Aplicação de redes neurais artificiais na modelagem matemática do comportamento dinâmico de um isolador viscoelástico

Abstract

Competitiveness in the automotive market means the equilibrium between quality and cost. Most of the automobiles in the Brazilian market use passive engine mounts. With the advent of electronics acting directly on component performance, the use of passive engine mounts is decreasing gradually in developed countries. Although recent research are focused on the use of active engine mounts, there is still a vast field to be explored in the development and use of passive engine mounts, whether elastomeric or hydraulic. The objective of the present study is to analyze the dynamics of passive engine mounts and to obtain mathematical models able to represent the restoration and dissipation forces of the elastic engine mounts as well as the hydraulic engine mounts. The dynamic analysis was made through the construction of physical samples of elastomeric and hydraulic engine mounts, with different geometric configurations of the decoupler. These samples were submitted to experimental tests in an appropriate test machine in order to obtain the dynamic characteristics inherent to the configuration under study. From the experimental data artificial neural networks were used to treat the data and to obtain the mathematical models, due to their versatility and accuracy of results. Obtaining of the mathematical models seeks to contribute, through numerical simulations, to the design of engine mounts that fulfill all the design requirements from the beginning of its conception. The mathematical models were used in a quarter-car simulation to quantitatively verify the reduction of the vibration transmitted from the engine, considered as ideal source of vibration, to the chassis of the car, through the use of the engine mounts.