Introduction to Finite Element Analysis
1.1 Methods to Solve Any Engineering Problem
|
Analytical Method Deed Experimental Method
= Classical approach
=100% accurate results
Closed form solution
- Applicable only for simple
problems like cantilever & simply
supported beams etc.
= Complete in itself
Though analytical methods could
also give approximate results if
the solution is not closed form,
but in general and broad sense,
‘analytical methods are considered
as closed form solutions ie.
100% accurate.
-Mathematical representation
~ Approximate, assumptions made
- Applicable even if physical
prototype not available (initial
design phase)
- Real life complicated problems
Results can not be believed
blindly & must be verified by
experimental method or hand
‘calculation for knowing the range
of results
Finite Element Method: Linear,
Nonlinear, Buckling, Thermal,
Dynamic & Fatigue analysis.
Boundary Element Method:
Acoustics /NVH
Finite Volume Method: CFD
(Computational Fluid Dynamics) &
‘Computational Electromagnetics
Finite Difference Method:
Thermal & Fluid flow analysis (in
combination with FVM)
~ Actual measurement
~Time consuming & needs
expensive set up
Applicable only if physical
prototypes available
-Results can not be believed
blindly & min. 3 to 5 prototypes
must be tested
Strain gauge
=Photo elasticity
-Vibration measurements
- Sensors for temp. & pressure etc,
~Fatigue test
1.2 Procedure for Solving Any Analytical or Numerical Problem
There are 2 steps
Step 1) Writing governing equation ~ Problem Definition or in other words formulating the
problem in the form of a mathematical equati
Step 2) Mathematical solution of governing equation
Final result is summation of step 1 & step 2. Result will be 100 % accurate when there is no
approximation at either of the steps (Analytical method).
Numerical methods make approximation at step 1 as well as at step 2 & hence all the numerical
methods are approximate.Introduction to Finite Element Analysis
Pee ee Dues ae icea
a, Cantilever beam deflection - Analytical approach
Aim :To determine deflection of cantilever beam (y)
Step 1: Governing equation
o/y= M/1=E/R
ay/dx?
WEL
Governing Equation (unknown = y))
Step 2: Mathematical solution of governing equation -
El d’y/dx? = M=-P x (Origin at free end, moment at a distance x from origin =-P x)
Integrating it w.rt.x, 1* integration
El dy/dx =-P x°/2+C,
Integration constant C, could be determined by substituting known boundary conditions ie. at
x=L dy/dx=0
= 0=-PL/2+C,.
=, =PL72
2nd integration -
El y=- Px?/6 + PL? x/2+ c
Integration constant C, could be determined by substituting
atx=L, y=0Practical Finite Element Analysis
= 0=-PL/6+PLY2+C,
=> =-PL3
Substituting values of constants C, andC,
Y= I/EI[-P °/6 + PL? x/2 - PL?/3]
For cantilever beam, deflection is maximum, when force is applied at free end of beam i.e. x = 0
& above equation reduces to
y=-PL'/3EI (-sign indicates deflection in downward direction)
Obvious question is, if analytical methods are accurate then why are they not used for solving
real life problems. Primary reason is step 1, the governing equation. For above problem beam
bending equation was readily available as starting point but this type of equations are not
available for solving real life problems like say for transmission casing or hydraulic housing. Basic
beam bending equation is also based on many assumptions such as small deflection, isotropic
material, c/s of the beam remains plane and perpendicular to the neutral axis and radius of
curvature is large in comparison to cross-sectional dimensions etc.
How does numerical methods solve the problem?
Numerical methods like FEM are based on discretization of integral form of equation. Basic
theme of all numerical methods is to make calculations at only limited number of points
& then interpolate the results for entire domain (surface or volume). Even before getting the
solution we assume how the unknown is going to vary over a domain. Say for example, when
meshing is carried out using linear quadrilateral elements, assumption made is linear variation
of displacement over the domain and for 8 noded quadrilateral element, assumption is parabolic
variation. This may or may not be the case in real life & hence all numerical methods are based
on an initial hypothetical assumption. After getting the results there are several ways to check
numerical as well as practical or field result correlation accuracy & minimization of errors.
1.3 Brief Introduction to Different Numerical Methods
1) Finite Element Method (FEM) :
FEM is the most popular numerical method.
Applications - Linear, Nonlinear, Buckling, Thermal, Dynamic & Fatigue analysis. FEM will be
discussed in detail at later stage.
Are FEA and FEM different?
Finite Element Method (FEM) and Finite Element Analysis (FEA) both are one & the same. Term
“FEA’ is more popular in industries while “FEM” at universities.
Many times there is confusion between FEA, FEM & one more similar but different term FMEA.Introduction to Finite Element Anal
(Failure Mode Effect Analysis). FEA/FEM is used by design or R&D department only while FMEA
is applicable to all the departments.
2) Boundary Element Method (BEM) :
Itis a very powerful and efficient technique to solve acoustics or NVH problems. Just like finite
element method it also requires nodes and elements but as the name suggest it considers only
outer boundary of the domain. So in case if the problem is of a volume, only outer surfaces are
considered. If the domain is area then only outer periphery is considered. This way it reduces
dimensionality of the problem by a degree of one & thus solving it faster.
3) Finite Volume Method (FVM) :
All Computational Eluic Dynamics (CFD) softwares are based on FVM. Unit volume is considered
in Finite Volume Method (similar to element in finite element analysis). Variable properties
at nodes are pressure, velocity, area, mass etc. It is based on Navier - Stokes equations (Mass,
Momentum and Energy conservation equilibrium equations).
4) Finite Difference Method (FDM) :
Finite Element and Finite difference share many common things. In general Finite Difference
Method is described as a way to solve differential equation. It uses Taylor's series to convert
differential equation to algebraic equation. In the conversion process higher order terms are
neglected. It is used in combination of BEM or otherwise FVM to solve Thermal and CFD coupled
problems.
Is it possible to use all the above listed methods (FEA, BEA, FVM, FDM) to solve same
problem (say cantilever problem)?
The answer is YES! But the difference is in accuracy achieved, programming ease & time required
to obtain the solution.
When internal details are required (such as stresses inside the 3-d object) BEM will lead to
Poor results (as it considers only outer boundary), while FEM or FDM or FVM are preferable.
FVM has been used for solving stress problem but it is well suited for computational fluid
dynamics problems where conservation & equilibrium is quite natural. FDM has limitations with
complicated geometry, assembly of different material components and combination of various
types of elements (1-D, 2-D & 3-0). For this type of problems FEM is far ahead of its competitors.
Discretization of problem:
All real life objects are continuous. Means there is no physical gap between any two consecutive
Particles. As per material science, any object is made up of small particles, particles of molecules,
molecules of atoms and so on and they are bonded together by force of attraction. Solving a realPractical Finite Element Analysis
life problem with continuous material approach is difficult and basic of all numerical methods
is to simplify the problem by discretizing (discontinuation) it. In simple words nodes work like
atoms and with gap in between filled by an entity called as element. Calculations are made at
nodes and results are interpolated for elements.
There are two approaches to solve any problem
Continuous approach Diserete approach
Allreal life components are Equivalent Mathematical
Continuous modeling
Discrete (mathematical equivalent) model,
chair represented by shell and beam
elements, person via lumped mass at C.G.
Real life continuous problem
From mechanical engineering point of view any component or system could be represented by
three basic elements -
Masser? > ~_ssas
Spring
Dampers cH :
All the numerical methods including Finite Element follows discrete approach. Meshing (nodes
and elements) is nothing but discretization of a continuous system with infinite degree of
freedoms to finite degree of freedoms.Introduction to Finite Element Analysis
1.4 What is DOF (Degree Of Freedom) ?
Force
“ x re
Object is fixed at one end. Force is applied at point“P”. Due to force object deforms and Point P
get shifted to new position P:
When can we say that we know solution of above problem?
Ifand only if we are able to define deformed position of each & every particle completely.
imum number of parameters (motion, coordinates, temp. etc.) required to define position of any
entity completely in the space is known as degree of freedom (dof).
Consider following 2-d (planer) problem. Suppose origin is at bottom left comer and is known.
To define position of point A completely with respect to the origin we need two parameters i.e.
x, and y,, in other words 2 dofs (translation x and y).Practical Finite Element Analysis
Consider that the point A is a part of line, now one angle should also be defined in addition to
two translations i.e. 3 dofs (two translations and one rotation).
\
Suppose points A and B are shifted out of the plane and the line is rotated arbitrarily wrt all the
three axes. Minimum parameters to define position of point A completely would be 6 dofs{3
translations (U,, U, , U,) and 3 rotations @,,@, 8}.
Dof is a very important concept. In FEA we use it for individual calculation points ie. nodes (total
dofs for a given mesh model = number of nodes x dof per node) while theory of machines and
mechanism uses this concept for body as a whole.
Itis not like always all the elements have 6 dofs per node. Degree of freedom depends on type
of element (1-d, 2-d, 3-d), family of element (thin shell, plane stress, plane strain, membrane etc.)
and type of analysis {for structural analysis a thin shell element has 6 dof/node (displacement
unknown, 3 translations and 3 rotations) while the same element when used for thermal analysis
has single dof /node (temperature unknown}.
Fora new user itis a bit confusing but there is lot of logical, engineering & mathematical thinking
behind assigning specific number of dofs to different element types & families
1.5 Why do We Carry Out Meshing, What is FEM
No. of Points = ¢2 No. of Nodes
dof per Point= 6 dof per Nod
Total equations = 22Introduction to Finite Element Analysis
FEM
-Anumerical method
- Mathematical representation of actual problem
- Approximate method
Definition of FEM is hidden in its words itself. Basic theme is to make calculations at only limited
(Finite) number of points and then interpolate the results for entire domain (surface or volume).
Finite - Any continuous object has infinite degrees of freedom &it's just not possible to solve the
problem in this format. Finite Element Method reduces degrees of freedom from Infinite to Finite
with the help of discretization i.e. meshing (nodes & elements)
Element - All the calculations are made at limited number of points known as nodes. Entity
joining nodes and forming a specific shape such as quadrilateral or triangular etc. is known as
Element. To get value of variable (say displacement) any where in between the calculation points,
interpolation function (as per the shape of element) is used.
Method - There are 3 methods to solve any engineering problem. Finite element analysis belongs
to numerical method category.
How the results are interpolated from few calculation points
Itis ok that FEA is making all the calculations at limited number of points, but the question is how
it calculates value of the unknown somewhere in between calculation points.
This is achieved by interpolation. Consider 4 noded quadrilateral element as shown in the figure.
Quad4 element uses following linear interpolation formula -
ay ta,xtay+a,xy
FEA calculates values at outer nodes 1,2, 3,4 i.e.a,,a,,a,,a, are known.
1 2
3 4
4noded (linear) quad
Value of the variable any where in between could be easily determined just by specifying x & y
coordinates in above equation.
For 8 noded quadrilateral, following parabolic interpolation function is used
Usa,taXtaytaxytaxtay+aryta, xy?Practical Finite Element Analysis
1 2 3
8 4
7 6 5
8 noded (parabolic) quad
How about accuracy if we increase number of calculation points (nodes & elements)
Yes, in general increasing calculation points improves accuracy.
Suppose somebody gives you 3 straight lines and ask to best fitit in the circle, find area of triangle
& compare it with circle and then repeat the exercise with 4, 6, 8,16, 32 & 64 lines.
OO
3lines lines
oo
6Lines BLines Shaded Area is Error
By increasing number of lines, error margin reduces. Number of straight lines are equivalent to
number of elements in Finite Element Analysis.
Assume exact answer for area of circle (rt) is 100. 3 lines give answer = 41 while 4 as 64 &so on.
Answer 41 or 64 is not at all acceptable but 80 or 90 is, considering time spent & relative design
concept.
Exact Answer 100
50
2
on
area
a
16 3
No.of inesIntroduction to Finite Element Analysis
If higher number of nodes and elements leads to higher accuracy then why not to always create
a very fine mesh with maximum possible nodes and elements ? The reason is solution time is
directly proportional to (dof). n= 1 to 4, depending on type of analyses and solver. Also large
size models are not easy to handle on the computer due to graphics card memory limitations.
Analyst has to make a fine balance between desired level of accuracy and element size (dof) that
could be handled satisfactorily using the available hardware resources.
Assume Analytical Method approach gives answers very close to 100 and time taken = 1 month
and Finite Element Analysis with reasonable mesh size gives answer 90 within 1 day. In industry
getting fast solutions with logical or reasonable accuracy is more important than absolute
accuracy.
1.6 Advantages of FEA
+ Visualization
+ Design cycle time |
+ No.of prototypes |
+ Testing )
+ Optimum design
Visualization of results: For simple geometries such as simply supported beam or cantilever
beam it is easy to visualize point of maximum stress and displacement. But in real life for parts
or assemblies with complex geometrical shapes, made up of different materials with many
discontinuities subjected to flexible constraints, complex loading varying wrt time and point
of application, further complicated by residual stresses and joints like spot and arc welds etc,, it
is not easy to predict failure location. Imagine someone shows you a complicated engine block
and ask to predict failure location for given set of forces. It is not easy to predict it successfully
unless and until you have years of experience in the similar field. But with tools like CAD & CAE,
if modelled in appropriate fashion, one can easily get stress contour plots clearly indicating
locations of high stress or displacement.
Previously components used to be designed by highly experienced engineers who had seen lot
of testing & failures of the components in real life. These days in most of the organizations design
engineers are very young, using tools like CAD / CAM / CAE and confident about their designs.
10Practical Finite Element Analysis
1.7 Design Cycles
Design Cycles
f OS
Chain Cycle
current Engineerin;
(Old or conventional “freer dean”
design cycle) oxcle & more efficient)
Follureof Ideas from features ji
Gustomer components inthe availableon—"Mnovative NZ
concept
foedbac lddeign compettors product °@
woN
ck
>> [Sereno
Tada
t Engineeing
BED (CAB)
RD [ Design
1 Maren ]
Test
Purchase,
1
Vendor (prototype)
}
Testing Dept
{———~-
} 4 Component Failed Test Successful
ualty
Process
xa
5 acide — See g
hal
Conventional or Traditional design cycle
Itwas used before 1980's in developed countries and in India up to late 90's.
Consider a simple bracket designed via conventional cycle, failed after conducting the test as
shown below
Failure Force
location
Original Design
What could be the probable solutions to above problem
1) Increase thickness : Design dept. would release the drawing, purchase engineer submit
it to vendor and then test the new prototype. If failure is reported again then increase the
thickness further by repeating the process. Say test is successful after three iterations but
the component becomes too heavy and not acceptable.
nIntroduction to Finite Element Analysis
Thickness Increased
2) Next alternative could be to add fillet at the sharp corner
Introduction of filet
assume this modification is also not working
3) Another alternative is to change the material (high strength material), say test is successful
but this suggestion is rejected by marketing dept. due to excessive cost.
4) Further tryouts could be carried out by adding stiffeners
Addition of Rib
&so0n .....
This is a very long & time consuming process. Before finalizing the design at component level
minimum 3 to 5 prototypes were required to be tested. This design cycle is called as a chain
design cycle because just like links of a chain all the departments are connected to each other
ina specific format & function similar to the chain i. e. when the force is applied at the last link it
will get transferred to topmost link in a sequential manner (from bottom to second last one & so
on one by one).
Total Design Time
Effective time + Idle time
In the conventional chain design cycle effective time as well idle time are very high. There is no
2Practical Finite Element Analysis
coordination among the departments. R&D department is not at all aware or concern about
the project unless & until drawings are delivered to them. Likewise Purchase dept. has no idea
about what is going on in design or status of the project at R&D. It gets involved only after R&D
approved drawings are handed over to them.
Concurrent Engineering Design Cycle
Itattacks simultaneously on idle as well effective time by using
-CAD/CAM/CAE softwares
-Smart management techniques
Why it is called Concurrent? In statics if you remember one of the force system is concurrent
forces; all the forces acting or emerging simultaneously from a single point.
oe
ge
In concurrent design cycle similar concept is used. Right from the first day of project a team
of representative from all the departments is formed. They seat & work together with specific
targets assigned to them. All the people have access to CAD software. Library or data storage
isa very useful and nice facility provided by the commercial softwares. It not only work as safe
backup, but also as fast & perfect tool to keep posted all the team members about current
status of design / project. Whenever design modifications are carried out, all team members are
intimated automatically.
Idle time reduction : Say after 2 weeks, design engineer is ready with primary design of a
component. Now while he is deciding tolerances, manufacturing process, surface finish methods
etc, CAE engineer starts analysis, simultaneously purchase engineer contacts vendor, transfer
basic CAD data with the advance information of the job order. So that by the time design is
finalized, vendor will also be prepared to start the job immediately. When prototype activity is in
the process, test engineer will schedule the test. This is how idle time is minimized,
Effective time reduction : CAE plays important role in reducing no. of prototypes. Test results
of the first prototype are compared with finite element model. 10 to 15 % difference in FEA &
experimental results is considered as good correlation. FE model behaving in the same way as
predicted by test is key to success, Now further permutations & combinations (like changing
thickness, material, fillet, addition of ribs etc.) could be performed very fast and in an optimum
way with the help of CAE. Say for example changing thickness of a sheet metal part from 1.5
mm to 2 mmis just a matter of literally one minute! CAE engineer has to create a new property
with thickness of 2 mm, assign it to mesh & run the analysis which will not take more than few
minutes. While earlier approach of conventional design would have consumed 3 to 6 months for
the same. Oh man, what a magic! The work of 3 months finished within ¥ day!! That sounds great
and is cost effective too. Its win-win situation for manufacturer as well as customers.
300K OO EE
Introduction to Finite Element Analysis
1.8 Absolute vs. Relative Design
Design
|
| “™~ ™~s
| Relative design Absolute design
1) Relative design : In industry usually basic design of a category of components remains same
over the years. Say for example existing vehicle power is to be increased from 100 hp to 125
hp. Basic design (shape and concept) of components would remain same with minor changes
like scaling the basic design in appropriate proportions. Suppose CAE model & analysis of the
Previous version which is performing satisfactorily in the field, is available. If Analysis of new
design (using same element type and size with appropriate loads) shows stress magnitude less
than or equal to previous model then it could be concluded that the new upgraded design is also
safe & will perform satisfactorily. This way one can also avoid test correlation for new model.
Some times too much emphasis is given to test correlation & accuracy of the FE model to minute
level. Too much attention to capture each and every detail complicates FE modeling & analysis
unnecessarily (such as modeling bolt threads when main objective is component design rather
than bolt, defining non linear contacts when simple linear connection can work or dense mesh in
the name of accuracy without due consideration for hardware and software capabilities etc )
2) Absolute design : This approach is useful when the product/component is designed for the
first time Le, innovative design and no previous record of similar product is available. The design
engineer himself not very sure about boundary conditions & various load cases, CAE results of
such a design must be verified properly via testing and FE model should be corrected in case of
variation in the test & FEA results.
1.9 Is FEA a Replacement for Costly and Time Consuming Testing
“Finite Element Analysis has minimized testing requirement but it will be wrong to assume that
it has or will totally replace testing”
In fact FEA and Testing are not rivals but friends. One thing is for sure that all the good designs
are product of excellent coordination between testing and CAE.
ACAE engineer depends on testing for following reasons
1) Input data for CAE (data acquisition)
2) Validation of the CAE results
Though in industry there are CAE experts who just by looking at FE results can tell whether the
design is acceptable or not and in 99% cases their statement matches with the test results. It
's possible for them because of their past experience say for example someone is analyzing
gear box for 15 years. In past he has calibrated many models and seen how it fails in the field.
But believing words of a novice CAE engineer who has no prior experience or just because the
software used is impressive and very costly would be a big mistake
4