ЗАСТОСУВАННЯ МЕТОДУ ГРАВІТАЦІЙНОГО ПОШУКУ ДЛЯ МІНІМІЗАЦІЇ ВАРТОСТІ ПРОВЕДЕННЯ БАГАТОФАКТОРНОГО ЕКСПЕРИМЕНТУ

Abstract

Experimental research methods are increasingly used in industry in the optimization of production processes. Experiments, as a rule, are multifactorial and are connected with optimization of quality of materials, search of optimum conditions of carrying out technological processes, development of the most rational designs of the equipment, etc. The use of experimental planning makes the behavior of the experimenter purposeful and organized, significantly increases productivity and reliability of the results. An important advantage is its versatility, suitability in the vast majority of research areas. When implementing an industrial experiment, the main task is to obtain the maximum amount of useful information about the influence of individual factors of the production process, provided that the minimum number of expensive observations in the shortest period of time. Therefore, it is important to increase the efficiency of experimental research with minimal time and cost. For this purpose, it is expedient to develop systems of automation of experiments which will allow to reduce terms of carrying out experimental researches and to reduce expenses for them. Object of research: processes of optimization of plans of multifactor experiment on cost and time expenses. Subject of research: an optimization method developed on the basis of the gravitational search algorithm, which consists in comparing the force of gravity (cost) of the first row of the planning matrix of the experiment to the next rows of the matrix. In the study of photoelectric transducers of angular displacements, the efficiency and effectiveness of the gravitational search method were analyzed in comparison with previously developed methods: analysis of line permutations, particle swarm, taboo search. The cost of carrying out the experiment plan and the efficiency for solving optimization problems in comparison with the original plan and the implementation of the above methods are shown.