Author Type

Graduate Student

Date of Award

1997

Document Type

Thesis

Publication Status

Version of Record

Submission Date

October 2026

Thesis/Dissertation Advisor [Chair]

Huang, Ming Z.

Abstract

Many complex engineering designs have conflicting requirements that must be compromised to effect a successful product. Traditionally, the engineering approach breaks up the complex problem into smaller sub-components in known areas of study. Tradeoffs occur between the conflicting requirements and a sub-optimal design results. A new computational approach based on the evolutionary processes observed in nature is explored in this dissertation. Evolutionary algorithms provide methods to solve complex engineering problems by optimizing the entire system, rather than sub-components of the system. Three standard forms of evolutionary algorithms have been developed: evolutionary programming, genetic algorithms and evolution strategies. Mathematical and algorithmic details are described for each of these methods. In this dissertation, four engineering problems are explored using evolutionary programming and genetic algorithms. Exploiting the inherently parallel nature of evolution, a parallel version of evolutionary programming is developed and implemented on the MasPar MP-1. This parallel version is compared to a serial version of the same algorithm in the solution of a trial set of unimodal and multi-modal functions. The parallel version had significantly improved performance over the serial version of evolutionary programming. An evolutionary programming algorithm is developed for the solution of electronic part placement problems with different assembly devices. The results are compared with previously published results for genetic algorithms and show that evolutionary programming can successfully solve this class of problem using fewer genetic operators. The finite element problem is cast into an optimization problem and an evolutionary programming algorithm is developed to solve 2-D truss problems. A comparison to LU-decomposition showed that evolutionary programming can solve these problems and that it has the capability to solve the more complex nonlinear problems. Finally, ordinary differential equations are discretized using finite difference representation and an objective function is formulated for the application of evolutionary programming and genetic algorithms. Evolutionary programming and genetic algorithms have the benefit of permitting over-constraining a problem to obtain a successful solution. In all of these engineering problems, evolutionary algorithms have been shown to offer a new solution method.

Extent

338 p.

Comments

College of Engineering and Computer Science|FAU Electronic Theses and Dissertations Collection|Thesis (Ph.D.)--Florida Atlantic University, 1997.|College of Engineering and Computer Science

Identifier

FADT12514

Format

pdf

Download

Share

COinS