Institute of Structural Engineering Page 1

Chapter 4

Isoparametric Elements

[O] V1/Ch3/77-​95
[F] Ch16

Method of Finite Elements I

 Institute of Structural Engineering Page 2

Today’s Lecture Contents

• The Shape Function: Reminder from the previous lecture

• Looking for a uniform mapping across element types

• Isoparametric elements

• 1D Demonstration: Bar elements

Method of Finite Elements I

30-Apr-10
 Institute of Structural Engineering Page 3

The Galerkin Method

In the previous lecture, we defined an approximation for the displacement field:

Note: this is just a vector of the

coefficients of the linear
approximation function.

u1, u2 do not correspond to

Displacement field displacements at the nodes

This also defined the first derivative of the approximation (strain)

Strain field

Method of Finite Elements I

 Institute of Structural Engineering Page 4

The Galerkin Method

where u1,u2 are so far random coefficients. Instead, we can choose to write the same
relationship using a different basis N(x):
Vector of degrees of

{} [ ]{ } freedom (DOFs) at the

u = N ( x) d element nodes

Displacement field Shape Function

Matrix

In this case we express u in terms of the degrees of freedom, i.e., the displacements at
the ends of the bar:

𝑢𝑢(𝑥𝑥 = 0) 𝑢𝑢0 1 0 𝑢𝑢1 𝑢𝑢1 1 0 −1 𝑢𝑢0

𝑑𝑑 = = 𝑢𝑢 = ⇔ =
𝑢𝑢(𝑥𝑥 = 𝐿𝐿) 𝐿𝐿 1 𝐿𝐿 𝑢𝑢2 𝑢𝑢2 1 𝐿𝐿 𝑢𝑢𝐿𝐿

u0, uL correspond to the displacements at the nodes (nodal DOFs), which are contained
in vector {d}.
So here, we expressed the coefficients in terms of the nodal DOFs.

Method of Finite Elements I

 Institute of Structural Engineering Page 5

The Galerkin Method

where u1,u2 are so far random coefficients. Instead, we can choose to write the same
relationship using a different basis N(x):
Vector of degrees

{} [ ]{ } of freedom at the
u = N ( x) d element nodes

Displacement field Shape Function

Matrix

In this case we express u in terms of the degrees of freedom, i.e., the displacements at
the ends of the bar:
𝑢𝑢(𝑥𝑥 = 0) 𝑢𝑢0 1 0 𝑢𝑢1 𝑢𝑢1 1 0 −1 𝑢𝑢0
𝑑𝑑 = = 𝑢𝑢 = ⇔ =
𝑢𝑢(𝑥𝑥 = 𝐿𝐿) 𝐿𝐿 1 𝐿𝐿 𝑢𝑢2 𝑢𝑢2 1 𝐿𝐿 𝑢𝑢𝐿𝐿

Substituting in our initial displacement approximation we obtain:

1 0 −1 𝑢𝑢0 𝐿𝐿 − 𝑥𝑥 𝑥𝑥 𝑢𝑢0 𝐿𝐿 − 𝑥𝑥 𝑥𝑥 shape

𝑢𝑢 = 1 𝑥𝑥 𝑢𝑢𝐿𝐿 = ⇒ 𝑁𝑁(𝑥𝑥) = functions
1 𝐿𝐿 𝐿𝐿 𝐿𝐿 𝑢𝑢𝐿𝐿 𝐿𝐿 𝐿𝐿
Method of Finite Elements I
 Institute of Structural Engineering Page 6

The Shape Functions for the generic bar (truss) Element

Truss Element Shape Functions

 u1 
u ( x ) = [ N1 N2 ]  
u2 
(in global coordinates)

Method of Finite Elements I

30-Apr-10
 Institute of Structural Engineering Page 7

The Galerkin Method

Then:
Vector of degrees
of freedom at the
u = [ N ( x ) ] {d } element nodes

Displacement field

The weak form also involves the first derivative of the approximation

d [ N ( x)]
=ε = {d } [ B ] {d }
Strain field dx
Strain
d [ N ( x) ] 1 Displacement
where [ B=] = [ −1 1] Matrix
dx L
Method of Finite Elements I
 Institute of Structural Engineering Page 8

Strain-Displacement relation

The truss element

define
stress displacement matrix

Constant Variation of strains along the element’s length

Method of Finite Elements I

30-Apr-10
 Institute of Structural Engineering Page 9

The Galerkin Method

The following FUNDAMENTAL FEM expression is derived

EA depends on material and cross-sectional properties

L

{ f } = ∫ [ N ] ( -αx ) dx
T
We call the nodal force vector for distributed loads
0

 f ( x = 0 ) 
This is defined as: { } 
f =  for concentrated loads* at the nodes
 (
f x = L )
*Notice the placement of the concentrated load on the corresponding degree of
freedom where it applies, i.e., x = 0 → Node 1, x = L → Node 2

Method of Finite Elements I

 Institute of Structural Engineering Page 10

The Galerkin Method

The FEM equation to solve is (as in the DSM):

[ K ]{d} = { f }

EA  1 −1 for a bar element with only two nodes

where [K ] = 
L  −1 1  (2-noded bar element)

We are however not restricted to only using 2-noded elements!

Method of Finite Elements I

 Institute of Structural Engineering Page 11

Generalized Elements
We are not restricted to only using 2-noded elements!

In order to generalize to
• elements of higher order (more
nodes) and
• generalized geometries
we need to generalize the shape
functions

Higher order
Why?

[O] Ch3§2,
Method of Finite Fig. 3.2 I
Elements
 Institute of Structural Engineering Page 12

Revisiting the Shape Function

Therefore: Shape functions will be defined as interpolation functions which relate
the variables in the finite element with their values in the element nodes. The
latter are obtained through solving the problem using finite element procedures.

In general, we may write:

=u ( x) ∑=
N ( x)u
i
i i Nu
where:
is the function under investigation (for example:
u ( x) displacement field)
is the shape functions matrix
N( x)
is the vector of unknowns in the nodes
u = [u1 u2  u N ]
T
(for example: displacements)

Method of Finite Elements I

30-Apr-10
 Institute of Structural Engineering Page 13

The Shape Function

Regardless of the dimension of the element used, we have to bear in mind that
Shape Functions need to satisfy the following constraints:
• Ni ( x ) in node i has a value of 1 and in all other nodes Why??
assumes a value of 0.
• Furthermore we have to satisfy the continuity between the adjoining
elements.
Example: rectangular element with 6 nodes

H4 H5

Method of Finite Elements I

30-Apr-10
 Institute of Structural Engineering Page 14

The Shape Function

Usually, polynomial functions are used as interpolation functions, for example:

N
Pn ( x ) = ∑ ai x i
i =0

where n is the order of the polynomial; is equal to the number of

unknowns in the nodes (degrees of freedom).

In the MFE we use three different polynomials:

 Lagrange
 Serendipity and
 Hermitian polynomials.

Method of Finite Elements I

30-Apr-10
 Institute of Structural Engineering Page 15

The Shape Function

Lagrange Polynomials

A function φ(x) can be approximated by a polynomial of order n and the

values of φi in those n+1 points:

n +1

∑ i
ϕ ( x) ≈ ϕ ( x) =L ( x ) ϕ
i =1
i

where the normalized Lagrange polynomial Lni of order n-1 is defined by n

points, as follows:

L ( x) ∏
= n
=
x − xj n
( x − x1 )( x − x2 ) ( x − xn )
i
j =1 xi − x j ( xi − x1 )( xi − x2 ) ( xi − xn )
i≠ j

 Note that Lni ( xi ) ≠ 0 & Lni ( x j ) =

0 for i ≠ j
Method of Finite Elements I
30-Apr-10
 Institute of Structural Engineering Page 16

The Shape Function

Lagrange Polynomials

 Note that

Lni ( xi ) ≠ 0 & Lni ( x j ) =

0 for i ≠ j

Method of Finite Elements I

30-Apr-10
 Institute of Structural Engineering Page 17

The Shape Function

Lagrange Polynomials

The shape function Ni of a Lagrange element with n nodes coincides

with the normalized Lagrange polynomial of degree n-1, i.e.

N i ( x ) = Lni ( x )

This explains why C0 continuous 1D elements are also called Lagrange

elements.

Method of Finite Elements I

30-Apr-10
 Institute of Structural Engineering Page 18

Lagrange Polynomials – 2D Example

4 ( x4 , y4 ) 3 ( x3 , y3 )
Let us break the process into steps:

Step 1: Interpolate along x

y
Interpolate along borders 1-2, 4-3:
1( x1 , y1 )
=ϕ1− 2 L11− 2 ( x ) ϕ1 + L12− 2 ( x ) ϕ2
2 ( x2 , y2 )
x =ϕ3− 4 L14−3 ( x ) ϕ3 + L42−3 ( x ) ϕ4
1 L1 1 L2

x2 − x 1− 2 x − x1
Therefore: L11− 2 ( x )
= = , L2 ( x )
x2 − x1 x2 − x1
x4 − x 4−3 x − x3
L ( x) =
= 4 −3
1 , L2 ( x )
x4 − x3 x4 − x3
Method of Finite Elements I
30-Apr-10
 Institute of Structural Engineering Page 19

Lagrange Polynomials – 2D Example

4 ( x4 , y4 ) 3 ( x3 , y3 )
Step 1: Interpolate along x
Interpolate along borders 1-2, 4-3:
=ϕ1− 2 L11− 2 ( x ) ϕ1 + L12− 2 ( x ) ϕ2
y
1( x1 , y1 ) =ϕ3− 4 L14−3 ( x ) ϕ3 + L42−3 ( x ) ϕ4
2 ( x2 , y2 )
x

1 L1 1 L2

Step 2: Interpolate along y

2. Interpolate along borders 1-2, 4-3: (one possible way is below)

ϕ = L1 ( y ) ϕ1− 2 + L2 ( y ) ϕ3− 4 =
= L11− 4 ( y ) L11− 2 ( x ) ϕ1 + L12−3 ( y ) L12− 2 ( x ) ϕ 2 + L22−3 ( y ) L14−3 ( x ) ϕ3 + L12− 4 ( y ) L42−3 ( x ) ϕ 4

Method of Finite Elements I

30-Apr-10
 Institute of Structural Engineering Page 20

Lagrange Polynomials – 2D Example

4 : u 4 , v4 3 : u 3 , v3

To interpolate the displacement field for

this 2D element, both
• the horizontal u(x,y) and
y
1: u1 , v1 • vertical displacements v(x,y)
should be interpolated
2 : u 2 , v2
x
we can use the following approximation:

( x2 − x )( y4 − y ) ( x − x1 )( y3 − y ) ( x4 − x )( y − y2 ) ( x − x3 )( y − y1 )
u ( x, y ) = u1 + u2 + u3 + u4
( x2 − x1 )( y4 − y1 ) ( x2 − x1 )( y3 − y2 ) ( x4 − x3 )( y3 − y2 ) ( x4 − x3 )( y4 − y1 )
( x2 − x )( y4 − y ) ( x − x1 )( y3 − y ) ( x4 − x )( y − y2 ) ( x − x3 )( y − y1 )
v ( x, y ) = v1 + v2 + v3 + v4
( x2 − x1 )( y4 − y1 ) ( x2 − x1 )( y3 − y2 ) ( x4 − x3 )( y3 − y2 ) ( x4 − x3 )( y4 − y1 )

ui, vi indicate the horizontal/vertical displacements at each one of the 4 nodes (i) of this
element
Method of Finite Elements I
30-Apr-10
 Institute of Structural Engineering Page 21

Lagrange Polynomials – 2D Example

4 : u 4 , v4 3 : u 3 , v3

The polynomials we used in this simple

example are of order 1
y
1: u1 , v1
2 : u 2 , v2
x

If polynomials of higher order are to be used,

further nodes should be typically introduced
within the elements.

[F] Ch16§8
 the nine-node biquadratic
Quadrilateral (Lagrangian)

Method of Finite Elements I

30-Apr-10
 Institute of Structural Engineering Page 22

Serendipity Polynomials
These functions are similar to Lagrangian polynomials, but are incomplete (they have
missing terms). Due to this fact we do not need to introduce additional inner nodes, as
for Lagrangian polynomials of higher order.

Lagrange Polynomials

Serendipity Polynomials

Method of Finite Elements I

30-Apr-10
 Institute of Structural Engineering Page 23

Serendipity Polynomials
These functions are similar to Lagrangian polynomials, but are incomplete (the have
missing terms). Due to this fact we do not need to introduce additional inner nodes, as
for Lagrangian polynomials of higher order.

Lagrange Polynomials Serendipity Polynomials

 the nine-node biquadratic  the eight-node “serendipity”

Quadrilateral (Lagrangian) quadrilateral.
[F] Ch16§8
Method of Finite Elements I
30-Apr-10
 Institute of Structural Engineering Page 24

Hermitian Polynomials
Lagrangian polynomials and serendipity functions provide a C0 continuity (when
similar shapes are adjoined).
If we additionally need continuity of the first derivatives between the finite
elements we use Hermitian polynomials.
The first few Hermitian polynomials:

Η0 ( x) =
1
Η1 ( x ) =
2x
Η 2 ( x ) =4 x 2 − 2
Η 3 ( x ) =8 x3 − 12 x
Η 4 ( x ) = 16 x 4 − 48 x 2 + 12
...

Source: Wolfram Mathworld

Method of Finite Elements I

30-Apr-10
 Institute of Structural Engineering Page 25

Hermitian Polynomials

Let us consider a uniaxial element with nodes on its ends. Unknowns are values of the
function φ at the end nodes 1 and 2, φ1 and φ2, and the first derivatives of φ with
respect to x , φ1,x and φ2,x

ϕ ( x) ϕ2,x Remember: The 1st derivative of vertical

ϕ1 displacement, corresponds to rotation
ϕ1,x ϕ2
1 ( x1 ) 2 ( x1 )

dϕ ( x ) dϕ ( x )
=ϕ1, x = , ϕ2, x
=
dx x x= dx x x2
1

Question: What well-known type of element will need to be represented by Hermitian

Polynomials?

Method of Finite Elements I

30-Apr-10
 Institute of Structural Engineering Page 26

Hermitian Polynomials
Hermitian shape functions relate not only the displacements at nodes to
displacements within the elements, but also to the first order derivatives
(e.g. rotational DOFs for a beam element).

2
 ∂ui ( x) 
=
Shape function u ( x) ∑  0i ( ) i
i =1 
N x u + N1i ( )
x
∂x 
 Shape function of
the derivative u´(x)
of u(x)

N 01 N 02
N 0i ( x ) = 1 at node i and 0 at other nodes
N 0′i ( x ) = 0 at all nodes
N1i ( x ) = 0 at all nodes
N1′i ( x ) = 1 at node i and 0 at other nodes N11
N12

Method of Finite Elements I

30-Apr-10
 Institute of Structural Engineering Page 27

Isoparametric Elements
So far we discussed the form of Shape Functions using a formulation of the
element in global coordinates
This element formulation may be extended to any type of element however,
but suffers from the following limitations:

1 The construction of C0 (or C1) shape functions for high order

elements with curved or arbitrary geometry is highly complicated.

2 The evaluation of integrals for the element stiffness matrix and force
vector is not possible in closed form.

These limitations may be overcome by combining

isoparametric representation & numerical quadrature.
current lecture next lecture & lab

Method of Finite Elements I

30-Apr-10
 Institute of Structural Engineering Page 28

Isoparametric Elements
In general, we would like to be able to represent any element in a
standardized manner – introducing a transformation between a set of
standardized (natural) coordinates and the real (global) coordinates

Geometry
description
x, y

Natural Shape
coordinates functions
r , s, t N i ( r , s, t )
Displacement
interpolation
ux , u y

[F] Ch16§2
Method of Finite Elements I
30-Apr-10
 Institute of Structural Engineering Page 29

Isoparametric Elements
In general, we would like to be able to represent any element in a
standardized manner – introducing a transformation between a set of
standardized (natural) coordinates and the real (global) coordinates
Different schemes exist for establishing such transformations:

1 sub parametric representations (less nodes for

Here we will use this one

geometric than for displacement representation)

2 isoparametric representations (same nodes for both

geometry and displacement representation)

3 super parametric representations (more nodes for

geometric than for displacement representation)

Method of Finite Elements I

30-Apr-10
 Institute of Structural Engineering Page 30

Isoparametric Elements
Displacement fields as well as the geometrical representation of the finite
elements are approximated using the same approximating functions – shape
functions

4 uˆ4 , vˆ4
s
y -1, 1 1, 1

uˆ1 , vˆ1 1 3 uˆ3 , vˆ3

r
2 uˆ2 , vˆ2
-1, -1 1, -1
x

This transformation allows us to refer to similar elements (eg. truss, beams,

2D elements) in a standard manner, using the natural coordinates (r,s),
without every time referring to the specific global coordinate system (x,y).

Method of Finite Elements I

30-Apr-10
 Institute of Structural Engineering Page 31

Isoparametric Elements
Isoparametric elements can be one-, two- or three-dimensional:

Geometric interpolation
n n n
= x ∑
= N i (r , s, t ) xi ; y ∑
= N i (r , s, t ) yi ; z ∑ N i (r , s, t ) zi
same approximating
=i 1 =i 1 =i 1
functions – shape
n n n
functions for
= u ∑
= N i (r , s, t )ui ; v ∑= N i (r , s, t )vi ; w ∑ N i (r , s, t )wi
=i 1 =i 1 =i 1

Displacement Interpolation

The principle is to assure that the value of the shape function Ni is

equal to one in node i and equals zero in other nodes.

Method of Finite Elements I

30-Apr-10
 Institute of Structural Engineering Page 32

Consider the bar element with two Bar Element

end nodes at points x1, x2 defined on
the Cartesian axis x.

Truss Element Shape Functions

 u1 
u ( x ) = [ N1 N2 ]  
u2 
(in global coordinates)

Method of Finite Elements I

30-Apr-10
 Institute of Structural Engineering Page 33

Strain-Displacement relation

The truss element

define
stress displacement matrix

Constant Variation of strains along the element’s length

Method of Finite Elements I

30-Apr-10
 Institute of Structural Engineering Page 34

Consider the bar element with two Bar Element

end nodes at points x1, x2 defined on
the Cartesian axis x.

and now consider the following standard

truss element defined on the natural axis ξ

ξ = -1 ξ=0 ξ=1

Is there a function M that can map each point on the x axis to a point on the ξ
axis, so that: 1. ξ = ξ x = M x = −1
@ x =0 ( 1) ( 1)
2. ξ=
@ x=L ξ=( x2 ) M=( x2 ) 1
Method of Finite Elements I
30-Apr-10
 Institute of Structural Engineering Page 35

Bar Element

x2
The relation between the x-coordinate and y
x1
the r-coordinate is given as:

x
1 1 2
x = (1 − ξ ) x1 + (1 + ξ ) x2 = ∑ N i (ξ ) xi û1 û2
2 2 i =1
ξr==-1−1 ξr ==00 rξ = 11

Then, relation between the displacement u and the nodal

displacements are defined in the same way:

1 1 2
u (ξ ) = (1 − ξ )uˆ1 + (1 + ξ )uˆ2 = ∑ N i (ξ ) uˆi
2 2 i =1

Method of Finite Elements I

30-Apr-10
 Institute of Structural Engineering Page 36

Finite Element Idealization

Isoparametric Coordinates ξ
A, l = 2

ξ = −1 ξ =0 ξ =1

1
N1 (=
ξ) (1 − ξ )
Truss Element Shape Functions 2

u 
u (ξ ) =  N1 (ξ ) N 2 (ξ )   1  -1 1
u2 
(in isoparametric “natural” 1
coordinates) N 2 (=
ξ) (1 + ξ )
2

-1 1

Method of Finite Elements I

30-Apr-10
 Institute of Structural Engineering Page 37

Isoparametric formulation
u 
ξ
u (ξ ) =  N1 (ξ ) N 2 (ξ )   1  A, l = 2
u2 

1 1 ξ = −1 ξ =0 ξ =1
ξ)
N1 (= (1 − ξ ) N 2 (=
ξ) (1 + ξ )
2 2

Strain Displacement Matrix

Chair rule of ∂N ( x ) ∂N (ξ ) ∂ξ ∂N (ξ ) −1 1
differentiation
B
= = = J= [ −1 1] Why?
∂x ∂ξ ∂x ∂ξ L

dx  ∂N1 (ξ ) ∂N 2 (ξ )   x1  L
−1
∂ξ dξ  dx  =x N1(ξ ) x1 + N 2 (ξ ) x2
= =   → =    x = = J
∂x dx  dξ  dξ  ∂ξ ∂ξ   2  2
dξ 2
This is termed the Jacobian, J ⇒ = =J −1 > 0
dx L
Method of Finite Elements I
30-Apr-10
 Institute of Structural Engineering Page 38

Isoparametric formulation
u 
ξ
u (ξ ) =  N1 (ξ ) N 2 (ξ )   1  A, l = 2
u2 

1 1 ξ = −1 ξ =0 ξ =1
ξ)
N1 (= (1 − ξ ) N 2 (=
ξ) (1 + ξ )
2 2

Strain Displacement Matrix

∂N ( x ) ∂N (ξ ) ∂ξ ∂N (ξ ) −1 1
Chair rule of B
= = = J= [ −1 1] Why?
differentiation ∂x ∂ξ ∂x ∂ξ L

Constant Variation of strains along the element’s length

Method of Finite Elements I

30-Apr-10
 Institute of Structural Engineering Page 39

Isoparametric Elements
Let us now consider the derivation of the stiffness matrix K, using the iso-
parametric expression.
*The strain-displacement matrix is the same for the case of the bar element
for both global and isoparametric coordinates:
1  uˆ1 
B= [ −1 1] with ε =
B 
L uˆ2 
We write up the integral for calculating the stiffness matrix:
1
K ∫ EAB dx
B ∫ EAB det J dξ
B
T T
= = Why?
L −1

dx
since=dx = d ξ Jd ξ
dξ

Method of Finite Elements I

30-Apr-10
 Institute of Structural Engineering Page 40

Bar Element
The strain-displacement matrix in this case remains unchanged:

1  uˆ1 
B= [ −1 1] with ε =
B 
L uˆ2 

and the stiffness matrix is calculated as:

AE  −1
1
dx L
2 ∫ 
K =−  [ 1 1] Jdξ , J= = ⇒
L −1  1 dξ 2
1
AE L  1 −1 AE  1 −1
K  =  ξ ⇒ K
L 2  −1 1 −1
2
L  −1 1

Method of Finite Elements I

30-Apr-10
 Institute of Structural Engineering Page 41

Isoparametric Elements in 3D
As pointed out, in order to evaluate the stiffness matrix we must differentiate the
displacements with respect to the coordinates (x, y, z).

The shape functions will be expressed in natural coordinates. Therefore the

coordinate transformation (Jacobian) between natural and cartesian coordinate
system should be used.
In the 3D domain we can express the derivative of a function 𝜑𝜑,which is
expressed in terms of coordinates (x,y,z), with respect to another set of
coordinates (r,s,t) as:
∂ϕ ∂ϕ ∂x ∂ϕ ∂y ∂ϕ ∂z
= + +
∂r ∂x ∂r ∂y ∂r ∂z ∂r
∂ϕ ∂ϕ ∂x ∂ϕ ∂y ∂ϕ ∂z
= + + Chain rule of differentiation!
∂s ∂x ∂s ∂y ∂s ∂z ∂s
∂ϕ ∂ϕ ∂x ∂ϕ ∂y ∂ϕ ∂z
= + +
∂t ∂x ∂t ∂y ∂t ∂z ∂t
Method of Finite Elements I
30-Apr-10
 Institute of Structural Engineering Page 42

Isoparametric Elements

Written in matrix notation:

∂φ ∂φ ∂x ∂φ ∂y ∂φ ∂z  ∂φ   ∂x ∂y ∂z   ∂φ 
= + +
∂r ∂x ∂r ∂y ∂r ∂z ∂r  ∂r   ∂r ∂r ∂r   ∂x 
   
∂φ ∂φ ∂x ∂φ ∂y ∂φ ∂z  ∂φ  =  ∂x ∂y ∂z   ∂φ 
= + + ⇒
∂s ∂x ∂s ∂y ∂s ∂y ∂s  ∂s   ∂s ∂s ∂s   ∂y 
    
∂φ ∂φ ∂x ∂φ ∂y ∂φ
= + +
∂z  ∂φ   ∂x ∂y ∂z   ∂φ 
∂t ∂x ∂t ∂y ∂t ∂z ∂t  ∂t   ∂t ∂t ∂t   ∂z 

Method of Finite Elements I

30-Apr-10
 Institute of Structural Engineering Page 43

Isoparametric Elements
This is the Jacobian operator J:

 ∂φ   ∂x ∂y ∂z   ∂φ 
 ∂r   ∂r ∂r ∂r   ∂x  ∂ ∂ ∂x ∂
    = = J ⇒
 ∂φ  =  ∂x ∂y ∂z   ∂φ 
∂r ∂x ∂r ∂x
 ∂s   ∂s ∂s ∂s   ∂y  ∂ ∂
     ⇒ J −1
=
 ∂φ   ∂x ∂y ∂z   ∂φ 
∂x ∂r
 ∂t   ∂t ∂t ∂t   ∂z 

J
which relates the natural coordinate derivatives to the Cartesian
coordinate derivatives.

Method of Finite Elements I

30-Apr-10
 Institute of Structural Engineering Page 44

Isoparametric Elements
The expression of element stiffness matrix

K = ∫ BT CB dV
V
The strain-displacement matrix B will be expressed in terms of the
natural coordinates (r,s,t) and the volume differential dV refers to the
Cartesian coordinate system.

Therefore: dV = det J dr ds dt

*The Jacobian determinant is used when making a change of variables and

evaluating a multiple integral of a function over a region within its domain. To
accommodate for the change of coordinates the magnitude of the Jacobian
determinant arises as a multiplicative factor within that integral.

Method of Finite Elements I

30-Apr-10
 Institute of Structural Engineering Page 45

Isoparametric Elements
The expression of element stiffness matrix

K = ∫ BT CB dV
V
The strain-displacement matrix B will be expressed in terms of the
natural coordinates and the volume differential dV refers to the
Cartesian coordinate system.

Therefore: dV = det J dr ds dt

K = ∫ BT CB det J dr ds dt
V

Method of Finite Elements I

30-Apr-10
 Institute of Structural Engineering Page 46

Isoparametric Elements
The expression of the equivalent nodal force vector is

f = ∫ NT q dV
V
The strain-displacement matrix N will be expressed in terms of the
natural coordinates (r,s,t) and the volume differential dV refers to the
Cartesian coordinate system.

Therefore: dV = det J dr ds dt

f = ∫ NT q det J dr ds dt
V

Method of Finite Elements I

30-Apr-10
 Institute of Structural Engineering Page 47

Example: Quadratic Bar Element with 3 nodes

Based on what we so far discussed, how can we define the equations
for a quadratic bar element with 3 nodes?
Define the higher order element by using the Lagrange Polynomial for
3 nodes, in isoparametric coordinates ξ

i (ξ ) i (ξ ) L3i (ξ ) ⇒
n
N= L=
( ξ − ξ 2 )(ξ − ξ3 ) ξ = 1
N= ( ξ ) −1,ξ =
 0
→1
N = ( ξ ) ξ (ξ − 1)
2
1
( 1 2 )( 1 3 )
ξ − ξ ξ − ξ ξ =1 1 3
2

N 2 (ξ ) =
(ξ − ξ1 )(ξ − ξ3 ) 
ξ =−1,ξ =0
→ N 1
( ξ ) = 1 − ξ 2
2

(ξ 2 − ξ1 )(ξ 2 − ξ3 ) ξ =1 2 3

( ξ − ξ1 )(ξ − ξ 2 ) ξ = 1
N= ( ξ ) −1,ξ =
0
→1
N= ( ξ ) ξ (ξ + 1)
2
3
(ξ3 − ξ1 )(ξ3 − ξ 2 ) ξ =1 3 3
2

Method of Finite Elements I

30-Apr-10
 Institute of Structural Engineering Page 48

Example: Quadratic Bar Element with 3 nodes

Self Study Task
Derive the stiffness matrix and the equivalent nodal force vector of the
quadratic 3-noded bar element!
Stiffness Matrix

Nodal force
vector

Solution: [O] Ch3§3.3

Method of Finite Elements I

30-Apr-10
 Institute of Structural Engineering Page 49

Higher order bar Elements

[O] Ch3§2, Fig. 3.2

Method of Finite Elements I