Introduction to Finite Elements

Shape functions in 1D
Reading assignment:

Lecture notes, Logan 2.2, 3.1

Summary:

• Linear shape functions in 1D

• Quadratic and higher order shape functions
• Approximation of strains and stresses in an element
 Axially loaded elastic bar
y A(x) = cross section at x
b(x) = body force distribution
F (force per unit length)
x E(x) = Young’s modulus
x
x=0 x=L

Potential energy of the axially loaded bar corresponding to the

exact solution u(x)

2
1  du 
L L
 (u)   EA  dx   bu dx  Fu(x  L)
2 0  dx  0
Finite element formulation, takes as its starting point, not the
strong formulation, but the Principle of Minimum Potential
Energy.

Task is to find the function ‘w’ that minimizes the potential energy
of the system
2
1  dw 
L L
 (w)   EA  dx   bw dx  Fw(x  L)
2 0  dx  0

From the Principle of Minimum Potential Energy, that function ‘w’

is the exact solution.
Rayleigh-Ritz Principle
Step 1. Assume a solution

w( x)  a 0 o ( x)  a11 ( x)  a 2 2 ( x)  ...

Where o(x), 1(x),… are “admissible” functions and ao, a1,

etc are constants to be determined.
Step 2. Plug the approximate solution into the potential energy
2
1  dw 
L L
 (w)   EA  dx   bw dx  Fw(x  L)
2 0  dx  0

Step 3. Obtain the coefficients ao, a1, etc by setting

 (w)
 0, i  0,1,2,...
ai
The approximate solution is

u ( x)  a 0 o ( x)  a11 ( x)  a 2 2 ( x)  ...
Where the coefficients have been obtained from step 3
Need to find a systematic way of choosing the approximation
functions.

One idea: Choose polynomials!

w( x)  a 0 Is this good? (Is ‘1’ an “admissible” function?)

w( x)  a1 x Is this good? (Is ‘x’ an “admissible” function?)

Finite element idea:

Step 1: Divide the truss into finite elements connected to each

other through special points (“nodes”)

1 2 3 4

El #1 El #2 El #3

Total potential energy=sum of potential energies of the elements

2
1  dw 
L L
 (w)   EA  dx  0 bw dx  Fw(x  L)
2 0
 dx 
 x1=0 x2 x3 x4=L

El #1 El #2 El #3

Total potential energy

2
1  dw 
L L
 (w)   EA  dx   bw dx  Fw(x  L)
2 0  dx  0

Potential energy of element 1:

2
1  dw 
x2 x2
 1 (w)   EA  dx  x bw dx
2 1
x
 dx  1

Potential energy of element 2:

2
1  dw 
x3 x3
 2 (w)   EA  dx  x bw dx
2 2
x
 dx  2
 x1=0 x2 x3 x4

El #1 El #2 El #3

Potential energy of element 3:

2
1  dw 
x4 x4
 3 (w)   EA  dx   bw dx  Fw(x  L)
2 x3  dx  x3

Total potential energy=sum of potential energies of the elements

 (w)   1 (w)   2 (w)   3 (w)

Step 2: Describe the behavior of each element

Recall that in the “direct stiffness” approach for a bar element, we

derived the stiffness matrix of each element directly (See lecture on
Trusses) using the following steps:

TASK 1: Approximate the displacement within each bar as a

straight line
TASK 2: Approximate the strains and stresses and realize that a bar
(with the approximation stated in Task 1) is exactly like a spring
with k=EA/L
TASK 3: Use the principle of force equilibrium to generate the
stiffness matrix
Now, we will show you a systematic way of deriving the stiffness
matrix (sections 2.2 and 3.1 of Logan).

TASK 1: APPROXIMATE THE DISPLACEMENT WITHIN

EACH ELEMENT
TASK 2: APPROXIMATE THE STRAIN and STRESS WITHIN
EACH ELEMENT
TASK 3: DERIVE THE STIFFNESS MATRIX OF EACH
ELEMENT (next class) USING THE PRINCIPLE OF MIN. POT
ENERGY

Notice that the first two tasks are similar in the two methods. The
only difference is that now we are going to use the principle of
minimum potential energy, rather than force equilibrium, to derive
the stiffness matrix.
TASK 1: APPROXIMATE THE DISPLACEMENT WITHIN
EACH ELEMENT

Simplest assumption: displacement varying linearly inside each bar

w(x)  a 0  a 1x d 2x

d 1x x El #1
x1 x2

How to obtain a0 and a1?

w(x1 )  a 0  a 1x1  d1x

w(x 2 )  a 0  a1x 2  d 2x
 w(x1 )  a 0  a 1x1  d1x
w(x 2 )  a 0  a 1x 2  d 2x

Solve simultaneously
x2 x1
a0  d 1x  d 2x
x 2  x1 x 2  x1
1 1
a1   d 1x  d 2x
x 2  x1 x 2  x1
Hence
x2 - x x - x1
w(x)  a 0  a 1x  d1x  d 2x  N1 (x)d1x  N 2 (x)d 2x
x 2  x1 x 2  x1
 
N1 (x) N 2 (x)

“Shape functions” N1(x) and N2(x)

In matrix notation, we write

w(x)  N d (1)

Vector of nodal shape functions

 x 2 - x x - x1 
N  N 1 (x) N 2 (x)   
x
 2  x 1 x 2  x 1 

Vector of nodal displacements

d 1x 
d 
d 2x 
NOTES: PROPERTIES OF THE SHAPE FUNCTIONS

1. Kronecker delta property: The shape function at any node

has a value of 1 at that node and a value of zero at ALL other
nodes. 1 1
x2 - x x - x1
N1 (x)  N 2 (x) 
x 2  x1 x 2  x1
x1 x2 x
El #1
Check
x2 - x
N1 (x) 
x 2  x1
x 2 - x1
 N1 (x  x1 )  1
x 2  x1
x2 - x2
and N1 (x  x 2 )  0
x 2  x1
2. Compatibility: The displacement approximation is continuous
across element boundaries x -x x-x
w (2) (x)  3
d 2x  2
d 3x
x2 - x x - x1 x3  x2 x3  x2
w (1) (x)  d1x  d 2x
x 2  x1 x 2  x1

x1 x2 x
El #1 El #2 x3
At x=x2
x2 - x2 x -x
w (1) (x  x 2 )  d1x  2 1 d 2x  d 2x
x 2  x1 x 2  x1
x3 - x2 x -x
w (2) (x  x 2 )  d 2x  2 2 d 3x  d 2x
x3  x2 x3  x2
Hence the displacement approximation is continuous across elements
3. Completeness

N1 (x)  N 2 (x)  1 for all x

N1 (x)x1  N 2 (x)x 2  x for all x
x2 - x
Use the expressions N1 (x)  ;
x 2  x1
x - x1
N 2 (x) 
x 2  x1
And check
x2 - x x - x1
N1 (x)  N 2 (x)   1
x 2  x1 x 2  x1
x2 - x x - x1
and N1 (x) x1  N 2 (x) x 2  x1  x2  x
x 2  x1 x 2  x1
Rigid body mode

N1 (x)  N 2 (x)  1 for all x

What do we mean by “rigid body modes”?

Assume that d1x=d2x=1, this means that the element should

translate in the positive x direction by 1. Hence ANY point
(x) on the bar should have unit displacement. Let us see
whether the displacement approximation allows this.

w(x)  N1 (x)d1x  N 2 (x)d 2x  N1 (x)  N 2 (x)  1

YES!
Constant strain states

N1 (x)x1  N 2 (x)x 2  x at all x

What do we mean by “constant strain states”?

Assume that d1x=x1 and d2x=x2. The strain at ANY point (x)
within the bar is
d 2x  d1x x 2  x1
 (x)   1
x 2  x1 x 2  x1
Let us see whether the displacement approximation allows this.
w(x)  N1 (x)d1x  N 2 (x)d 2x  N1 (x)x1  N 2 (x)x 2  x
dw(x)
Hence,  (x)  1
dx YES!
Completeness = Rigid body modes + Constant Strain states

Compatibility + Completeness  Convergence

Ensure that the solution gets better as more elements are introduced
and, in the limit, approaches the exact answer.
4. How to write the expressions for the shape functions easily
(without having to derive them each time):
Start with the Kronecker delta property (the shape function at
any node has value of 1 at that node and a value of zero at all
other nodes) 1 1
x2 - x x - x1
N1 (x)  N 2 (x) 
x 2  x1 x 2  x1
x1 x2 x
El #1
Node at which N1 is 0
N1 (x) 
x 2 - x  Notice that the length of the element = x -x
x 2 - x 1  2 1

x1 - x   x - x1 
The denominator is
N 2 (x)   the numerator evaluated at
x1 - x 2  x 2 - x1 
the node itself
A slightly fancier assumption:
displacement varying quadratically inside each bar
N1 (x) N 2 (x) N 3 (x)

x1 x2 x
x3
El #1
N1 (x) 
x 2 - x x 3 - x 
x 2 - x1 x 3 - x1  w(x)  N1 (x)d 1x  N 2 (x)d 2x  N 3 (x)d 3x
N 2 (x) 
x1 - x x 3 - x 
x1 - x 2 x 3 - x 2  This is a quadratic finite element in
N 3 (x) 
x1 - x x 2 - x  1D and it has three nodes and three
x1 - x 3 x 2 - x 3  associated shape functions per element.
TASK 2: APPROXIMATE THE STRAIN and STRESS
WITHIN EACH ELEMENT
From equation (1), the displacement within each element

w(x)  N d
Recall that the strain in the bar dw
ε
dx
Hence
dN 
ε  dBd (2)
 dx 
The matrix B is known as the “strain-displacement matrix”
dN 
B
 dx 
For a linear finite element

 x 2 - x x - x1 
N  N 1 (x) N 2 (x)   
x
 2  x 1 x 2  x 1 

Hence
 -1 1  1
B    1 1
 x 2  x1 x 2  x1  x 2  x1
 -1 1  d 1x 
εBd  d 
 x 2  x 1 x 2  x 1   2x 
d 2x - d 1x

x 2  x1
Hence, strain is a constant within each element (only for a
linear element)!
 Displacement is linear
w(x)  a 0  a 1x
d 2x

d 1x x El #1
x1 x2
Strain is constant
d 2x - d 1x
ε
x 2  x1

x El #1
x1 x2
 du
Recall that the stress in the bar   Eε  E
dx

Hence, inside the element, the approximate stress is

  EB d (3)

For a linear element the stress is also constant inside each element.
This has the implication that the stress (and strain) is discontinuous
across element boundaries in general.
 Summary

Inside an element, the three most important approximations in

terms of the nodal displacements (d) are:

Displacement approximation in terms of shape functions

u(x)  N d (1)

Strain approximation in terms of strain-displacement matrix

ε(x)  B d (2)

Stress approximation in terms of strain-displacement matrix and

Young’s modulus
  EB d (3)
 Summary

For a linear element

Displacement approximation in terms of shape functions

 x 2 - x x - x 1  d 1x 
u(x)    d 
 x 2  x 1 x 2  x 1   2x 
Strain approximation
1 d 1x 
ε  1 1  
x 2  x1 d 2x 
Stress approximation
E d 1x 
  1 1  
x 2  x1 d 2x 