Diversity-controlled adaptive stochastic learning based particle swarm optimization for multimodal emerging engineering problems

• Proposes a diverse-controlled adaptive PSO for multimodal optimization problems. • Uses stochastic learning with social-best and average global-best updates. • Adapts control parameters using Euclidean distance–based population diversity. • Achieves superior results on pressure vessel and welded beam design problems. • Demonstrates robustness on real-world engineering design benchmarks. This study introduces the diversity-controlled adaptive parameter control mechanism with stochastic learning-based position update policy in particle swarm optimization (PSO) algorithm to gain sustainable solutions for emerging multimodal single objective problems. The stochastic learning mechanism introduces better exploration and exploitation capabilities through the randomly chosen social best and average global best solutions for velocity update. A parameter adaptation strategy that controls three control parameters independently depending on the population’s Euclidean distance-based genotypic diversity measure, which the diversity of the solution, also helps in maintaining better exploration and exploitation trade-off. It results in the proposed method outperforming the other 8 PSO variations for 20 classical multimodal single objective benchmark functions on the basis of convergence rate and accuracy level for dimensions D = 10 , 30 , and 50. On the basis of total rank the proposed algorithm has gained an increase of 39.68%, 57.75%, and 39.71% over others to secure top rank for considered three dimensions. Also, a simulation study of the proposed algorithm is observed on D = 30 , and 50 for CEC 2017 single objective multimodal optimization benchmark problem and found that the proposed one is comparable and consistent with respect to the other PSO variants despite its competitive time complexity. In this case, the proposed algorithm has achieved 1 st and 4 th position in terms of mean results respectively. Finally, the competitive performance of the proposed algorithm on three engineering design problems (for pressure vessel design and welded beam design problems it stood first with an increase of performance 1.8% and 0.73% respectively and for tension-compression spring design marginal lack of performance of 0.88% with respect to 2nd best performance) shows how the proposed PSO variant is efficient to solve various similar types of real-world multimodal problems.