Utilizing fuzzy logic, particle swarm optimization, desirability function and NSGA-II for multi-objective optimization of sponge Gourd–Bagasse polymer composite properties

Abstract

Listen

Translate article

The process of developing mathematical models to forecast material properties can be intricate and time-consuming. Consequently, there has been a shift towards simpler and more precise computational models founded on artificial intelligence. In this particular research, fuzzy logic (FL) was employed to anticipate the mechanical properties of a newly formulated hybrid composite material, which combines sponge gourd, bagasse, and epoxy resin for golf club applications. Furthermore, a multi-objective optimization of these properties was conducted using a modified desirability function (DF) and the Non-Dominated Sorting Genetic Algorithm (NSGA-II). The FL model took into account three input factors: the weight percentage of bagasse, the weight percentage of sponge gourd, and the fiber size measured in micrometers. The response variables examined were tensile strength, hardness, flexural strength, modulus, elongation, and impact strength. The FL model was combined with the modified DF algorithm and the NSGA-II algorithm, respectively. To optimize the DF, the particle swarm optimization (PSO) algorithm was employed. The results demonstrated that the FL model accurately predicted the mechanical properties of the hybrid composite material. The lowest correlation coefficient (R) between experimental responses and FL predictions was found to be 0.9529. The modified algorithms addressed specific characteristics in the desirability properties, such as elongation, where the desirability remained constant within a range. When the modified algorithm was utilized, the optimized properties exhibited a desirability value of 0.84, surpassing the values of 0.83 and 0.82 obtained without modifying the optimization algorithms. In conclusion, the FL model in conjunction with the modified DF algorithm and the NSGA-II algorithm successfully predicted the mechanical properties of the hybrid composite material. The modifications made to the algorithms contributed to enhancing the optimization process of these properties.