Basic Concepts of FEM

Ali Karrech
School of Civil Environmental, and Mining
Engineering

Learning resources: Chapter 1 + Lecture notes

+ references mentioned in the slides
Outline (what are we expected to learn)

Where this course fit in engineering sciences

What is FEM? (layman definition)
The process of FE Modeling and Simulation
Engineer vs computer (safety, responsibility, verification and validation)

2
Where this course fits

Engineering sciences can be subdivided into three areas of competence

Theoretical
Applied
Computational
Nano and micromechanics
Continuum mechanics Linear
Solids and structures
Multi-physics Nonlinear
Fluids
Systems

Computational multi-physics of solids and

3
structures in the linear regime
Where this course fits
Computational multi-physics of solids and structures in the linear regime
Finite Element Method
Finite Volume
Displacement
Finite Difference Method Formulation Equilibrium
of FEM
Boundary Element Method Mixed
Models
Hybrid
Spectral Method
Mesh-Free Method
Solution of Stiffness
FEM Flexibility
Models Mixed

4
Where this course fits - Course Coverage
This course consists of 4 components
Basic concepts
Finite Elements for Structures
Finite Element for heat transfer / poromechanics
Introduction to the finite volume method

5
What is FEM?
layman definition
A method that solves field problems numerically.
Description:
Subdivides a domain into several elements (pieces)
Reconnects the elements through nodes
Transforms the field variables into degrees of freedom (DOF)
attached to the nodes via algebraic equations
Produces an approximate solution to a field problem.

6
What is FEM?
approximation T T

Field variables can be Exact Analytical

Approximate
Piecewise Linear
approximated by a discrete Solution Solution

model composed piecewise

continuous functions
defined over a finite number x x
of subdomains.
examples: distance,
surface area,
volume,
stress,
displacement,
temperature,
pressure, etc.

7
What is FEM?
Advantages
Handles bodies of complex shapes;
Takes into account general loading/boundary conditions;
Delivers approximate solutions to problems that are otherwise
difficult to solve analytically;
Can produce increasingly accurate solution if element sizes
are varied;
Can tackle dynamic/static problems;

8
 The process of FE Modeling and Simulation

9
Source: V. BONGALE
 The process of FE Modeling and Simulation

10
Source Introduction to Finite Element Methods (ASEN 5007), University of Colorado at Boulder Carlos Felippa:
The process of FE Modeling and Simulation
discretisation examples

Three-
One- Two-Dimensional Dimensional
Dimensional Triangular Brick Elements
Frame Elements Elements
The process of FE Modeling and Simulation
discretisation examples

Copied from Development and Validation of a

Finite Element Simulation of a Vertical Drop
Test of an ATR 42 Regional Transport Airplane,
2008, DOT/FAA/AR-08/19, U.S. Department of
Transportation, Federal Aviation Administration
http://www.tc.faa.gov/its/worldpac/techrpt/
ar0819.pdf
Total Human Model for Safety THUMS
by Toyota Central R&D Labs. Adapted from
http://www.tytlabs.co.jp/tech/thums/thums04.html

12
The process of FE Modeling and Simulation
element dimension and approx. order
Approximation order
Space dimension

13
Source: modified from Introduction to Finite Element Methods (ASEN 5007), University of Colorado at Boulder Carlos Felippa:
The process of FE Modeling and Simulation
measures of performance
Accuracy
Error = |(Exact Solution)-(FEM Solution)|

(Discretization with 228

Convergence Elements)

Limit of error as
Number of elements (h-convergence) or
Approximation order (p-convergence)
increase
Ideally, error tends to 0 as number of elements or (Discretization with 912
Elements)
approximation order tend to infinity
Flow chart of problem solving
using FEM

15
Based on Finite Element Procedures by Bathe (1996)
Engineer Vs Computer, safety issues, responsibility

Model is created/formulated/chosen by an Analyst/You (not a commercial finite

element code)
You judge whether the mathematical model has been solved to a sufficient
accuracy and whether the chosen model was appropriate (reliable) for a
question asked

The mathematical model must depend on the analysis question asked

FEM in Design
Design Engineer is responsible for Ensuring product safety (product liability
issue)

10
Engineer Vs Computer, safety issues, responsibility

Code of Hammurabi, 229 (Babylon, ca 2000 BC):

If a builder builds a house for someone, and does not construct it properly,
and the house which he built falls and kills its owner, then that builder shall be
put to death.

From R.F. Harper (1904) The Code of Hammurabi. King of the Babylonia,
The University of Chicago Press; downloaded from Cornell University Library

17
 Geometry Computational Grid Solution and
(finite element mesh)+ Results
Boundary Conditions and
Load+
Constitutive Properties and
EXAMPLE: CONNECTING LUG
Constitutive Models EXAMPLE: CONNECTING LUG

EXAMPLE: CONNECTING LUG S, Mises

Finite
(Avg: 75%)
+4.262e+08

Meshing
+3.910e+08
+3.558e+08

Element
+3.206e+08
+2.854e+08
EXAMPLE: CONNECTING +2.502e+08LUG
+2.150e+08
+1.797e+08
Y
X ? Algorithms +1.445e+08
+1.093e+08
+7.410e+07
+3.888e+07
Z 2. Select the surface associated with the bottom half of the hole using the cursor; the region is
+3.666e+06
highlighted in Figure 420. When the appropriate surface is selected, click Done in the prompt
area. EXAMPLE: CONNECTING LUG

1 3 1 S, Mises
3 (Avg: 75%)
2 +4.262e+08
+3.910e+08
Figure 418 Built-in end of the connecting lug. Figure 421 Suggested mesh of C3D20R elements for the connecting lug model. +3.558e+08
1 +3.206e+08
3 Step: LugLoad, Apply uniform pressure to the hole
+2.854e+08
Increment 1: Step Time =
+2.502e+08 1.000
and drag the cursor over the virtual trackball in the viewport. The view rotates Primary Var: S, Mises+2.150e+08
Y which is what you were asked to determine. The inuence
ely; try dragging the cursor inside and outsidein thethe lug trackball
under the applied loads, of Var: U Deformation
Deformed Scale Factor: +2.964e+01
virtual to see the +1.797e+08
X simulation is discussed in Mesh convergence, Section 4.4. +1.445e+08
increasing
e in behavior. Click mouse button 2 to exit the rotate the mesh
view tool before density usedZin this
proceeding. +1.093e+08
e left end of the lug (indicated in Figure 418) using the Abaqus/CAE Figure
offersin a variety of meshing techniques to mesh models of different topologies.
cursor. Click Done 431 Filled contour plot of Mises stress.
+7.410e+07
+3.888e+07

Fixed
pt area when the appropriate region is highlighted in the viewport, and toggle on +3.666e+06
The different meshing techniques provide varying levels of automation and user control. The
TRE in the Edit Boundary Condition dialog box that appears. Click OK to apply range between these values into 12 intervals. You can control the minimum and ma
following three types of mesh generation techniques are available:
2
dary condition.
appear on the face indicating the constrained degrees of freedom. The encastre3
Structured meshing
Pressure 1
Abaqus/CAE displays (for example, to examine
as the number of intervals.
variations within a xed set of bo
condition constrains all active structural degrees of freedom in the region specied;
part is meshed and the job is created, this constraint will Structured
be applied to allmeshing
the nodes applies preestablished mesh patterns to particular model topologies.
Complex models must generally be partitioned into simpler regions toTo
18 usegenerate a customized contour plot:
All figures copied from Abaqus 6.14 Getting Started with Abaqus: Interactive Edition Options
py the region.
arries a pressure of 50 MPa distributed around the bottom half of the hole. To applyFigure
the by dialog
420 Surface to which pressure will be applied.
box, drag
this technique.
1. In the Basic tabbed page of the Contour Plot
Dassault System
however, the part must rst be partitioned (i.e., divided)
s: a top half and a bottom half.
Swept so that the hole is composed
meshing to nine. Intervals slider to change the2 number of intervals
3. Specify a uniform pressure of 5.0E7 in the Edit Load dialog box, accept the default
Swept meshing extrudes an internally generated mesh along a sweep path or revolves it around
Many days of tedious work

Source =: Joldes et al. (2009), MICCAI 2009, Part II, LNCS 5762, pp. 300-307
How do we know that solutions we obtain using FEM/FVM are accurate?
Validation and Verification in computational engineering (V&V)

Read
Oden, J. T., Belytscho, T., Babuska, I., Hughes, T.J.R. (2003)
"Research directions in computational mechanics", Computer
Methods in Applied Mechanics and Engineering, vol. 192, pp. 913-922
(available through CMO)

20
 Basics (5): Verification
To assess the difference between results produced by the computational model
(discretisation of the mathematical model) and the mathematical model (according to
Oden et al. 2003)
Code Verification
The code may not be accurate implementation of the discretised model
Solution Verification
The discretised model may not be an accurate representation of the
mathematical model.

21
 Engineer Vs Computer, safety issues, responsibility
Verification
Against benchmark problems (specific, simplified model problems for which accurate
solutions or analytical solutions are known).

22
Joldes et al. (2009) Medical Image Analysis, 13: 912919
Engineer Vs Computer, safety issues, responsibility
Validation

The process of determining if a mathematical model of a physical event represents the

actual physical event with sufficient accuracy.
(according to Oden et al. 2003)

Figures copied from The National Crash Analysis

Center of The George Washington University website
http://www.ncac.gwu.edu/research/infrastructure.html.
The model shown in the figure has been developed by
The National Crash Analysis Center (NCAC) of The
George Washington University under a contract with
the FHWA and NHTSA of the US DOT.

23