Personal Computer Based Distributed Genetic Algorithm for Optimization Problems

Abstract

Genetic Afgorithms (GAs) have proven to be a well-suited technique for solving structural optimization problems. Due to large computational time requirement even with latest desktop computing power, however, it becomes difficult to opply this technique to large-scale complex problems. To speed up the computations for such real life problems, GAs have been implemented by researchers in parallel using supercomputers, transputers or network of distributed computing workstations. These machines are either not so easily accessible or too complicated to use for the want of theoretical background of porallel processing. This paper reports a study carried out for implementation of distributed genetic algorithm over a network of persona! computers using Window based WebDedip environment, which is developed using Java technology. As WebDedip eliminates the use of message passing functions for communication of data between computers, development of distributed applications becomes quite easy. This approach is illustrated here with an example of weight optimization of a truss. The observed speedup is compared with the ideal speedup for finding the efficiency of the proposed methodology.