Hybrid harmony search algorithm for continuous optimization problems

Harmony Search (HS) algorithm has been extensively adopted in the literature to address optimization problems in many different fields, such as industrial design, civil engineering, electrical and mechanical engineering problems. In order to ensure its search performance, HS requires extensive tuning of its four parameters control namely harmony memory size (HMS), harmony memory consideration rate (HMCR), pitch adjustment rate (PAR), and bandwidth (BW). However, tuning process is often cumbersome and is problem dependent. Furthermore, there is no one size fits all problems. Additionally, despite many useful works, HS and its variant still suffer from weak exploitation which can lead to poor convergence problem. Addressing these aforementioned issues, this thesis proposes to augment HS with adaptive tuning using Grey Wolf Optimizer (GWO). Meanwhile, to enhance its exploitation, this thesis also proposes to adopt a new variant of the opposition-based learning technique (OBL). Taken together, the proposed hybrid algorithm, called IHS-GWO, aims to address continuous optimization problems. The IHS-GWO is evaluated using two standard benchmarking sets and two real-world optimization problems. The first benchmarking set consists of 24 classical benchmark unimodal and multimodal functions whilst the second benchmark set contains 30 state-of-the-art benchmark functions from the Congress on Evolutionary Computation (CEC). The two real-world optimization problems involved the three-bar truss and spring design. Statistical analysis using Wilcoxon rank-sum and Friedman of IHS-GWO’s results with recent HS variants and other metaheuristic demonstrate superior performance