Health monitoring of triboelectric self-sensing bearings through deep learning

Abstract

Traditional sensing techniques suffer from additional installation space and experience complex transmission paths. Intelligent bearings based on triboelectric nanogenerators (TENGs) are attracting increasing attention for self-sensing working conditions, which is of great significance to realize in-situ measurements; however, its ability of monitoring health condition still remains a virgin to be investigated. Hence, we develop a novel monitoring means for bearings using TENG-based capacity, and the sensing principle of fault is analyzed. Then, four state-of-the-art deep learning models are evaluated to identify fault type and degree of the bearing and high identifying accuracy is achieved. Such marriage between TENG self-sensing and deep learning is shown to individually monitor health condition of the target bearing out of the coupling effect from other bearings in a multiple-bearing mechanical system. This work shows the potential of deep learning in enhancing the precision of TENG-based sensors, providing a self-sensing method to monitor the health condition.