You are on page 1of 2

1.

0 The Finite Element Method: Introduction


The finite element method (FEM) may be broadly defined as a numerical technique for
obtaining approximate solutions to differential equations. And since many of the mathe-
matical models employed by engineers contain differential equations this method is of in-
terest to engineers in a wide variety of disciplines. Of course there are many methods for
solving differential equations approximately. Some of the more familiar ones are: finite
difference method; boundary integral equation method; shooting methods; etc. No method
is perfect; and in a particular situation any one of the methods mentioned may be the most
effective. However the principal claim to fame of the FEM is its generality - and this has
allowed it to revolutionize the way we approach many problems in engineering today.

Although the FEM can be thought of as a topic in numerical analysis, i.e. mathematics, it
was developed by structural engineers interested in solving actual problems and not con-
cerned with solving differential equations. In fact early in its history it generality was not
recognized. It was a method for solving certain kinds of problems in structural mechanics.

The exact origins of the FEM are a little vague since many researchers were converging on
a similar approach about the same time. But there are certainly a few seminal works which
should be mentioned. Richard Courant, the famous mathematician, published a paper in
1943 on a proposed method for approximately solving the problem of a vibrating mem-
brane. The method converted the partial differential equation into a large set of algebraic
equations and since computers were not available at this time no solutions were given and
basically this approach ended here. The basic developmental work leading to the FEM in
structural engineering (actually aircraft engineering) is normally attributed to Kelsey and
Argyris in England and Germany in the mid-fifties (See their book on Energy Theorems.
Note. Sydney Kelsey emigrated to the USA and taught at Notre Dame in Civil Engineer-
ing from 1967 until his retirement in 1987. He is still at the University.) and in the USA in
a paper by Clough, Martin and Topp in 1956. Professor Clough of Berkeley shortly there-
after dubbed the technique the “Finite Element Method.”

The early work was a natural progression whereby the methods employed by structural
engineers to solve so-called discrete systems (trusses, frames, etc.) were extended to find
approximate solutions to continuum problems (plane elasticity) by a process of “discreti-
zation.;” that is by converting the continuous problem into a discrete problem, a problem
governed by a system of algebraic equations. Within a couple of years however the under-
lying mathematical techniques were uncovered and it was recognized that the new FEM
was actually an implementation of the proposed method of Courant in his 1943 paper. Ex-
tensions to other fields such as heat and mass transfer, fluid mechanics and electrostatics
soon followed. For some time most of the advances were made by researchers who “grew
up” in structural mechanics and courses could only be found in civil engineering depart-
ments and based on structural mechanics applications. However after ten years or so the
method was firmly rooted in many disciplines and today courses on the FEM can be found
at all universities in departments of civil engineering, mechanical engineering, electrical
engineering, chemical engineering and mathematics.

January 22, 1999 1


We may categorize our investigation of the FEM into two topics. We are concerned with
the ideas or notions of the FEM; how does it work, does it converge to the correct solution;
how is it applied to different kinds of problems; what are the different techniques within
the FEM. These are the conceptual underpinnings of the method. However there are also
the mechanics; how does an FEM program work; what is a good program architecture. We
will be concerned with both of these topics. We (and by we, of course, I mean you) will do
a fair amount of programming. At the end of this course you should be able to say the fol-
lowing.
1. I understand the notion of “weak” solution to a differential equation.
2. I can take any of the classic, linear partial differential equations (Laplace’s equation,
heat equation, wave equation, etc.) and discretize the weak form of the problem using
the FEM.
3. I can write a matlab program to solve the problems formulated in 2.
4. I am familiar with the basic element types in one and two spatial dimensions including
the isoparametric family of quadrilaterals.
5. I am familiar with basic time-marching methods for problems of parabolic and hyper-
bolic type.
6. I understand how to use the FEM to solve eigenvalue problems associated with the
classical problems mentioned in 2.
7. I understand the relationship of the FEM to the classical Rayleigh-Ritz and Galerkin
methods.
8. I have an introductory knowledge about applying the FEM to nonlinear problems.

The are many ways to begin our study. The approach we will use mimics the historical de-
velopment. We will examine program structure for solving so-called discrete problems.
Then we will study the FEM solution of several one-dimensional differential equations.
We will see that the codes we developed for discrete problems will, with little modifica-
tion, serve our needs for the continuous problems. We will then advance to two-dimen-
sional problems and look at different element types, time-marching methods etc.

January 22, 1999 2

You might also like