Classificação de falhas com o perfil mecânico de um analisador de eletrólitos utilizando redes neurais

Abstract

Preventing system failures is a challenge for researchers and engineering professionals. Several techniques for the loss minimization related to the system were developed, using the analysis of signals from the studied system, it’s possible to arrive at the discovery of unexpected wear and tear. Examples of such signals are vibration, acoustic pressure, temperature, particulates in lubricating oils, and others. Electromechanical systems present themselves as a challenge apart, with unexpected behaviours, requiring that their maintenance is based on the reliability of technological resources, robust and efficient to guarantee their efficiency and effectiveness. Using artificial neural networks (ANNs) for signal analysis, used in large amounts of data we can achieve an excellent degree of reliability by joining with statistics and computation in the optimization of dynamic processes. This work tries to apply and to demonstrate the effectiveness of the method proposed by Cabral (2017) in the detection and classification of failures in electromechanical systems, specifically an electrolyte analyzer, using artificial neural networks with the simulation of failures in a test bench.