Development of a Hybrid Harmony Search for Water Distribution System Design

Abstract

Metaheuristic Optimization Algorithms (MOA) have been widely employed for optimal designs of Water Distribution System (WDS). Generally, the size, capacity, and location of WDS components (e.g., pipes, pumps, and tanks) are determined in the WDS design. Therefore, solutions with good fitness have engineeringly sound value of decision variables and share commonality. For example, a solution with transmission pipes bigger than distribution pipes is better with respect to fitness than that with the opposite case in the pipe sizing problem. However, few efforts have been made to consider good fitness rules in the optimization. In this study, we develop a hybrid harmony search (HyHS) algorithm that combines Harmony Search (HS) and a rule induction algorithm. The proposed HyHS algorithm's performance is compared with an improved GA, improved HS, and hybrid GA through optimizations of well-known benchmark functions and two WDS designs. The four algorithms are first applied to the optimization of the De Jong functions. Then, the least-cost and robustness-based design problems of New York tunnel system are solved using the four algorithms. The application results confirm that the proposed HyHS outperforms the other algorithms in terms of computational speed and effectiveness.