AI-driven fault diagnosis in smart bearings using mosaic-patterned electret for multi-modal motion sensing

Smart bearings offer a promising approach for real-time motion and condition monitoring in rotating machinery by embedding sensing functionality directly into the bearing structure. However, most existing smart bearing designs remain constrained to single-modal motion sensing, typically capturing only rotational or vibrational motion. In this paper, a novel smart thrust ball bearing (STBB) is proposed, incorporating a mosaic-patterned polarized electret for enhanced multi-modal motion sensing. Charges are injected into the electret film in a patterned manner using silicon shadow masks fabricated via deep reactive ion etching (DRIE). The coupled signal from simultaneous rotational and vibrational motions is fully decoupled through the dual Fourier transform (DFT) algorithm. The results show that the STBB exhibits high linearity, with R² = 0.9999 and R² = 0.9998 for rotation and vibration, respectively, and a rotational sensitivity of 7.48 rpm·Hz⁻¹. The STBB is further integrated into a six-joint robotic arm for dynamic monitoring. Combined with a long short-term memory (LSTM) neural network, the STBB achieves 90.33% fault classification accuracy. These findings highlight its potential as a scalable and multi-modal sensing solution for intelligent maintenance systems.