Performance of a novel hybrid algorithm in the design of water networks

Abstract

This paper proposes a novel hybrid evolutionary-based approach, named the modified cellular automaton network design algorithm–honey-bee mating optimisation (Mcanda–HBMO) algorithm, to the optimal design of water distribution systems. HBMO is a meta-heuristic approach for solving discrete and continuous optimisation problems, and Mcanda is a powerful local search algorithm that can help HBMO in better evolution. Two benchmark water distribution networks are considered in which the design variables are pipe diameters. Subsequently, the HBMO algorithm is applied to determine optimal discrete and continuous pipe sizes for the networks. Furthermore, to determine the effects of the roughness coefficient and minimum required nodal pressure on the HBMO algorithm performance, some additional information is presented to demonstrate those influences in the value of the final objective function as well as optimal pipe diameters. Moreover, in order to improve the performance of the search method, a combination of a local search algorithm with HBMO is tested in both examples. The hybrid algorithm consists of the modified Canda algorithm and pure HBMO which has had a better performance in both convergence rate and average objective functions found. The final results yielded by both HBMO and hybrid Mcanda–HBMO show the capability of these algorithms and also present some improvements in previously reported results.