Fault Detection for Displacement Sensors in Vehicle-Mounted Flywheel Battery Based on Dynamic Self-Adjusting LMS Optimized by Deep Learning

To address the safety issue that the displacement sensors of the vehicle-mounted flywheel battery may mechanically loosen due to strong vibration in complex road conditions. A dynamic self-adjusting LMS algorithm based on deep learning optimization for detecting faults in the displacement sensor of vehicle-mounted flywheel batteries is proposed in this paper. Firstly, a self-adjusting fault detection method based on the least mean square (LMS) algorithm is proposed. This method can accurately detect sensor faults and identify faulty sensors. Secondly, to further improve the accuracy and adaptive performance of the detection, deep learning technology is adopted to extract and preprocess the sensor signal features, and a cascaded detection model of LMS-CNN-LSTM is built. By establishing a dual-threshold dynamic adjustment mechanism, the fault detection of displacement sensors under continuous disturbances can achieve balanced convergence and significantly improve robustness. Finally, the detection method is quantitatively and qualitatively analysed through an experimental platform and compared with other improved LMS algorithms. The application of this method reduces the fault detection response time by 48% and increases the accuracy of dual-fault identification by 61%. The results show that this detection strategy has higher accuracy, robustness, and fast convergence characteristics under time-varying disturbances.