Flexible Self-Powered Triboelectric Nanogenerator Sensor With Machine Learning for Tightness Recognition of Axle Box Bolt

Any loosening of bolts may lead to a potential threat to the running stability of trains. Therefore, new sensors should be developed to the loosening status monitoring of bolts under the limited space and complicated environment for enhancing the operation safety of trains. Flexible self-powered vibration triboelectric nanogenerator (V-TENG) sensor based on conductive hydrogel is designed to the axle box bolt tightness monitoring without external power supply. The flexible materials are selected to V-TENG sensor for the adaptability to the curved surface of the axle box. By the optimal resistance, the V-TENG sensor, equipped with conductive hydrogel electrodes, achieves a maximum power output of 0.435 μW, which is enough to axle box bolt tightness monitoring. Self-powered V-TENG sensor produced an open circuit voltage of 9 V and current of 0.32 μA. The output voltage of V-TENG sensor has a linear correlation with the vibration frequency, which could be applied to tightening or loosening monitoring effectively. Deep Convolutional Neural Networks (DCNN) algorithms are adopted to the tightness state recognition of bolts with the self-powered V-TENG sensors efficiently with 90.38% success recognition accuracy. Self-powered V-TENG sensors have excellent performance in the tightness detection and early fault diagnosis of train axle box bolts, which also has great potential in the field of structural health monitoring (SHM) for trains, enhancing the safety and reliability of train operations.