Application of the Modified Genetic Algorithm for Optimization of Plasma Coatings Grinding Process

Abstract

The problem of defining optimal conditions for grinding plasma coatings may be considered as multi-objective optimization problem with a system of bounding inequalities that contain surface roughness, temperature, local and residual stresses as well as intrinsic defects size. This approach contributes to applying evolutionary algorithms such as genetic algorithm to solve the stated problem. Taking into account special characteristics of technological process, modification of the classical genetic algorithm has been carried out in the presented research. The combined method of selection based on the mitosis and meiosis operators makes it possible to increase fitness of a population ensuring its diversity during the following iterations. It is particularly important to maintain population diversity in genetic algorithm. The reason for that is preventing premature convergence which causes the obtained solution to be far from optimal. Another way to ensure population diversity is applying the developed mutation domain model that allows to alter random genes in chromosomes with the lowest value of the fitness function. The presented algorithm is based on both the combined method of selection and the mutation domain model. In order to compare the results of solving the problem of optimization of plasma coatings grinding process using modified genetic algorithm with other evolutionary algorithms, solutions performed by the classical genetic algorithm, ant colony optimization, particle swarm optimization and scatter search algorithm are presented. It was found that applying modified genetic algorithm provides high efficiency of solving process and reliability of the obtained results.