Comparative Study of PSO, GA, and ACO for Optimizing Dielectric Performance in Fly Ash Filled Silicone Rubber

Abstract

This study investigates the optimization of coal fly ash composition as a filler in Silicone Rubber (SiR) insulator materials, aiming to enhance their dielectric characteristics. Compositional optimization was achieved by evaluating and comparing three advanced meta-heuristic algorithms Particle Swarm Optimization (PSO), Genetic Algorithm (GA), and Ant Colony Optimization (ACO), using the Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE) as performance metrics. The utilized fly ash, containing dominant silica, alumina, and iron oxides, was directly incorporated into the SiR matrix. Results indicate that, compared to PSO, GA and ACO exhibited superior performance and consistency. Specifically, for Relative Permittivity, the optimal composition of 80% yielded the lowest errors with GA and ACO (RMSE = 0.0751; MAPE = 0.9044). For Hydrophobicity, these two algorithms showed superior accuracy in the RMSE metric (RMSE = 0.8883) at 15.39% loading. These findings underscore the scientific contribution of this study by establishing the superior reliability of GA and ACO for optimizing fly ash composition in SiR, thus providing a robust analytical methodology to advance the use of industrial waste for high-performance dielectric materials.