Using Deep Learning Algorithms (CNN and LTSM) to Detect Faults in Electric Motors

The detection of faults in electric motors is vital to the maintenance of reliability, safety, as well as minimal maintenance in the industrial environment. Conventional diagnostics methods are based on signal processing and domain experience, and are perhaps unable to cope with complicated and non-stationary data. Over the past few years, the deep learning algorithms have shown better results in automatic feature extraction and classification. In this paper, the author has conducted an in-depth research into the evaluation of Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (LTSMs) in detecting the faults of electric motors. As input data, motor condition monitoring signals are used, including the vibration and current signals. CNNs can be utilized to obtain spatial information on the time-frequency representations whereas LTSMs and especially Long Short-Term Memory (LSTM) networks are applied to capture temporal information on the serial data. The experimental findings provide that deep learning-based techniques are more effective at detecting faults than the traditional ones and can therefore be used in real-time motor health monitoring systems.