Noncrossover dither creeping mutation-based genetic algorithm for pipe network optimization

Abstract

A non-crossover dither creeping mutation-based genetic algorithm (CMBGA) for pipe network optimization has been developed and is analyzed. This CMBGA differs from the classic GA optimization in that it does not utilize the crossover operator, but instead it only uses selection and a proposed dither creeping mutation operator. The creeping mutation rate in the proposed dither creeping mutation operator is randomly generated in a range throughout a GA run rather than being set to a fixed value. In addition, the dither mutation rate is applied at an individual chromosome level rather than at the generation level. The dither creeping mutation probability is set to take values from a small range that is centered about 1/ND (where ND=number of decision variables of the optimization problem being considered). This is motivated by the fact that a mutation probability of approximately 1/ND has been previously demonstrated to be an effective value and is commonly used for the GA. Two case studies are used to investigate the effectiveness of the proposed CMBGA. An objective of this paper is to compare the performance of the proposed CMBGA with four other GA variants, and other published results. The results show that the proposed CMBGA exhibits considerable improvement over the considered GA variants, and comparable performance with respect to other previously published results. A big advantage of CMBGA is its simplicity and that it requires the tuning of fewer parameters compared with other GA variants.