Predictive maintenance of production equipment based on neural network autoregression and ARIMA

This paper presents a predictive study applied to a manufacturing equipment in order to predict malfunctions, and consequently enabling predictive maintenance practices. ARIMA forecasting methods are successfully compared with neural networks models, both used over data obtained from a monitoring system that continuously keeps track of the relevant equipment parameters. The results show that both models could detect the discs replacement, however The ARIMA model forecasts quite well the increasing of the distance between the discs before and after the replacement which is not the case for the NN model.