are thin (those with a small z dimension compared to the in-plane x and y
dimensions) and whose loads act only in the x-y plane can be considered to be under plane stress.

5.4.2 Plane Strain

Plane strain is defined to be a state of strain in which the strain normal to the x-y plane and the shear
strains are assumed to be zero.

 z =  yz =  zx = 0

The assumptions of plane strain are realistic for long bodies (say, in the z direction) with constant
cross-sectional area subjected to loads that act only in the x and/or y directions and do not vary in the
z direction. Some plane strain examples are shown in Figure 5.4-2. In these examples, only a unit
thickness (1 in. or 1 ft) of the structure is considered because each unit thickness behaves identically
(except near the ends). The finite element models of the structures in Figure 6–2 consist of
appropriately discretized cross sections in the x-y plane with the loads acting over unit thicknesses in
the x and/or y directions only.

by Dr. Md. Jahidul Islam Page 11

CE 6109: Finite Element Methods I Chapter 5

Figure 5.4-2 Plane strain problems: (a) dam subjected to horizontal loading; (b) pipe subjected to a
vertical load.

5.4.3 Material Constitutive (Stress-Strain) Relationships

For linear isotropic elasticity,

 xx − ( yy +  zz ) 
1
 xx =
E 
1
 yy =  yy − ( xx +  zz ) 
E
 zz =  zz − ( xx +  yy ) 
1
E

1 E
 xy = 2 xy =  xy where, G =
G 2 (1 + )

In matrix form,

by Dr. Md. Jahidul Islam Page 12

CE 6109: Finite Element Methods I Chapter 5

 1 − − 0 0 0 
 E E E 
 xx     
   − E
1 − 0 0 0   xx 
E E   yy
 yy     
 zz   − E − 1 0 0 0   
E E  zz
 =  
 xy   0 0 0 1
G
0 0    xy 
 yz      yz 
   0 0 0 0 1 0  
 zx   G 
  zx 
 0 0 0 0 0 1 
G

The above can be re-written in terms of stress components as,

2G
 xx =
1 − 2
( xx +  yy +  zz ) + 2G xx
2G
 yy =
1 − 2
( xx +  yy +  zz ) + 2G yy
2G
 xx =
1 − 2
( xx +  yy +  zz ) + 2G zz
E
 xy =  xy = 2G xy where, G =
2 (1 + )

In matrix form,
1 −   0 0 0 
 xx      xx 
   1 −  0 0 0   
  yy
   1 − 0 0 0   yy 
 zz  E    zz 
 =  0 0 0 1 − 2 0 0   
  xy (1 +  )(1 − 2 ) 
2
  xy 
  yz   0 0 0 0 1 − 2 0   yz 
   2  
  zx   0 0 0 0 0 1 − 2   zx 
 2
or,  = E

5.4.4 Two-Dimensional Specialization

by Dr. Md. Jahidul Islam Page 13

CE 6109: Finite Element Methods I Chapter 5

5.4.4.1 Plain Strain

For plane strain case where,  zz =  yz =  zx = 0

( )
This implies,  zz =   xx +  yy ,  xz = 0 and  yz = 0

 1 − 2   − (1 +  ) 
 xx =    xx +    yy
 E   E 
 − (1 +  )   1 − 2 
 yy =   +
 xx    yy
 E   E 
1
 xy =  xy
2G

  xx  1 −  0  
  E    xx 
  yy  = 1 + 1 − 2   1 − 0    yy 
  ( )( )  
 xy  0 1 − 2  xy
 0 2   
or,  = E

5.4.4.2 Plane Stress

For plane stress case where,  zz =  zx =  yz = 0


 zz = −
1 −
( xx +  yy )

1  − 
 xx =    xx +    yy
E E
 −  1
 yy =    xx +    yy
E E
1
 xy =  xy
2G

  xx  1  0  
  E    xx 
  yy  =  1 0    yy 
  (1 −  2
)   
 xy   0 0 1 −  xy
2   
or,  = E

by Dr. Md. Jahidul Islam Page 14

CE 6109: Finite Element Methods I Chapter 5

5.5 Example
5.5.1 Example 5.1
Evaluate the stiffness matrix for the element shown in figure below. The coordinates are shown in
units of inches. Assume plane stress conditions. Let E = 30 x 106 psi, v = 0.25, and thickness t = 1 in.
Assume the element nodal displacements have been determined to be u1 = 0.0, v1 = 0.0025 in., u2 =
0.0012 in., v2 = 0.0, u3 = 0.0, and v3 = 0.0025 in. Determine the element stresses.

Solution:

Stress-strain relationship

  xx  1  0  
  E    xx 
  yy  =  1 0    yy 
  (1 −  2
)   
 xy   0 0 1 −  xy
2   
or,  = E

Stiffness matrix:

K =  BT EBdV = tA ( BT EB )
V

Strain-displacement relationship,

ε = Bq
Where,

 y23 0 y31 0 y12 0

1 
B= 0 x32 0 x13 0 x21 
2A 
 x32 y23 x13 y31 x21 y12 

where, xij = xi − x j and yij = yi − y j (i, j = 1, 2,3)

by Dr. Md. Jahidul Islam Page 15

CE 6109: Finite Element Methods I Chapter 5

Area of the triangle:

1 x1 y1  1 0 −1
A = det 1 x2 y2  = det 1 2 0  = ( 2 + 2 ) = 2

1 1 1
2 2 2
1 x3 y3  1 0 1 

Step 1:

x13 = x1 − x3 = 0 − 0 = 0
x32 = x3 − x2 = 0 − 2 = −2
x21 = x2 − x1 = 2 − 0 = 2

y23 = y2 − y3 = 0 − 1 = −1
y31 = y3 − y1 = 1 + 1 = 2
y12 = y1 − y2 = −1 − 0 = −1

Step 2: Strain-displacement relationship:

ε = Bq
 −1 0 2 0 −1 0 
1 
B=  0 −2 0 0 0 2 
2 ( 2)
 −2 −1 0 2 2 −1

Step 3: for plane stress conditions:

1  0   1 0.25 0 
E   30  10 6  
E=  1 0 =  0.25 1 0  psi
(1 − 2 )   (1 − 0.25 ) 
2

 0 0 1 − 
2 

0 0 1 − 0.25 
2

Step 4: element stiffness matrix:

K =  BT EBdV = tA ( BT EB )
V

 −1 0 −2 
 0 −2 1 
  1 0.25 0 
 −1 0 2 0 −1 0 
1 2  30 10  2 0 0 
6   1 
K= 2   0.25 1 0  0 −2 0 0 0 2 
2 ( 2 ) (1 − 0.25 )  0 0 2   ( )
2 2 
0 1 − 0.25   −2 −1 0 2 2 −1
 −1 0 2   0 2
 
 0 2 −1

by Dr. Md. Jahidul Islam Page 16

CE 6109: Finite Element Methods I Chapter 5

Step 5: stress-strain relationship:

 = E = EBq
 0.0 
 
  xx   1  − −  0.0025 
0.25 0  1 0 2 0 1 0 
  30 106   1    0.0012 

 yy  =  0.25 1 0   0 −2 0 0 0 2   0.0 
  (1 − 0.25 2
)   2 ( 2 ) 
 −2 −1 0 2 2 −1  
 xy   0 0 1 − 0.25  
2 0.0 
 
 0.0025 

Performing the matrix triple product,

  xx   19, 200 
   
  yy  =  4,800  psi
    −15, 000 
 xy   

by Dr. Md. Jahidul Islam Page 17

CE 6109: Finite Element Methods I Chapter 5

5.5.2 Example 5.2

For a thin plate subjected to the surface traction shown in figure below, determine the nodal
displacements and the element stresses. The plate thickness t = 1 in., E = 30 x 106 psi,  = 0.30.

Solution:

• To illustrate the finite element method solution for the plate, we first discretize the plate into two
elements.
• Coarseness of the mesh will not yield as true a predicted behavior of the plate as would a finer
mesh, particularly near the fixed edge.
• However, we will use a coarse discretization for simplicity.
• The original tensile surface traction has been converted to nodal forces as follows:
1
F = TA
2
F = (1000 psi ) (110 in 2 )
1
2
F = 5, 000 lb

The governing global matrix equation is

F = Kq

by Dr. Md. Jahidul Islam Page 18

CE 6109: Finite Element Methods I Chapter 5

 F1x   d1x   R1x   0 

F  d     0 
 1y   1 y   R1 y   
 F2 x   d 2 x   R2 x   0 
       
 F2 y  d 2 y   R2 y   0 
  = K      = K  
 F3 x   d3 x  5000   d3 x 
 F3 y   d3 y   0   d3 y 
       
 F4 x   d 4 x  5000  d4 x 
 F4 y  d 4 y   0  d 4 y 
       

Assemblage of the stiffness matrix (Element 1):

x13 = x1 − x3 = 0 − 20 = −20
x32 = x3 − x2 = 20 − 0 = 20
x21 = x2 − x1 = 0 − 0 = 0

y23 = y2 − y3 = 10 − 10 = 0
y31 = y3 − y1 = 10 − 0 = 10
y12 = y1 − y2 = 0 − 10 = −10

 y23 0 y31 0 y12 0

1 
B= 0 x32 0 x13 0 x21 
2A 
 x32 y23 x13 y31 x21 y12 

 0 0 10 0 −10 0 
1 
B=  0 −20 0 0 0 20 
2 (100 )
 −20 0 0 10 20 −10 

1 1
Area: A = bh = ( 20 in )(10 in ) = 100 in 2
2 2

by Dr. Md. Jahidul Islam Page 19

CE 6109: Finite Element Methods I Chapter 5

For plane stress conditions,

1  
0  1 0.30 0 
E   30 106 
E=  1 0 = 0.30 1 0  psi
(1 − 2 )   (1 − 0.30 )
2 

 0 0 1 − 2 
 0 0 0.35

Now,

K = tA ( BT EB )

 0 0 −20   0 0 −7 
 0 −20 0   −6 −20 0 
   1 0.3 0  
30  10 6
 10 0 0   = 0.15 10
6
 10 3 0 
B E=
T
  0.3 1 0  
200 ( 0.91)  0 0 10    0.91  0 0 3.5 
 0 0 0.35
 −10 0 20   −10 −3 7 
   
 0 20 −10   6 20 −3.5

 0 0 −7 
 −6 −20 0 
  0 0 10 0 −10 0 
0.15 106  10 0  1 
20 
3
K = (1)(100 )    0 −20 0 0 0
0.91  0 0 3.5  2 (100 )
 −20 0 0 10 20 −10 
 −10 −3 7 
 
 6 20 −3.5

 140 0 0 −70 −140 70 

 0 400 −60 0 60 −400 

75000  0 −60 100 0 −100 60 
K=  
0.91  −70 0 0 35 70 −35 
 −140 60 −100 70 240 −130 
 
 70 −400 60 −35 −130 435 

Assemblage of the stiffness matrix (Element 2):

by Dr. Md. Jahidul Islam Page 20

CE 6109: Finite Element Methods I Chapter 5

x13 = x1 − x3 = 0 − 20 = −20
x32 = x3 − x2 = 20 − 20 = 0
x21 = x2 − x1 = 20 − 0 = 20

y23 = y2 − y3 = 10 − 0 = 10
y31 = y3 − y1 = 0 − 0 = 0
y12 = y1 − y2 = 0 − 10 = −10

 −10 0 10 0 0 0
1 
B=  0 0 0 −20 0 20 
2 (100 )
 0 −10 −20 10 20 0 

1 1
Area: A = bh = ( 20 in )(10 in ) = 100 in 2
2 2

For plane stress conditions:

1  
0  1 0.30 0 
E   30 106 
E=  1 0 = 0.30 1 0  psi
(1 − 2 )   (1 − 0.30 )
2 

 0 0 1 − 2 
 0 0 0.35

(
Now, K = tA BT EB )
 −10 0 0   −10 −3 0 
 0 0 −10   0 0 −3.5
  1 0.3 0  
30 106  10 0 −20    = 0.15 10
6
 10 3 −7 
B E=
T
  0.3 1 0  
200 ( 0.91)  0 −20 10    0.91  −6 −20 3.5 
 0 0 0.35
 0 0 20   0 0 7 
   
 0 20 0   6 20 0 

 −10 −3 0 
 0 0 0 
  0 0 10 0 −10 0 
0.15 106  10 0  1 
20 
3
K = (1)(100 )   0 −20 0 0 0
0.91  −6 −20 3.5  2 (100 ) 
 −20 0 0 10 20 −10 
 0 0 7 
 
 6 20 −3.5

by Dr. Md. Jahidul Islam Page 21

CE 6109: Finite Element Methods I Chapter 5

 100 0 −100 60 0 −60 

 0 35 70 −35 −70 0 

75000  −100 70 240 −130 −140 60 
K=  
0.91  60 −35 −130 435 70 −400 
 0 −70 −140 70 140 0 
 
 −60 0 60 −400 0 400 

Using superposition of the element stiffness matrices, we obtain the total global stiffness matrix:

 48 0 −28 14 0 −26 −20 12 

 0 87 12 −80 −26 0 14 −7 

 −28 12 48 −26 −20 14 0 0 
 
375000  14 −80 −26 87 12 −7 0 0  lb
K=
0.91  0 −26 −20 12 48 0 −28 14  in
 
 −26 0 14 −7 0 87 12 −80 
 −20 14 0 0 −28 12 48 −26 
 
 12 −7 0 0 14 −80 −26 87 

The governing global matrix equation is

F = Kq

 R1x   48 0 −28 14 0 −26 −20 12   0 

 R   0  
 1y   87 12 −80 −26 0 14 −7   0 
 R2 x   −28 12 48 −26 −20 14 0 0  0 
    
 R2 y  375000  14 −80 −26 87 12 −7 0 0  0 
 =  
5000  0.91  0 −26 −20 12 48 0 −28 14   d3 x 
 
 0   −26 0 14 −7 0 87 12 −80   d3 y 
   −20 14  
5000  0 0 −28 12 48 −26   d 4 x 
 
 0  −7 14 −80 −26 87  d 4 y 
   12 0 0

Applying the support or boundary conditions by eliminating rows and columns corresponding to
displacement matrix rows and columns equal to zero [namely, rows and columns 1–4, we obtain

5000   48 0 −28 14   d3 x 
 0   0  
  375000  87 12 −80   d3 y 
 =  
5000  0.91  −28 12 48 −26   d 4 x 
 0   
 14 −80 −26 87  d 4 y 

Solving for the displacements, we obtain

by Dr. Md. Jahidul Islam Page 22

CE 6109: Finite Element Methods I Chapter 5

 d 3 x  609.6 
d   
 3 y   4.2  −6
  =    10 in.
 d 4 x  663.7 
 d 4 y  104.1 

Stresses in element 1:

 = E = EBq
 0.0 
 
 0.0 
  xx   1 0.30 0   0 0 10 0 −10 0 



=
30 106   
10−6 
−   609.6 
 yy 
( − )
2 
0.30 1 0  2 (100 )  0 20 0 0 0 20   4.2 
0 10 20 −10   
  1 0.30
 xy   0 0 0.35  −20 0
 0.0 
 
 0.0 

  xx  1005 
   
   yy  =  301  psi
    2.4 
 xy   

by Dr. Md. Jahidul Islam Page 23

CE 6109: Finite Element Methods I Chapter 5

5.6 BILINEAR QUADRILATERAL (Q4) ELEMENT

 4 node, 8 dofs

 Bi-linear interpolation function

u ( x, y ) = 1 +  2 x +  3 y +  4 xy (5.26)
v ( x, y ) = 1 +  2 x + 3 y +  4 xy (5.27)

u
x = = 2 + 4 y
x
v
 y = =  2 +  4 x, (5.28)
y
u v
 xy = + =  3 +  4 x +  2 +  4 y
y x

 For Q4, plane section remains plane under deformation.

 Curvature is not model within one element.

❖ Right angles in the element are not preserved under moment loading.

❖ Spurious shear stress (if few elements are used).

❖ Too stiff in bending.

 However, Q4 element converges properly with mesh refinement and for most problems is better
than CST.

by Dr. Md. Jahidul Islam Page 24

CE 6109: Finite Element Methods I Chapter 5

5.6.1 Displacement Shape Function

 u1 
 
 v1 
 u2 
 
 u ( x, y )   N1 0 N2 0 N3 0 N4 0   v2 
 = = Nq
N 4   u3 
(5.29)
 v ( x, y )   0 N1 0 N2 0 N3 0
 
 v3 
u 
 4
v 
 4

N1 =
( a − x )( b − y ) N2 =
( a + x )( b − y )
4ab 4ab

N3 =
( a + x )( b + y ) N4 =
( a − x )( b + y )
4ab 4ab

 u1 
 
 v1 
 u2 
  x    x 0   
   N 0   v2 
 y   1
0 N2 0 N3 0 N4
 y  =  0 N 4   u3 
(5.30)
    y  x  
0 N1 0 N2 0 N3 0
 xy     
 v3 
u 
 4
v 
 4

by Dr. Md. Jahidul Islam Page 25

CE 6109: Finite Element Methods I Chapter 5

 u1 
 
 N1 N 2 N 3 N 4   v1 
 0 0 0 0 u 
  x   x x x x  2 
   N1 N 2 N 3 N 4   v2 
 y  =  0 y
0
y
0
y
0 
y   u3 
= Bq (5.31)
    
 xy   N N1 N 2 N 2 N 3 N 3 N 4 N 4   v3 
1
 
 y x y x y x y x   u4 
 
v 
 4

5.6.2 Iso-paramatric Q4 Element

 Global coordinates:

4 4
X =  Ni X i , Y =  NiYi (5.32)
i =1 i =1

 Displacements in global X, Y directions:

4 4
u =  Ni ui , v =  Ni vi (5.33)
i =1 i =1

N1 =
( a − x )( b − y ) N2 =
( a + x )( b − y )
4ab 4ab

N3 =
( a + x )( b + y ) N4 =
( a − x )( b + y )
4ab 4ab

For a = 1 and b = 1, the displacement shape functions in natural coordinate are,

1 1
N1 = (1 − s )(1 − t ) N 2 = (1 + s )(1 − t )
4 4
1 1
N 3 = (1 + s )(1 + t ) N 4 = (1 − s )(1 + t )
4 4

by Dr. Md. Jahidul Islam Page 26

CE 6109: Finite Element Methods I Chapter 5

Displacement:

 x1 
 
 y1 
 x2 
 
 x   N1 0 N2 0 N3 0 N4 0   y2 
 = 0 N 4   x3 
(5.34)
 y  N1 0 N2 0 N3 0
 
 y3 
x 
 4
y 
 4

Strain-Displacement Relationship:

 u   y u y u   y  y  
 x  1  −  −
1  
  =  t s s t  =  t s s t  u (5.35)
 u  J  x u x u  J  x  x  
 y   −   − 
   s t t s   s t t s 

Where, J is Jacobian.

v v
Similar expression can be obtained for and .
x y

   y  y  
 0  t s − s t 0 
  x   x   
    u  1  x  x    u 
 y  =  0   =
y   v  J 
0 −   = Bq
s t t s   v 
(5.36)
    
 xy      x  − x  y  y  
− 
   s t t s t s s t 
 y x 

The strain shape function:

 B11 0 B13 0 B15 0 B17 0 

B =  0 B22 0 B24 0 B26 0 B28  (5.37)
 B22 B11 B24 B13 B26 B15 B28 B17 

by Dr. Md. Jahidul Islam Page 27

CE 6109: Finite Element Methods I Chapter 5

1 1
B11 = B32 = ( y24 + sy43 + ty32 ) B13 = B34 = ( y31 + sy34 + ty14 )
8J 8J
1 1
B15 = B36 = ( y42 + sy12 + ty41 ) B17 = B38 = ( y13 + sy21 + ty23 )
8J 8J
1 1
B22 = B31 = ( x42 + sx34 + tx23 ) B24 = B33 = ( x13 + sx43 + tx41 )
8J 8J
1 1
B26 = B35 = ( x24 + sx21 + tx14 ) B28 = B37 = ( x31 + sx12 + tx32 )
8J 8J

Where, xij = xi − x j yij = yi − y j

The Jacobian:

 x y x y  4 N i 4 N i 4
N i 4 N i
J = −  
= xi y −  xi  yi
 s t t s  i =1 s i =1 t i =1 t i =1 s
i

4 
4
 N N j N i N j  
=   yi  i −  xj 
i =1 j =1   s t s t  
 0 1− t t − s s − 1   x1 
t −1  
1 0 s + 1 − s − t   x2 
= y1 y2 y3 y4  (5.38)
8  s − t −s − 1 0 t + 1   x3 
  
1 − s s + t −t − 1 0   x4 

Element Stiffness Matrix:

K =  BT EB.dV =  BT EB.hdA =  BT EB.h J ds.dt (5.39)

V A A

5.7 Quadrilateral 8-Node (Q8) Element

 8 node: 4 nodes at the corner and 4 nodes at the middle of the side.

 Global coordinates and displacements in global direction.

by Dr. Md. Jahidul Islam Page 28

CE 6109: Finite Element Methods I Chapter 5

8 8
X =  Ni X i , Y =  NiYi
i =1 i =1

8 8
u =  Ni ui , v =  Ni vi
i =1 i =1

Displacement Shape Functions:

1 1
N1 = − (1 − s )(1 − t )(1 + s + t ) N 2 = − (1 + s )(1 − t )(1 − s + t )
4 4
1 1
N 3 = − (1 + s )(1 + t )(1 − s − t ) N 4 = − (1 − s )(1 + t )(1 + s − t )
4 4
N 5 = (1 − s 2 ) (1 − t ) N 6 = (1 + s ) (1 − t 2 )
1 1
2 2
N 7 = (1 − s 2 ) (1 + t ) N 8 = (1 − s ) (1 − t 2 )
1 1
2 2

 Lines of zero method to obtain shape functions:

For N1 : 1 − s = 0 1 − t = 0 s + t + 1 = 0
1
N1 = − (1 − s )(1 − t )(1 + s + t )
4

For N 3 : 1 + s = 0 1 + t = 0 s + t − 1 = 0
1
N3 = − (1 + s )(1 + t )(1 − s − t )
4

by Dr. Md. Jahidul Islam Page 29

CE 6109: Finite Element Methods I Chapter 5

5.8 Numerical Integration

In this section, we will describe a method for numerical evaluation of definite integrals, because it has
proven most useful for finite element work.

Consider the problem of numerically evaluating a one-dimensional integral of the form,

I =  f ( s ) ds
1 (5.40)
−1

The Gaussian quadrature approach for evaluating I is given subsequently. This method has proven
most useful in finite element work. Extension to integrals in two and three dimensions follows readily.

Consider the n-point approximation,

I =  f ( s ) ds  w1 f ( s1 ) + w2 f ( s2 ) + + wn f ( sn )
1 (5.41)
−1

Where, w1, w2, …, wn are the weights and s1, s2, …, and sn are the sampling points or Gauss points.
The idea behind Gaussian quadrature is to select the n Gauss points and n weights such that the above
Eqn. provides an exact answer for polynomials f(s) of as large a degree as possible. In other words,
the idea is that if the n-point integration formula is exact for all polynomials up to as high a degree as
possible, then the formula will work well even if f is not a polynomial. Sampling points are located
symmetrically with respect to the center of the interval. Symmetrically paired points are given the
same weight wi.

5.8.1 One-Point Formula

Consider the formula with n = 1 as,

 f ( s ) ds  w f ( s )
1 (5.42)
−1 1 1

Since there are two parameters, w1 and s1, we consider requiring the formula in Eqn. 5.42 to be exact
when f(s) is a polynomial of order 1. Thus, if f(s) = a0 + a1s, then we require

( a0 + a1s ) ds − w1 f ( s1 ) = 0 (5.43a)
1
Error = 
−1

Error = 2a0 − w1 ( a0 + a1s ) = 0 (5.43b)

Error = a0 ( 2 − w1 ) − w1a1s1 = 0 (5.43c)

From Eqn. 5.43c, we see that the error is zeroed if,

w1 = 2; s1 = 0 (5.44)

For any general f, then, we have,

 f ( s ) ds  2 f ( 0 )
1 (5.45)
−1
Which is seen to be familiar midpoint rule (Fig. 5.8).

by Dr. Md. Jahidul Islam Page 30

CE 6109: Finite Element Methods I Chapter 5

Figure 5.8 Gaussian quadrature using one sampling point

5.8.2 Two-Point Formula

Consider the formula with n = 2 as,

 f ( s ) ds  w f ( s ) + w f ( s )
1 (5.46)
−1 1 1 2 2

We have four parameters to choose: w1, w2, s1 and s2. We can therefore expect the formula in Eqn.
5.46 to be exact when f(s) is a cubic polynomial. Thus, if f(s) = a0 + a1s+ a2s2+ a3s3, then we require

(a + a1s + a2 s 2 + a3 s 3 ) ds −  w1 f ( s1 ) + w2 f ( s2 ) 
(5.47)
1
Error =  0
−1

Requiring zero error yields,

w1 + w2 = 2 (5.48)
w1s1 + w2 s2 = 0
w1s12 + w2 s22 = 2
3
ws +w s =0
3
1 1
3
2 2

These nonlinear equations have the unique solution:

w1 = w2 = 1 (5.49)
− s1 = s2 = 1 3 = 0.5773502691
From this solution, we can conclude that n-point Gaussian quadrature will provide an exact answer if f
is a polynomial of order (2n – 1) or less.

Table 5–1 gives appropriate sampling points and weighting coefficients. For example, using two
points (Figure 5–9), we simply have I = s1 + s2 because w1 = w2 = 1.000. This is the exact result if f(s)
is a polynomial containing terms up to and including x3. In general, Gaussian quadrature using n
points (Gauss points) is exact if the integrand is a polynomial of degree 2n - 1 or less. In using n
points, we effectively replace the given function f(s) by a polynomial of degree 2n - 1. The accuracy
of the numerical integration depends on how well the polynomial fits the given curve.

by Dr. Md. Jahidul Islam Page 31

CE 6109: Finite Element Methods I Chapter 5

If the function f(s) is not a polynomial, Gaussian quadrature is inexact, but it becomes more accurate
as more Gauss points are used. Also, it is important to understand that the ratio of two polynomials is,
in general, not a polynomial; therefore, Gaussian quadrature will not yield exact integration of the
ratio.

Table 5-1 Gauss points and weights for Gaussian Quadrature

n
f ( s ) ds   wi f ( si )
1
−1
i =1

Number of points, n Location, si Weights, wi

1 0.0 2.0
2 ±0.5773502692 1.0
3 ±0.7745966692 0.5555555556
0.0 0.8888888889
4 ±0.8611363116 0.3478548451
±0.3399810436 0.6521451549
5 ±0.9061798459 0.2369268851
±0.5384693101 0.4786286705
0.0 0.5688888889

Figure 5.9 Gaussian quadrature using two sampling points

5.8.3 Example
Evaluate,

 s 2
1 1 
−1  3e + s +  ds
( s + 2) 


Using one-point Gauss quadrature.

Solution:

For n = 1, we have w1 = 2, x1 = 0, and

by Dr. Md. Jahidul Islam Page 32

CE 6109: Finite Element Methods I Chapter 5

I  2 f ( 0)
= 7.0

For n = 2, we have w1 = 2, w2 = 1, x1 = - 0.57735…, x2 = +0.57735…, and I  8.7857.

This may be compared with exat solution:

Iexact = 8.8165

5.8.4 Two-Dimensional Integrals

The extension of Gaussian quadrature to two-dimensional integrals of the form,

 f ( s, t ) dsdt
1 1 (5.50)
I =
−1 −1

Follows readily, since

I   i =1  j =1 wi w j f ( s, t ) (5.51)
n n

by Dr. Md. Jahidul Islam Page 33