City Research Online

Abstract

Considerable use of finite element methods is currently made in obtaining numerical solutions of problems in mathematics, science and engineering. The computation is often performed using a software package specially designed for a class of applications or by using a program, written in a general purpose high-level language such as Fortran, for a particular problem.

In an environment of research into finite element methods, it is often desirable to investigate different algorithms for solving a problem. To make such investigations one cannot usually modify any of the programs in a package and so a large number of Fortran programs have been developed for such problems. There is, however, an underlying framework common to finite element methods and a language such as Fortran does not specifically assist the programmer in the task.

A language, which is based on Algol 68, has been developed which facilitates the programming of a wide variety of finite element methods. A series of extensions to Algol 68, which allow a concise representation of finite element solution algorithms, is described in the thesis. Particular emphasis is given to iterative methods for non-linear problems, though linear and eigenvalue problems are also considered.

An implementation of the language features is reported. This utilises a preprocessor written in the general purpose macro processor ML/1 which converts programs written in the language into Algol 68.

Publication Type:	Thesis (Doctoral)
Subjects:	Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Departments:	School of Science & Technology > Department of Computer Science School of Science & Technology > School of Science & Technology Doctoral Theses Doctoral Theses

Export

Downloads