Abstract

In the era of artificial intelligence (AI) and digitization, developing self-monitoring and smart-diagnosis bearings has become a meaningful yet challenging problem. This study investigates an AI-enabled bearing-structural rolling triboelectric nanogenerator (B-TENG), which can achieve condition monitoring and fault diagnosis for bearing wear. The geometrical structure of B-TENG is designed to directly use rolling balls as the freestanding layer. Besides, the sensing principle of triboelectric signal waveforms and the mapping mechanism of wear faults are firstly revealed through a signal decomposition method. Furthermore, a deep learning algorithm can classify different wear types, degrees and positions on rolling balls, with higher accuracies of 95.20∼98.40 % for the feature components. The detection of wear degree related to bearing health and failure evolution is realized for the first time. The proposed B-TENG has the potential for digital twin application via interaction with professional simulation software according to the real-time diagnosis classified by AI.