Hybrid approach of optimization applied to an inverse problem in the dynamic modeling of a three-floor structure

Abstract

In the vast area of Mechanical Sciences, heuristic optimization techniques play a fundamental role in solving problems using the inverse approach. However, for methods where this solution arises in terms of population, depending on the complexity of the model, several evaluations of the objective function are required for each iteration, resulting in a very large processing time. Thus, in the context of this following work, we discuss the problem of identifying the input parameters of the dynamic model of a three-floor structure. Thus, a solution to the methodology is proposed combining the Genetic Algorithm, Firefly Colony and Differential Evolution techniques in a hybrid procedure, aiming to increase the performance, reduce the total number of evaluations of the objective function and obtain a quality solution. Three different levels of random noises were inserted simulating experimental errors in order to assess the capacity and the quality of the solutions provided by the procedural proposal compared to the results obtained from the same techniques used independently and from the Gravitational Search method.