A REMAINING USEFUL LIFE PREDICTION METHOD OF ROTATING ELECTRICAL MACHINES BASED DEEP LEARNING TO SUPPORT MAINTENANCE DECISION MAKING

Abstract

Predicting the remaining useful life (RUL) of equipment plays an important role in developing maintenance methods based on equipment condition,
reducing system downtime, improving reliability and safety. This study proposes a model to predict the remaining lifetime of rotating electrical machines using
a deep learning approach. First, time series data are obtained from sensors attached to the electrical machine that measure variation of the vibration. This dataset
is used to train a convolutional neural network hybridized with long short-term memory (CNN-LSTM) architecture. Applying a model based on a hybrid of CNN
and LSTM has yielded superior results compared to traditional models. These results can be beneöcial for optimizing maintenance schedules and improving the
overall efficiency of systems consisting of multiple rotating electrical machines.