Metaheuristic Algorithms in Industrial Process Optimisation: Performance, Comparison and Recommendations

Abstract

The process parameters design is one of the most demanding tasks in a modern industry, aiming to find optimal process parameters that meet strict requirements for process responses. Various methods have been employed to tackle this problem, including metaheuristic algorithms. The objective of this paper is twofold. Firstly, it presents a review analysis and comparison of metaheuristics’ performance in optimising industrial processes as evidenced from the literature, with a particular focus on the most commonly used algorithms: genetic algorithm (GA), simulated annealing (SA), and particle swarm optimisation (PSO). Secondly, an intelligent method for parametric multiresponse process design, based on the soft computing techniques, is presented. GA, SA and PSO are used as an optimisation tool, and comparison of their results in real case studies is performed including two criteria: (i) accuracy of an obtained optimum; (ii) a number of objective function evaluations needed to reach an optimum. Tuning of the algorithms’ own parameters is also discussed, which is especially interesting due to a different nature of three algorithms: a single point-based algorithm (SA), a population-based algorithm (GA), and a population-based algorithm with swarm intelligence (PSO). The algorithms’ robustness, i.e. sensitivity in respect to the algorithm-specific parameters tuning is studied and introduced as the third criterion in comparing the algorithms’ performances. The concluding remarks are drawn from this analysis, followed by the recommendations for an efficient metaheuristics’ application in optimising industrial processes.