THE EVOLUTIONARY METHOD BASED ON PARTICLE SWARM OPTIMIZATION AND ARTIFICIAL IMMUNE SYSTEMS MODELLING

Abstract

An optimization method, which embodies the approaches of particle swarm methods and artificial immune system modeling, for a random multidimensional continuous function is proposed. Research was conducted to determine the optimal settings of the algorithm that implements the proposed method. The proposed method can be applied as a component of information technology to support management decisions and solve technological processes optimizing problems in metallurgical production. The purpose of the work is to develop an original hybrid method of global optimization based on the artificial immune system method and the particle swarm method, as well as to formulate recommendations for setting its parameters. The proposed algorithm should speed up decision-making in the management of technological processes of metallurgical production and increase their accuracy. The methodology of providing a solution consists in the combination of basic search operators of evolutionary optimization methods based on the particle swarms and the human artificial immune system modeling. Traditional swarm search steps of finding a solution and exchanging information about found local optima are used. The search method is supplemented by the principle of competition, borrowed from the method of the artificial immune system, for which the population is divided into smaller swarms or search teams. Steps to control the diversity in the swarm and the particle scattering radius are also proposed. This hybridization allows solving both unconditional and conditional optimization problems in a continuous high-dimensional space. The scientific novelty of the results obtained in the work is that a new evolutionary optimization method in a continuous space based on swarm intelligence is proposed. This method, in contrast to the previously known swarm optimization methods, uses the principle of swarm subgroup competition and the population compression operator, which are characteristic of the method of modeling artificial immune systems. Also, the parameters of the method that maximize its effectiveness are empirically determined. Conclusions. The application of the proposed evolutionary search optimization method to the minimization of test functions in a continuous space with up to 20 dimensions showed its effectiveness in comparison with previously known ones. The application of the described algorithm for solving the optimization problems of technological processes in metallurgical production is considered relevant in further. This method is considered effective for smelting charging task, for using of ferroalloys in foundry production optimization, as well as for predicting the mechanical properties of finished melting products.