Enhanced Beluga Whale Optimization Algorithm for Handling Global Optimization and Engineering Applications

The swarm-based beluga whale optimization (BWO) algorithm draws its inspiration from beluga whale behaviors, including swimming, seeking to catch food, and falling into the water to drown. BWO’s weakness is that its exploration and exploitation phases contain certain random parameters, which introduces unpredictability of the algorithm’s performance. Randomness enables the algorithm to investigate more areas during exploration and to have a slower convergence during exploitation. In this study, we introduce three enhancements of the original BWO algorithm. The enhancements of BWO (EBWO) are expressed by EBWO1, EBWO2, and EBWO3, which are based on the two parameters, the reduction parameter and the touring parameter. The modifications of reduction parameter make the rapid convergence rate happen during the exploitation phase, which is utilized to restrict the way in which search agents move inside the area of search. On the other hand, modifications of the touring factor aid search agents in locating areas with the best outcomes in search, conducting worldwide searches, and improving algorithm exploration skills. 24 benchmark functions, CEC2019 functions, and many problems with engineering design are used to evaluate the proposed algorithms. From results of the 24 benchmark functions, EBWO3 obtained the best solutions in 18 functions with a percentage of 75%, while the original BWO obtained the best solutions in 6 functions only with a percentage of 25%, which indicates the importance of the proposed modifications to the original BWO. While results of CEC2019 benchmarks problems show that EBWO1 and EBWO2 perform better than the regular BWO in all 10 tasks, indicating greater global search capabilities. Finally, to manufacture certain products with particular chemical qualities and to meet the goal costs, the proposed methodologies are used to solve the petrochemical engineering application of blending four ingredients, three feed streams, one pool, and two products. Results have shown that the suggested approaches are preferable for locating the best solution overall. Finally, the findings of the EBWO Algorithms were then contrasted with those from past studies, and Friedman and Wilcoxon’s tests were used in statistical analysis to show the system’s efficiency and capability in dealing with such problems.