Indirect Dynamic Stress Measurement of Metro Bogie Using LSTM Network in Frequency Domain

The wheel–rail excitation has made the fatigue failure problem worse caused by the modal vibration of the metro bogie components. It is of great significance to monitor the state of stress and treat fatigue failure problems in time. Due to the inconvenience of sensor arrangement in the metro bogie system and the high cost of daily maintenance, the direct measurement of dynamic stress is unsuitable for structural health monitoring (SHM) of the metro bogie system. Therefore, it is necessary to provide an indirect stress measurement method for the metro vehicle at risk of abnormal wheel–rail wear. In this article, we propose a data-driven LSTM-based method and spectrum analysis to establish mapping relations between the axle box accelerations and the uniaxial stress of the metro bogie in the frequency domain. Then, the dynamic stress could be measured indirectly by the axle box accelerations. The method was applied to the two cases of metro bogie frame and the bogie cowcatcher, which are both influenced by the bending modal vibration caused by wheel polygonal wear and rail corrugation, respectively. The results show that the proposed method shows a high accuracy for the estimation of dynamic stress in metro bogie. Therefore, the indirect measurement strategy makes stress monitoring convenient in daily maintenance for the real-time alarm of abnormal vibration and early treatment of fatigue failure.