LECTURE

FINITE ELEMENTS FOR PLANE SOLIDS

1. Shape Function – 1D, 2D (Natural Coordinate)

2. Constant Strain triangle
3. Isoperametric formulation

1
 INTRODUCTION
• Plane Solids
– All engineering problems are 3-D. It is the engineer who
approximates the problem using 1-D (beam or truss) or 2-
D (plane stress or strain).
– Stress and strain are either zero or constant in the
direction of the thickness.
– System of coupled second-order partial differential
equation
– Plane stress and plane strain: different constraints
imposed in the thickness direction
– Plane stress: zero stresses in the thickness direction (thin
plate with in-plane forces)
– Plane strain: zero strains in the thickness direction (thick
solid with constant thickness, gun barrel)
– Main variables: u (x-displacement) and v (y-displacement) 2
 TYPES OF 2D PROBLEMS
 yy y+
dy

• Governing D.E. 2

  xx  xy  yx y+
dy

 + + bx = 0 2
 x y y  xy dx
 b x+

  xy +  yy + b = 0  xx  xx
2
dx dx
x− y x+
 x y
y 2 bx x 2
 xy dx
• Definition of strain x−
2

u v  u v   yx
 xx = ,  yy =  xy = +
dy
, y−
x y 
 y x 
2

 yy
• Stress-Strain Relation y−
dy
2
 xx   C11 C12 C13    xx 
     
 yy  = C21 C22 C23  yy   { } = [C]{ }

  C C33    xy 
  xy   31
C32
 
– Since stress involves first-order derivative of displacements, the
governing differential equation is the second-order

3
 TYPES OF 2D PROBLEMS cont.
• Boundary Conditions
– All differential equations must be accompanied by boundary conditions
u = g, on Sg
s n = T, on ST

– Sg is the essential boundary and ST is the natural boundary

– g: prescribed (specified) displacement (usually zero for linear problem)
– T: prescribed (specified) surface traction force
• Objective: to determine the displacement fields u(x, y) and
v(x, y) that satisfy the D.E. and the B.C.

4
 PLANE STRESS PROBLEM
• Plane Stress Problem:
– Thickness is much smaller than the length and width dimensions
– Thin plate or disk with applied in-plane forces
– z-directional stresses are zero at the top and bottom surfaces
– Thus, it is safe to assume that they are also zero along the thickness

 zz =  xz =  yz = 0
fy
– Non-zero stress components:
σxx, σyy, τxy

– Non-zero strain components:

εxx, εyy, εxy, εzz
y fx
x

5
 PLANE STRESS PROBLEM cont.
• Stress-strain relation
 xx  1  0   xx 
  E  1   

 yy  =   {} = [C ]{}
  yy 
0
1 −  2 
  0 0  
  xy  2 (1 −  )    xy 
1

– Even if εzz is not zero, it is not included in the stress-strain relation

because it can be calculated from the following relation:

 zz = − ( xx +  yy )
E

• How to derive plane stress relation?

– Solve for zz in terms of xx and yy from the relation of zz = 0 and Eq.
(1.57)
– Write xx and yy in terms of xx and yy

6
 PLANE STRAIN PROBLEM
• Plane Strain Problem
– Thickness dimension is much larger than other two dimensions.
– Deformation in the thickness direction is constrained.
– Strain in z-dir is zero
 zz = 0,  xz = 0,  yz = 0

– Non-zero stress components: σxx, σyy, τxy, σzz.

– Non-zero strain components: εxx, εyy, εxy.

Plane strain model

7
 PLANE STRAIN PROBLEM cont.
• Plan Strain Problem
– Stress-strain relation

 xx  1 −   0   xx 
     
0   yy   {} = [C ]{}
E

 yy  = 1− 
 
  (1 +  )(1 − 2 )  0  
  xy  2 −     xy 
1
 0

– Even if σzz is not zero, it is not included in the stress-strain relation

because it can be calculated from the following relation:

E
 zz = ( xx +  yy )
(1 +  )(1 − 2 )

8
 EQUIVALENCE
• A single program can be used to solve both the plane stress
and plane strain problems by converting material properties.

From → To E 

   2  
Plane strain → Plane E 1 −   
stress   1 +   1 +

E
Plane stress → Plane 
  
2
strain 1−   1 −
 1 − 

9
 PRINCIPLE OF MINIMUM POTENTIAL ENERGY
• Strain Energy
– energy that is stored in the structure due to the elastic deformation

1 h
U=    =   {} dA
T T
{ } { } dV { }
2 volume 2 area
h
=   [C]{} dA
T
{ }
2 area

– h: thickness, [C] = [Cσ] for plane stress, and [C] = [Cε] for plane strain.
– stress and strain are constant throughout the thickness.
– The linear elastic relation {σ} = [C]{ε} has been used in the last relation.

10
PRINCIPLE OF MINIMUM POTENTIAL ENERGY cont.
• Potential Energy of Applied Loads
– Force acting on a body reduces potential to do additional work.
– Negative of product of the force and corresponding displacement
– Concentrated forces
ND
V = − Fq
i i
i=1

– Fi and qi are in the same direction

– Reaction force does not have any potential (because qi = 0)
– Distributed forces (pressure load) acting on the edge

V = −h  (Txu + Ty v)dS
ST
A
Tx  h
= −h  [u v]   dS
ST Ty  z ST

 −h  {u} T {T } dS y
ST
{Tx,Ty}
x
11
PRINCIPLE OF MINIMUM POTENTIAL ENERGY cont.
• Total Potential Energy
– Net energy contained in the structure
– Sum of the strain energy and the potential energy of applied loads

 =U+ V

• Principle of Minimum Potential Energy

– The structure is in equilibrium status when the potential energy has a
minimum value.
   
=0 = 0, =0 =0
 {u} u1 u2 uN

Finite Element Equation

12
 CST ELEMENT
• Constant Strain Triangular Element
– Decompose two-dimensional domain by a set of triangles.
– Each triangular element is composed by three corner nodes.
– Each element shares its edge and two corner nodes with an adjacent
element
– Counter-clockwise or clockwise node numbering
– Each node has two DOFs: u and v
– displacements interpolation using the shape functions and nodal
displacements.
– Displacement is linear because three nodal data are available.
– Stress & strain are constant.
y v3
3 u3
v1

1 u1 v2
u2
2 x 13
 CST ELEMENT cont.
• Displacement Interpolation
– Since two-coordinates are perpendicular, u(x,y) and v(x,y) are
separated.
– u(x,y) needs to be interpolated in terms of u1, u2, and u3,
and v(x,y) in terms of v1, v2, and v3.
– interpolation function must be a three term polynomial in x and y.
– Since we must have rigid body displacements and constant strain
terms in the interpolation function, the displacement interpolation must
be of the form
u(x,y) = 1 +  2 x +  3 y

v(x,y) = 1 + 2 x + 3 y

– The goal is how to calculate unknown coefficients αi and βi, i = 1, 2, 3,

in terms of nodal displacements.
u(x,y) = N1(x,y)u1 + N2 (x,y)u2 + N3 (x,y)u3

14
 CST ELEMENT cont.
• Displacement Interpolation
– x-displacement: Evaluate displacement at each node
 u(x1,y1 )  u1 = 1 +  2 x1 +  3 y1 v3

u(x 2 ,y 2 )  u2 = 1 +  2 x 2 +  3 y 2 3 u3
u(x ,y )  u =  +  x +  y
 3 3 3 1 2 3 3 3 v1

u1 v2
– In matrix notation 1
 u1  1 x1 y1   1  u2
     2
u2  = 1 x 2 y 2   2 

u  1 x y 3  3 
 3  3

– Is the coefficient matrix singular?

15
 CST ELEMENT cont.
• Displacement Interpolation
−1
 1  1 x1 y1   u1   f1 f2 f3   u1 
      
y2 
1 
 2  = 1 x 2 u2  = b1 b2 b3  u2 
  
  1 x y 3  u  2A c c c  u 
 3  3  3  1 2 3 3

– where
 f1 = x 2 y 3 − x 3 y 2 , b1 = y 2 − y 3 , c1 = x 3 − x 2

f2 = x 3 y1 − x1y 3 , b2 = y 3 − y1, c 2 = x1 − x 3
f = x y − x y , b = y − y , c = x − x
3 1 2 2 1 3 1 2 3 2 1

– Area:
1 x1 y1
1
A= det 1 x 2 y2
2
1 x3 y3

16
 CST ELEMENT cont.
1
1 = (f1u1 + f2u2 + f3u3 )
2A
1
2 = (b1u1 + b2u2 + b3u3 )
2A
1
3 = (c1u1 + c 2u2 + c 3u3 )
2A
• Insert to the interpolation equation
u(x,y) = 1 +  2 x +  3 y
1
=
2A
 (f1u1 + f2u2 + f3u3 ) + (b1u1 + b2u2 + b3u3 )x + (c1u1 + c 2u2 + c 3u3 )y 

1
= (f1 + b1x + c1y) u1 N1(x,y)
2A
1
+ (f2 + b2 x + c 2 y) u2 N2(x,y)
2A
1
+ (f3 + b3 x + c 3 y) u3
2A N3(x,y)
17
 CST ELEMENT cont.
• Displacement Interpolation
– A similar procedure can be applied for y-displacement v(x, y).
 u1 
   1
u(x,y) = [N1 N2 N3 ] u2  N
 1 (x,y) = (f1 + b1x + c1y)
2A
u  
 3  1
 2
N (x,y) = (f2 + b2 x + c 2 y)
 v1   2A
  
v(x,y) = [N1 N2 N3 ] v 2  N (x,y) =
1
(f3 + b3 x + c 3 y)
v   3
 3  2A
Shape Function

– N1, N2, and N3 are linear functions of x- and y-coordinates.

– Interpolated displacement changes linearly along the each coordinate
direction.

18
 CST ELEMENT cont.
• Displacement Interpolation
– Matrix Notation u1 
v 
 1
u  N 0 N2 0 N3 0  u2 
{u}    =  1  
v  0 N1 0 N2 0 N3  v 2 
u3 
 
v 3 

{u(x,y)} = [N(x,y)]{q}

– [N]: 2×6 matrix, {q}: 6×1 vector.

– For a given point (x,y) within element, calculate [N] and multiply it with
{q} to evaluate displacement at the point (x,y).

19
 CST ELEMENT cont.
• Strain Interpolation
– differentiating the displacement in x- and y-directions.
– differentiating shape function [N] because {q} is constant.

u   3  3 Ni 3
bi
 xx  =  
x x  i=1
Ni (x,y)ui = 
 i=1 x
ui = 
i=1 2A
ui

v   3  3 Ni 3
ci
 yy  =  
y y  i=1
Ni (x,y)v i = 
 i=1 y
v i = 
i=1 2A
vi

u v 3
ci 3
bi
 xy  + = ui +  vi
y x i=1 2A i=1 2A
 1
 N1(x,y) = 2A
(f1 + b1x + c1y)

 1
N2 (x,y) = (f2 + b2 x + c 2 y)
 2A
 1
N3 (x,y) = (f3 + b3 x + c 3 y)
 2A

20
 CST ELEMENT cont.
• Strain Interpolation

 u  u1 
  v 
 x  b1 0 b2 0 b3 0  1
 v  1   u2 
{ } =   = 0 c1 0 c 2 0 c 3    [B]{q}
  v
 y  2A c b c b c b3   2 
 u v   1 1 2 2 3
u3 
 +   
 y x  v 3 
– [B] matrix is a constant matrix and depends only on the coordinates of
the three nodes of the triangular element.
– the strains will be constant over a given element

21
 CST ELEMENT cont.
• Property of CST Element
– Since displacement is linear in x and y, the triangular element deforms
into another triangle when forces are applied.
– an imaginary straight line drawn within an element before deformation
becomes another straight line after deformation.
– Consider a local coordinate x such that x = 0 at Node 1 and x = a at
Node 2.
3
– Displacement on the edge 1-2:
u(x) = 1 +  2x

v(x) =  3 +  4 x
1 a
– Since the variation of displacement is linear, x
the displacements should depend only on 2
u1 and u2, and not on u3.

22
 CST ELEMENT cont. 3

• Property of CST Element Element

1
  x x 1
 a  1 a u2 = H1 ( x ) u1 + H2 ( x ) u2
a
 u( x ) = 1 − u +
   1 x
 2
v(x) =  1 − x  v + x v = H ( x ) v + H ( x ) v Element 3
  a 1 a 2 1 1 2 2
  2

• Inter-element Displacement Compatibility 2

– Displacements at any point in an element can be computed from nodal

displacements of that particular element and the shape functions.
– Consider a point on a common edge of two adjacent elements, which
can be considered as belonging to either of the elements.
– Then the nodes of either triangle can be used in interpolating the
displacements of this point.
– However, one must obtain a unique set of displacements independent
of the choice of the element.
– This can be true only if the displacements of the points depend only on
the nodes common to both elements. 23
 EXAMPLE - Interpolation
• nodal displacements
{u1, v1, u2, v2, u3, v3, u4, v4} = {−0.1, 0, 0.1, 0, −0.1, 0, 0.1, 0}
y
• Element 1: Nodes 1-2-4
(0,1) (1,1)
x1 = 0 x2 = 1 x3 = 0 4 3
y1 = 0 y2 = 0 y3 = 1
f1 = 1 f2 = 0 f3 = 0 2
b1 = −1 b2 = 1 b3 = 0
c1 = −1 c2 = 0 c3 = 1 1
N1(x,y) = 1 − x − y 1 2
x
N2 (x,y) = x (0,0) (1,0)

N3 (x,y) = y u(1)
 (1)
= = 0.2
x
xx
3
u (x,y) =  NI (x,y)uI = 0.1(2x + 2y − 1)
(1)
v (1)
I=1 (1) = = 0.0
y
yy
3
v (x,y) =  NI (x,y)v I = 0.0
(1)
u(1) v (1)
I=1  (1)
= + = 0.2
y x
xy
24
 EXAMPLE – Interpolation cont.
y
• Element 2: Nodes 2-3-4 (1,1)
(0,1)
x1 = 1 x2 = 1 x3 = 0 4 3
y1 = 0 y2 = 1 y3 = 1
f1 = 1 f2 = −1 f3 = 1 2
b1 = 0 b2 = 1 b3 = −1
c1 = −1 c2 = 1 c3 = 0 1
N1(x,y) = 1 − y 1 2
x
N2 (x,y) = x + y − 1 (0,0) (1,0)
N3 (x,y) = 1 − x
u(2)
3  (2)
= = −0.2
x
xx
u (x,y) =  Ni (x,y)ui = 0.1(3 − 2x − 2y)
(2)

i=1 v (2)
(2) = = 0.0
y
3 yy
v (2) (x,y) =  Ni (x,y)v i = 0.0
i=1 u(2) v (2)
 (2)
= + = −0.2
y x
xy

Strains are discontinuous along the element boundary

25
 CST ELEMENT cont.
• Strain Energy: h
2 
U(e) = {  } T
[ C]{  } dA (e)

h (e) T
= {q }  [B]T63 [C]33 [B]36 dA{q(e) }
2 A

1 (e) T (e)
 {q } [k ]66 {q(e) }
2

– Element Stiffness Matrix: [k (e) ] = hA[B]T [C][B]

– Different from the truss and beam elements, transformation matrix [T]
is not required in the two-dimensional element because [k] is
constructed in the global coordinates.
• The strain energy of the entire solid is simply the sum of the
element strain energies
NE
1 NE (e) T (e) (e) assembly 1
U = U (e)
=  {q } [k ]{q } U= {Q s } T [K s ]{Q s }
e =1 2 e=1 2
26
 CST ELEMENT cont.
• Potential energy of concentrated forces at nodes
ND
V = − (Fixui + Fiy v i )  −{Q s } T {FN }
i=1
{FN } = [F1x F1y FNDx FNDy ]T

• Potential energy of distributed forces along element edges

– Surface traction force {T} = [Tx, Ty]T is applied on the element edge 1-2

V (e) = −h  {u(s)} T {T(s)} ds = −{d} T h  [ H(s)]T {T(s)} ds

{d} T = u1 v1 u2 v2 
ST ST
y
3 H1 0 H2 0 
H =  0 H 0 H 
 1 2
Ty
1
s Tx
2 x
{T}={Tx,Ty} 27
 CST ELEMENT cont.
• Rewrite with all 6 DOFs
V (e) = − {q(e) } T h  [N(s)]T {T(s)} ds = −{q(e) } T {fT(e) }
ST
l − s s 
 l 0 0 0 0  Work-equivalent nodal forces
l
[N] =  
0 l−s s
0 0 0
 l l 

• Constant surface traction

(l − s) / l 0  Tx 
 0 T 
(l − s) / l  y
 
l l s/l 0  Tx  hl Tx 
{fT } = h [N] {T } ds = h  
(e) T
 T  ds =  
0 0
 0 s /l  y 2 Ty 
 0 0  0
   
 0 0   0 
Equally divided to two nodes 28
 CST ELEMENT cont.
3 3
hlTy/2
Ty
1 1 hlTx/2 hlTy/2
S
Tx
hlTx/2
2 2

• Potential energy of distributed forces of all elements

NS
V = − {q(e) } T {fT(e) } = −{Q s } T {FT }
e =1

29
 CST ELEMENT cont.
• Potential energy of body forces
– distributed over the entire element (e.g. gravity or inertia forces).
b x  b x  1
V (e)
= −h [u v]   dA = −{q } h [N] dA   2
(e) T T

A b y  A b y 
 − {q(e) } T {fb(e) } 1 0 b x 
0  b 

1
  y
hA 1 0  b x  hA b x 
{fb } =
(e)
  b  =  
3 0 1   y  3 b y 
• Potential energy of body 1 b 
0
forces for all elements    x
0 1 b y 
NE
V = − {q(e) } T {fb(e) } = −{Q s } T {FB }
e =1

30
 CST ELEMENT cont.
• Total Potential Energy
1
 =U+ V = {Q s } T [K s ]{Q s } − {Q s } T {FN + FT + FB }
2

• Principle of Minimum Potential Energy


= 0  [K s ]{Q s } = {FN + FT + FB }
 {Q s }
Finite Element Matrix Equation for CST Element

• Assembly and applying boundary conditions are identical to

other elements (beam and truss).
• Stress and Strain Calculation
– Nodal displacement {q(e)} for the element of interest needs to be
extracted

{ } = [B]{q(e) } { } = [C]{} = [C][B]{q(e) }

Stress and strain are constant for CST element 31
 EXAMPLE 1
20 N4
• Cantilevered Plate
– Thickness h = 0.1 in, N3
50,000 lbs
15
E = 30×106 psi and
ν = 0.3. E2
10

• Element 1 5
E1
N2
– Area = 0.5×10×10 = 50. 50,000 lbs
N1
10
x1 = 0, y1 = 0 x 2 = 10, y 2 = 5 x 3 = 10, y 3 = 15
b1 = y 2 − y 3 = −10 b 2 = y 3 − y1 = 15 b3 = y1 − y 2 = −5
c1 = x 3 − x 2 = 0 c 2 = x1 − x 3 = −10 c 3 = x 2 − x1 = 10

32
 EXAMPLE 1 cont.
• Matrix [B]
b1 0 b2 0 b3 0  −10 0 15 0 −5 0 
1 
c3  =
1 
[B] = 0 c1 0 c 2 0 0 0 0 −10 0 10 
2A   100  
c1 b1 c 2 b2 c 3 b3   0 −10 −10 15 10 −5 

• Plane Stress Condition

1  0   1 .3 0 
E 
[C  ] =  1 0  = 3.297  107 .3 1 0 
1 − 2    
0 0 
2 (1 −  ) 
1
 0 0 .35 

33
 EXAMPLE 1 cont.
• Stiffness Matrix for Element 1
.5 0. −.75 .15 .25 −.15 
 .175 .175 −.263 −.175 .088 
 
 1.3 −.488 −.55 .313 
[k (1) ] = hA[B]T [C ][B] = 3.297  10 6  
 .894 .338 −.631 
 .3 −.163 
 
 .544 

• Element 2: Nodes 1-3-4

x1 = 0, y1 = 0 x 2 = 10, y 2 = 15 x 3 = 0, y 3 = 20
b1 = y 2 − y 3 = −5 b 2 = y 3 − y1 = 20 b3 = y1 − y 2 = −15
c1 = x 3 − x 2 = −10 c 2 = x1 − x 3 = 0 c 3 = x 2 − x1 = 10

34
 EXAMPLE 1 cont.
• Matrix [B]
 −5 0 20 0 −15 0 
1 
[B] = 0 −10 0 0 0 10 
200  
 −10 −5 0 20 10 −15 

• Stiffness Matrix
.15 .081 −.25 −.175 .1 .094 
 .272 −.15 −.088 .069 −.184 
 
 1. 0. −.75 .15 
[k (2) ] = 3.297  106  
 .35 .175 −.263 
 .65 −.244 
 
 .447 

35
 EXAMPLE 1 cont.
• Assembly
.65 .081 −.75 .15 .0 −.325 .1 .094   u1   R x1 
 
 .447 .175 −.263 −.325 .0 .069 −.184   v1   R y1 
 
 1.3 −.488 −.55 .313 .0 .0  u2   0 
    
.894 .338 − .631 .0 .0   2  = 
v −50,000 
3.297  106  
 1.3 −.163 −.75 .15  u3   50,000 
    
 .894 .175 −.263  v 3   0

 .65 −.244  u4   R x 4 
 Symmetric   
 .447  v 4   R y 4 

• Rx1, Ry1, Rx4, and Ry4 are unknown reactions at nodes 1 and 4
• displacement boundary condition u1 = v1 = u4 = v4 = 0

36
 EXAMPLE 1 cont.
• Reduced Matrix Equation and Solution
1.3 −.488 −.55 .313  u2   0
 −  v  −50,000 
  2  =  
.894 .338 .631
3.297  106  
 1.3 −.163  u3   50,000 
   
 .894  v 3   0 

u2 = −2.147  10 −3
v 2 = −4.455  10 −2
u3 = 1.891 10 −2
v 3 = −2.727  10 −2

37
 EXAMPLE 1 cont.
• Element Results
– Element 1  0 
 0 
 xx   −10 0 15 0 −5 0    −1.268  10 −3 
  1 
−3
 −2.147  10  =  1.727  10 −3 

 yy  = −
  −4.455  10 −2   
0 0 0 10 0 10
100 
   − − −     −3.212  10 −3 

 xy   0 10 10 15 10 5   1.891 10  
−2 
 −2 
 −2.727  10 
 xx   1 .3 0  −1.268  10  −24,709 
−3

  7    1.727  10 −3  =  44,406  psi


 yy  = 
   
3.297 10 .3 1 0

     −3.212  10 −3   −37,063 

 xy   0 0 .35    

38
 EXAMPLE 1 cont.
• Element Results
– Element 2
 0 
 0 
 xx   −5 0 20 0 −15 0     1.891 10 −3 
  1    1.891 10 −2   
 yy  = 
0 −10 0 0 0 10 
 −2  =  0 
  200  −10 −5 0 20 10 −15   −2.727  10  −2.727  10 −3 
  xy    0   
 
 0 

 xx   1 .3 0   1.891 10   62,354 
−3

  7     

 yy  =  =
    psi
3.297 10 .3 1 0 0 18,706

     −2.727  10 −3   −31,469 

 xy   0 0 .35    

39
 DISCUSSION
• These stresses are constant
over respective elements.
• large discontinuity in stresses  62,354 
 
across element boundaries { } =  18,706 
−31, 469 
 

−24,709 
 
{ } =  44, 406 
 −37,063
 

40
 BEAM BENDING EXAMPLE
-F
2 4 6 8 10
1m
1 3 5 7 9

5m F

• xx is constant Max v = 0.0018

along the x-axis and
linear along y-axis
• Exact Solution:
xx = 60 MPa
• Max deflection
vmax = 0.0075 m
xx

41
 BEAM BENDING EXAMPLE cont.
• y-normal stress and shear stress are supposed to be zero.

yy Plot xy Plot

42
 CST ELEMENT cont.
• Discussions
– CST element performs well when strain gradient is small.
– In pure bending problem, xx in the neutral axis should be zero.
Instead, CST elements show oscillating pattern of stress.
– CST elements predict stress and deflection about ¼ of the exact
values.
• Strain along y-axis is supposed to be linear. But, CST elements
can only have constant strain in y-direction.
• CST elements also have spurious shear strain.
u
u ( x, y ) = a0 + a1 x + a2 y  xx = = a1 3
x
v( x, y ) = b0 + b1 x + b2 y
v
 yy = = b2 v2
y 1 2
How can we improve accuracy?
u v u2
What direction?
 xy = + = a2 + b1
y x 43
 CST ELEMENT cont.
• Two-Layer Model
– xx = 2.32 × 107
– vmax = 0.0028

44
 RECTANGULAR ELEMENT
• Each edge is parallel to the coordinate direction (not practical)
• Lagrange interpolation for shape function calculation
• Interpolation:
 u1 = 1 +  2 x1 +  3 y1 +  4 x1y1
u = 1 +  2 x +  3 y +  4 xy u =  +  x +  y +  x y
 2 1 2 2 3 2 4 2 2
v = 1 + 2 x + 3 y + 4 xy 
u3 = 1 +  2 x 3 +  3 y 3 +  4 x 3 y 3
u4 = 1 +  2 x 4 +  3 y 4 +  4 x 4 y 4
4 3
y
?
 u1 
u 
1 2  
x u(x,y) = [N1 N2 N3 N4 ]  2 
u3 
u4 

45
 LAGRANGE INTERPOLATION
• Interpolation

x1 x x2
u1 u(x) u2

u ( x) = N1 ( x)u1 + N 2 ( x)u2

x − x2 x − x1
N1 ( x) = , N 2 ( x) =
x1 − x2 x2 − x1

l−x x
u ( x) = u1 + u2
l l

46
 RECTANGULAR ELEMENT cont.
uII
• Lagrange Interpolation 4 3
– Along edge 1-2, y = y1 (constant) y

 u1 
uI (x,y1 ) = [n1(x) n2 (x)]  
u2  1 uI x 2

x − x2 x − x1
n1(x) = , n2 (x) =
x1 − x 2 x 2 − x1

– Along edge 4-3, y = y3 (constant)

u4 
uII (x,y 3 ) = [n4 (x) n3 (x)]  
u3 

x − x3 x − x4
n4 (x) = , n3 (x) =
x 4 − x3 x3 − x 4

47
 RECTANGULAR ELEMENT cont.
uII
• Lagrange Interpolation 4 3
– Y-direction y

 uI (x,y1 ) 
u(x,y) = [n1(y) n4 (y)]  
 II
u (x,y 3 
)
1 uI x 2
y − y4 y − y1
n1(y) = , n4 (y) =
y1 − y 4 y 4 − y1

– Combine together

  u1  
 1
[n (x) n 2 (x)]  
 u2  
u(x,y) = [n1(y) n4 (y)]  
[n (x) n (x)] u4 
 4 3  
 u3 

48
 RECTANGULAR ELEMENT cont.
• Lagrange Interpolation
 u1 
u 
 
u(x,y) = [n1(x)n1(y) n2 (x)n1(y) n3 (x)n 4 (y) n 4 (x)n 4 (y)]  2 
u3 
u4 
 1
 N1  n1(x)n1(y) = A (x 3 − x)(y 3 − y)

N  n (x)n (y) = − 1 (x − x)(y − y)
 2 2 1
A
1 3


N  n (x)n (y) = 1 (x − x)(y − y)
 3 3 4
A
1 1

 1
N4  n4 (x)n4 (y) = − (x 3 − x)(y1 − y)
 A

49
 RECTANGULAR ELEMENT cont.
• Shape functions for rectangular elements are product of
Lagrange interpolations in the two coordinate directions.
• Note that N1(x, y) is:
– 1 at node 1 and 0 at other nodes.
– Linear function of x along edge 1-2 and linear function of y along edge
1-4.
– Zero along edge 2-3 and 3-4.
• Other shape functions have similar behavior.

N1

4 3

1 2
50
 RECTANGULAR ELEMENT cont.
• Displacement interpolation
– Same interpolation for both u and v.
u1 
v 
 1
u2 
 
u  N 0 N2 0 N3 0 N4 0  v 2 
{u}    =  1  
v  0 N1 0 N2 0 N3 0 N4  u3 
v 3 
 
u4 
v 4 

{u} = [N]28 {q} 81

51
 EXAMPLE 2
• Shape Functions y

( x − 3)( y − 2) − x( y − 2)
N1 = N2 = 4 (0,2) 3 (3,2)
6 6
xy − y ( x − 3)
N3 = N4 =
6 6
x

1 (0,0) 2 (3,0)

N1 N2

52
 RECTANGULAR ELEMENT cont.
• Strain-displacement relation
– Similar to CST element

u   4  4 Ni
 xx  =  
x x  i=1
Ni (x,y)ui  = 
 i=1 x
ui u1 
v 
 1
u2 
 y − y3 0 y3 − y 0 y − y1 0 y1 − y 0  
v 2 
{ } =  0 
1
x − x3 0 x1 − x 0 x − x1 0 x3 − x  
A  u
 x − x 3 y − y3 x1 − x y 3 − y x − x1 y − y1 x 3 − x y1 − y   3 
v 3 
 
u4 
v 4 
 [B]{q}
– [B] is a linear function of x and y.
– Strain will change linearly within the element (not completely linear in
both directions)

53
 RECTANGULAR ELEMENT cont.
• Element stiffness matrix (from strain energy)
h h (e) T
U(e) =    = 
T (e) T (e)
{ } [ C]{ } dA { q } [B] 83 [ C] 33 [B] 38 dA{ q }
2 A 2 A

1 (e) T (e)
 {q } [k ]88 {q(e) }
2

– Generally needs numerical integration

– A square, plane-stress, rectangular element:

 36− 1+
8 − 312+ −1+ 3 
8
−3 +
12 − 1+
8

6
1−3 
8 
 1+ 3 − 1−3  
− 1+ −3 + −1+ 3 
− 312+ 
 8 6 8 6 8 12 8

 − 312+ 1−3 
8
3 −
6 − 1+
8

6
−1+ 3 
8
−3 +
12
1+
8

 −3 + 
Eh  −1+83  
− 1+ 3 − 1−3 
− 312+ 1+
12 
[k ] =
(e) 6 8 6 8 8

1 −  2  −312+ − 1+  1−3  3 − 1+

− 312+ −1+ 3  

 1+  
8 6 8 6 8 8

− 8 − 312+
−3 + −1+ 3  1+ 3 − 1−3 
12 8 8 6 8 6 
 6 −1+ 3  −3 + 1+
− 312+ 1−3  3 −
− 1+ 
 1−3  
8 12 8 8 6 8

 8 − 312+ 1+
8
−3 +
12
−1+ 3 
8

6 − 1+
8
3 −
6  
54
 RECTANGULAR ELEMENT cont.
• Nodal and distributed forces are the same with CST element
• Body force (constant body force b = {bx, by}T)
b x  b x 
V (e)
= −h [u v]   dA = −{q } h  [N] dA  
(e) T T

A b y  A b y 
 {q(e) } T {fb(e) }

1 0 b x 
0  b 

1
  y
1 0 b x 
  b  
hA  0 1   x  hA b y 
{fb } =
(e)
 =  
4 1 
0 b y  4 b x 
  b y 
0 1
  Equally divide the total magnitude
1 0  b x  of the body force to the four nodes
  b 
0 1
 y
55
 EXAMPLE – SIMPLE SHEAR
y
• E = 10 GPa, v = 0.25, h=0.1m
• F = 100 kN/m2
• {Qs} = {u1, v1, u2, v2, u3, v3, u4, v4}T. 4 f 3

• Non-zero DOFs: u3 and u4.

• Stiffness matrix 1m
1m
Eh  36− − 312+  u3
[K] =  − 3+ 3 − 
1 − 2  12 6  u4 1 2
x
• FEM equation (after applying BC)
 4.88 −2.88  u3  5,000 
 =
8
10   
 −2.88 4.88  4 
u 5,000 
• Nodal displacements
u3 = u4 = 0.025mm
56
 EXAMPLE – SIMPLE SHEAR cont.
0 
• Strain & Stress 0 
 
0 
 y − 1 0 1 − y 0 y 0 − y 0    0 
   
{ } =  0 1− x 
0
x −1 0 −x 0 x =
 2.5  10 −5   
0 0

 x − 1 y − 1 − x 1 − y x y 1 − x − y    2.5  10 −5 
0   
 −5 
 2.5  10 
0 

 xx   1 0.25 0  0   0 
  1010 0.25    
 yy  = 1 0  0  =  0  Pa
1 − 0.25 2  
   0 0.375  2.5  10 −5  105 
  xy 
0

gxy Deformed
Undeformed shape
shape

57
 EXAMPLE – PURE BENDING
y
• Couple M = 100 kN.m f
• Analytical solution 4 3
M( − h2 )
( xx )max =− = 6.0MPa 1m
I 1m
 xx = 6.0(1 − 2y) MPa
1 2 f
x
• FEM solution
– Non-zero DOFs: u2, v2, u3, and v3.
 4.89 −1.67 0.44 −0.33  u2   100,000 
 −1.67 4.89   
0.33 −2.89  v 2   0 

108    =  
 0.44 0.33 4.89 1.67  u3   −100,000 
 
 −0.33 −2.89 1.67 4.89  v 3  
 0 


u2 = 0.4091mm, v 2 = 0.4091mm
u3 = −0.4091mm, v 3 = 0.4091mm
58
 EXAMPLE – PURE BENDING cont.
• Strain & Stress 0 
0 
 
0.41 
y − 1 0 1− y 0 y 0 −y 0   0.41 10 −3 (1 − 2y) 
0.41   
{ } =  0 x −1 0 −x 0 x 0 1− x   −3
=
 −0.41  
10 0

 x − 1 y − 1 − x 1 − y x y 1 − x − y    0.41  10 −3
(1 − 2x) 
0.41   
 
 0 
0 
 xx   1 0.25 0  0.41 10 −3 (1 − 2y)  4.4(1 − 2y) 
  1010
0.25    
 yy  = 2 
1 0
  0  =  1.1(1 − 2y)  MPa
  1 − 0.25
  0 0 0.375  0.41 10 −3 (1 − 2x) 1.6(1 − 2x) 
 xy

Undeformed
shape
Unable to make curvature
Deformed Trapezoidal shape -> non-zero shear stress
shape
(xx)max /(xx)exact = 4.364/6.0 (73%)

59
 BEAM BENDING PROBLEM cont.
• Sxx Plot Max v = 0.0051

• Stress is constant along the x-axis (pure bending)

• linear through the height of the beam
• Deflection is much higher than CST element. In fact, CST
element is too stiff. However, stress is inaccurate.
60
 BEAM BENDING PROBLEM cont.
• Caution:
– In numerical integration, we did not calculate stress at node points.
Instead, we calculate stress at integration points.
– Let’s calculate stress at the bottom surface for 4(0,1) 3(1,1)
element 1 in the beam bending problem.
– Nodal Coordinates:1(0,0), 2(1,0), 3(1,1), 4(0,1)
– Nodal Displacements:
u = [0, 0.0002022, -0.0002022, 0]
v = [0, 0.0002022, 0.0002022, 0]
1(0,0) 2(1,0)

– Shape functions and derivatives

N1 = ( x − 1)( y − 1) N1 / x = ( y − 1) N1 / y = ( x − 1)
N 2 = − x( y − 1) N 2 / x = −( y − 1) N 2 / y = − x
N 3 = xy N 3 / x = y N 3 / y = x
N 4 = −( x − 1) y N 4 / x = − y N 4 / y = −( x − 1)
61
 BEAM BENDING PROBLEM cont.
• At bottom surface, y = 0 4 3

N1 / x = −1 N1 / y = x − 1
N 2 / x = 1 N 2 / y = − x
N 3 / x = 0 N 3 / y = x 2
1
N 4 / x = 0 N 4 / y = −( x − 1)
• Strain u = [0, 0.0002022, -0.0002022, 0]
4
NI v = [0, 0.0002022, 0.0002022, 0]
 xx =  uI = 1 0.0002022
I=1 x
4
NI
 yy =  vI = −0.0002022  x + 0.0002022  x = 0
I=1 y
4
 N N 
 xy =   I vI + I uI  = 0.0002022 − 0.0004044x
I=1  x y 
• Stress:
{ } = [C]{ } = {4.44, 1.33, 1.55}107
62
 RECTANGULAR ELEMENT
• y-normal stress and shear stress are supposed to be zero.

yy Plot xy Plot

xx is a linear function of y alone N1 / x = ( y − 1) N1 / y = ( x − 1)

yy is a linear function of x alone N 2 / x = −( y − 1) N 2 / y = − x
xy is a linear function of x and y
N 3 / x = y N 3 / y = x
4
N 4
N N 4 / x = − y N 4 / y = −( x − 1)
 xx =  I uI  yy =  I vI
I=1 x I=1 y

63
 RECTANGULAR ELEMENT
• Discussions
– Can’t represent constant shear force problem because xx must be a
linear function of x.
– Even if xx can represent linear strain in y-direction, the rectangular
element can’t represent pure bending problem accurately.
– Spurious shear strain makes the element too stiff.

u = 1 +  2 x +  3 y +  4 xy
Exact
v = 1 +  2 x +  3 y +  4 xy

 xx =  2 +  4 y
 yy = 3 +  4 x Rectangular
element
 xy = ( 3 +  2 ) +  4 x +  4 y 4  0
64
 RECTANGULAR ELEMENT
• Two-Layer Model
– xx = 3.48×107
– vmax = 0.0053

65
 BEAM BENDING PROBLEM cont.
• Distorted Element
Max v = 0.004

• As element is distorted, the solution is not accurate any more.

66
 BEAM BENDING PROBLEM cont.
• Constant Shear Force Problem
Max v = 0.0035

• Sxx is supposed to change linearly along x-axis. But, the

element is unable to represent linear change of stress along
x-axis. Why?
• Exact solution: v = 0.005 m and xx = 6e7 Pa.
67
 BEAM BENDING PROBLEM cont.
4 7 3
• Higher-Order Element?
– 8-Node Rectangular Element
8 6
u ( x, y ) = a0 + a1 x + a2 y + a3 x + a4 xy
2

+ a5 y 2 + a6 x 2 y + a7 xy 2 1 5 2
– Strain
u ( x, y )
= a1 + 2a3 x + a4 y + 2a6 xy + a7 y 2
x
– Can this element accurately represent pure bending and constant
shear force problem?
1
x y
x2 xy y2
x3 x2y xy 2 y3

68
 BEAM BENDING PROBLEM cont.
• 8-Node Rectangular Elements

• Tip Displacement = 0.0075 m, Exact!

69
 BEAM BENDING PROBLEM cont.
• If the stress at the bottom surface is calculated, it will be the
exact stress value.

Sxx Syy

70
 ISOPARAMETRIC ELEMENT
• Quadrilateral Shape
– Most commonly used element (irregular shape)
– Generalization of rectangular element
– Use mapping to transform into a square (Reference element).
– The relationship between (x, y) and (s, t) must be obtained.
– All formulations are done in the reference element.
v3
y v4 t
3 u3
4 4 (-1,1) 3 (1,1)
u4

v1
s

1 u1 v2

u2 1 (-1,-1) 2 (1,-1)
2 x

Actual Element (x, y) Parent Element (s, t)

71
 ISOPARAMETRIC MAPPING
• Definition
– the same interpolation method is used for displacement and geometry.
• Procedure
– Construct the shape functions N1, N2, N3, and N4 at the reference
element
s −1 t −1
N1(s,t) = 
−1 − 1 −1 − 1 t
 1 4 (-1,1)
N
 1 (s,t) = (1 − s)(1 − t) 3 (1,1)
4

N (s,t) = 1 (1 + s)(1 − t) s
 2
4

N (s,t) = 1 (1 + s)(1 + t)
 3 4 1 (-1,-1) 2 (1,-1)
 1
N4 (s,t) = (1 − s)(1 + t)
 4

72
 ISOPARAMETRIC MAPPING cont.

y t
3
4 4 (-1,1) 3 (1,1)

1
1 (-1,-1) 2 (1,-1)
2 x

Proportional mapping
For a given (x,y), find corresponding (s,t).
For a given (s,t), find corresponding (x,y).

73
 ISOPARAMETRIC MAPPING cont.
– Use the shape functions for interpolating displacement and geometry.
– For a given value of (s,t) in the parent element, the corresponding point
(x,y) in the actual element and displacement at that point can be
obtained using the mapping relationship.

Displacement interpolation Geometry interpolation

 u1   x1 
u  x 
   
u(s,t) = [N1 N2 N3 N4 ]  2  x(s,t) = [N1 N2 N3 N4 ]  2 
u3   x3 
u4   x 4 

 v1   y1 
v  y 
   
v(s,t) = [N1 N2 N3 N4 ]  2  y(s,t) = [N1 N2 N3 N4 ]  2 
v 3   y3 
v 4   y 4 
74
 EXAMPLE
• Find mapping point of A in the physical element
– At point A, (s, t) = (0.5, 0.5)
1 3 9 3
N1( 21 , 21 ) = , N2 ( 21 , 21 ) = , N3 ( 21 , 21 ) = , N4 ( 21 , 21 ) =
16 16 16 16
– Physical coord
4
1 3 9 3
x( 21 , 21 ) =  NI ( 21 , 21 )xI = 6 + 4+ 2+  0 = 2.25
I=1 16 16 16 16
4
1 3 9 3
y( , ) =  NI ( 21 , 21 )yI =
1 1
2 2 0 + 4+ 4+ 0 = 3
I=1 16 16 16 16

y t
3 (2,4) 4 (−1,1) 3 (1,1)
2 (4,4)
A (.5,.5)
B (1,2) s

x
4 (0,0) 1 (6,0) 1 (−1,−1) 2 (1,−1)
75
 EXAMPLE cont.
• Find mapping point of B in the reference element
– At point B, (x, y) = (1, 2)
4
x = 1 =  NI (s,t)xI = 41 (1 − s)(1 − t)  6 + 41 (1 + s)(1 − t)  4
I=1

+ 41 (1 + s)(1 + t)  2 + 41 (1 − s)(1 + t)  0
= st − 2t + 3
4
y = 2 =  NI (s,t)yI = 41 (1 − s)(1 − t)  0 + 41 (1 + s)(1 − t)  4
I=1

+ 41 (1 + s)(1 + t)  4 + 41 (1 − s)(1 + t)  0
= 2 + 2s

– Thus, (s, t) = (0, 1)

76
 JACOBIAN OF MAPPING
Shape functions are given in (s,t). But, we want to differentiate w.r.t. (x,y) in
order to calculate strain and stress. Use chain rule of differentiation.
NI NI x NI y NI NI x NI y
= + = +
s x s y s t x t y t
In Matrix Form
 NI   x y   NI   NI 
 s   s s   x   x 
 =   = [ J]  

 I 
N  x y 
 I N 
 I N
 t   t t   y   y 
Matrix [J] is called the Jacobian matrix of mapping.
How to calculate matrix [J]?

x 4
NI 1 t
= xI = ( − x1 + x 2 + x 3 − x 4 ) + (x1 − x 2 + x 3 − x 4 )
s I=1 s 4 4
x 4
N 1 s
=  I xI = ( − x1 − x 2 + x 3 + x 4 ) + (x1 − x 2 + x 3 − x 4 )
t I=1 t 4 4
77
 JACOBIAN OF MAPPING cont.
• Derivatives of shape functions w.r.t. (x,y) coordinates:
 NI   NI   y y   NI 
 x    1  t −
−1  s  s   s 
  = [ J]  =   
 NI   NI  J  − x x   NI 
 y   t   t s   t 

• Determinant |J|: Jacobian

x y x y
J= −
s t t s
• What happen if |J| = 0 or |J| < 0?
– shape function derivative cannot be obtained if the |J| = 0 anywhere in
the element
– Mapping relation between (x, y) and (s, t) is not valid if |J| = 0 or |J| < 0
anywhere in the element (–1 ≤ s, t ≤ 1).

78
 JACOBIAN OF MAPPING cont.
• Jacobian is an important criterion for evaluating the validity of
mapping, as well as the quality of element
• Every point in the reference element should be mapped into
the interior of the physical element
• When an interior point in (s, t) coord. is mapped into an
exterior point in the (x, y) coord., the Jacobian becomes
negative
• If multiple points in (s, t) coordinates are mapped into a single
point in (x, y) coordinates, the Jacobian becomes zero at that
point
• It is important to maintain the element shape so that the
Jacobian is positive everywhere in the element

79
 EXAMPLE (JACOBIAN)
Jacobian must not be zero anywhere in the domain (-1 ≤ s, t ≤ 1)
y
• Nodal Coordinates 3(2, 2)

x1 = 0, x 2 = 1, x 3 = 2, x 4 = 0
y1 = 0, y 2 = 0, y 3 = 2, y 4 = 1 4(0, 1)

• Iso-Parametric Mapping x
4 1(0, 0) 2(1, 0)
1
x =  NI xI = N2 + 2N3 = (3 + 3s + t + st)
I=1 4
4
1
y =  NI yI = 2N3 + N4 = (3 + s + 3t + st)
I=1 4

• Jacobian Matrix  x y 
 s s  1 3 + t 1 + t 
[ J] =  = 
 x y  4 1 + s 3 + s 
 t t 
80
 EXAMPLE (JACOBIAN) cont.
• Jacobian
1 1 1 1
J= [(3 + t)(3 + s) − (1 + t)(1 + s)] = + s + t
4 2 8 8
– It is clear that |J| > 0 for –1 ≤ s ≤ 1 and –1 ≤ t ≤ 1.

Constant s
t
4 (-1,1) 3 (1,1)

Constant t
1 (-1,-1) 2 (1,-1)

81
 EXAMPLE (JACOBIAN) cont.
y
• Nodal Coordinates 4(0, 5) 3(5, 5)
x1 = 0, x 2 = 1, x 3 = 5, x 4 = 0
2(1, 4)
y1 = 0, y 2 = 4, y 3 = 5, y 4 = 5

• Mapping
4
1
x =  NI xI = (1 + s)(3 + 2t)
I=1 2 1(0, 0)
4
1 x
y =  NI yI = (7 + 2s + 3t − 2st)
I=1 2

• Jacobian
1
J= (5 − 10s + 10t) |J| = 0 at 5 – 10s + 10t = 0; i.e., s – t = 1/2
4

82
 EXAMPLE 3 (JACOBIAN) cont.
Constant s
• In general the element geometry
is invalid if the Jacobian is either
zero or negative anywhere in the Constant t
element.
• Problems also arise when the
Invalid mapping
Jacobian matrix is nearly singular
either due to round-off errors or
due to badly shaped elements.
• To avoid problems due to badly > 15o
shaped elements, it is suggested
< 165o
that the inside angles in quadrilateral
elements be > 15˚ and < 165˚

83
 INTERPOLATION
 u1 
• Displacement Interpolation (8-DOF) v 
 1
 u2 
 
u  N1 0 N2 0 N3 0 N4 0  v 2 
 =    = [N]{q}
v   0 N1 0 N2 0 N3 0 N4  u3 
v 3 
 
u 4 
v 4 

– the interpolation is done in the reference coordinates (s, t)

– The behavior of the element is similar to that of the rectangular
element because both of them are based on the bilinear Lagrange
interpolation

84
 INTERPOLATION cont.
• Strain
 xx   u x 
u x   1 0 0 0 
      u y 
{ } =  yy  =  v y  = 0 0 0 1   
  u y + v x  0 1 1 0  v x 
  xy      v
 y 

u x   y t − y s 0 0  u s 
u y  1  − x t x s 0  u t 
 0
 =   
v x  J  0 0 y t − y s  v s 
v  
y   0 0 − x t x s  v t 

 xx   y t − y s 0 0  u s  u s 
1 0 0 0    u
  1  − x t x s 0 0  t  u
 t 

 yy  = 0 0 0 1     [ A ]  
  0   −    v s 
  J
 
0 y t y s  v s 
  xy  0 1 1 0   0
 0

− x t x s  v t  v t 

85
 INTERPOLATION cont.
 u1 
• Strain cont. v 
 1
u s   −1 + t 0 1− t 0 1+ t 0 −1 − t 0   u2 
u t    
  1  −1 + s 0 −1 − s 0 1+ s 0 1− s 0  v 2 
 =     [G]{q}
v s  4  0 −1 + t 0 1− t 0 1+ t 0 −1 − t  u3 
v t   
 0 −1 + s 0 −1 − s 0 1+ s 0 1 − s  v 3 
 
u 4 
v 4 
 xx  u s 
  u t 


 yy  = [ A ]   = [ A][G]{q}  [B]{q} Strain-displacement matrix
  v s 
  xy  v t 

– The expression of [B] is not readily available because the matrix [A]
involves the inverse of Jacobian matrix
– The strain-displacement matrix [B] is not constant

86
 EXAMPLE
• {u1, v1, u2, v2, u3, v3, u4, v4} = {0, 0, 1, 0, 2, 1, 0, 2}
• Displacement and strain at (s,t)=(1/3, 0)?
• Shape Functions
1 1
N1(s,t) = (1 − s)(1 − t) N2 (s,t) = (1 + s)(1 − t)
4 4
1 1
N3 (s,t) = (1 + s)(1 + t) N4 (s,t) = (1 − s)(1 + t)
4 4
1 1 1 1
• At (s,t)=(1/3, 0) N1 = , N2 = , N3 = , N4 =
6 3 3 6

y t
4 (0,2) 3 (3,2) 4 (-1,1) 3 (1,1)

s
x
1 (0,0) 2 (3,0) 1 (-1,-1) 2 (1,-1)
87
 EXAMPLE cont.
• Location at the Actual Element
 4
1 1 1 1


x = 
I=1
N x
I I =
6
 0 +
3
 3 +
3
 3 +
6
0 = 2
 4
y = N y = 1  0 + 1  0 + 1  2 + 1  2 = 1
 
I=1
I I
6 3 3 6

• Displacement at (s,t) = (1/3,0)

 4
1 1 1 1


u = 
I=1
Nu
I I =
6
 0 +
3
 1 +
3
 2 +
6
0 =1
 4
v = N v = 1  0 + 1  0 + 1  1 + 1  2 = 2
 
I=1
I I
6 3 3 6 3

88
 EXAMPLE cont.
• Derivatives of the shape functions w.r.t. s and t.
 N1 1 1  N1 1 1
 s = − (1 − t) = −  t = − (1 − s) = −
4 4 4 6
 
 N2 1
= (1 − t) =
1  N2 1
= − (1 + s) = −
1
 s 4 4  t 4 3
 
 N3 1
= (1 + t) =
1  N3 1
= (1 + s) =
1
 s 4 4  t 4 3
 N 1 1  N 1 1
 4 = − (1 + t) = −  4 = (1 − s) =
 s 4 4  t 4 6

• But, we need the derivatives w.r.t. x and y. How to convert?

89
 EXAMPLE cont.
• Jacobian Matrix
 x 1 1 1 1 3
 s =− 0 + 3 + 3 − 0 =
4 4 4 4 2

 y 1 1 1 1
= − 0 + 0 + 2 − 2 = 0
 s 4 4 4 4

 x 1 1 1 1
= − 0 − 3 + 3 + 0 = 0
 t 6 3 3 6
 y 1 1 1 1
 = − 0 − 0 + 2 + 2 =1
 t 6 3 3 6
 x y 
 s s   3 0  2 
0
[ J] =   = 2 , [ J]−1 =  3 
 x y   0 1  
 t t     
0 1
– Jacobian is positive, and the mapping is valid at this point
– Jacobian matrix is constant throughout the element
– Jacobian matrix only has diagonal components, which means that the
physical element is a rectangle
90
 EXAMPLE cont.
• Derivative of the shape functions w.r.t. x and y.
 NI NI x NI y  NI   NI 
 s = x s + y s  s   x 

    = [ J] 


 NI = NI x + NI y 
 IN  IN
 t x t y t  t   y 
 NI   NI   NI   2 NI 
 x     2 
−1  s   0   s   3 s 
  = [ J]  = 3  = 

 IN  N   
 I   0 1  I  N NI 
 y   t   t   t 

• Strain
u 4
NI 4
2 NI 2 1 1 1 1 1
 xx = = uI =  uI = ( −  0 +  1 +  2 −  0) =
x I=1 x I=1 3 s 3 4 4 4 4 2

v 4
NI 4
N 1 1 1 1 2
 yy = = vI =  I vI = −  0 −  0 +  1 +  2 =
y I=1 y I=1 t 6 3 3 6 3
91
 FINITE ELEMENT EQUATION
• Element stiffness matrix from strain energy expression
h
2 
U(e) = {  } T
[ C]{  } dA (e)

h (e) T
= {q }  [B]8T3 [C]33 [B]38 dA{q(e) }
2 A

1 (e) T (e)
 {q } [k ]88 {q(e) }
2
– [k(e)] is the element stiffness matrix
– Integration domain is a general quadrilateral shape
– Displacement–strain matrix [B] is written in (s, t) coordinates
– we can perform the integration in the reference element
1 1
[k (e) ] = h [B]T [C][B]dA  h   [B]T [C][B] J dsdt
A −1 −1

dA = J dsdt
92
 NUMERICAL INTEGRATION
• Stiffness matrix and distributed load calculations involve
integration over the domain
• In many cases, analytical integration is very difficult
• Numerical integration based on Gauss Quadrature is
commonly used in finite element programs
• Gauss Quadrature:
1 n
I =  f(s)ds   w i f(si )
−1 i=1

– Integral is evaluated using function values and weights.

– si: Gauss integration points, wi: integration weights
– f(si): function value at the Gauss point
– n: number of integration points.

93
 NUMERICAL INTEGRATION cont.
• Constant Function: f(s) = 4
– Use one integration point s1 = 0 and weight w1 = 2
1
I =  4ds = w 1f(s1 ) = 2  4 = 8
−1

– The numerical integration is exact.

• Linear Function: f(s) = 2s + 1
– Use one integration point s1 = 0 and weight w1 = 2
1
I =  (2s + 1)ds = w 1f(s1 ) = 2  1 = 2
−1

– The numerical integration is exact.

• One-point Gauss Quadrature can integrate constant and
linear functions exactly.

94
 NUMERICAL INTEGRATION cont.
• Quadratic Function: f(s) = 3s2 + 2s + 1
– Let’s use one-point Gauss Quadrature
1
I =  (3s2 + 2s + 1)ds = 4
−1

w1f(s1 ) = 2  1 = 2

– One-point integration is not accurate for quadratic function

– Let’s use two-point integration with w1 = w2 = 1 and -s1 = s2 = 1/ 3

w1f(s1 ) + w 2 f(s2 ) = 1 f( − 1
3
) + 1 f( 13 )
= 3  31 − 2
3
+ 1 + 3  31 + 2
3
+1= 4

• Gauss Quadrature points and weights are selected such that

n integration points can integrate (2n – 1)-order polynomial
exactly.
95
 NUMERICAL INTEGRATION cont.
• Gauss Quadrature Points and Weights

Exact for polynomial

n Integration Points (si) Weights (wi)
of degree
1 0.0 2.0 1
2 .5773502692 1.0 3
.7745966692 .5555555556 5
3
0.0 .8888888889
.8611363116 .3478546451 7
4
.3399810436 .6521451549
.9061798459 .2369268851 9
5 .5384693101 .4786286705
0.0 .5688888889

96
 NUMERICAL INTEGRATION cont.
• 2-Dimensional Integration
– multiplying two one-dimensional Gauss integration formulas

1 1 1 m n m
I=   f(s,t)dsdt    w f ( s ,t ) dt =  w w f(s ,t )
−1 −1 −1 i=1
i i
j=1 i=1
i j i j

– Total number of integration points = m×n.

t t t

s s s

(a) 11 (b) 22 (c) 33

97
 NUMERICAL INTEGRATION EXAMPLE
• Integrate the following polynomial:
1
I =  (8x 7 + 7x 6 )dx = 2
−1

– One-point formula

s1 = 0, f(s1 ) = 0, w1 = 2
I = w 1f(s1 ) = 2  0 = 0

– Two-point formula

s1 = −.577, f(s1 ) = 8( −.577)7 + 7( −.577)6 = .0882, w 1 = 1

s2 = .577, f(s2 ) = 8(.577)7 + 7(.577)6 = .4303, w2 = 1
I = w 1f(x1 ) + w 2 f(x 2 ) = .0882 + .4303 = .5185

98
 NUMERICAL INTEGRATION EXAMPLE cont.
– 3-point formula

s1 = −.577, f(s1 ) = 8( −.577)7 + 7( −.577)6 = .0882, w1 = 1

s2 = .577, f(s2 ) = 8(.577)7 + 7(.577)6 = .4303, w2 = 1
I = w 1f(x1 ) + w 2 f(x 2 ) = .0882 + .4303 = .5185

– 4-point formula
s1 = −.8611, f(s1 ) = .0452, w 1 = .3479
s2 = −.3400, f(s2 ) = .0066, w 2 = .6521
s3 = .3400, f(s3 ) = .0150, w 3 = .6521
s3 = .8611, f(s3 ) = 5.6638, w 3 = .3479
I = w 1f(s1 ) + w 2 f(s2 ) + w 3 f(s3 ) + w 4 f(s 4 ) = 2.0

– 4-point formula yields the exact solution. Why?

99
 NUMERICAL INTEGRATION
• Application to Stiffness Matrix Integral
1 1
[k (e) ] = h   [B]T [C][B] J dsdt
−1 −1
2 2
 h w i w j [B(si ,t j )]T [C][B(si,t j )] J(si,t j )
i=1 j=1

100
 Assignment
• Analysis and Design of Torque Arm
– Material properties: E = 206.8 GPa, Poisson’s ratio = 0.29, thickness =
1.0 cm, mass density = 7850 kg/m3
– Preliminary analysis: Knowledge from Mechanics of Materials
– Convergence study on vertical displacement at the load application pt
– Modify the shape of the torque arm to reduce weight

All dimensions in cm
2.5 4.2
4
1.0 2789N
Fixed
5

12 15 5066N
42

101