A comparative study of the influence of codification on discrete optimum design of frame structures

Abstract

This paper is concerned with the influence of the codification type on the optimization of frame structures for constrained minimum weight, when using discrete variables: cross-section types corresponding to standard profiles. Three different codifications are compared: standard binary, standard reflected gray, and other binary codification with a greater number of different bits between consecutive integers. The focus of the work described is on determining how the codification selection influences the final performance of the algorithm, considering in addition two population sizes and two crossover types. A generational evolutionary algorithm with ranking selection and low generation gap is tested in a well known fifty-five bar sized frame structural test case. Results have been obtained from executing one hundred independent runs of each parameter/codification combination, which means twelve different algorithm outcomes. The gray code obtains the best results, and codification influence has a greater impact than the other considered factors (population size or crossover type). The experimentation performed in this single optimization work points out that the quality of the solutions is greatly affected by the codification, being the selection of gray codification of major importance in the algorithmic tuning of parameters in the evolutionary optimization of the structural optimum design problem. © Civil-Comp Press, 2013.