Computational Engineering ENG435

Introduction to Finite Element Analysis and Procedure

Eur Ing Dr Ali Tolooiyan
BSc(Eng), MSc, PhD, FEANI, CEng MIEI, CPEng, FIEAust, NER, APEC Engineer, IntPE(Aus)
[email protected]

Finite Element Method

The Finite Element Method (FEM) is widely used for analysis of

structural engineering problems.

It is frequently used in civil, mechanical, aerospace ocean, mining,

biomechanical, nuclear,… engineering.

1
 In general we use FEM in

• Linear analysis
• Nonlinear analysis
• Static analysis
• Dynamic analysis

FEM is applied in various computer programs

• Ansys
• Abaqus
• Plaxis
• Comsol
• …

2
 The FEM is …

The Finite Element Method (FEM) is a numerical technique for

finding approximate solutions to Partial Differential Equations
(PDE) and their systems.

In simple terms, FEM is a method for dividing up a very

complicated problem into small elements that can be solved in
relation to each other.

Basic concept

Objective: What is the area of this

geometry?

3
 Basic concept

C Objective: What is the area of this

geometry?

Problem: Geometry is complicated

A B D
Solution:
‐ Divide the geometry to A, B, …, F
F
E ‐ Calculate the area of A, B, …, F
‐ Assemble the calculated areas

FEM in Civil and Mechanical Engineering

• To simulate stress distribution in structures

• To predict failure of structures
• To predict the displacement and judge the allowable
displacement
• To define the stable and optimum design
• and etc…

4
 Objectives in this lecture

To introduce to you FEM for the linear analysis of solids (by linear I mean
small displacements according to Hooke’s law.

We will consider:
• The formulation of the finite element equilibrium equations.
• The calculation of finite element matrices
• Methods for solution of the governing equations
• Computer implementations using a FEM software

Remarks

In this lecture the emphasis is given to physical explanation rather than

mathematical derivations.

The techniques we discuss in this lecture are employed in several

commercial FEM programs such as Plaxis, Ansys…

All the FEM lectures in this unit represent a very brief and compact
introduction to the field of FEM in civil and mechanical engineering.
Additional recourses should be used if you want to use FEM in your
future engineering career, as a professional FEM user.

10

5
 Constitutive Models

When designing a racing car, we would prefer rubber for tires and steel for the
frame, but why? What is the advantage of using one rather the other?

11

Why Study Materials?

 Selecting materials is part of the design process.

 The choice of materials is driven by several factors such as

Cost
Manufacturability
Performance
Availability

 A design engineer must consider all the above in picking the correct
material.

12

6
 Why Study Materials?

• Performance may include

1.Mass‐to‐weight ratio
2.Thermal and electrical conductivity
3.Mechanical strength and behaviour
4.Corrosion and wear resistance

• All the above depend on the material's behaviour.

13

Material Behaviour

• For instance, rubber offers smaller resistance to deformation

compared to structural steel.
• So, rubber must stretch longer compared to structural steel before it
can support the same dead weight.
• Quantifying this behaviour in a mathematical equation is the essence
of a material model.

14

7
 Choosing Material

Material criteria: Material criteria:

• Support the whole structure of the • Deform and provide traction
car • Absorb small impacts and reduce the
• Protect the car during a crash bumpiness
• Material chosen: • Material chosen:
‐ Structural steel ✓ ‐ Structural steel
‐ Rubber ‐ Rubber ✓

15

Need for a Model

• Therefore we need a constitutive model!

• It’ll help us both characterise and quantify material properties.

• Such a model helps engineers make informed decision on choosing

materials.

16

8
 Constitutive Model

• A constitutive model is a mathematical relation between stress and

strain.

• The coefficients of the function are called material properties.

• Two different materials may have the same function form, but they’ll
have different material properties.

17

Constitutive Model

• The linear elastic model is the simplest form of constitutive model.

• It’s used to represent most metals.

• It may be limiting but it is very useful for most common engineering

applications.
• It is governed by Hooke’s law and it will be discussed in later sections.

18

9
 Constitutive Model

• There are many classes of nonlinear constitutive models available to

model more complex behaviour.

• The choice of the material model should also depend on the physical
phenomenon that we are trying to capture.

19

Elastic Behaviour
• Elastic deformation no permanent changes
are developed in the microstructure.
• All work done on the system is stored as
internal energy and is dissipated as heat
energy upon unloading.
• Upon unloading, the structure returns to its
original undeformed state.
• The choice of the material model should
also depend on the physical phenomenon
that we are trying to capture.

20

10
 Hooke’s Law

• Hooke’s Law simplest form of

elastic deformation.

• Most metals undergo elastic

deformation up to a limit in strain the
elastic limit.

• When metals are loaded beyond

this point, they sustain permanent
deformation due to changes made to
their microstructure.

21

Hooke’s Law

• So, energy is not conserved beyond

this point.

• Loading and unloading paths in the

curve are not the same.

• But the unloading curve is parallel

to the loading curve.

• Hooke’s law is applicable only up to

the elastic limit.

22

11
 Hooke’s Law
• Hooke’s law states that within the
elastic limit, the stress developed in
the material is linearly proportional to
its strain.

• The constant of proportionality is the

elastic modulus also known as Young’s
modulus.

• In this case, the linear equation form

is called the constitutive model and
the Young’s modulus is the material
constant.

• Hooke’s law is applicable to 3D cases

as well.

23

Poisson’s Ratio
• Poisson’s effect: when a material is
stretched/compressed in one
direction, it deforms in the opposite
direction in other perpendicular
directions.

• Poisson’s ratio is defined as

• The negative sign ensures that the

value of Poisson’s ratio is positive.
• Typically, the value ranges from 0 to
0.5.

24

12
 Hooke’s Law 3D ‐ 1 stress state

25

Hooke’s Law 3D ‐ 2 stress state

same

26

13
 Hooke’s Law 3D ‐ 3 stress state

27

Hooke’s Law 3D ‐ 3 stress state

28

14
 Engineering Shear Strains

29

Hooke’s Law 3D

30

15
 Strain Vector in Tensor Form

31

Strain Vector in Tensor Form

32

16
 Anisotropic Materials

33

Orthotropic Materials

34

17
 FEM Solution Process

Physical problem

Establish finite element model of the

problem

Solve the model Revise the model ?

Interpret the results

35

Element Assemblage

36

18
 Element Assemblage – Various geometries

37

Element Assemblage – Crash Barrier

38

19
 Element Assemblage – Liquefaction

39

Steps Involved in Analysis of a FE System

• System idealisation into elements

• Evaluation of element equilibrium requirements
• Element assemblage
• Solution of response

40

20
 Physical Layout of Element System

Steps in analysis of this System:

u1, R1 u3, R3
o Equilibrium requirements of each
k4
spring
u2, R2
o Interconnection requirements
k3
k1 between springs (constitutive
k5
relations)
k2
o Compatibility requirement
between springs 

41

Element Equilibrium Requirements

u1
k1
F1(1) 𝐹 𝑘 𝑢

u1 u2
k2 𝐹
1 1 𝑢
𝑘 𝑢 =
F2(2) 1 1 𝐹
F1(2)

42

21
 Element Equilibrium Requirements

u1 u2
k3 𝐹
1 1 𝑢
𝑘 𝑢 =
F2(3) 1 1 𝐹
F1(3)

u1 u3
k4 𝐹
1 1 𝑢
𝑘 =
F3(4) 1 1 𝑢 𝐹
F1(4)

43

Element Equilibrium Requirements

u2 u3
k5 𝐹
1 1 𝑢
𝑘 =
F3(5) 1 1 𝑢 𝐹
F2(5)

44

22
 Element Interconnection Requirements

u1, R1 u3, R3
k4

u2, R2 𝐹 𝐹 𝐹 𝐹 𝑅

k3
k1 𝐹 𝐹 𝐹 𝑅
k5
k2
𝐹 𝐹 𝑅

45

Element Interconnection Requirements

3×3 3×1
𝐹 𝐹 𝐹 𝐹 𝑅 Matrix Vector

𝐹 𝐹 𝐹 𝑅 𝑲𝑼 𝑹

𝐹 𝐹 𝑅 3×1
Vector

46

23
 Element Interconnection Requirements

𝐾𝑈 𝑅
𝑘 𝑘 𝑘 𝑘 𝑘 𝑘 𝑘

𝑢
𝑈= 𝑢 𝐾= 𝑘 𝑘 𝑘 𝑘 𝑘 𝑘
𝑢

𝑘 𝑘 𝑘 𝑘
𝑅
𝑅= 𝑅
𝑅

47

Element Interconnection Requirements

where :

𝑘 0 0
we know :
𝐾 = 0 0 0
0 0 0
𝐾 𝐾
𝑘 𝑘 0

𝐾 = 𝑘 𝑘 0

0 0 0

similarly we can obtain 𝐾 and so on …

This process is called the Direct Stiffness Method

48

24
 Element Interconnection Requirements

Direct Stiffness Method

An extremely important concept of FEM, which is very well

implemented in FEM codes and represent the basis of the
implementation of the finite element method.

It is not difficult but needs well physical and mathematical

explanation to be understood …

49

Direct Stiffness Method

𝐾=
u1 u3

0 0 0
u2

0 0 0

0 0 0

50

25
 Direct Stiffness Method

𝐾=
u1 u3

𝑘 0 0
u2

k1
0 0 0

0 0 0

51

Direct Stiffness Method

𝐾=
u1 u3

𝑘 𝑘 𝑘 0
u2

k1
𝑘 𝑘 0

k2
0 0 0

52

26
 Direct Stiffness Method

𝐾=
u1 u3

𝑘 𝑘 𝑘 𝑘 𝑘 0
u2

k3 𝑘 𝑘 𝑘 𝑘
k1 0

k2
0 0 0

53

Direct Stiffness Method

𝐾=
u1 u3
k4
𝑘 𝑘 𝑘 𝑘 𝑘 𝑘 𝑘
u2

k3 𝑘 𝑘 𝑘 𝑘
k1 0

k2
𝑘 0 𝑘

54

27
 Direct Stiffness Method

𝐾=
u1 u3
k4
𝑘 𝑘 𝑘 𝑘 𝑘 𝑘 𝑘
u2

k3 𝑘 𝑘 𝑘 𝑘 𝑘 𝑘
k1
k5
k2
𝑘 𝑘 𝑘 𝑘

55

Elements, Nodes and Stress Points

Element

Node

Stress
point

56

28
 Displacement Interpolation

3 𝐴 𝐴 𝐴
y ξ ξ ξ
𝐴 𝐴 𝐴

where ξ is area coordinate, and

A2 P A1
ξ ξ ξ 1
A3
2
1

57

Displacement Interpolation

3
y

ξ =1 ξ =0 ξ =0
ξ =0 ξ =1 ξ =0
ξ =0 ξ =0 ξ =1

2
1

58

29
 Displacement Interpolation

3
y

ξ =1/2 ξ =0 ξ =1/2
5 ξ =1/2 ξ =1/2 ξ =0
6 ξ =0 ξ =1/2 ξ =1/2

2
1 4
x

59

Displacement Interpolation

𝑁 𝑎 ξ ξ ξ

Where 𝑁 is shape function for node 𝑖 and 𝑞, 𝑟, and 𝑠 are nonnegative

integers that range over the 𝑛 possible combinations for which 𝑞 𝑟
𝑠 𝑝. Thus 𝑁 is a complete polynomial of degree 𝑝 in Cartesian
coordinates

For 6 nodes quadratic triangle element; 𝑛 6 and 𝑝 2, so:

𝑁 𝑎 ξ 𝑎 ξ 𝑎 ξ 𝑎 ξ ξ 𝑎 ξ ξ 𝑎 ξ ξ

60

30
 Displacement Interpolation

𝑁 ξ (2 ξ ‐ 1)
3
𝑁 ξ (2 ξ ‐ 1)

𝑁 ξ (2 ξ ‐ 1)
5
6
𝑁 4ξ ξ

2 𝑁 4ξ ξ
1 4
𝑁 4ξ ξ

61

Displacement Interpolation
3
y v5y
u x ( x, y )  N1v1x  N 2 v2 x  ...  N 6 v6 x
6 5 v5x
u y ( x, y )  N1v1 y  N 2 v2 y  ...  N 6 v6 y

2
1 4  v1x 
uNv
e v 
x  1y 
 v2 x 
 
v2 y
v  
e
 ... 
u x ( x , y )   N1x 0 N2x 0 ... ... N6x 0   
u  N
u y ( x, y )  0 N1 y 0 N2 y ... ... 0 N 6 y   ... 
v 
 6x 
v6 y 

62

31
 Strain Interpolation

3
y v5y

6 5 v5x

du x dN1 dN dN 2
 xx   v1x  2 v2 x  ...  6 v6 x 1 4
dx dx dx dx
x
du y
dN dN dN
 yy   1 v1 y  2 v2 y  ...  6 v6 y
dy dy dy dy

du x du y dN1 dN dN dN
 xy    v1x  1 v1 y  2 v2 x  ...  6 v6 y
dy dx dy dx dy dx

63

Strain Interpolation

 xx ( x, y ) 
 Bv e  
   yy ( x, y )  v1x 
 xy ( x, y ) v 
   1y 
B : Strain interpolation matrix
 v2 x 
 
v2 y
v  
e
 N1 N 2 N 6   ... 
0 0 ... ... 0 
 x x x  
 
N1 N 2 N 6   ... 
B 0 0 ... ... 0
 y y y  v 
 N
 1
N1 N 2 N 2 N 6 N 6 

 6x 
... ...
 y x y x y x  v6 y 

64

32
 Nodal Forces

 f1x 
f 
3  1y 
y f5y  f2x 
 
f2 y
f  
e
5 f5x
6  ... 
 
2  ... 
1 4 f 
x  6x 
 f 6 y 

65

Element Stiffness Matrix

Ke
 B T D Β dv

B : Strain interpolation matrix

D : Material stiffness matrix (here is according
to Hooke’s law)
1   0 
E  
M  D 1  0 
(1  2 )(1  ) 
 0 0 1
2
 

66

33
 Element Stiffness Matrix

 K1x1x K1x1 y K1 x 2 x K1 x 2 y ... ... K1 x 6 x K1 x 6 y 

K K1 y1 y K1 y 2 x K1 y 2 y ... ... K1 y 6 x K1 y 6 y 
 1 y1x
 K 2 x1x K 2 x1 y K2x2x K2x2 y ... ... K 2 x6 x K 2 x6 y 
 
K 2 y1x K 2 y1 y K2 y2x K2 y2 y ... ... K2 y6x K 2 y6 y 
K 
e
 ... ... ... ... ... ... ... ... 
 
 ... ... ... ... ... ... ... ... 
K K 6 x1 y K6x2x K6x2 y ... ... K6 x6 x K 6 x6 y 
 6 x1x 
 K 6 y1x K 6 y1 y K6 y2x K6 y2 y ... ... K6 y6x K 6 y 6 y 

67

Lets Solve Strain‐Stress System

Strain interpolation matrix

Interpolated Strain   B ve Nodal displacements

Element stress  M

Element stiffness matrix

68

34
 Errors in FEM

What type of errors can happen in a FEM solution?

Source of Error 1: Number of Element

It can also be explained as and error introduced due to the
discretisation of geometry

i.e. Higher number of elements (finer discretisation) lead to less error.

How accurately we capture the geometry of a problem?

Proper size and type of element.

69

Errors in FEM

Source of Error 2: Choice of interpolation functions

i.e. higher order of interpolation function leads to less error
How accurately we capture the variation of the employed function?
Polynomial order

Source of Error 3: Numerical error

i.e. accuracy of operation in numerical analysis of finite element
procedure.

70

35
 Errors in FEM

k1 k2

A beam made of multiple materials

Element level equations => Assembly => Boundary conditions =>

Solving algebraic equations => Solution

71

Errors in FEM

Accuracy (absolute) = Exact solution – FE solution

Accuracy (%) = (Exact solution – FE solution) / Exact solution × 100

72

36
 Errors in FEM

k1 k2

Exact solution
Convergence
Deflection

Number of element

73

Boundary Conditions (BC)

74

37
 Boundary Conditions (BC)

Force / Stress FEM Body Deformation / Strain

Deformation / Strain FEM Body Force / Stress

How about a combination of both?

75

Boundary Conditions (BC)

Force

Force Boundary
conditions.
Kinematic or
Static or Dynamic
geometric boundary
conditions
Disp.

Linear
Non‐linear
Time‐dependant
….

76

38
 Boundary Conditions (BC)

Symmetry Constraints
• Needed to stabilise the model
• Can be used to reduce the model size
• You should not use symmetry constraints for non‐symmetrical
problem (unless you are sure that it will not affect the results)

77

Boundary Conditions (BC)

Symmetry Constraints
Two common types of nodal symmetry constraint:

78

39
 Boundary Conditions (BC)

Force
Under Constrained Model
Over Constrained Model

Better Constrained Model

79

Contact Mechanics

80

40
 Types of Contact

81

Bonded Contact

82

41
 Frictionless Contact

83

Frictional Contact

84

42
 Frictional Force Balancing in Contact

85

Frictional Force Balancing in Contact

86

43
 Frictional Force Balancing in Contact

87

Frictional Force Balancing in Contact

88

44
 Frictional Force Balancing in Contact

89

Calculation of Contact Forces

90

45
 Nonlinearities

In most real life and industrial applications, nonlinearities

are too important to be neglected.

91

Nonlinearities

There are three factors that introduce nonlinearities into mechanical

systems:
• Material nonlinearity.
• Boundary nonlinearity.
• Geometric nonlinearity.

92

46
 Material nonlinearity

This type of nonlinearity is probably the one that you are most familiar with.
Most metals have a fairly linear stress/strain relationship at low strain values;
but at higher strains the material yields, at which point the response becomes
nonlinear and irreversible.
Material nonlinearity may be
related to factors other than strain.
Strain‐rate‐dependent material data
and material failure are both forms
of material nonlinearity. Material
properties can also be a function of
temperature and other predefined
Stress-strain curve for an elastic-plastic
material under uniaxial tension fields.

93

Boundary nonlinearity

Boundary nonlinearity occurs if the boundary conditions change during the analysis.
Consider the cantilever beam, that deflects under an applied load until it hits a “stop.”

The vertical deflection of the tip is linearly related to the load (if the deflection is
small) until it contacts the stop. There is then a sudden change in the boundary
condition at the tip of the beam, preventing any further vertical deflection, and so the
response of the beam is no longer linear. Boundary nonlinearities are extremely
discontinuous: when contact occurs during a simulation, there is a large and
instantaneous change in the response of the structure.

94

47
 Geometric nonlinearity

Geometric nonlinearity occurs whenever the magnitude of the displacements affects

the response of the structure. This may be caused by:
• Large deflections or rotations.
• Snap through.
• Initial stresses or load stiffening.

For example, consider a cantilever beam loaded vertically at the tip

95

Geometric nonlinearity

If the tip deflection is small, the analysis can be considered as being approximately
linear. However, if the tip deflections are large, the shape of the structure and, hence,
its stiffness changes.
In addition, if the load does not remain perpendicular to the beam, the action of the
load on the structure changes significantly. As the cantilever beam deflects, the load
can be resolved into a component perpendicular to the beam and a component acting
along the length of the beam.
Both of these effects contribute to the nonlinear response of the cantilever beam.

96

48
 Application of Nonlinearities

CAE University
97

Calibration and Validation

Calibration of an FE model is usually undertaken to ensure that specific features which

have been modelled provide a realistic estimate of the model stiffness or other
behaviour.

‐ Material information provided in data sheets

‐ Collecting additional experimental data
‐ FE analysis of simple loading scenarios (including curve fitting)
‐ Often an iterative process

98

49
 Calibration and Validation

99

Calibration and Validation

Tolooiyan and Gavin (2011)

100

50
 Calibration and Validation

Tolooiyan and Gavin (2011)

101

Calibration and Validation

Validation of an FE model is ensuring that the model as a whole predicts measured

behaviour properly.

102

51