Intelligent Health Monitoring of Machinery Using Multi-Channel Time Series Compressed Data and Deep Reconstruction Network

Nowadays, the safety and reliability of operating equipment are critical in various sectors such as wind energy, manufacturing, transport, and the chemical industry. Intelligent condition monitoring (CM) tracks the operating status of machinery, ensuring effectiveness, maintaining production quality, reducing unplanned breakdowns, and minimizing economic losses due to equipment failures. Rolling element bearings, widely used in almost all rotating machinery, contribute significantly to many machinery failures. However, the operation of rotating machinery spans from months to years, generating vast amounts of data from installation to failure. Moreover, it is common to collect multi-channel data such as vibration, acoustic, and temperature, etc., presenting a challenge in selecting the appropriate data source for optimal tracking degradation. This paper developed an intelligent approach, utilizing a novel deep learning model, named CLSTM, multi-channel and compressed data, for tracking the degradation process of bearing. CLSTM first extracts hidden features from the convolution neural network (CNN) and then uses long short term memory (LSTM) network to learn dependencies between samples. Experimental validation demonstrates that the developed method is effective and outperforms other statistics indicators and single-channel strategy.