Selection of Genetic Algorithm Parameters for Optimization of Storm-Sewer Networks Using Taguchi Method

Authors : Tülin Çetin, Mehmet Ali Ilgın, Mehmet Ali Yurdusev

Pages : 1-10

View : 204 | Download : 157

Publication Date : 2024-04-29

Article Type : Research Paper

Abstract :Genetic algorithm is one of the most referred optimization techniques for the hydraulic optimization of stormwater and sewer system design. The genetic algorithm has different parameters affecting its performance such as population sizes, crossover methods, crossover rates, elitism rates, mutation rates and comparative values for the tournament selection algorithm. By the nature of the genetic algorithm, different values of these parameters should be tried to get the optimum values for them. But checking all possible values of these parameters that is a full factorial design is too much time consuming. Therefore, in this study, the Taguchi method is proposed for the first time in the literature to determine the most suitable parameter values of the genetic algorithm used for the hydraulic optimization of stormwater and sewage system design. The L16 Taguchi orthogonal array that involves 16 experiments was employed. Each experiment was replicated six times. Only 16x6=96 runs were carried out to determine the significance of the factors instead of running a full factorial design with 44x6=1536 runs. As a result, the cost value obtained with the Taguchi method is very close (0,15 %) to the cost value obtained with the full factorial design. Moreover, with the Taguchi method, the results are obtained in a shorter time (7 days instead of 4 months) with much fewer attempts.
Keywords : Yağmur suyu kanalizasyon şebekeleri, optimizasyon, genetik algoritma, taguchi

ORIGINAL ARTICLE URL