You are on page 1of 57

Dr Andrew Cowell

Computer Aided Engineering


MHH113285
Introduction to Finite Element Analysis
Module Introduction
• Practical Module
• Assessed by Coursework only
– Please note the following statement on
plagiarism:- 'The deliberate and substantial
unacknowledged incorporation in your submission
of material derived from work (published or
unpublished) of another is prohibited. In such
cases the work submitted is classified as illegal and
will result in further action being taken by the
University which may include recording a failure.'
Brief History
• Method around since 1950’s
• 1960’s
– Clough coined the term ‘finite element’
– Boeing developed method for structural
analysis using beams
– Large main frame digital computers
required
Brief History
• 1960’s
– Software written in-house
• Boeing – air frames
• Rolls Royce – aero-engine
• NASA – space travel
• CEGB – nuclear
• Westinghouse - nuclear
Brief History
• 1960’s
– Large Fortran code needing linked together
• Used as a research tool
– Used only by PhD’s
• 1970’s
– Software becoming a ‘little’ more user friendly
• Commercial software now available
– NASA’s FE Code became MSC – Nastran
– ANSYS
– BERSAFE
– Patran
Brief History
• 1970’s
– Large mini-computers still needed
– Still experienced mathematicians and
engineers using method
– Large companies starting to use method,
having access to large computers
– Software approx. cost £50K - £100K
– Computer approx. cost £100K
Brief History
• 1980’s
– Small mini-computers (Prime 550’s, Micro Vax) of
approx cost £50K
– Software commercially available and more user
friendly
– Computers now in most large companies, more
powerful and can cope with large/real problems
– 3rd Party software for pre/post processing, e.g.
Patran, FEMGEN
Brief History
• 1990’s
– Workstations very powerful and reasonable cost ~
£20K
– FE software – more robust and cost ~ £20K – £10K
– Most companies have access to hardware/
software
– MSc and BEng/BSc graduates asked to use method
– Software ported to Windows operating system full
version
Brief History
• 2000
– High specification PC’s available (128 Mb
RAM, 17” Colour Monitor, 13Gb Hard
Disc, CD Drive) for under £900
– Software ‘user friendly’, relatively low
cost ~ £5K – £10K
– Now accessible to everyone, even the
individual
– Advent of good solid modelers with FE
links make FE modelling easy
Brief History
• 2012
– Dell Precision T3400 PCs
– 2 x Dell 1908FP-BLK (19in 1908FP-BLK
European Black UltraSharp (1280 x 1024)
TCO99 DVI-D HT) monitors
– Intel Core 2 Quad CPU, Q6600 @ 2.4 GHz
– 8 Gb RAM
– Windows 7 64-bit Operating System
– NVIDIA Quadro FX 1700 Graphics Card
Summary of Analysis Types
• Large class of problems solvable
– Statics – small and large displacement,
plasticity, creep, thermal, plus non-linear
(gaps), fatigue
– Buckling – linear, non-linear
– Vibrations – modal, harmonic, transient
– Thermal – conduction, convection,
radiation
– CFD – velocity, pressure, temperature
Elements and Terminology

• Nodes
– Points in geometric space
– Numbered with x, y, & z co-ordinates
• Elements
– Have a number, topology (27, 127, 280, 290) and
material
Elements and Terminology

• Elements
– Can be quadrilaterals or triangular
– Can have straight or curved sides
• Elements with curved sides have mid-side
node
Elements and Terminology

• Nodes
– Element type determines the number of
degrees of freedom (d.o.f.), i.e. Ux, Uy,
Uz, Rotx, Roty, Rotz
Elements and Terminology

• Elements are joined at the nodes only


– Corner nodes to corner nodes
– Mid-side node to mid-side node
Finite Element Method – Process Outline
• Pre-processing Phase
– Sub-divide the problem into nodes and elements
– Assume a shape function to represent the
physical behaviour of an element
– Develop equations for an element
– Assemble the elements to represent the
complete problem
• Construct the global stiffness matrix
– Apply boundary conditions, initial conditions and
loading
Finite Element Method – Process Outline
• Solution Phase
– Solve a set of equations (linear or non-
linear) simultaneously to obtain the
nodal results, e.g.
• Displacement values
• Temperature (for a heat transfer problem)
Finite Element Method – Process Outline
• Post-processing Phase
– Obtain additional important
information, e.g.
• Principal stresses
• Heat flux
• Thermal expansion
Finite Element Solution Approaches
• Direct Formulation
• Minimum Total Potential Energy
Formulation
• Weighted Residual Formulations
Finite Element Solution Approaches –
Direct Formulation
• Example
– Bar of variable cross-
section
– Supports load, P
– Fixed at one end
– Width of bar at top, w1
– Width of bar at bottom,
w2
– Thickness, t
– Length, L
– Modulus of Elasticity, E
Finite Element Solution Approaches –
Direct Formulation
• Aim
– Approximation
of deflection at
various points
along length
• Assumption
– Neglect the
weight of bar as
negligible
compared to
load
Direct Formulation – Pre-processing
• Discretize the domain into finite elements
• Simple model with 5 nodes and 4 elements
– Increased accuracy is possible with a greater
number of nodes and elements
Direct Formulation – Pre-processing
• Discretize the domain into finite elements
– Each element has a uniform cross-section
– Cross-sectional area represented by average area
at the cross-sections of the nodes
Direct Formulation – Pre-processing
• Assume a solution that approximates the
behaviour of an element
– Typical element of cross-section, A, length, l,
subjected to a force, F
– Average stress, , given by:

F
 (1)
A
Direct Formulation – Pre-processing
• Average normal strain, , defined by change in
length, l, per unit of original length, l:
l
 (2)
l
• From Hooke’s Law:

  E (3)
Direct Formulation – Pre-processing
• Combining Equations (1) – (3):
AE
F l (4)
l
• And modelling as a spring with equivalent stiffness:

AE
k eq  (5)
l
Direct Formulation – Pre-processing
• Model as a series of centrally loaded members with
different cross-sections
– Bar can be represented by model consisting of four elastic
springs in series
– Elastic behaviour of element modelled by equivalent linear
springs

f  k eq ui1 - ui  
A avgE
ui1 -u 
A i1  A i E
u - ui 
i i1
l 2l
(6)
Direct Formulation – Pre-processing
• Free body diagram
of forces acting
through nodes 1 –
5
• Static equilibrium
requires sum of
forces at each
node to be zero
Direct Formulation – Pre-processing
• Which results in five equations:

k1u1 - k 1u2  -R1


- k1u1  k 1u2  k 2u2 - k 2u3 0
- k 2u 2  k 2u 3  k 3 u 3 - k 3 u 4 0
- k 3 u 3  k 3 u 4  k 4u 4 - k 4 u 5 0
- k 4u 4  k 4 u 5 P
Direct Formulation – Pre-processing
• And in matrix form:

 k1 - k1   u1   R1 
 -k k1  k 2 - k 2  u   0 
 1
  2   
 - k2 k2  k3 - k3  u3    0 
 - k3 k3  k4 - k4  u   0 
  4   
 - k4 k4  u5   P 
Direct Formulation – Pre-processing
• Which can be re-written as:
  R1   k 1 - k1   u1  0 
 0   -k k k -k  u  0 
   1 1 2 2   2   
 0  - k2 k2  k3 - k3  u3   0 
 0   - k3 k3  k4 - k 4  u   
     4  0 
 0   - k4 k4  u5  P

• Or in general form:R  Ku- F (7)

• Which stands for:


reaction matrix 
stiffness matrixdisplacement matrix
- load matrix
Direct Formulation – Pre-processing
• Boundary condition
– Displacement at Node 1 is zero
• Unknowns
– Four nodal displacements
– 1 reaction force
Direct Formulation – Pre-processing
• Application of boundary condition allows the
displacements to be considered independent
of the reaction force, resulting in the following
matrix  1 0 0 0 0  u1  0 
 -k k k -k 0 0  u  0 
 1 1 2 2
  2   
 0 - k2 k2  k3 - k3 0  u3   0 
 0 0 - k3 k3  k 4 - k 4  u   
   4  0 
 0 0 0 - k4 k4  u5  P
• the solution of which gives the nodal
displacements
Direct Formulation – Pre-processing
• The application of boundary condition results
in a new general equation for solid mechanics
problems that contains the stiffness matrix,
the displacement matrix and the load matrix
stiffness matrixdisplacement matrix
 load matrix

• Once the displacement matrix is known,


Equation (7) can be used to solve for the
reaction force(s)
Direct Formulation – Pre-processing
• Develop equations for an element
– Each element has two nodes
– Each node has a displacement
• Thus, two equations for each element, each involving
nodal displacements and the element stiffness
Direct Formulation – Pre-processing
• Static equilibrium requires that sum of forces on
element is zero
• Using fi and fi+1 both in positive direction:

• or in matrix form:
fi  k eq ui - ui1 
fi1  k eq ui1 - ui  (8)

 fi   k eq - k eq   ui 
     (9)
fi1  - k eq k eq  ui1 
Direct Formulation – Pre-processing
• Assemble the elements to present
the entire problem
– Apply Equation (9) to all the elements
 k1 - k1 
– The stiffness matrix for element (1) is: K   
1

- k 1 k1 

and the position in the global stiffness


matrix is:  k1  k1 0 0 0  u1
 k 1 k1 0 0 0  u2
K1G   0 0 0 0 0 u
  3
 0 0 0 0 0  u4
 0 0 0 0 0  u5
Direct Formulation – Pre-processing
• This is then repeated for the other three
elements resulting in:
KG   K1G   K2G   K3G  K4G
or k k 0 0 0
 1 1 
  k1 k1  k 2 - k 2 0 0 
KG   0 - k2 k2  k3 - k3 0 
 
 0 0 - k3 k3  k4 - k4 
 0 0 0 - k4 k4 
• which is identical to the matrix
developed from the nodal analysis
Direct Formulation – Pre-processing
• Application of boundary conditions
and loads results in the same matrix
as generated previously:
 1 0 0 0 0   u1  0 
 -k k k -k 0 0  u  0 
 1 1 2 2
  2   
 0 - k2 k2  k3 - k3 0  u3   0 
 0 0 - k3 k3  k4 - k4  u   
   4  0 
 0 0 0 - k4 k4  u5  P

• In the general form of:


stiffness matrixdisplacement matrix
 load matrix
Direct Formulation – Practical Example
• Aluminium bar
– E = 71.0 GPa
– w1 = 50 mm
– w2 = 25 mm
– t = 5 mm
– L = 250 mm
– P = 500 kg
Direct Formulation – Practical Example
• Spreadsheet Formulae
  w - w1  
A y    w 1   2  y t (10)
  L  
k eq 
 A i1  A i E A avgE
 (11)
2l l
 ui1 - ui 
  E  E  (12)
 l 
f
 (13)
A avg
R  Ku- F (14)
Minimum Total Potential Energy
Formulation
• Common approach
• External loads cause deformation of body
• Work done by external forces is stored as elastic
energy, called strain energy
• General strain energy equation for an element under
axial load
y y
d   Fdy   kydy  21 ky2  21 kyy
0 0

  y dxdzdy 
1
2 dV
2
 E 2
e    d   dV   dV (15)
V 2 V 2
Minimum Total Potential Energy
Formulation
Minimum Total Potential Energy
Formulation
• Strain energy equation for axially
loaded beam element
E 2 A avgE 2
e  
V 2
dV  ui1  ui2 - 2ui1ui  (16)
2l

• Minimising this equation with respect


to the two deflections, ui and ui+1
 e  
 u   k
eq - k eq   ui 
 e    - k k  u 
i
(17)
    eq eq   i1 
 ui1 
Minimum Total Potential Energy
• This equation is Formulation
very similar in form
to Equation (9) for the Direct
Formulation and leads to the same
stiffness, reaction and load matrices
as this method
Exact Solution
PL    w 2 - w1   
uy   ln w 1    y  - lnw 1  (18)
Et w 2 - w 1     L   
Location of a Results from Results from Results from
Point Along the Exact the Direct the Energy
the Bar (mm) Displacement Method (mm) Method (mm)
Method (mm) (from CAE –
(Eq. 18) Level 3)

y=0 0 0 0
y = 70 0.017220 0.017194 0.017194
y = 140 0.037099 0.037034 0.037034
y = 210 0.060611 0.060481 0.060481
y = 280 0.089387 0.089138 0.089138
Weighted Residual Formulations
• Assume approximate solution for the
governing differential equation
– Must satisfy the initial and boundary conditions
for the given problem
– Not an exact solution, hence will lead to some
residuals or errors
• Each residual method requires the error to
vanish over some selected interval
Weighted Residual Formulations
• Apply principle to axially loaded beam element
– Governing differential equation is:

du
A y E -P  0 (19)
subject todythe boundary condition u(0) = 0
– Approximate solution is:

which
uy   ccertainly
y  c y 2 satisfies
 c y 3 the boundary
(20) condition
1 2 3
Weighted Residual Formulations – Axially
Loaded Beam Element
• Error Function (R) (gained by
substituting the assumed solution
into the governing differential
equation): du
    Ay   
    dy    
  w 2 - w1  
 y tEc 1  2c 2 y  3c 3 y  - P  R
2
w 
 1  (21)
  L  
Weighted Residual Formulations
• Collocation Method
– Error function forced to zero at as many
points as unknown coefficients, i.e. three in
this case
– ANSYS uses with Galerkin method
• Sub-domain Method
– Integral of error function over some
selected sub-intervals forced to zero
– Number of sub-intervals equal to number of
unknown coefficients, i.e. three (again)
Weighted Residual Formulations
• Galerkin Method
– The error is required to be orthogonal to some
weighting functions i:
b
a
Rdy  0 i  1, 2, ..., N (22)
– Weighting functions chosen to be members of
the approximate solution, i.e. y, y2, y3
– ANSYS uses for CFD, including thermal analysis
(Streamline Upwind/Petrov-Galerkin &
Collocated Galerkin), Electro-mechanics,
Magnetics
– One of the most commonly used procedures
Weighted Residual Formulations
• Least Squares Method
– Requires the error to be minimised with
respect to the unknown coefficients in
the assumed solution, according to the
relationship:
b
Minimise  R 2dy 
 (23)
 a 
which leads to:
b R
a R c i dy  0 i  1, 2, ..., N (24)
Comparison of Methods

Displacement (mm)
Location Exact Collocation Sub- Galerkin Least
on Bar Solution Method domain Method Squares
(mm) Method Method

y=0 0 0 0 0 0
y = 70 0.017220 0.018034 0.017319 0.015381 0.017217
y = 140 0.037099 0.037881 0.036801 0.035164 0.037025
y = 210 0.060611 0.061356 0.060111 0.059345 0.060664
y = 280 0.089387 0.090270 0.088914 0.087921 0.089374
Launching ANSYS
ANSYS GUI
ANSYS Help
ANSYS Main Menu

You might also like