Avaliação de métodos numéricos aplicada à identificação de parâmetros e de perturbações externas em sistemas mecânicos.

Abstract

n this work, the Central Difference, Houbolt, Wilson θ and Newmark types of numeric integration methods are applied in the simulation process of mechanical systems and the results are used in a identification process, which is used to verify and compare their performance with the 4th order Runge Kutta method. Those methods have been applied in the simulation of the physical behavior of a Single Degree of Freedom (SDOF) system, which is represented by a mass-spring-damper system, and a Multi Degree of Freedom (MDOF) system, which is represented by the two degree of freedom axis-plain cylindrical journal bearing system. To better analyze the performance of the simulation method, during the search of the best solution for the problem, some internal variables were investigated and analyzed, such as: the integration time-step, the number of points, the choice of the excitment signal, and so on. On the other hand, two types of formulations for the identification process were used: a discrete and a continuous version of the state-space approach. The attained results have shown that the simulation accuracy and performance of the Newmark, Central Difference, Houbolt and Wilson θ methods, when comparing the identified parameters and excitment forces with the ones of the model, are dependent upon the choice of the integration time-step, dynamical properties of the system and numerical stability of the method.