A hybrid optimization algorithm of differential evolution and sine cosine algorithm

Abstract

To address the inherent limitations of the sine cosine algorithm (SCA), including slow convergence, limited optimization accuracy, and a tendency to become trapped in local optima, a hybrid optimization algorithm of differential evolution and sine cosine algorithm (DESCA) is proposed. Instead of simply combining different operators, DESCA is designed as a coordinated two-phase search framework with complementary functional roles. First, identical parameter settings for r2, r3, and r4 are adopted for all dimensions of the same individual to reduce randomness and improve search efficiency. Second, the spiral update strategy from the whale optimization algorithm (WOA) is introduced as a local refinement mechanism to further improve the solutions generated by SCA and strengthen local exploitation. Third, within the two-phase framework, SCA primarily performs global exploration in the early stage, while DE with the DE/best/1 mutation strategy is activated in the later stage to improve convergence accuracy and local exploitation capability. The performance of DESCA is comprehensively evaluated on 23 classical benchmark functions, the IEEE CEC2014 benchmark functions, and the IEEE CEC2020 benchmark functions. Experimental results show that DESCA achieves strong optimization performance on the 23 classical benchmark functions and the CEC2014 benchmark functions, while maintaining competitive overall performance on the CEC2020 benchmark functions. In addition, Wilcoxon rank-sum tests and Friedman tests are conducted to further evaluate statistical significance and overall ranking performance. To verify its practical applicability, DESCA is further applied to four mechanical design optimization problems. The results demonstrate that DESCA is an effective and competitive optimization method with favorable convergence behavior, high-quality solutions, and strong engineering applicability.