Title: The Finite Element Method
1The Finite Element Method
2General Overview
- widespread use in many engineering applications
- Applications of FEM in Engineering
- Mechanical/Aerospace/Civil/Automobile
Engineering - Structure analysis (static/dynamic,linear/nonlin
ear) - Thermal/fluid flows
- Electromagnetics
- Geomechanics
- Biomechanics
- ...
3General Overview
- examples
- conduction heat transfer, solve for the
temperature distribution throughout the body with
known boundary conditions and material properties - fluid mechanics problems range from steady
inviscid incompressible flow to complex viscous
compressible flow,
4General Overview
- acoustics uses finite element and boundary
element numerical methods - electromagnetic solution for magnetic field
strength provide insight to the design of
electromagnetic devices - capabilities extended to include fluid-structure
interactions, convective heat transfer - Bio-mechanics-bone structural analysis, blood
flow in blood vessels
5General Overview
- Finite element method is a numerical method of
solving a system of governing equations over the
domain of a continuous physical system - method applies the many fields of science and
engineering - for engineering use, fields of continuum
mechanics and the theory of elasticity provide
the governing equations
6Why numerical method
Most engineering problem involve solution of
governing differential equations.
7For heat transfer, torsion of shafts,
irrotational flow, seepage through porous media
8Solution of differential equation is tedious and
some times impossible
Complex geometry, boundary conditions, loading
conditions and material
9General Overview
- Finite element method can be summarized in the
following steps - small parts called elements subdivide the domain
of the solid structure - elements assemble through interconnections at a
finite number of points (nodes) on each element - assembly provides a model of the structure
10General Overview
- within each small domain, we assume a simple
general solution to the governing equations - solution for each element is a function of the
unknown solutions at the nodes
11Fundamental concept of FEM
The fundamental concept of FEM is that continuous
function of a continuum (given domain ?) having
infinite degrees of freedom is replaced by a
discrete model, approximated by a set of
piecewise continuous function having a finite
degree of freedom.
12(No Transcript)
13- since the continuum domain is divided into finite
elements with nodal values as the unknowns, the
loads and boundary conditions must be converted
to nodal quantities - single forces F apply to the nodes directly
- distributed forces are converted to equivalent
nodal values - supports, e.g. grounds are converted to specified
displacements for the supported nodes
14General Overview
- sources of error
- assumed solution within the element is rarely the
exact solution - error between exact and assumed solution
- magnitude depends on the size of the elements
relative to the solution variation - in most cases, assumed solution converges to the
correct as element size decreases
15General Overview
- precision of the algebraic equation solution
- function of computer accuracy, algorithm used,
number of equations and element size
16General Overview
- all solid structures could be modeled with
three-dimensional solid elements, but for many
cases this is overkill - many structures can be simplified by making some
assumptions e.g. plane stress and plane strain
assumptions, simple beam theory
17General Overview
- different types of elements are formulated for
these simplified cases - lead to simpler and faster solutions
18General Overview
- elements are categorized as either structural or
continuum - structural elements include trusses, beams,
plates and shells - formulations are based on same assumptions as in
their structural theories - finite element solution is no more accurate than
a solution using conventional beam or plate theory
19General Overview
- continuum elements are two- and three dimensional
solid elements - formulation based on the theory of elasticity
(provides the governing equations for deformation
and stress) - Few closed form or numerical solutions exist for
these problems
20One-Dimensional Spring System
- Basic finite element method can be simply
illustrated by a one-dimensional spring system - the subscripted u values are the nodal
displacements - there is an applied force F at node 3
- we want to find the nodal displacements and
spring forces
21One-Dimensional Spring System
- first formulate a general element
- each spring is identified by the element number
in the box - spring elements have a node at each end and
connect at a common node - consider a single spring element p with nodes i
and j
22One-Dimensional Spring System
- assume positive displacements ui at node i and uj
at node j - element has spring constant k so nodal forces
occur when we have displacements - define fip as the force acting on node i due to
the nodal displacements of element p - equilibrium gives
23One-Dimensional Spring System
In the matrix form
24One-Dimensional Spring System
25One-Dimensional Spring System
- Let the external forces be Fi , where i
represents the node, so that the equilibrium
equations are - at node 1
- at node 2
- at node 3
26One-Dimensional Spring System
- substitute the element equations into these
(structure assembly process) - the resulting equation is
- or
27- Solve the matrix equation
- if the applied forces are known, and the nodal
displacements are unknown, we have 3 simultaneous
equations to solve - the stiffness matrix above is singular so we do
not have a unique solution - this means the structure is at equilibrium at any
location in x space i.e. we have rigid body motion
28- a unique solution requires must have some part of
the structure grounded - i.e. we must apply some
boundary conditions such as a fixed displacement
of one of the nodes - If we apply an external force F at node 3 and we
attach the spring to ground at node 1, we can set
the displacement at node 1 to zero.
29One-Dimensional Spring System
- If the external force on node 2 is 0 then
- this effectively zeros the first column of the
stiffness matrix, and gives three equations in
two unknowns - If the value of the reaction force at node 1 is
unknown, we can also skip the first equation, and
consider only the last two - Now we can solve this for the unknown us
30Using a Computer Program
- 3 stages
- preprocessing
- processor
- postprocessing
31Using a Computer Program
32 Using a Computer Program
- Engineer
- responsible for interpreting results
- must ensure results are valid
33Analysis Step-by-Step
- procedure includes
- initial planning
- decide if FEA is needed
- doing the analysis
- presenting the results
34Analysis Step-by-Step
35Using a Computer Program
- preprocessing
- create model
- nodal point locations
- element selection
- nodal connectivities
- material properties
- displacement boundary conditions
- loads and load cases
- preprocessor assembles data into a format for
execution
36 Using a Computer Program
- processor
- code that solves the system equations
- generates element stiffness matrices
- stores data in files
- assembles the structure stiffness matrix
- must provide enough displacement boundary
conditions to prevent rigid body motion - solution gives nodal displacements
- with element information, get strain and stress
37 Using a Computer Program
- postprocessing
- numeric output data difficult to use
- reduces data to graphic displays (contour plots,
graphs) - magnifies nodal displacements
- nodal displacements are single valued
- stress at a node can be multivalued if multiple
elements are attached to the node - (stress is found from within each element)
38 Analysis Step-by-Step
- Start by having some conceptual design or some
existing design failure for analysis - define the analysis problem
- what type of analysis (static, dynamic, etc.)
- 2-d or 3-d solution
- criteria for analysis
- important variables (maximum stress, average
stress, strain, deformation, fracture load, yield
load, critical stress location, etc.) - most critical guide the modeling and presentation
of results
39 Analysis Step-by-Step
- time allowed for analysis determines whether an
adequate analysis can be done - tight time constraints may cause a sacrifice in
accuracy - analyses may only show qualitative aspects of
behavior - required solution accuracy helps estimate the
number of analysis cycles
40 Analysis Step-by-Step
- approximate engineering analysis
- determine the load conditions
- estimate the accuracy of loading
- get material property data and its statistical
variation - (may need to check for sensitivity to material
property variations)
41 Analysis Step-by-Step
- use approximate structure made up of similar
simpler structures with known solutions (beams,
cylinders, simple plates, etc.) - use approximate solutions to estimate value and
locations of critical variables - now decide if you need FEA
42 Analysis Step-by-Step
- if Yes
- develop conceptual model
- geometry
- element types
- plan mesh
- look for symmetries that can simplify model
- determine if a 2-d or 3-d solution is needed
- if 3-d, try a 2-d simplified analysis first
- try to lay out mesh based on expected solution
43 Analysis Step-by-Step
- choose a computer program to use
- many different programs available for use,
sometimes you have no choice - make sure the program has the correct element
types for the model - consider its mesh generation and pre- and
postprocessing capabilities
44Analysis Step-by-Step
- Prepare model
- detailed mesh plan that includes the degree of
refinement desired at critical locations - apply boundary conditions
- apply load and load cases
45 Analysis Step-by-Step
- Run analysis and check for errors
46 Analysis Step-by-Step
- Postprocessing
- look at deformed displacements and check for
consistency with expected results - look at stresses and compare to approximate
solution
47 Analysis Step-by-Step
- Refine model by considering the results of the
first analysis - high stress and rapid variations Þ reduce element
size - low stress Þ increase element size
- Redo analysis and check if results are converging
48(No Transcript)
49(No Transcript)
50 Analysis Step-by-Step
- Figure 1-7 is a refined model of 1-6
- note how the maximum stress has increased
- convergence has not yet been achieved
- Serious mistake if only one model is analyzed
- Figure 1-6 is in error by 23, while Figure 1-7
is in error by 19 - There is no guarantee that results will be
accurate
51 Analysis Step-by-Step
- Use results from the analysis to estimate the
converged solution - rate the analysis by estimating the accuracy
achieved - determine if important criteria identified
earlier are satisfied
52 Analysis Step-by-Step
- When finished, prepare a report which highlights
the significant results