Eventos Anais de eventos
COBEM 2019
25th International Congress of Mechanical Engineering
Development of a Methodology for Prediction of Unbalance Using Vibration Analysis and Artificial Neural Networks
Submission Author:
bruna mariano , MG
Co-Authors:
bruna mariano, Sandro Izidoro, Glauber Zerbini Costal Costal
Presenter: bruna mariano
doi://10.26678/ABCM.COBEM2019.COB2019-0503
Abstract
Nowadays the high competitive scenario, predictive techniques for diagnosing the mechanical failure increased. These techniques, when correctly applied, they can be the solution to increase equipment availability, decrease the number of assets in the warehouse and the cost with maintenance. In this context, this paper studies vibration analysis with Artificial Neural Network (ANN) to help diagnosing unbalance motor at low speed of rotation. Thus, it was used an electric motor connect into frequency inverter and an unbalanced rotor, in which it was possible to simulate different level of unbalance. The algorithm purposed is capable of identify four different levels of unbalance, that are: not exist (balanced), low, medium, and high. A Multilayer Perceptron (MLP), a type of ANN, with fitness function as sigmoid and radial basis, was used in this classification. Inputs were: only radial, and radial and axial simultaneously. ANN that had inputs radial and axial at the same time got better accuracy than ANN that had only radial data. At the literature, it is said that for the unbalance failure is used just radial data, however, training network with addition of token data from axial direction brought accuracy to in terms of classification.
Keywords
artificial neural network, Multilayer Perceptron, Predictive maintenance, Vibration Analysis, Severity in Unbalance Failure
