COMPARATIVE ANALYSIS OF OPTIMIZATION METHODS IN THE INVESTIGATION OF A WEIGHMEASURING SYSTEM AND THERMOREGULATOR

Abstract

Context. For the first time, the use of taboo-search methods, random search, a swarm of particles for the construction of costeffectiveexperiment plans for the study of a weighing system and a temperature regulator was proposed.Objective – to carry out a comparative analysis of the developed optimization methods, such as taboo search, random search,particle swarm when searching for the optimal plans for the experiment during the study of the weighing system and thermostat.Method. Methods for constructing the experimentally optimal implementation matrix for the experiment using algorithms of aswarm of particles, taboo search and random search are proposed. In the beginning, the number of factors and cost of transitions foreach level of factors is introduced. Then, taking into account the input data, the initial experimental design matrix is formed. Whenusing the taboo search algorithm at each iteration step, the best solution in the neighborhood of the current solution is chosen as thenew current solution and the check is made whether it is in the taboo list. Thus, calculations occur until the algorithm reaches thespecified number of iterations. The list of taboos is formed from decisions that have a minimum cost. The random search method isbased on permuting the columns of the planning matrix. The number of iterations of the algorithm is specified by the user. Themethod of the particle swarm is based on modeling the behavior of the particle population. At each point where the particle visited,the value of the experiment is calculated. In this case, each particle remembers which (and where) the best value of the cost of theexperiment, she personally found and where the point is located, which is the best among all the points that explored the particles. Ateach iteration, the particles correct their velocity (modulus and direction). After a certain number of iterations, the particles are collectednear the best point. Then, among all the new points, we check whether we have found a new globally better point, and iffound, remember its coordinates and the value of the cost of conducting the experiment in it. Then the gain is calculated in comparisonwith the initial cost of the experiment.Results. The software that implements the proposed methods was developed, which was used to conduct computational experimentsto study the properties of these methods in the study of a weighing system and a temperature regulator. Optimized for the costof implementation of the experiment plans were synthesized, as well as the gains in optimization results as compared to the initialand maximum costs of the experiment.Conclusions. The conducted experiments confirmed the efficiency of the proposed methods and the software that implementsthem, and also allow them to be recommended for application in practice when constructing optimal experimental design matrices.