Implementation of Particle Swarm Algorithm to Sine-Cosine Optimization: Case Study of Plant-Wide Chemical Processes

Abstract

The sine-cosine algorithm (SCA) is a population-based optimization approach that relies on mathematical models incorporating sine and cosine functions to effectively obtain the global optimum. The SCA necessitates multiple initial solutions to fluctuate outward and toward the best solution. In this study, the particle swarm algorithm was replaced with randomization and selection of the best fit to enhance SCA operations, resulting in faster convergence and improved accuracy, thereby termed the modified SCA (MSCA). The proposed algorithm's performance was investigated and compared using 7 general problems, 19 common benchmarks, and 5 industrial-scale chemical processes. These benchmarks varied in complexity, with and without constraints. For comparative purposes regarding the algorithm's effectiveness, the results were averaged based on a maximum of 1,000 iterations, utilizing 500 populations from 5 replications. Subsequently, 5 common chemical processes were used to demonstrate the implementation using commercial simulation software (Aspen Plus), seamlessly integrated with both optimization algorithms to adjust operational conditions and design parameters. MSCA produced a lower error percentage and CPU time than SCA.