ABSTRACT
Genetic algorithms are a global optimization method that is ideally suited for implementation on parallel computers. In this article, a genetic algorithm developed to solve problems in heat conduction is parallelized and run on a parallel computer (IBM SP2). The algorithm presented here contains a novel local search operator that greatly improves its accuracy. The algorithm itself and the parallelization strategy are discussed in the first part of the article, followed by a detailed discussion of the algorithm's results and its performance vis-à-vis its serial version.
The authors wish to thank the UC Berkeley Block Grant program for facilitating the necessary computing resources for this project, as well as the staff at the San Diego Supercomputing Center for their valuable assistance during the course of this work.