Abstract

Smart fault diagnosis of bearings is of great significance due to their extensive applications on various occasions. Recently, self-powered sensing technology based on triboelectric nanogenerators promotes the development of intelligent bearings. However, the effective detection and recognition of the rolling element defects of bearings need to be investigated further. This study proposes a triboelectric sensor-embedded rolling bearing (T-bearing) to monitor the working conditions and conduct the defect diagnosis of rolling balls. The interdigitated copper electrode covered by polytetrafluoroethylene film is attached to the inner surface of the outer ring of a commercial bearing. Such a design not only directly forms the TENG with rolling balls to obtain the contact-sensing signals, but also successfully achieves the diagnosis of rolling ball defects with similar triboelectric signals through a novel analysis and prediction paradigm combining signal decomposition and automated machine learning. Finally, a recognition accuracy of 99.48% with five different conditions of bearing balls is reached, which is extremely superior to the highest accuracy of 78.34% without signal decomposition. Thus, this study provides a new strategy for the defect diagnosis and the intelligent application of triboelectric bearings.