Experimental Study of Convergence and Stability of a Particle Swarm Optimization Algorithm: Application to the Vessel Design Optimization Problem

Particle Swarm Optimization (PSO) is a metaheuristic optimization algorithm inspired by collective behaviors in nature. This article examines the performance of PSO by considering three methods for adapting the inertia weight: the constriction method, the Random Inertia Weight Method (RIWM), and the Linearly Decreasing Inertia Weight Method (LDIWM). The study addresses a complex optimization problem due to its constraints, specifically focusing on optimizing the manufacturing cost of a pressure vessel. The performance of PSO is measured in terms of convergence and stability. In this way, it is determined which of the three methods achieves greater precision and how often this precision level can be consistently reached. The results demonstrate that the inertia weight is a hyperparameter that significantly impacts the convergence of the PSO algorithm. Therefore, for a given problem, a thorough analysis must be conducted to achieve optimal results.