Comparison in the application of Multi-objective Genetic Algorithm, Particle Swarm Optimization and Firefly Algorithm in unbalance identification in rotating machinery.

Andrei Bavaresco Rezende¹; Cassio Pereira de Paula¹; Helio Fiori de Castro¹

¹ UNICAMP

doi:10.20906/CPS/COB-2015-0660

Resumo

Rotating machines operate important functions in the industries today. It is very useful to detect problems and correct them before failures event, preventing possible material damages and financial lost. Because of that, the study of rotordynamic models occupies a prominent position in the rotating machinery context due to the significant amount of phenomena that can occur during operation of such equipment. The main fault that occurs in rotating systems is unbalance. So it is interesting to identify the failure parameters from the minimization of the differences between the model and the experimental responses. Besides, as rotating system vibration should be monitoring different position and directions, a multi-objective problem is characterized, because the difference between model and experimental response in each monitoring degree of freedom is an objective function to be minimized. Because of that, meta-heuristics search methods are interesting tools for solving this problem. This work propose a comparison of the performance between the Multi-objective methods based on Genetic Algorithm, Firefly Algorithm and Particle Swarm Optimization. In the first method, the optimization methodology is based on evolutionary genetic algorithm and is intended to evolve a set uniformly distributed solutions belonging to the Pareto optimal set. The solutions are represented by individuals that belong to a population, and this population generates a new population through the evolution operators. The second method, the algorithm is based on the bioluminescent behavior of fireflies colonies. The method is organized considering that a firefly is a possible solution of the problem and moves randomly in the sample space, attracting or being attracted to other fireflies. Each firefly has a certain attraction for the intensity brightness issued for him, and this intensity is related to the objective function. And last method is the multi-objective Particle Swarm Optimization, which is a population based stochastic optimization te

Palavras-chave: Multi-objective Firefly Algoritm; Multi-objective Genetic Algorithm; Multi-objective Particle Swarm Optimization; rotating system; unbalance identification