FEM - 6 Isoparametric Elements

Finite Element Course
6 Isoparametric elements
Prof. Raul Durand

Finite Local and global coordinate systems
Elements
Local and global coordinate systems
Raul
Durand 1
Finite Interpolation
Elements
Interpolation
In the finite element method, continuous models are approxi-
mated using information at a finite number of discrete points
(nodes).
Interpolation of nodal values within elements is achieved
through the use of shape functions. For example, displace-
ment at point x along a 1D element can be estimated by
interpolation of its nodal displacements:
u = N1 (x)u1 + N2 (x)u2
where Ni (x) is the shape function of node i.

For convenience, shape functions can be written in terms of
natural (local) coordinates, in this case ξ:
u = N1 (ξ)u1 + N2 (ξ)u2
Similarly, global coordinates can also be interpolated as:
x = N1 (ξ)x1 + N2 (ξ)x2
Raul
Durand 2
Finite Element mapping
Elements
Element mapping
The mapping process allows to calculate global coordinates from
local ones and vice-versa. Thus, the real coordinates of a point
within an element can be found by interpolation of nodal coor-
dinates. For example, if a point P has local coordinate ξ in a
three-node 1D element, the global coordinate x is found by:
x = N1 (ξ)x1 + N2 (ξ)x2 + N3 (ξ)x3
If the same element is in 3D space, the global coordinates are

given by:
 
h i h i x1 y1 z1
x y z = N1 (ξ) N2 (ξ) N3 (ξ) x2 y2 z2 
 
x3 y3 z3
or xT = NX(e) , where x is a vector with global coordinates and

X(e) a matrix comprising all nodal element coordinates.
The inverse process, finding local coordinates from global ones,
is called inverse mapping and need to solve a system of nonlinear
equations.
Raul
Durand 3
Finite Isoparametric elements
Elements
Isoparametric elements
Finite elements that use the same set of shape functions to represent both the element geom-
etry and displacement are called isoparametric elements. Shape functions of isoparametric
elements need to attend two relations. For example, in a 2D element we have:
Geometric relations:
n
X n
X n
X
1= Ni ; x= Ni xi ; y= Ni yi
i=1 i=1 i=1
Nodal displacement interpolation:

n
X n
X
ux = Ni uxi ; uy = Ni uyi
i=1 i=1
Other variables, e.g. temperature, along the element can be interpolated the same way.
Isoparametric element equations are formulated using the natural coordinate system that is
defined according to the element geometry and not by the element position in the global
coordinate system.
The isoparametric formulation allows formulting elements with curved shapes. Also, isopara-
metric elements are the standard in finite element codes.
Raul
Durand 4
Finite Shape functions for 1D elements
Elements
Shape functions for 1D elements

A 1D element is defined with the natural coordinate ξ. The basic
functions to interpolate two points are:
N1 (ξ) = a1 + b1 ξ N2 (ξ) = a2 + b2 ξ
Shape function should return 1 if evaluated at the correspond-

ing point and 0 otherwise. Thus, a and b coefficients are found
substituting ξ = −1 and ξ = 1:
a1 + b1 (−1) = 1 a1 + b1 (+1) = 0
a2 + b2 (−1) = 0 a2 + b2 (+1) = 1
which leads to the following systems: 1.0

" # ! ! " # ! ! 0.8
1 −1 a1 1 1 −1 a2 0 0.6
= and =
1 1 b1 0 1 1 b2 1
Ni
0.4
0.2
Solving systems above, shape functions are:
0.0
1.0 0.5 0.0 0.5 1.0
1 ξ 1 ξ ξ
N1 = 2 − 2 N2 = 2 + 2 N1 N2
Since interpolation is linear, two-node isoparametric elements are Shape functions for two-node
called linear elements. elements
Raul
Durand 5
Elements
For a three-node 1D element, also called quadratic ele-

ment, shape functions have the form:
Ni (ξ) = ai + bi ξ + ci ξ 2
To find a, b and c coefficients three systems of equations

are defined:
    
1 −1 1 a1 1
1 1 1  b1  = 0
    
1 0 0 c1 0
    
1 −1 1 a2 0
1 1 1  b2  = 1
    
1 0 0 c2 0
    
1 −1 1 a3 0
1 1 1  b3  = 0
    
1 0 0 c3 1
As can be seen, all systems have the same coefficients

matrix.
Raul
Durand 6
Elements
So, they can be condensed into the following matrix

equation:
    
1 −1 1 a1 a2 a3 1 0 0
1 1 1  b1 b2 b3  = 0 1 0
    
1 0 0 c1 c2 c3 0 0 1
1.0
0.8
Then, all shape functions coefficients are found by in-
0.6
version of the coefficients matrix in the equation above:
0.4
Ni
0.2
   −1 0.0
a1 a2 a3 1 −1 1 0.2
1.0 0.5 0.0 0.5 1.0
 b1 b2 b3  = 1 1 1
    ξ
N1 N2 N3
c1 c2 c3 1 0 0
Shape functions for three-node elements
Finally, the three shape functions are:
N1 = 21 ξ 2 − 12 ξ
N2 = 21 ξ 2 + 12 ξ
N3 = 1 − ξ 2
Raul
Durand 7
Finite Exercises
Elements
1D Lagrangian shape functions

1.0
For n-node 1D elements, Lagrangian polynomials that 0.8
interpolates points in the interval [−1, 1] can be used to 0.6
generate shape functions. In this case these polynomials 0.4
Ni
take the form: 0.2
n 0.0
ξ−ξj 0.2
Y
Ni (x) = ξi −ξj 1.0 0.5 0.0 0.5 1.0
j=0, j6=i ξ
N1 N2 N3
For example, for a two-node element we have: Shape functions for three-node elements
ξ−1 1 ξ
N1 = −1−1 = 2 − 2 1.2
ξ−(−1) 1 ξ
1.0
N2 = 1−(−1) = 2 + 2 0.8
0.6
0.4
Ni
and for a three-node element it gives: 0.2
0.0
(ξ−1)(ξ−0) 0.2
N1 = = 12 ξ 2 − 12 ξ 0.4
(−1−1)(−1−0) 1.0 0.5 0.0 0.5 1.0
(ξ−(−1))(ξ−0) 1 2 1
ξ
N2 = (1−(−1))(1−0) = 2ξ + 2ξ N1 N3
(ξ−(−1))(ξ−1) 2
N2 N4
N3 = (0−(−1))(0−1) =1−ξ
Shape functions for four-node elements
Raul
Durand 8
Finite Exercises
Elements
Exercises
1. Determine the shape functions for the four-node and five-node 1D by calculating the
polynomial coefficients using a system of equations. Later, compare the result with
the shape functions obtained using Lagrange polynomials. Plot the obtained shape
functions.
2. Nodal displacements at the three nodes of a 1D element are shown in the table below.
Find the displacement at ξ = 0.5. Explain why the result of interpolation is a negative
value if all nodal values are positive?
u1 u2 u3
0.99 0.11 0.05
3. Values of a variable u found at the five nodes of a 1D element are shown in the table
below. Plot the u field along the element by using shape functions.
u1 u2 u3 u4 u5
0.0 0.22 −0.15 0.12 0.15
4. Using computer software elaborate a routine that computes the shape function coeffi-
cients of a n-node 1D element, with n = 2, 3, 4, 5.
Raul
Durand 9
Finite Linear triangular element
Elements
Linear triangular element

A three-node triangles is the simplest 2D solid finite element. It is a linear element and
frequently referred as constant strain triangle (CST).
Conventionally, 2D elements have the nodes num-

bered counterclockwise, so for this triangle element
we have:
Since two local coordinates are required for shape
functions, they assume the following form:
Ni (ξ, η) = ai + bi ξ + ci η
Substituting all local coordinates points in Ni the

following matrix equation is generated:
    
1 0 0 a1 a2 a3 1 0 0
1 1 0  b1 b2 b3  = 0 1 0
    
1 0 1 c1 c2 c3 0 0 1
Raul
Durand 10
Finite Linear triangular element
Elements
Since the right-hand-side of the equation is equal to the

identity matrix I, this equation can be written incon-
densed form as:
AC = I
where A contains the systems coefficients and the 1.0

0.8
columns of C contains the shape functions coefficients 0.6
for each node. C matrix is easily found by: 0.4N
0.2 1
C = A−1 0.0
0.2
0.4
Extracting coefficients a, b and c from matrix C, the
shape functions for a linear triangle element are found 0.0 0.2
0.60.81.0
0.4 0.6 0.4
as: ξ 0.8 1.00.0 0.2 η
N1 = 1 − ξ − η Shape function of node 1

N2 = ξ
N3 = η
Raul
Durand 11
Finite Shape functions requirements
Elements
Shape functions requirements

The construction of shape functions need to meet certain conditions:
1. Interpolation
Shape function is continuous and derivable. A function
Ni is evaluated as 1 at node i, and as zero at all other
nodes.
2. Local support
A function Ni vanishes over any element side that does
not include node i.
3. Interelement compatibility
Shape functions should provide displacements continu-
ity between elements. Elements that satisfy this condi-
tion are called conforming or compatible.
Example of non-conforming elements
4. Completeness
Shape functions should contain all polynomials terms
up to the maximum degree term.
Raul
Durand 12
Finite Quadratic triangular element
Elements
Completeness of polynomials
The polynomial terms of shape functions can be obtained from Pascal’s triangle.
For triangle elements, for example, terms can be taken from the top using triangular shapes.
For lagrangian square elements, terms are taken using diamond shapes. For Serendipity
elements some terms are avoided keeping symmetry.
Raul
Durand 13
Finite Quadratic triangular element
Elements
Quadratic triangular element

This element is one of the most versatile elements since
it can be used in the generation of irregular meshes and
1.0
its quadratic interpolation leads to a variable strain field 0.8
inside the element. 0.6
0.4N
Nodal numbering and local coordinates are specified in 0.2 1
0.0
the figure below: 0.2
0.4
0.0 0.2
0.60.81.0
0.4 0.6 0.4
ξ 0.8 1.00.0 0.2 η
1.0
0.8
0.6
0.4N
0.2 6
Shape functions have the form: 0.0
0.2
Ni (ξ, η) = ai + bi ξ + ci η + di ξ 2 + ei η 2 + fi ξη 0.4
0.0 0.2
0.60.81.0
0.4 0.6 0.4
All coefficients from the shape functions can be found ξ 0.8 1.00.0 0.2 η
following the same procedure used for triangular ele-
Shape functions for nodes 1 and 6
ments.
Raul
Durand 14
Finite Bilinear quadrilateral element
Elements
Bilinear quadrilateral element

This element contains only four nodes as shown below:
1.0
0.8
0.6
0.4N
0.2 1
0.0
0.2
0.4
1.0 1.0
0.5 0.5
0.0
ξ0.0 0.5 0.5 η
1.0 1.0
Shape functions have the form:
1.0
Ni (ξ, η) = ai + bi ξ + ci η + di ξη 0.8
0.6
Using Lagrange polynomial, the first shape function can 0.4N
0.2 3
be written as: 0.0
0.2
N1 (ξ, η) = 12 (ξ − 1) 12 (η − 1) 0.4
1.0 1.0
0.5 0.5
0.0
A general form for the shape functions of this element is ξ0.0 0.5 0.5 η
1.0 1.0
given by:
Ni (ξ, η) = 14 (1 + ξi ξ)(1 + ηi η)
Raul
Durand 15
Finite Serendipity quadrilateral element
Elements
Biquadratic quadrilateral element
This element contains 9 nodes and its numbering is de- 1.0

fined as follows: 0.8
0.6
0.4N
0.2 1
0.0
0.2
0.4
1.0 1.0
0.5 0.5
0.0
ξ0.0 0.5 0.5 η
1.0 1.0
1.0
0.8
0.6
0.4N
It is called biquadratic since the highest degree term in 0.2 9
0.0
shape functions is ξ 2 η 2 . 0.2
0.4
This element is one from the set of lagrangian elements, 1.0 1.0
0.5 0.5
0.0
ξ0.0 0.5
therefore shape functions can be easily found by La-
0.5 η
grange polynomials products. 1.0 1.0
Raul
Durand 16
Finite Serendipity quadrilateral element
Elements
Serendipity quadrilateral element

This quadratic quadrilateral element contains just 8 nodes.
It is part of a set called Serendipity elements that have the
1.0
nodes distributed in the element boundary. 0.8
0.6
0.4N
0.2 1
0.0
0.2
0.4
1.0 1.0
0.5 0.5
0.0
ξ0.0 0.5 0.5 η
1.0 1.0
It is one of the most used elements in 2D domains with 1.0

regular meshes. 0.8
0.6
Shape functions for corner nodes are: 0.4N
0.2 8
0.0
Ni (ξ, η) = 41 (1 + ξξi )(1 + ηηi )(ξξi + ηηi − 1) 0.2
0.4
and for mid-side nodes are: 1.0 1.0
0.5 0.5
0.0
ξ0.0 0.5 0.5 η
Ni (ξ, η) = 21 (1 − ξ 2 )(1 + ηηi ) for ξi = 0 1.0 1.0
1 2 Shape functions for nodes 1 and 8
Ni (ξ, η) = 2 (1 + ξξi )(1 − η ) for ηi = 0
Raul
Durand 17
Finite Trilinear hexahedron element
Elements
Linear tetrahedron element

This tetrahedron with 4 nodes is the simplest 3D solid element. It is widely used in unstruc-
tured 3D meshes.
The shape functions are given by:
N1 (ξ, η, ζ) = 1 − ξ − η − ζ
N2 (ξ, η, ζ) = ξ
N3 (ξ, η, ζ) = η
N4 (ξ, η, ζ) = ζ
As can be seen all functions are linear, thus that is the reason why this element is called
linear.
Raul
Durand 18
Finite Trilinear hexahedron element
Elements
Trilinear hexahedron element

This hexahedron element contains 8 nodes. It is the simplest member of the hexahedron
family.
It is usually referred to as brick element and is extensively used in 3D analysis.

The shape functions are given by:
Ni (ξ, η, ζ) = 18 (1 + ξi ξ)(1 + ηi η)(1 + ζi ζ)
This element is referred as trilinear hexahedron since shape functions highest term is in ξηζ.
Sometimes it is incorrectly called linear.
Raul
Durand 19
Finite Other 3D elements
Elements
Other 3D elements
The 10-node serendipity tetrahedron:
The 20-node serendipity hexahedron:
Raul
Durand 20
Finite Remarks
Elements
Remarks
Although elements with few nodes have the advantage of requiring fewer calculations they
could not show an accurate interpolation field.
In general, it is recommended to avoid linear elements in favor of elements with at least
quadratic terms.
Serendipity elements are very convenient since they use fewer nodes to represent the same
area or volume. However lagrangian elements can provide a better interpolation field due to
intermediary nodes.
Extrapolation, estimating values outside element geometry, is highly not recommended be-
cause shape functions vary severely outside element boundaries.
Raul
Durand 21
Finite Exercises
Elements
Exercises
1. Why the first three shape functions of a quadratic triangle element cannot be used to
interpolate nodal values of a linear triangle element?
2. List all terms needed to generate the shape functions for a serendipity quadrilateral
element with 12 nodes. Use computer software to compute all polynomials coefficients.
3. Compute all polynomial coefficients of the shape functions of a 10-node tetrahedron
element.
4. Find all shape functions for the transition element shown below.
5. Using computer software plot the shape functions of nodes 1 and 5 for the element in
the last exercise.
Raul
Durand 22
Finite Jacobian matrix
Elements
Jacobian matrix
In multivariable calculus, the Jacobian matrix is the matrix of all first-order partial deriva-
tives of a vector function.
For example, for a vector function f : R → R that comprises two functions f1 (x, y) and
f2 (x, y), the Jacobian matrix is given by:
" ∂f1 ∂f1 #
∂(f1 ,f2 ) ∂x ∂y
J= ∂(x,y) = ∂f2 ∂f2
∂x ∂y
A more general expression for n functions that depend on m variables is:

 ∂f ∂f1

1
∂x1 ··· ∂xm
 . .. .. 
 ..
J(x1 , x2 , . . . , xm ) =  . . 

∂fn ∂fn
∂x1 ··· ∂xm
If m = n the matrix is square and the determinant J is often called just as the Jacobian.
One use of the Jacobian determinant is in the change of variables when evaluating a multiple
integral over a region. It is also useful to relate the changes between two coordinate systems,
for example the system defined by f1 , f2 , . . . , fn and the system defined by x1 , x2 , . . . , xm .
Raul
Durand 23
Elements
The simplest use of a Jacobian is for the integration of

a one-variable function. For example, consider:
Z 1√
2x + 1dx
0
which can be solved by using the substitution u = 2x+1.

0 1
In this case, x and u define two 1D coordinate systems
and both can be related by the Jacobian:
du
J= dx 1 3
then du = 2dx and the Jacobian J = 2 represents the Intervals in x and u spaces
length ratio between u and x systems.
By substitution, the integral can be redefined as:
Z 3
√
u 12 du
1
Raul
Durand 24
Elements
Now consider a multiple integral:

2
ZZ 4
4x2 − y 2 dxdy
R
which can be rewritten as: (1/2, 1)

ZZ
(2x − y)4 (2x + y )4 dxdy
R
1/2
Making the substitutions u = 2x − y and v = 2x + y the
Jacobian matrix between systems uv and xy is found:
" ∂u ∂u
# " #
∂x ∂y 2 −1 2
J= ∂v ∂v =
∂x ∂y
2 1
The determinant is J = 4 and it can be proved that

dudv = 4dxdy. After substitution, the multiple integral
turns to:
-2
Z 2Z 0
u4 v 4 41 dudv
0 −v Region R in spaces xy and uv
Raul
Durand 25
Finite Area ratio
Elements
Area ratio
Consider two 2D coordinate systems, say a global xy
and a local ξη systems, where x = x(ξ, η) and y =
y(ξ, η). Total differentials are:
∂x ∂x
dx = ∂ξ dξ + ∂η dη
∂y ∂y
dy = ∂ξ dξ + ∂η dη
and lead to the following relation between differentials:

! " ∂x ∂x
# !
dx ∂ξ ∂η dξ
= ∂y ∂y Area differential
dy ∂ξ ∂η
dη
A differential area dA equivalent to dxdy is shown in dx

!
dξ dξ
the figure at right. The corresponding differential area a= dy
in local coordinates is dA0 = dξdη. dξ dξ
dx
!
dη dη
Assuming vectors a and b, dA can be calculated by b= dy
the vectorial product dA = |a × b| and renders to the dη dη
following area ratio: Auxiliary vectors
dA = |J(ξ, η)|dξdη
Raul
Durand 26
Finite Volume ratio
Elements
Volume ratio
For the systems xyz and ξηζ we have x =
x(ξ, η, ζ), y = y(ξ, η, ζ) and z = z(ξ, η, ζ).
Total differentials lead to the following differ-
ential relation that is in terms on the Jacobian 𝑧
matrix:
   ∂x ∂x ∂x   
dx ∂ξ ∂η ∂ζ dξ
   ∂y ∂y ∂y   
dy =
   ∂ξ ∂η dη
∂ζ   
dz ∂z ∂z ∂z dζ
∂ξ ∂η ∂ζ 𝑦
A volume differential dV equivalent to dxdydz 𝑥

is shown in the figure. The corresponding vol-
Volume differential
ume differential in ξηζ system will be dV 0 =
dξdηdζ.
 dx   dx   dx 
Defining three auxiliary vectors as shown be- dξ dξ dη dη dζ dζ
 dy   dy   dy 
side the volume differential dV can be calcu- a=  dξ dξ  b=  dη dη  c=  dζ dζ 
dz dz dz
lated by vector operations as dV = |a × b| · c. dξ dξ dη dη dζ dζ
The volume ratio between both systems is: Auxiliary vectors
dV = |J(ξ, η, ζ)|dξdηdζ
Raul
Durand 27
Finite Inverse of Jacobian matrix
Elements
Inverse of Jacobian matrix

For two systems, say xyz and ξηζ the following relationships can be established:
Direct relationship:
∂x ∂x ∂x
dx = ∂ξ dξ + ∂η dη + ∂ζ dζ
∂y ∂y ∂y
dy = ∂ξ dξ + ∂η dη + ∂ζ dζ
∂z ∂z ∂z
dz = ∂ξ dξ + ∂η dη + ∂ζ dζ
Inverse relationship:
∂ξ ∂ξ ∂ξ
dξ = ∂x dx + ∂y dy + ∂z dz
∂η ∂η ∂η
dη = ∂x dx + ∂y dy + ∂z dz
∂ζ ∂ζ ∂ζ
dζ = ∂x dx + ∂y dy + ∂z dz
If the first set of equations is used to define a Jacobian matrix J, then the second set defines
the inverse Jacobian matrix of the first set and it is given by:
 
  ∂ξ ∂ξ ∂ξ  
dξ ∂x ∂y ∂z  dx
   ∂η ∂η ∂η   
dη  = 
 ∂x ∂y ∂z   dy 
dζ ∂ζ ∂ζ ∂ζ dz
∂x ∂y ∂z
If J is available, the inverse Jacobian matrix can be obtained directly by matrix inversion.
Raul
Durand 28
Finite Jacobian in finite elements
Elements
Jacobian in finite elements

In finite elements, frequently the Jacobian matrix is defined as the transposed of its mathe-
matical definition. Thus, for 2D, the following Jacobian definition is used:
∂y
" ∂x #
∂ξ ∂ξ
J= ∂x ∂y
∂η ∂η
and for 3D:

 
∂x ∂y ∂z
∂ξ ∂ξ ∂ξ
 ∂x ∂y ∂z

J=
 ∂η ∂η ∂η


∂x ∂y ∂z
∂ζ ∂ζ ∂ζ
Since the determinant of a matrix is the same as the determinant of the transposed matrix,
then area and volume relationships already given are still valid.
Z Z
f (x) dx = f (ξ) J dξ
0
The Jacobian matrix is useful for Z L ZL
the integration of functions over el- f (x, y) dA = f (ξ, η) Jdξdη
A 0
ements, so the integral transforma- Z ZA
tions shown beside apply. f (x, y, z) dV = f (ξ, η, ζ) Jdξdηdζ
V V0
Change of variables for integration

Raul
Durand 29
Finite Computing the Jacobian matrix
Elements
Computing the Jacobian matrix

∂x ∂y
To construct a Jacobian matrix, all terms such as ∂ξ , ∂ξ , etc.
need to be calculated.
For isoparametric elements, coordinates x, y and z can be ex-
pressed as functions of natural coordinates by means of shape
functions and nodal coordinates, as shown at right.
∂x
For instance, to get ∂ξ , the coordinate x is represented by inter- n
X
polation as: x= Ni (ξ, η, ζ) xi
i=1
n
!
X n
∂x ∂ X
∂ξ = ∂ξ Ni (ξ, η, ζ) xi y= Ni (ξ, η, ζ) yi
i=1 i=1
n
Since the element nodal coordinates are constant (xi ), the equa-
X
z= Ni (ξ, η, ζ) zi
tion above can be rewritten as: i=1
n
∂x
X
∂
Coordinates interpolation
∂ξ = ∂ξ (Ni (ξ, η, ζ)) xi
i=1
h ih iT
∂N1 ∂N2 ∂Nn
= ∂ξ ∂ξ ··· ∂ξ x1 x2 · · · xn
The same procedure is used for all required terms that make up
the Jacobian matrix.
Raul
Durand 30
Finite Computing the Jacobian matrix
Elements
For a 3D element, the Jacobian matrix can be calculated by:

P ∂Ni P ∂Ni P ∂Ni 
∂ξ xi ∂ξ yi ∂ξ zi
P ∂N P ∂Ni P ∂Ni 
∂η xi yi ∂η zi 
i
J(ξ, η, ζ) = 
 ∂η

P ∂Ni P ∂Ni P ∂Ni
∂ζ xi ∂ζ yi ∂ζ zi
For convenience, derivatives of shape functions and nodal coordinates can be organized in
matrix form as:
 ∂N1 ∂N2 ∂Nn 
∂ξ ∂ξ ··· ∂ξ
∂N
 ∂N1 ∂N2 ∂Nn 
∂(ξ,η,ζ) = 
 ∂η ∂η ··· 
∂η 
∂N1 ∂N2 ∂Nn
∂ζ ∂ζ ··· ∂ζ
and
 
x1 y1 z1
 x2 y2 z2 


X=
 .. .. .. 
 . . .

x n yn zn
Finally, the Jacobian matrix can be calculated by the following matrix product:
∂N
J= ∂(ξ,η,ζ) X
Raul
Durand 31
Finite Exercises
Elements
Exercises
1. Determine the Jacobian determinant for the transformation from Cartesian coordinates
to polar coordinates:
x = r cos θ, y = r sin θ
and from Cartesian coordinates to spherical coordinates:
x = ρ sin φ cos θ, y = ρ sin φ sin θ, z = ρ cos φ
2. Calculate the integral below and show the Jacobian matrix used.
Z +∞
2 −y 2
I= e−x dx
−∞
3. Nodal coordinates of a 1D element are: x1 = 2.0, x2 = 10.0 and x3 = 9.0. Plot the
Jacobian J on the natural coordinates space and explain the behavior between nodes
2 and 3.
4. Write a routine that given the coordinates of a quadrilateral element with four nodes
calculates the Jacobian at given coordinates (ξ, η).
5. If the area of a triangular element is equal to A, show that the Jacobian determinant
is equal to 2A.
Raul
Durand 32
Finite Jacobian restrictions
Elements
Jacobian restrictions
3
(3,3)
It is required Jacobian positiveness for all elements in
a mesh. Thus, for each element, the determinant J = 4
det J must be positive at any interior point. (1,2)
Two-node line elements and three-node triangle ele-

ments have always constant Jacobians. Also, any el- 1 2
(1,1) (2,1)
ement can have constant Jacobian along its domain if
its shape and orientation in global coordinates are the
same as in natural coordinates. Global coordinates
To comply the requirement J > 0 all nodes must be
properly positioned and numbered. This is called con-
vexity condition. For example, the convex four-node el- 1.0
0.8
ement shown at right J is positive over whole element. 0.6
0.4 J
The scalar field defined by the Jacobian is linear show- 0.2
ing that the area ratio between local and global coordi- 0.0
0.2
nates is variable and positive for the element domain. 0.4
1.0 1.0
Finite element codes should check node numbering and 0.5 0.5
0.0
Jacobian positiveness for all elements prior analysis. ξ0.0 0.5 0.5 η
1.0 1.0
Jacobian determinant field
Raul
Durand 33
Elements
For 2D elements with more than three nodes distortions

may render regions with negative Jacobian determinant. 4
(1,2)
For example, figure at right shows a quadrilateral ele-
ment where node 3 was dislocated inwards in order to 3 (1.4,1.4)
get a concave shape. 1 2
(1,1) (2,1)
The Jacobian along the element is variable as show in
the second figure; however, near the node 3 there is
a region where the Jacobian is negative. In this case
Jacobian positiveness is not complied, thus the element Global coordinates
is invalid for finite element analysis.
In general, zero or negative Jacobians render the ele-
ment unacceptable for finite element analyses. Loca- 0.4
tions where Jacobian determinant is zero imply that 0.2
there is not an inverse Jacobian relation for the whole J
0.0
element. 0.2
In locations with negative Jacobians may not be unique 0.4
correspondence between points in local and global co- 1.0 1.0
0.5 0.5
0.0
ordinates. This means that some points in natural co- ξ0.0 0.5 0.5 η
1.0 1.0
ordinates can be mapped to global space but not all of
Jacobian determinant field
them can be mapped back.
showing negative values
Raul
Durand 34
Elements
This other example shows the effect of gradually moving a node of a quadrilateral element.
The top left figure shows the initial geometry while the other figures show the distortions
obtained after displacing node 4.
The top right figure is a triangle; however, the Jacobian determinant is zero at node 4.
Raul
Durand 35
Finite Restriction on mid-side nodes
Elements
Restriction on mid-side nodes

For higher order elements proper location of corner nodes
is not enough. Mid-side and interior nodes must be placed
sufficiently close to their natural locations in order to keep
positive Jacobians.
For instance, for a three-node bar element, consider the
following nodal coordinates: x1 = −L/2, x2 = L/2 and Local coordinate system
x3 = αL as shown in the figure. In this case the Jacobian
determinant is given by:

1
J(ξ) = 2 − 2ξα L
Global coordinate system

1
where the expression 2− 2ξα must be greater that zero.
Considering the element domain in natural coordinates ξ ∈
[−1, +1], it can be shown that α can vary along the interval
Position of mid-side node
[− 14 + 14 ]. That means that for three-node bars the mid-node
must be located along the central region with length L2 .
This result can be extended for the case of mid-side nodes on
2D and 3D elements, where their locations must be within
the central part of the side with length L2 , where, in this
case, L represents the side length.
Raul
Durand 36
Elements
The following figures show the distortions in a nine-node Lagrangian element by displacing
mid-side node 5.
Figures taken from “FEM Convergence Requirements”. C.A. Felippa.
Raul
Durand 37
Elements
Figures below show distortions in a quadratic triangle element.

Displacing mid-side nodes inwards:
Moving mid-side nodes outwards:
Figures taken from “FEM Convergence Requirements”. C.A. Felippa.
Raul
Durand 38
Finite Non-square Jacobian matrices
Elements
Non-square Jacobian matrices

One case of non-square Jacobians is observed when
mapping of 1D elements in two or three dimensions.
These Jacobians are useful to perform integrations while
working with reinforcement, joint elements and to cal-
culate edge traction forces.
If a 1D element is placed in the 3D space, the Jacobian
is given by:
h i
∂x ∂y ∂z
J(ξ) = ∂ξ ∂η ∂ζ
Three-node line element in 2D space
Using coordinates interpolation leads to:

h P i 𝑧
∂Ni P ∂Ni P ∂Ni
J (ξ) = ∂ξ xi ∂ξ yi ∂ξ zi = ∂N
∂ξ X
To relate lengths beetween global and natural coordi- 𝑦

nates, the Euclidean norm is used:
√ 𝑥
J (ξ) = J · J
dL
Thus, dL = J (ξ) dξ, or alternatively J (ξ) = dξ where
p
dL = dx2 + dy 2 . Three-node line element in 3D space
Raul
Durand 39
Elements
Bar element orientation

When studying 1D elements in 2D or 3D space the el-
ement orientation at a point is frequently needed. The
orientation can be defined by a unitary vector r as
shown in the figure. For a given coordinate ξ, it can
be expressed by:
h iT
∂x ∂y ∂z ∂x
r(ξ) = r(x, y, z) = ∂L ∂L ∂L
= ∂L 𝑧
where L is the curvilinear coordinate along the bar path.

Using chain rule, we have: 𝑦
∂x ∂ξ
r(ξ) = ∂ξ ∂L 𝑥
∂x
The term ∂ξ is a vector equal to JT . In turn, the term Bar element orientation in 3D
∂ξ
∂L represent the inverse of the Jacobian. Finally, the
vector r is written as:
JT
r(ξ) = JT |J|
1
= J
This results indicates that, in this case, JT is a vector

parallel to the orientation of a line element at point ξ.
Raul
Durand 40
Elements
Surface mapping
Surface elements can also generate non-square Jaco-
bians. These Jacobians are useful when working with 𝑧
joint elements and when integrating traction forces on
the faces of 3D elements.
For this case the Jacobian matrix is given by: 𝑦
" ∂x ∂y ∂z
# 𝑥
∂ξ ∂ξ ∂ξ
J= ∂x ∂y ∂z
∂η ∂η ∂η
Using coordinates interpolation we have:

P P ∂Ni 
∂Ni P ∂Ni
∂ξ xi ∂ξ yi ∂ξ zi
J(ξ, η) = P ∂N P ∂Ni P ∂Ni 
∂η
i
xi ∂η yi ∂η zi
∂N
= ∂(ξ,η) X
∂N
where ∂(ξ,η) is the matrix of derivatives of shape func-
tions evaluated at coordinates ξ and η and X is the Surface element
matrix of coordinates in global space.
Raul
Durand 41
Elements
For the case of surface elements, the Jacobian J relates

the surface differential area dA and the differential area 𝑧
dξdη in natural coordinates:
dA = Jdξdη
𝑦
To find dA two auxiliary vectors, a and b, are used, 𝑥
thus:
Surface area differential
dA = |a × b|
 dx 
This leads to the following expression for the Jacobian:  dξ dξ 
 
dy
a= dξ dξ 
v  dz

dξ dξ
u 2 2 2 
u J J12 J J13 J J11
J = t 11
12 13
+ +
 dx 
J21 J22 J22 J23 J23 J21  dη dη 

 
dy
b= dη dη 
where Jij are components of the Jacobian matrix J.  dz

dη dη

Using this non-square Jacobian it is possible to integrate Auxiliary vectors

functions in a region defined by a curved surface.
Raul
Durand 42
Finite Remarks
Elements
Remarks
Elements should be as regular as possible. It is recommended that aspect ratio, the ratio of
the longest edge to the shortest edge of an element, be at least 5:1. It is also recommended
that element skew be less than 60◦ .
Quadratic elements with straight faces should be preferred to elements with curved faces
when possible.
Abrupt changes in element sizes should be avoided.
Raul
Durand 43
Finite Remarks
Elements
Element size should be uniform whenever possible.
There should be a smooth transition between coarse to fine mesh.
Raul
Durand 44
Finite Exercises
Elements
Exercises
1. For a three-node bar element with coordinates (x1 , y1 ) = (1, 1), (x2 , y2 ) = (2, 1) and
(x3 , y3 ) = (3, 1), determine an expression for its orientation as a function of the local
coordinate ξ.
2. Using integration, calculate the length of the element in the last exercise.
3. Study the limitations for the location of intermediate nodes in a four-node bar element.
4. By inspection determine the Jacobian norm of the four-node element shown below.
Then, compare the result with the one obtained by using the Jacobian matrix.
5. Compute the Jacobian norm as function of ξ and η of the element in the last exercise
changing the coordinates of node 3 to x3 = (1, 1, 0). Later calculate the surface area
by integration.
Raul
Durand 45
Finite Numerical integration
Elements
Numerical integration
In numerical analysis a quadrature rule is an approximation of the definite integral function,
frequently stated as a weighted sum of function values at sampling points within the domain
of integration:
Z b n
X
f (x)dx = wi f (xi )
a i=1
where xi are the sampling points and wi are corresponding weights.

The basic idea of numerical integration is to approximate a given function f (x) by a poly-
nomial, then perform an analytical integration.
Two widely used methods to perform numerical integration are the Newton-Cotes and Gauss-
Legendre formulas.
In finite elements, numerical integration is used to perform integrals along the element do-
main. For example numerical integration is used to approximate lengths, areas and volumes
of elements. Also, it is used to find stiffness matrices, nodal boundary conditions, internal
forces, etc.
Raul
Durand 46
Finite Newton-Cotes formulae
Elements
Newton-Cotes formulae
Newton-Cotes formulae, are a group of rules for numerical integration based on the evaluation
of the integrand at equally spaced points.
There are open and closed Newton-Cotes formulae regarding if they evaluate the function at
endpoints or not. For a closed Newton-Cotes formulae with n sampling points an integral is
estimated as:
Z xn n
X
f (x)dx ≈ wi f (xi ) i = 1, . . . , n
x1 i=1
xn −x1
with xi = x1 + ih and h = n−1 .
The weights are derived from the integration of Lagrange basis polynomials Li (x). That is,
approximating the integrand using Lagrange interpolation as f (x) ≈ ni=1 f (xi )Li (x), we
P
have:
Z xn n
Z xn X
f (x)dx ≈ f (xi )Li (x)dx
x1 x1 i=1
n
X Z xn
≈ f (xi ) Li (x)dx
i=1 x1
| {z }
wi
Raul
Durand 47
Finite Newton-Cotes formulae
Elements
The two-point Newton-Cotes rule approximates the integral as:

Z xn
b−a
f (x)dx ≈ 2 (f (x1 ) + f (x2 ))
x1
≈ h2 f (x1 ) + h2 f (x2 )
and the 3-point Newton-Cotes rule is given by:

Z xn
b−a
f (x)dx ≈ 6 (f (x1 ) + 4f (x2 ) + f (x3 ))
x1
≈ h3 f (x1 ) + 4h
3 f (x2 ) + h3 f (x3 )
The weights for the first four Newton-Cotes rules are listed below.
wi
n degree rule h
1 1
1 1 Trapezoid 2, 2
1 4 1
2 3 Simpson’s 3, 3, 3
3 9 9 3
3 3 Simpson’s 3/8 8, 8, 8, 8
14 64 8 64 14
4 5 Boole’s 45 , 45 , 45 , 45 , 45
Raul
Durand 48
Finite Gauss-Legendre quadrature
Elements
Gauss-Legendre quadrature
The Gauss-Legendre quadrature is formulae to yield the exact integral for polynomials of
degree up to 2n − 1, where n is the quadrature order (number of sampling points), by a
suitable choice of points xi and weights wi .
Z +1 n
X
f (x)dx ≈ wi f (xi ) i = 1, . . . , n
−1 i=1
The Gauss-Legendre quadrature only produce accurate results if an arbitrary function f (x)
is well approximated by a polynomial.
Although this quadrature can be applied for any integration limits, the standard range
[−1, 1] is defined to stablish fixed sets of integration points (sampling points) and weights.
The required integration points xi for a quadrature order n are given by the roots of the
Legendre polynomials Pn (x).
For an integral over an arbitrary [a, b], the limits must be changed to the range [−1, 1] before
the use the Gaussian quadrature. This is achieved using the substitution:
b−a a+b
x= 2 t + 2
thus the integral becomes:

Z b Z +1
b−a b−a a+b
f (x)dx ≈ 2 f 2 t + 2 dt
a −1
Raul
Durand 49
Elements
Sampling points and weights

To exactly integrate a third degree polynomial, 2 integration points and 2 corresponding
weights are required. In this case the quadrature will be able to integrate accurately mono-
mials as 1, x, x2 and x3 . This gives the following systems of non-linear equations:
Z +1
w1 + w2 = 1 dx = 2
−1
Z +1
w1 x 1 + w2 x 2 = x dx = 0
−1
Z +1
w1 x21 + w2 x22 = x2 dx = 2
3
−1
Z +1
w1 x31 + w2 x32 = x3 dx = 0
−1
Solving the systems we get the following sampling points:
x1 = − √13 and x2 = + √13
with the corresponding weights:
w1 = 1 and w2 = 1
Raul
Durand 50
Elements
The first four sets of sampling points and weights for Gauss-Legendre quadrature on [−1, 1]
are:
n p ξi wi
1 1 0 2
p
2 3 ± 1/3 = ±0.577350269 1
3 5 0 8/9
p
± 3/5 = ±0.774596669 5/9
4 7 ±0.339981044 0.6521452
±0.861136312 0.3478548
Raul
Durand 51
Elements
Multiple integrals
Gauss rules for multiple integrals are called product rules. They are obtained by applying
one-dimensional rules to each independent variable. For example, a 2D integral becomes:
Z +1 Z +1 nξ nη
X X
f (ξ, η) ≈ wi wj f (ξi , ηj )
−1 −1 i=1 j=1
where nξ and nη are the number of Gauss points in the ξ and η directions, respectively.
All weights and point coordinates can be condensed into n = nξ nη integration points and
weights, thus:
Z +1 Z +1 n
X
f (ξ, η) ≈ wi f (ξi , ηi )
−1 −1 i=1
Raul
Durand 52
Elements
Quadrature for triangular elements

Integration points and weights to integrate triangle elements are:
n p i ξi ηi wi
1 1 1 1/3 1/3 1/2
3 2 1 1/6 1/6 1/6
2 2/3 1/6 1/6
3 1/6 2/3 1/6
4 3 1 1/3 1/3 −9/32
2 3/5 1/5 25/96
3 1/5 3/5 25/96
4 1/5 1/5 25/96
Raul
Durand 53
Elements
Quadrature for quadrilateral elements
n p i ξi ηi wi
1 1 1 0 0 4
√ √
4 3 1 −1/ 3 −1/ 3 1
√ √
2 1/ 3 −1/ 3 1
√ √
3 −1/ 3 1/ 3 1
√ √
4 1/ 3 1/ 3 1
p p
9 5 1 − 3/5 − 3/5 25/81
p
2 0 − 3/5 40/81
p p
3 3/5 − 3/5 25/81
p
4 − 3/5 0 40/81
5 0 0 64/81
p
6 3/5 0 40/81
p p
7 − 3/5 3/5 25/81
p
8 0 3/5 40/81
p p
9 3/5 3/5 25/81
Raul
Durand 54
Elements
Quadrature for tetrahedral elements
n degree i ξi ηi ζi wi
1 1 1 1/4 1/4 1/4 1/6
4 1 1 a b b 1/24
2 b a b 1/24
3 b b a 1/24
4 b b b 1/24
5 3 1 1/4 1/4 1/4 −2/15
2 1/2 1/6 1/6 3/40
3 1/6 1/2 1/6 3/40
4 1/6 1/6 1/2 3/40
5 1/6 1/6 1/6 3/40
√
a = (5 + 3 5)/20 = 0.5854101966249685
√
b = (5 − 5)/20 = 0.1381966011250105
Raul
Durand 55
Elements
Quadrature for hexahedral elements
n i ξi ηi ζi wi
1 1 0 0 0 8
√ √ √
8 1 −1/ 3 −1/ 3 −1/ 3 1
√ √ √
2 1/ 3 −1/ 3 −1/ 3 1
√ √ √
3 −1/ 3 1/ 3 −1/ 3 1
.. .. .. .. ..
. . . . .
√ √ √
8 1/ 3 1/ 3 1/ 3 1
p p p
27 1 − 3/5 − 3/5 − 3/5 125/243
p p
2 0 − 3/5 − 3/5 200/243
p p p
3 3/5 − 3/5 − 3/5 125/243
p p
4 − 3/5 0 − 3/5 200/243
.. .. .. ..
. . . .
p p p
27 3/5 3/5 3/5 125/243
Raul
Durand 56
Finite Exercises
Elements
Exercises
1. Evaluate the integral 01 √1+x
x
R
2
dx using the Trapezoid and Simpson’s rules with 4
subdivisions of the integration limits. Later use the Gauss quadrature rule with 4
integration points. Compare the results.
2. The nodal coordinates of a 3-node bar element are (x1 , y1 ) = (−1, −1), (x2 , y2 ) = (1, 1)
and (x3 , y3 ) = (0, 1). Determine the element length analytically; then use the Gauss
quadrature testing different numbers of integration points. How many integration
points are required to find the element length accurately?
3. The nodal coordinates of a 4-node quadrilateral element are (x1 , y1 ) = (1, 1), (x2 , y2 ) =
(5, 2), (x3 , y3 ) = (6, 5) and (x4 , y4 ) = (2, 4). Determine the element area using the
Gauss quadrature with 1, 4, and 9 integration points. Compare the results.
4. Compute the surface area shown below. How many integration points are required to
find the area accurately?
(0,0,1) (0,1,1)
(1,0,1)
(1,1,0)
Raul
Durand 57
Finite Integration of boundary conditions
Elements
Integration of boundary conditions

In finite elements, in general, there are two types of boundary conditions, essential and
natural. When related with the underlining differential equation, they are known as as
Dirichlet and Neumann boundary conditions, respectively.
In an statistic or dynamic analysis essential boundary conditions are given by prescribed
displacements, as in supports, for example. In turn, natural boundary conditions are given
by external forces: concentrated forces and surface forces. Also, body forces can conveniently
be considered as natural boundary conditions.
Concentrated forces are applied directly at nodes. For this reason, the mesh should be
compatible with the locations of concentrated forces. For surface tractions and body forces,
equivalent nodal forces need to be calculated and applied at nodes.
For other problems, other than mechanical analyses, boundary conditions are applied in a
similar way. For example, in a heat transfer problem, prescribed temperatures are applied
directly at nodes while heat flux along a surface is interpreted as a set of nodal “heat fluxes”.
Raul
Durand 58
Finite Body forces
Elements
Body forces
Body forces represent the forces due self-weight and act over the element volume.
For a given element with n nodes in 2D or 3D space, the

vector containing the nodal forces equivalent to the body
forces is defined as:
h iT
Fb = F1x F1y F1z F2x . . . Fny Fnz
This vector is calculated by:

Z
Fb = N̄T b dV
V
where b = [bx by bz ]T is a vector with the material unit-

weight at each direction and
h i
N̄ = IN1 IN2 IN3 . . . INn
D×Dn
is a matrix containing the element shape functions. I is the

D × D identity matrix where D is the space dimension.
Raul
Durand 59
Finite Body forces
Elements
Using numerical integration, the vector of nodal forces can be approximated by:
m
X
Fb ≈ N̄T bJ w
i=1
where m is the number of integration points of the element and J the Jacobian and w is the
integration point weight. N̄, J and w must be evaluated at each integration point.
Raul
Durand 60
Finite Body forces
Elements
In most finite element analysis, body forces act only in

the vertical direction. For example, in a 3D analysis
self-weight acts in the z direction, thus the vector of
nodal forces can be defined as:
h iT
Fz = F1z F2z F3z . . . Fnz
and calculated by:

Z m
X
Fz = NT bz dV ≈ NT bz J w
V i=1
where N is the row vector:
N = [N1 N2 N3 . . . Nn ]
Each component of force can be simply calculated by:

Z m
X
Fzi = Ni bz dV ≈ Ni bz Jw
V i=1
Raul
Durand 61
Finite Body forces
Elements
The finite element formulation for the vector with equiv-

alent nodal forces can be derived by the use of the Vir-
tual Work Principle.
Equating the work done by body forces Wb and the work
done by the equivalent nodal forces WF , we have:
Z
δu · b dV = δU · Fb
V
where δu is a virtual displacement vector of a point

inside the volume and δU is a vector with virtual nodal
displacements.
Since the displacements if a point inside the element
can be estimated by interpolation of nodal values, δu =
N̄δU, then the equation can be rewritten as:
Z
T
δU Fb = δUT N̄T b dV
V
Later, taking out from the equation the virtual displace-

ments vector we arrive to:
Z
Fb = N̄T b dV
V
Raul
Durand 62
Finite Surface tractions
Elements
Surface tractions
Surface tractions are applied on elements faces and equiv-
alent forces are calculated at the faces nodes. For a vector
field t = [tx ty tz ]T of traction forces applied at the face of
an element, the vector of equivalent nodal forces is defined
as:
h iT
Ft = F1x F1y F1z F2x . . . Fny Fnz
This vector is calculated by the area integral:

Z
Ft = N̄T t dA
A
where A is the face area and N̄ is given by:
N̄ = [IN1 IN2 IN3 . . . INn ]
Using numerical integration we have:

m
X T
Ft ≈ N̄ tJw
i=1
where J is the element face Jacobian.

Raul
Durand 63
Finite Surface tractions
Elements
Similarly, the equation for nodal forces for surface trac-

tions can be derived from the Virtual Work Principle,
thus:
Z
δu · t dV = δU · Ft
A
Substituting vector δu in terms of virtual nodal dis-

placements δU, the equation above becomes:
Z
δUT Ft = δUT N̄T t dV
A
and finally:
Z
Ft = N̄T t dV
A
Raul
Durand 64
Finite Normal surface traction
Elements
Normal surface traction

Eventually, there is interest in applying a normal trac-
tion tn over a surface. In this case the traction vector
field t is always parallel to the surface normal vector n̂:
t = tn n̂
When an element has curved faces the vector n̂ is vari-

able; thus, while using numerical integration, a normal
vector should be calculated for each integration point.
Fortunately, the Jacobian matrix from a face or edge
of an element carries useful geometric information to
determine n̂. For example, in 2D, the Jacobian matrix
for an element edge represent a vector parallel the edge
itself, then the edge normal vector is calculated by:
n = [J2 − J1 ]
In 3D, the rows from the face Jacobian matrix represent

tangent vectors to the surface. Using the cross product,
the normal vector is calculated by:
n = [J11 J12 J13 ] × [J21 J22 J23 ]

Raul
Durand 65
Finite Normal surface traction
Elements
If the element faces are numbered using the right-hand convention, vector n̂ should point
outwards the elements. Therefore, tn should be a positive number if the normal surface
traction pulls the face outwards the element and negative otherwise.
Finally, the equivalent nodal forces are calculated by:
Z
Fn = N̄T tn n̂ dA
A
m
X
≈ tn N̄T n̂Jw
i=1
Raul
Durand 66
Finite Exercises
Elements
Exercises
1. The nodal coordinates of a 3-node bar element are (x1 , y1 ) = (−1, −1), (x2 , y2 ) = (1, 1)
and (x3 , y3 ) = (0, 1). If the material unit weight is γ = 70 kN/m3 and the section area
is A = 0.02 m2 , calculate the equivalent nodal forces. Use numerical integration.
2. For the elements below, determine nodal forces equivalent to body forces. Consider
the element volume as V and the unit weight as γ.
3. Determine the equivalent nodal forces to the distributed load q.
Raul
Durand 67
Finite Exercises
Elements
4. For the element below, determine the corresponding nodal forces if a normal traction
tn = 1 kN/m is applied at the face specified in the figure.
5. For the element below, determine the corresponding nodal forces if a normal traction
tn = 1 kPa is applied in the shaded face.
𝑧
(0,0,2)
(2,0,2) (0,2,1)
𝑦
(2,2,1)
𝑥 (2,2,0)
(2,0,0)
6. Find out what are the boundary conditions analogous to displacements, concentrated
forces and surface forces in the seepage problem.
Raul
Durand 68
Finite Suggested bibliography
Elements
Suggested bibliography
1. K. J. Bathe, Finite Element Procedures, Pearson, 2006.
2. T.J.R. Hughes, The Finite Element Method, Dover, 2000.
3. J.N. Reddy, An Introduction to the Finite Element Method, Third edition, McGraw
Hill, 2005.
4. C.A. Felippa, Introduction to Finite Element Methods, Lecture notes, Colorado, 2001.
Beauty is the first test:

there is no permanent place in the world
for ugly mathematics.
G. H. Hardy
Raul
Durand 69

FEM - 6 Isoparametric Elements

Hochgeladen von

Dokumentinformationen

Originalbeschreibung:

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

FEM - 6 Isoparametric Elements

Hochgeladen von

Copyright:

Verfügbare Formate

Finite Element Course

Prof. Raul Durand

Local and global coordinate systems

where Ni (x) is the shape function of node i.

Similarly, global coordinates can also be interpolated as:

x = N1 (ξ)x1 + N2 (ξ)x2 + N3 (ξ)x3

If the same element is in 3D space, the global coordinates are

or xT = NX(e) , where x is a vector with global coordinates and

Nodal displacement interpolation:

Shape functions for 1D elements

Shape function should return 1 if evaluated at the correspond-

which leads to the following systems: 1.0

For a three-node 1D element, also called quadratic ele-

To find a, b and c coefficients three systems of equations

As can be seen, all systems have the same coefficients

So, they can be condensed into the following matrix

Finally, the three shape functions are:

1D Lagrangian shape functions

Linear triangular element

Conventionally, 2D elements have the nodes num-

Substituting all local coordinates points in Ni the

Since the right-hand-side of the equation is equal to the

where A contains the systems coefficients and the 1.0

N1 = 1 − ξ − η Shape function of node 1

Shape functions requirements

Quadratic triangular element

Bilinear quadrilateral element

Biquadratic quadrilateral element

This element contains 9 nodes and its numbering is de- 1.0

Serendipity quadrilateral element

It is one of the most used elements in 2D domains with 1.0

Linear tetrahedron element

The shape functions are given by:

Trilinear hexahedron element

It is usually referred to as brick element and is extensively used in 3D analysis.

Ni (ξ, η, ζ) = 18 (1 + ξi ξ)(1 + ηi η)(1 + ζi ζ)

The 20-node serendipity hexahedron:

A more general expression for n functions that depend on m variables is:

The simplest use of a Jacobian is for the integration of

which can be solved by using the substitution u = 2x+1.

Now consider a multiple integral:

which can be rewritten as: (1/2, 1)

The determinant is J = 4 and it can be proved that

and lead to the following relation between differentials:

A differential area dA equivalent to dxdy is shown in dx

following area ratio: Auxiliary vectors

A volume differential dV equivalent to dxdydz 𝑥

The volume ratio between both systems is: Auxiliary vectors

Inverse of Jacobian matrix

Jacobian in finite elements

and for 3D:

Change of variables for integration

Computing the Jacobian matrix

For a 3D element, the Jacobian matrix can be calculated by:

and from Cartesian coordinates to spherical coordinates:

x = ρ sin φ cos θ, y = ρ sin φ sin θ, z = ρ cos φ

Two-node line elements and three-node triangle ele-

For 2D elements with more than three nodes distortions

Restriction on mid-side nodes

Figures taken from “FEM Convergence Requirements”. C.A. Felippa.

Figures below show distortions in a quadratic triangle element.

Moving mid-side nodes outwards:

Figures taken from “FEM Convergence Requirements”. C.A. Felippa.