Documente Academic
Documente Profesional
Documente Cultură
June 2019
Course Contents
Background to Finite Element Analysis
Introduction to Finite Element Analysis
Foundational Theory of Elasticity
Variational Calculus
The Basic Raleigh-Ritz Method
The Raleigh-Ritz Method Applied to Structural Analysis
1D Finite Element Analysis
2D Finite Element Analysis
FEA Using Abaqus
M6FEA_9 DM Madyira
Course Objectives
At the end of this module you should be able to:
Formulate the basic finite element governing equations for
general 1D & 2D stress analysis
Formulate the governing equations for 3D cases
Use variational calculus to formulate 1D, 2D and 3D element
equations
Use the matrix method to analyse basic 2D and 3D structural
problems
Develop a numerical model in Abaqus to solve basic structural
analysis problems
M6FEA_9 DM Madyira
Course Prerequisites
The key pre-requisites for this course are:
M6FEA_9 DM Madyira
Course Assessment
Exercises Date Venue
1 TBA TBA
M6FEA_9 DM Madyira
Background to Finite Element Analysis
M6FEA_9 DM Madyira
Background to Finite Element Analysis
• In 1909 Ritz developed an effective method for the
approximate solution of problems in the mechanics of
deformable solids based on Rayleigh’s formulations
M6FEA_9 DM Madyira 7
Background to Finite Element Analysis
• In 1943, Courant considerably increased possibilities of the
Ritz method by introduction of the special linear functions
defined over triangular regions and applied the method for
the solution of torsion problems.
M6FEA_9 DM Madyira 8
Background to Finite Element Analysis
• The Finite Elements Method (FEM) is a powerful numerical analysis tool used to
solve differential equations that describe most physical problems
• FEM originated from the need to solve complex elasticity and structural analysis
problems where no analytical solution could be used
• Its roots are in aircraft and aerospace structural analysis as an extension of the
force and displacement methods
• The term “finite element” was coined by Clough in the paper “The finite element
method in plane stress analysis”, 1960
• The first book on FEM, which can be considered a textbook, was published in
1967 by Zienkiewicz and Cheung and called “The finite element method in
structural and continuum mechanics”
M6FEA_9 DM Madyira 9
Background to Finite Element Analysis
• Thus, FEM represents the convergence of Matrix
Structural Analysis (MSA), Variational Approximation
Theory (VAT) and the convenience of powerful digital
computers
► Transient problems
M6FEA_9 DM Madyira 11
Background to Finite Element Analysis
• In early 1960s, the MacNeal-Schwendle Corporation (MSC)
developed the first general FEA code which was limited to
68000 degrees of freedom.
• Solution
→computes the unknown values of the primary field variable(s)
→computed values are then used by back substitution to compute additional, derived variables, such as reaction forces, element
stresses, and heat flow
• Post-processing
→Postprocessor software contains sophisticated routines used for sorting, printing, and plotting selected results from a finite
element solution
M6FEA_9 DM Madyira 13
Background to Finite Element Analysis
M6FEA_9 DM Madyira 14
Background to Finite Element Analysis: Direct Heated Rotary Kiln
M6FEA_9 DM Madyira 15
Background to Finite Element Analysis: Pipework T-junction
M6FEA_9 DM Madyira 16
Background to Finite Element Analysis: Pipework T-junction Reinforcement
M6FEA_9 DM Madyira 17
Introduction to Finite Element Analysis
M6FEA_9
Introduction to Finite Element Analysis
Common Methods for Solving General Field Problems
M6FEA_9 DM Madyira 19
Introduction to Finite Element Analysis
• Therefore, a gap exists between what is needed in applications and what can be
solved by analytical closed-form methods
M6FEA_9 DM Madyira 21
Introduction to Finite Element Analysis
M6FEA_9 DM Madyira 22
Introduction to Finite Element Analysis
M6FEA_9 DM Madyira 23
Introduction to Finite Element Analysis
M6FEA_9 DM Madyira 24
Introduction to Finite Element Analysis
• The fundamental concept involves dividing the body under study into a finite
number of pieces (subdomains) called elements
M6FEA_9 DM Madyira 25
Introduction to Finite Element Analysis
• Because element size, shape and approximating scheme can be varied to suit
the problem, the method can accurately simulate solutions to problems of
complex geometry and loading and thus this technique has become a very
useful and practical tool
M6FEA_9 DM Madyira 26
Introduction to Finite Element Analysis
Example of Discretized Domain
M6FEA_9 DM Madyira 27
Introduction to Finite Element Analysis
Example of Discretized Domain
M6FEA_9 DM Madyira 28
Introduction to Finite Element Analysis
M6FEA_9 DM Madyira 29
Introductory Theory of Elasticity
M6FEA_9
Introductory Theory of Elasticity
M6FEA_9 DM Madyira 31
Introductory Theory of Elasticity
Strength of Materials generally aims for approximate solutions which are easy
to use in general applications. Such solutions are obtained by invoking
simplifying assumptions.
M6FEA_9 DM Madyira 32
Introductory Theory of Elasticity
M6FEA_9 DM Madyira 33
Introductory Theory of Elasticity
What is Solid Mechanics?
Solid Mechanics is a more general term encompassing both strength of
materials and theory of elasticity. Areas of application include:
1. Geomechanics - modeling of the shape of planets; tectonics; and
earthquake prediction
2. Civil engineering - designing foundations or structures
3. Mechanical engineering - designing load bearing components for vehicles
power generation and transmission;
4. Manufacturing engineering - designing metal and polymer forming
processes, machining etc
5. Biomechanics - designing implants; bone mechanics
6. Materials Science - designing composites; alloy microstructures, thin films,
and developing materials processing methods
7. Microelectronics - designing failure resistant components
8. Nanotechnology - stress driven self-assembly on surfaces
M6FEA_9 DM Madyira 34
Introductory Theory of Elasticity
Defining a problem in Solid Mechanics
M6FEA_9 DM Madyira 35
Introductory Theory of Elasticity
What can be computed using Solid Mechanics?
M6FEA_9 DM Madyira 37
Introductory Theory of Elasticity
M6FEA_9 DM Madyira 38
Introductory Theory of Elasticity
M6FEA_9 DM Madyira 39
Introductory Theory of Elasticity
Equilibrium
F2
• At any point in time, the forces acting on a F3
deformable body must obey Newton's laws of F1
motion
• Resultant force acting on the body should be Deformable
zero unless the body is undergoing Body
acceleration
• Resultant moment about any point should be
zero unless the body is undergoing angular R1
acceleration R2
M6FEA_9 DM Madyira 40
Introductory Theory of Elasticity
Compatibility
M6FEA_9 DM Madyira 41
Introductory Theory of Elasticity
Constitutive Relations
The stresses and strains in a body subjected to some form of loading are
related in some way, which is governed by the properties of the material that
the body is made from
In elastic analysis we assume that the materials obey Hooke's law
This implies that the induced stress is always proportional to the induced
strain
Boundary Conditions F2
F3
F1
•Any real engineering problem will consist of a set
of forces acting on a body Deformable
Body
•These forces can act on the boundary or the
interior of the body
R1
•The stresses and strains that we seek to find are a R2
M6FEA_9 DM Madyira 43
Introductory Theory of Elasticity
System Approach LOADING
AND B.C.
KINETIC
RELATIONS
RESPONSE
GENERATED GENERATED
DEFORMATION FORCES
KINEMATIC CONSTITUTIVE
RELATIONS RELATIONS
M6FEA_9 DM Madyira 44
Introductory Theory of Elasticity
M6FEA_9 DM Madyira 45
Introductory Theory of Elasticity
M6FEA_9 DM Madyira 46
Introductory Theory of Elasticity
Stress Equations of Equilibrium
For equilibrium of forces in the x-direction we have:
σ x τ yx τ zx
σx + dx - σx dy dz + τ yx + dy - τ yx dxdz + τ zx + dz - τ zy dxdy + Fxdxdydz = 0
x y z
Simplifying this equation and dividing through by the volume of the element
(dx,dy,dz) leads to :
σ xx τ τ
+ yx + zx + Fx = 0
x y z
In a similar manner, it can be shown that the equilibrium of forces in the y and z
directions are described by:
σ yy τ τ
+ xy + zy + Fy = 0
y x z
σ zz τ τ yz
+ xz + + Fz = 0
z x y
M6FEA_9 DM Madyira 47
Introductory Theory of Elasticity
Stress Equations of Equilibrium
For a complete check on equilibrium, consider equilibrium of moments about
any given point yields:
τ yx = τ xy
τ zy = τ yz
τ xz = τ zx
Moment balance implies that shear stresses and their complements are equal
and opposite. The immediate mathematical implication is that the nine stress
components required to fully define the state of stress at a point are reduced to
six.
These sets of equations give us the general equations that a loaded solid has
to satisfy to maintain equilibrium.
M6FEA_9 DM Madyira 48
Introductory Theory of Elasticity
Strain-displacement Relations
Consider the deformation of two initially perpendicular lines in a deformable body as shown in
the figure below.
Y u
y dy
v B
v + dy
y
B
C
v
dy x dx
u
A v
C
A dx u
u + dx
x
X
M6FEA_9 DM Madyira 49
Introductory Theory of Elasticity
Strain-displacement Relations
In a similar manner, the direct strains in the other Cartesian directions (y and z)
can be shown to be:
v
εy =
y
w
εz =
z
M6FEA_9 DM Madyira 51
Introductory Theory of Elasticity
Strain-displacement Relations
Shear strain is the change in the 90o angle and hence is given by:
γ xy = α+β
v u
dx dy
x y
α = β =
dx dy
v u
= =
x y
M6FEA_9 DM Madyira 52
Introductory Theory of Elasticity
Strain-displacement Relations
It follows that:
u v
γ xy = +
y x
And similarly:
v w
γ yz = +
z y
u w
γ zx = +
z x
M6FEA_9 DM Madyira 53
Introductory Theory of Elasticity
Compatibility Equations
Thus in all, we have six strain equations relating six strains to three
displacements
It is clear that the problem is over determined
We therefore need to have another set of equations if our solution is to be non-
trivial and unique
Assume that by some practical approach we have managed to determine the 6
strain components and have to solve for the three displacement components
With the derived system of equations we have an excess of 3 equations since
we only need three equations to find the displacement components u, v and w.
M6FEA_9 DM Madyira 54
Introductory Theory of Elasticity
Strain-displacement Relations
u εx u
2 3
εx = =
x y 2
x y 2
v εy
2
v
3
εy = =
y x 2
x 2y
M6FEA_9 DM Madyira 56
Introductory Theory of Elasticity
Compatibility Equations
γ xy u v
2
2
xy y y
= +
xy
u v
3 3
= +
x y 2 x 2y
γ xy
2
εx εy
2 2
= +
xy y 2
x2
M6FEA_9 DM Madyira 57
Introductory Theory of Elasticity
Compatibility Equations
Similarly, we have:
γ yz
2
εy εz
2 2
= +
yz z 2
y2
γ zx εz εx
2 2 2
= +
zx x 2
z2
εx
2
γ xy γ zx γ yz
2 = + -
yz x z y x
εy
2
γ yz γ yx γ xz
2 = + -
xz y x z y
εz
2
γ yz γ zx γ xy
2 = + -
yx z x y z
M6FEA_9 DM Madyira 58
Introductory Theory of Elasticity
Compatibility Equations
Therefore, the complete set of compatibility relations is:
γ xy
2
εx
2
εy
2
= +
xy y 2
x2
γ zx
2
εz
2
εx
2
= +
zx x2 z2
γ zy
2
εy
εz
2 2
= +
zy y2 z2
εx
2
γ xy γ zx γ yz
2 = + -
yz x z y x
εy
2
γ xy γ yz γ xz
2 = + -
xz y z x y
εz
2
γ zx γ zy γ xy
2 = + -
xy z y x z
M6FEA_9 DM Madyira 59
Introductory Theory of Elasticity
Constitutive Relations
Constitutive relations introduce the material properties into the analysis i.e.
1
εx = σ x - ν( σ y + σ z )
E
εy =
1
σ y - ν( σ x + σ z )
E
1
εz = σ z - ν( σ x + σ y )
E
τ xy 2(1 + ν)
γ xy = = τ xy
G E
τ yz 2(1 + ν)
γ yz = = τ yz
G E
τ zx 2(1 + ν)
γ zx = = τ zx
G E
M6FEA_9 DM Madyira 60
Introductory Theory of Elasticity
Boundary Conditions
M6FEA_9 DM Madyira 61
Introductory Theory of Elasticity
Boundary Conditions
If
l = cos (x)
m = cos (y)
n = cos (z)
then area:
M6FEA_9 DM Madyira 62
Introductory Theory of Elasticity
Boundary Conditions
Equating forces in the x-direction and assuming that the body force per unit volume is
Fx + εx we have:
σ1 = σx l + σyx m + σzx n
σ2 = σy m + σxy l + σzy n
σ3 = σz n + σyz m + σxz l
M6FEA_9 DM Madyira 63
Introductory Theory of Elasticity
Boundary Conditions
M6FEA_9 DM Madyira 64
Introductory Theory of Elasticity
Boundary Conditions
X = σ xl + σ yx m + σ zx n
Y = σ xyl + σ y m + σ zy n
Z = σ xzl + σ yz m + σ z n
M6FEA_9 DM Madyira 65
Variational Calculus
• M6FEA_9 DM Madyira
Variational Calculus
• One of the questions one might ask is the difference between maxima and
minima in ordinary calculus and that in the calculus of variations
M6FEA_9 DM Madyira 67
Variational Calculus
M6FEA_9 DM Madyira 68
Variational Calculus
M6FEA_9 DM Madyira 69
Variational Calculus
In order to extremise the functional א, its first variation should vanish, i.e.
δχ = 0
This condition has an associated equivalent differential equation called the
Euler, or Euler-Lagrange equation, which can be expressed as follows:
F d F d2 F
- ' + 2 "
- .K .K = 0
y dx y dx y
M6FEA_9 DM Madyira 70
The Basic Raleigh-Ritz Method
M6FEA_9 DM Madyira
The Basic Raleigh-Ritz Method
• The Rayleigh-Ritz method can also be used for the solution of differential
equations, provided that an equivalent variational statement can be found
M6FEA_9 DM Madyira 72
The Basic Raleigh-Ritz Method
Consider that it is required to extremise the following functional:
1
1 dy 2
0
xy dx
2 dx
with the boundary conditions:
i.) y(0) 0
i.) y(1) 0
Note that the corresponding differential equation can be deduced from the
Euler-Lagrange equation as follows:
1 ' 2
1
χ = 0 2 (y ) + xy dx
Then:
F x, y, y1 =
1
y '2 + xy
2
M6FEA_9 DM Madyira 73
The Basic Raleigh-Ritz Method
i.e.
F d F d '
- = x - y = 0
Y dx y '
dx
or
d2 y
2
- x = 0
dx
The procedure for using the variational solution technique is outlined in the
following steps.
M6FEA_9 DM Madyira 74
The Basic Raleigh-Ritz Method
Step 1
~ m
y(x) = y (x) = α ψ x
j=1
j j
ψ j x = x
j-1
M6FEA_9 DM Madyira 75
The Basic Raleigh-Ritz Method
Step 2
Verify that the assumed solution satisfies the given boundary conditions.
The given boundary conditions (say t conditions) are to be substituted into the
previous solution and t coefficients are to be eliminated to reduce the solution
into the form:
~ n
y (x) = α
j=1
j Φj x
where n = m – t. For the given problem the assumed solution would be:
y x = α1 + α 2 x + α3 x 2 + α 4 x 3 + .....
M6FEA_9 DM Madyira 76
The Basic Raleigh-Ritz Method
That is:
α1 = 0
α 2 = - α3 + α 4 + .....
M6FEA_9 DM Madyira 77
The Basic Raleigh-Ritz Method
y x = a1 φ1 x + a2 φ2 x + a3 φ3 x +
φj x = x j+1 − x
M6FEA_9 DM Madyira 78
The Basic Raleigh-Ritz Method
Step 3
where
y x = a1 x 2 - x + a 2 x 3
- x + ....
and
= a1 2x -1 + a 2 3x 2 -1 + ....
dy
dx
M6FEA_9 DM Madyira 79
The Basic Raleigh-Ritz Method
Hence:
a1 2x - 1 + a 2 3x - 1 + ....
1
F y =
2
2
2
+ a1 x 3 - x 2 + a 2 x 4 - x 2 + ....
Substituting into the variational statement yields:
1
χ = F y dx
0
1
- 1 + a 2 3x - 1 + .... dx
1
2x
2
a1 2
=
2 0
1
+ a1
x 3
- x2 + a2 x 4
- x 2 + .... dx
0
M6FEA_9 DM Madyira 80
The Basic Raleigh-Ritz Method
Step 4
or
χ
= 0, i = 1, 2, ..., n
a i
M6FEA_9 DM Madyira 81
The Basic Raleigh-Ritz Method
i.e.
1 1
a1 2x - 1 + a 2 2x - 1 3x 2 - 1 dx+ x - x 2 dx = 0
2 3
0 0
2x - 1 dx
2
c11 =
0
1
dx
1
c12 = 2x - 1 3x 2
- 1 dx b1 = -
0
x 3
- x 2
0
M6FEA_9 DM Madyira 82
The Basic Raleigh-Ritz Method
Now the variation with respect to a2 gives:
1 1
3x 1 a1 2x 1 a2 3x 1 ... dx x x 2 dx 0
2 2 4
a2 0 0
i.e.
c21 a1 + c22 a 2 + .... = b 2
In this case the constants of integration are given by:
1
c 21 3x 2 1 2 x 1 dx c12
0
1
c 22 3x 2 1 dx
2
1
b2 x 4 x 2 dx
0
M6FEA_9 DM Madyira 83
The Basic Raleigh-Ritz Method
Step 5
where
c11 c12 ... c1 n
c c 22 ... c 2 n
c = 21
a = a1 a 2 .. a n
b = b1 b2 .. bn
M6FEA_9 DM Madyira 84
The Basic Raleigh-Ritz Method
y = a1 x 2 - x + a 2 x 3 - x
6x - 3x 2 - 2x + 1 dx
3
3 1 =
2x 1 0
.
2 3 6 3 2
0 = - - + 1
1 4 3 2
1
3 =
2
M6FEA_9 DM Madyira 85
The Basic Raleigh-Ritz Method
and:
1 1
3x 2 - 1 dx dx
2 3 2
c 22 = b1 = - x - x
0 0
1
1 1
9x + 1 dx = - -
4 2
= - 6x
0
4 3
9 6 1
= - + 1 =
5 3 12
1
x - x 2 dx
4 4
= b2 = -
5 0
1 1
= - -
5 3
2
=
15
M6FEA_9 DM Madyira 86
The Basic Raleigh-Ritz Method
1 1 1
a1 a 2
3 2 12
1 4 2
a1 a 2
2 5 15
4a1 6a 2 1
15a1 24a 2 4
M6FEA_9 DM Madyira 87
The Basic Raleigh-Ritz Method
which can be solved to yield:
1 6 4 1
4 24 15 4
a1 = a2 =
4 6 6
15 24 16 - 15
=
24 - 24 6
= 1
96 - 90 =
6
= 0
Substituting these constants into the assumed solution we have:
y =
1
6
x3 - x
M6FEA_9 DM Madyira 88
The Raleigh-Ritz Method Applied to Structural
Analysis
M6FEA_9 DM Madyira
The Raleigh-Ritz Method Applied to Structural Analysis
Introductory definitions:
Vector
A vector is a column matrix, for example:
a1
a
A = 2
..
a n
Row Matrix
A row matrix is a matrix, which contains one row. The transpose of a vector is a
row matrix, for example:
A t = a1 a 2 .. a n 1×n
M6FEA_9 DM Madyira 90
The Raleigh-Ritz Method Applied to Structural Analysis
Introductory definitions:
Scalar Product
Let A be defined as above and B be:
B = b1 b2 .. bn
The scalar product of two vectors can be defined as follows:
n
t
AB = BA = t
i=1
a i bi
Stress Vector
The state of stress at any point in a continuum, is defined by a stress tensor
which can be expressed as follows:
σx τx y τx z
τij = τ y x σy y τy z
τz x τz y σ z
M6FEA_9 DM Madyira 91
The Raleigh-Ritz Method Applied to Structural Analysis
Introductory definitions:
which is symmetric. For the finite element method it is preferable to represent
the stress tensor with the vector:
σ = σ x σy σz τx y τy z τz x
Strain Vector
The state of strain at any point in a continuum may be defined in terms of the
following strain tensor:
ε xx ε xy ε xz
ε ij = ε yx ε yy ε yz
ε zx ε zy ε zz
The strain vector can be defined as follows:
ε = ε x εy εz γ xy γ yz γ zx
M6FEA_9 DM Madyira 92
The Raleigh-Ritz Method Applied to Structural Analysis
Variational Statement
For many problems of continuum mechanics, there are forms of energy
balance theorems which provide variational statements directly
There are many energy theorems that can be used for structural analysis
The basic energy theorem employed in this course is the minimum total
potential energy theorem, which states that:
M6FEA_9 DM Madyira 93
The Raleigh-Ritz Method Applied to Structural Analysis
Variational Statement
The total potential energy of the structure can be expressed as follows:
χ = U + V
where U is the strain energy of the structure which is given by:
U =
structure
σ dε
t
d Vol.
a Pa
x, u P(x)
L
b
Pb
M6FEA_9 DM Madyira 95
The Raleigh-Ritz Method Applied to Structural Analysis
EXAMPLE
Suppose that it is required to determine the deformation and stress-strain
distribution for the heavy bar, which is hung vertically from one end, and
carries a heavy rigid mass at the other. The problem can be modelled as a bar
a-b fixed at end ‘a’ and subjected to a linearly distributed thrust.
x
P x = Pa + Pb - Pa
L
du
u + Δ x + higher order terms
dx
M6FEA_9 DM Madyira 97
The Raleigh-Ritz Method Applied to Structural Analysis
EXAMPLE
Now we can derive the variational statement for the problem. The strain energy
is given by:
1
U =
2 σ ε dxdydz
2
1 du
= Ε dxdydz
2 dx
dydz dx
2
1 du
U =
2 dx
E
M6FEA_9 DM Madyira 99
The Raleigh-Ritz Method Applied to Structural Analysis
EXAMPLE
From the above figure it can be shown that the double integral is cross sectional area A of the
section, hence:
1 L 2
du
U = AE dx
2 0 dx
du
L
W = Pdu = P dx
Bar 0 dx
1 L 2 L du
du
χ = U - W = AE dx - P dx
2 0 dx 0 dx
Raleigh-Ritz Solution
Step 1
Step 2
Verify that the assumed solution satisfies the given boundary conditions.
The only boundary condition we have is that u = 0 at x = 0 implying that α1 = 0. Hence the
assumed solution reduces to:
~
u(x) a1x a 2 x 2
du
= a1 + 2a 2 x
dx
This gives the variational statement as:
2
du du
L L
1
χ =
2 0 dx dx -
AE 0 dx dx
L L
1 x
AE a1 + 2a 2 x dx - 0 a L b a a1 + 2a 2 x dx
2
= p + p - p
2 0
a1
= AE a1 + 2 a 2 x dx -
0
pa +
0
X
L pb - p a dx = 0
AE a1x + a 2 x pb - p a = 0
L
2
- p a x + X2
2L
o
a1 + L a 2 = 1
2AE pa + pb
M6FEA_9 DM Madyira 104
The Raleigh-Ritz Method Applied to Structural Analysis
EXAMPLE
Raleigh-Ritz Solution
For variation with respect to a2 we have:
χ
L L
= 2 xAE a1 + 2 a 2 x dx - 2 x pa + X
pb - p a dx = 0
a 2 0 0
L
L L
AE a1x + 2 a 2 x dx - x pa + X2
pb - p a dx = 0
2
L
0 0
AE a1 X2 + 2 a32 X
2 3
- X2 p a +
2
X3
3L pb - p a L
0 = 0
3a1 4 La 2 1
AE
pa 2 pb
Solution of these equations yields:
a1
pa
AE
pb p a
a2 2 LAE
For exact graphical representation of y(x), a theoretically infinite number of points is required.
An acceptable approximation is to represent y(x) in terms of a finite number of points i.e. y1, y2,.
. . . . . . . yn at x = x1, x2,. . . . . . . . xn.
This concept is known as ‘Pointwise Discretisation’.
An approximate continuous model can be obtained from the discretised model by fitting a
continuous curve, which passes through the points (xi, yi), i = 1, 2,. . . . ., n
M6FEA_9 DM Madyira 108
The Raleigh-Ritz Method Applied to Structural Analysis
DISCRETIZATION CONCEPTS
Point Wise Discretisation
This can be achieved generally by employing Lagrangian interpolation as
follows:
~ n
yx = L x y
n
i i
i =1
Using pointwise discretisation, for the previous example, the function y(x) can
be approximated to a very high degree, but the resulting integral expression is
not simple.
The second approach is to divide the whole domain into a finite number of
subdomains.
The function y(x) does not require to be approximated accurately for each sub
domain.
Recalling that:
yi + 1 - yi
y x = yi + x
xi + 1 - xi
xi + 1
- x i yi + 1 + yi
Ii =
2
Any boundary value problem, which has a variational statement, can be solved
by means of the Raleigh–Ritz method as described earlier.
b. The need to make the assumed solution valid for the whole domain.
u x = u1 1 - 3 + 2 + u 2 4 - 4 + u 3 - + 2
L L L L L L
N 2 x = 4 ξ 1 - ξ
N3 x = ξ 2 ξ - 1
where
x
ξ =
L
Points 1, 2, and 3 are called nodes. N1, N2, and N3 are known as shape
functions.
M6FEA_9 DM Madyira 119
The Raleigh-Ritz Method Applied to Structural Analysis
DISCRETIZATION OF THE RALEIG-RITZ METHOD
Discretised Rayleigh-Ritz Solution
STEP 1
u u1 1 1 2
4 u 2 1
u 3 2 1
STEP 2
u1 0
u 4u 2 1 u 3 2 1
0
u3
STEP 5
u2
L
3 Pa Pb
8 AE
u3
L
Pa Pb
2 AE
M6FEA_9 DM Madyira 121
The Raleigh-Ritz Method Applied to Structural Analysis
DISCRETIZATION OF THE RALEIG-RITZ METHOD
Discretised Rayleigh-Ritz Solution
This process implies that the whole domain is to be divided into a number of
sub domains, which are known as finite elements.
Each sub domain should obey the same physical principles applied to the
whole domain, and can be treated as a separate problem.
The equations for the whole domain can be obtained by summing the sub
domain equations.
The Rayleigh-Ritz method with piecewise and point wise discretization is the
well-known finite element method, as illustrated next.
M6FEA_9 DM Madyira
The Finite Element Method
SIMPLIFY SIMPLIFY
RAYLEIGH-RITZ
SATISFACTION ASSUMED
METHOD
OF BC’s SOLUTION
FEM
This is the part that may differ from one problem to another. It may be deduced
directly from the discretised Rayleigh-Ritz method.
For some applications the finite element equations are deduced from the
weighted-residual approach or the Garlekin approach.
This part involves the derivation of element equations. This exercise normally
results in the element stiffness matrix.
In order to understand each part in more detail, let us apply the method to the
previous bar problem.
The steps used in the formulation of element equations are outlined next.
STEP 2
This step must be postponed because the sub domain boundary conditions
are not defined generally. Boundary conditions can be applied only on the
global system equations.
STEP 3
χ = U - W
xj 2
u j - ui
= 1
2 x AE L dx
i
AE
u - ui
2
1
= 2 j
L
And work done is:
xj
du
W = Fi u i + Fj u j P x dx
xi dx
M6FEA_9 DM Madyira 130
The Finite Element Method
FORMULATION OF ELEMENT EQUATIONS
Therefore the variational functional is:
AE
u j - ui - F u + Fj u j
2
χ = U - W = 1
2 i i
L
STEP 4
χ AE
uj
=
L
u j - u i - Fj = 0
AE
L
- u i + u j = Fj
K δ
e e
= F
e
EA 1 1
K (e)
L 1 1
M6FEA_9 DM Madyira 133
The Finite Element Method
STEPS FOR THE SOLUTION OF THE WHOLE DOMAIN
The domain can be divided into two subdomains and each subdomain can be
considered as a 2-node element.
For the problem at hand, a logical solution is to assume that the nodal forces
are the average values of the thrust distribution for each element.
For Element 1
AE 1 -1 u1 - 3 Pa + Pb 4
=
L -1 1 u 2 + 3 Pa + P b 4
It is useful to note, in the above element equations that ui is the total or actual
displacement at the ith node and Fi is a partial local force, for the considered
sub-domain. Hence, the law of the assembly is based upon the mechanical
principle.
M6FEA_9 DM Madyira 138
The Finite Element Method
STEPS FOR THE SOLUTION OF THE WHOLE DOMAIN
STEP 3: Assembly of sub-domain equations to obtain the equations for the
whole domain
The total force at each node is equal to the sum of the contributions from the
local element forces for elements connected to that node.
For the first element, inserting a correct number of zeros, the equations for the
element can be expressed as follows.
0 -
L L
M6FEA_9 DM Madyira 140
The Finite Element Method
STEPS FOR THE SOLUTION OF THE WHOLE DOMAIN
STEP 3: Assembly of sub-domain equations to obtain the equations for the
whole domain
The two matrices can be added together to yield:
AE AE
L - 0
L
u1 - 3Pa +Pb 4
AE AE AE
- 2 - u 2 = 2Pa - 2Pb 4
L L L Pa + 3Pb 4
u3
AE AE
0 -
L L
This can be written in the form:
K F
where K is the global stiffness matrix, δ is the global nodal displacement vector
and F is the global nodal force vector.
M6FEA_9 DM Madyira 141
The Finite Element Method
STEPS FOR THE SOLUTION OF THE WHOLE DOMAIN
The first equation and first column of stiffness matrix are eliminated, and a
reduced system of equations will be obtained, as follows:
AE AE 2Pa - 2Pb
2 u2 - u3 =
L L 4
AE AE pa + 3p b
- u2 + u3 =
L L 4
Noting that:
L = S2
S
u2 = 3Pa + Pb
8AE
S
u3 = Pa + Pb
2AE
M6FEA_9 DM Madyira 144
General 1D Finite Element Modelling
M6FEA_9 DM Madyira
General 1D Finite Element Modelling
Coordinate Transformation
The local stiffness matrix for a pin-jointed bar was derived as:
AE 1 1 ui Fi
u F
L 1 1 j j
However, this is not very useful because in general, trusses are inclined to the
horizontal.
This is done by first establishing the relationship between the local and global
coordinate systems.
u l m u o
o
v m l v
This may be expressed as:
Ui ξ Uio
if the displacements are those of node i in a truss. Since the element has two
nodes, the total displacement vector is given by:
ui u 0
v 0
i
i 0 v
0
i
u j 0 u j
v j v 0
j
λF0 k λU0
M6FEA_9 DM Madyira 149
General 1D Finite Element Modelling
Coordinate Transformation
However, λ is orthogonal, so that λT= λ-1, and we have:
F
0
λ k λU
T 0
k 0
k
T
λU0
u i l m 0 0 v
0
i
u j 0 0 l m u j
v0
j
l2 lm l2 lm
2
m2 lm m
k
AE lm
L l 2 lm l2 lm
2
lm m m
2
lm
Bar Element
Fi EA 1 1 u i
F 1 1 u
j L j
Beam Element
Fiy 12 6L 12 6L δiy
6L 4L2 6L 2L2 θ
M i EI i
Fjy L3 12 6L 12 6L δ jy
2
6L 2L 6L 4L θi
2
M i
M6FEA_9 DM Madyira 152
2D Finite Element Modelling
M6FEA_9 DM Madyira
2D Finite Element Modelling
x y xy
x y xy
and the strain energy for a linear elastic material can be expressed as follows:
dxdydz
1 t
U 2
t dxdy
1 t
2
The element has 4 corner nodes and 4 straight-line sides. Rectangular and
parallelogram elements are special cases.
Using a local node numbering system, the element can be defined as shown in
the above figure.
where N1, N2, N3 and N4 are shape functions. Hence, it can be deduced that:
u x , y u1 N 1 u 2 N 2 u 3 N 3 u 4 N 4
v x , y v1 N 1 v 2 N 2 v 3 N 3 v 4 N 4
u x , y 1 2 x 3 y 4 xy
u1 1 2 x1 3 y1 4 x1 y1
u2 1 2 x2 3 y2 4 x2 y 2
u3 1 2 x3 3 y 3 4 x3 y 3
u4 1 2 x4 3 y4 4 x4 y 4
1 x1 y1 x1 y1 1 u1
1 x2 y2 x 2 y 2 2 u
2
1 x3 y3 x 3 y 3 3 u 3
1 x4 y4 x 4 y 4 4 u 4
The major difficulty is that these shape functions depend upon the nodal global
co-ordinates and for every different sub domain the previous procedure of
shape function derivation should be repeated.
Such a system is known as an intrinsic system and its co-ordinates are the
intrinsic co-ordinates and .
Let us discuss the unique intrinsic element in the - system and then try to
correlate the - system with the x – y system.
One very good idea is to transform the quadrilateral element into a square of
unit length as shown next.
N2 1
N3
N4 1
This leads to assumed solutions of the form:
4
u , u i N i ,
i 1
4
v , v i N i ,
i 1
The problem now becomes one of obtaining the equation of the transformation:
x x ,
y y ,
One approach is to assume that x and y are field functions defined within the
element domain in terms of their nodal values and that they obey the same
interpolation formula, i.e.
x , x1 N 1 x 2 N 2 x 3 N 3 x 4 N 4
y , y1 N 1 y 2 N 2 y 3 N 3 y 4 N 4
Step 3
Express the strain component at any point in terms of nodal displacements and
shape functions.
u x, y N1 0 N 2 0 N3 0 N4 0
v x, y 0 N1 0 N 2 0 N3 0 N4
We already know from theory of elasticity that the strains are related to
displacements through kinematics relations:
u N1 N 2 N 3 N 4
x 0 0 0 0
x x x x x
u N1 N 2 N 3 N 4
y 0 0 0 0
y y y y y
u v N1 N1 N 2 N 2 N 3 N 3 N 4 N 4
xy
y x y x y x y x y x
M6FEA_9 DM Madyira 166
2D Finite Element Modelling
This can be reduced to:
B
3 1 3 8 8 1
where:
x y xy
N1 N 2
0 0
x x
N1 N 2
B 0 0
y y
N1 N1 N 2 N 2
y x y x
Consider the shape function, and apply the chain rule of partial differentiation
i.e.
N i N i x N i y
x y
N i N i x N i y
x y
In matrix form, this can be expressed as:
N i N i
x, y x
N J N
i , i
y
M6FEA_9 DM Madyira 168
2D Finite Element Modelling
where:
x y
x, y
J
, x y
N I N i
x 1 x, y
N J
i , i
N
y
Hence:
σ D B δ
In this case note that D is the material stiffness matrix given by:
1 1 0
E
D 1 1 0
1
2
1
0 0
2
M6FEA_9 DM Madyira 170
2D Finite Element Modelling
where:
y y
1
J 1
J x x
x y
J
x y
x y x y
Step 5
1
χ σ t ε dVol Fδ t
2
1
D B δ t
Bδ dVol Fδ t
2
1
δt Bt D B δ dVol Fδ t
2
1 t
2
δ Bt D B dVol δ Fδ t
Step 6
Extremise the variational functional i.e. minimise total potential energy which is
satisfied for:
χ
t
0
δ
Bt D B dVolδ F
M6FEA_9 DM Madyira
3D Finite Element Modelling
The main aim of this section is to introduce 3D elements in the same way done
for 2D.
δ u1 v1 w1 u2 v2 w2 u8 v8 w8
8
v x, y, z vi Ni x, y, z
i2
8
w x, y, z w i Ni x, y, z
i 1
u u1 at ξ 0, η 0 ζ 0
u u2 at ξ 1, η 0 ζ 0
u u8 at ξ 0, η 1 ζ 1
N 2 ξ, η, ζ ξ 1 η1 ζ
N3 ξ, η, ζ ξ η 1 ζ
N 4 ξ, η, ζ 1 ξ η 1 ζ
N5 ξ, η, ζ 1 ξ 1 η ζ
N 6 ξ, η, ζ ξ 1 η ζ
N 7 ξ, η, ζ ξ η ζ
N8 ξ, η, ζ 1 ξ η ζ
M6FEA_9 DM Madyira 180
3D Finite Element Modelling
The relationship between the Cartesian system (x, y, z) and the intrinsic
system (ξ, η, ζ) can be obtained through isoparametric transformation i.e.
8
x ξ, η, ζ x i Ni ξ, η, ζ
i 1
8
y ξ, η, ζ yi Ni ξ, η, ζ
i 1
8
z ξ, η, ζ zi Ni ξ, η, ζ
i 1
ε ε x εy εz γ xy γ yz γ zx
It can also be shown that:
ε 61 B624 δ 241
where:
B b1 b2 b8
Ni
x 0 0
0 Ni
0
y
Ni
0 0
z
bi
Ni Ni
y 0
x
0 Ni Ni
z y
Ni Ni
0
z x
f f x f y f z
ξ x ξ y ξ z ξ
f f x f y f z
η x η y η z η
f f x f y f z
ζ x ζ y ζ z ζ
f f x y z
ξ x
f f x y z
x, y, z x, y , z
η J ξ, η, ζ y J
, ,
f f x y z
ζ z
It has also been shown that | J | = The volume of the parallelopiped element.
and shape functions
σ σ x σ y σ z τ xy τ yz τ zx
σ D ε DB δ
1- ν ν ν 0 0 0
ν 1- ν ν 0 0 0
ν ν 1- ν 0 0 0
D
E
1 ν 1 ν 0 0 0
1 - 2ν
0 0
2
1- 2ν
0 0 0 0 0
2
1 - 2ν
0 0 0 0 0
2
M6FEA_9 DM Madyira 188
3D Finite Element Modelling
STEP 5:
Express the total potential energy of the element in terms of nodal
displacements
χ U-W
The strain energy is given by:
ε t
U σ d ε dxdydz
element ε o
U 1
2
δ B DB dxdydz δ
t t
element
The work done by external forces is given by:
W δ t FE
Thus the total potential energy of the element is given by:
χ 1
2
δ B DB dxdydz δ δ t FE
t t
element
K Bt
DB dxdydz
element
M6FEA_9 DM Madyira
Practical Implementation of FEM
Reminder of Basic Facts
1. Finite element modelling is an approximation
2. The approximations include:
a. Geometry
b. Material properties
c. Loading
d. Method of discretization (type and number of elements)
3. It is important to understand how these affect your solution
4. Strategies must be implemented to ensure that the model produces the most
accurate solution
Mesh sensitivity study can be used to selected the most optimum element and element density
M6FEA_9 DM Madyira
Abaqus
Abaqus is a suite of powerful simulation programs
They are all based on the finite element method
It allows general purpose simulation
It covers both linear and nonlinear simulations
The range of available elements allows modelling of very complex geometries
A wide range of material models are available that can be used to simulate a
wide range of materials
Common problems that can be simulated include structural, heat transfer, mass
diffusion, acoustics, solid mechanics, piezoelectric, electromagnetic and fluid
dynamics analyses
M6FEA_9 DM Madyira
Closure
This course has presented the background, development and applications of
finite element analysis
The Raleigh-Ritz method was used to demonstrate the complete finite
element analysis process
Element stiffness matrices were developed starting with 1D, 2D and 3D
elements
The use of the FEM was demonstrated using the Abaqus commercial FEA
software produced by Dassault Systems
Course participants should have developed the knowledge to analyse basic
structural problems
M6FEA_9 DM Madyira 208
Bibliography
M6FEA_9 DM Madyira
Bibliography
1. Zienkiewicz, O. C. and Taylor, R. L., “The Finite Element Method for Solid and Structural
Mechanics”, 7th Edition, Elsevier, London, 2014
2. Rieg, F., Hackenschmidt, R. and Alber-Laukant, B., “Finite Element Analysis for Engineers”,
Carl Hanser Verlag Munich, Munich, 2014
3. Kim, N.-H. and Sankar, B. V., “Introduction to Finite Element Analysis and Design”, John
Wiley & Sons, 2008
4. Fagan, M.J., “Finite Element Analysis: Theory and practice”, Longman Scientific and
Technical, Singapore, 1992
5. Clough, R. W., “The finite element method in plane stress analysis”, Proc. American
Society of Civil Engineers (2nd Conference on Electronic Computation, Pittsburgh,
Pennsylvania), Vol. 23, P. 345-378, 1960
6. Ellobody, E., Feng, R. and Young B., “Finite Element Analysis and Design of Metal
Structures”, Elsevier, London, 2014
M6FEA_9 DM Madyira 210
Bibliography
1. Logan, D. L., “A first course in the Finite Element Method”, 3rd Edition, Brooks/Cole –
Thompson Learning, Pacific Grove, California, USA pp2-3, 2002
2. Moaveni, S., “Finite Element Analysis – Theory and applications with ANSYS”, 2nd Edition,
Pearson Education, Upper Saddle River, New Jersey, USA, 2003