Diversity-enhanced modified sine cosine algorithm and its application in solving engineering design problems

Abstract

The sine cosine algorithm (SCA) is a recently developed metaheuristic algorithm that is inspired by the characteristics of sine and cosine trigonometric functions. Although it has successfully solved several benchmark and real-life problems, it experiences the issues of slow convergence rate, premature convergence and local optima stagnation. To overcome these shortcomings, this paper introduces an improved version of the SCA named ISCA, in which four different search strategies namely, population division, modification of the original SCA search scheme, hybridization of the modified SCA with differential evolution (DE), and additional mutation phase are integrated. In the first stage, the population division is performed to pass candidate solutions through different levels of exploration and exploitation. The second stage modifies the original SCA to exploit the main features of the SCA related to exploring the elite area of the search space. In the third stage, the proposed modified SCA is hybridized with DE to maintain the diversity of the population. The fourth and last stage embeds the Cauchy mutation-based search scheme for failure candidates to prevent them from stagnation at local optima. To validate the performance of the ISCA, it has been tested on 30 standard benchmark problems including unimodal, multimodal and composite problems and its comparison is performed with the original SCA and other metaheuristic algorithms. Its performance on the scalability of optimization problems is also verified by increasing the dimensions of test problems from 30 to 1000. Various performance measures such as average and standard deviation of optimization results, statistical analysis, and convergence analysis conclude the better search efficiency of the ISCA. Furthermore, six engineering application problems are solved using ISCA to demonstrate its real-world applicability. The experimental results attest that the proposed ISCA is highly competitive with the other metaheuristics.