Abordagem de detecção de avarias em sistema dinâmico utilizando de técnica de inteligência artificial

Abstract

When analyzing an industrial environment, it is possible to observe that it is composed of the most diverse types of equipment, some of which are dynamic systems. It turns out that equipment in general tends to have faults or malfunctions with the time of operation, and this may prove to be unavailable, thus causing a reduction or break production. To minimize as much as possible those stops that add costs and no added value to the product, predictive maintenance techniques such as vibration analysis, thermography, oil analysis and others are used. The work in question aims to use Artificial Neural Network (ANN) techniques to analyze a signal from a dynamic system to detect what type of damage the system has. Later, this technique and Differential Evolution (ED) were used to determine the physical parameters of the system. Both studies were done computationally, representing a real system of two degrees of freedom, where after its validation, damage was added with the aim that later they could identify changes in the real. For the simulations, changes were made in the properties of mass and stiffness and the excitation in the system was performed by a sinusoidal force. Then, with experimental data, tests were made on the systems built to see their robustness. The results of the study proved to be promising so that one can follow up on-line and assist in decision making, mainly due to the results obtained experimentally. In addition, by analyzing the parameters, it is possible to have a better idea of the equipment's life cycle and management.