Wheel-Rail Force Inversion Via Transfer Learning-Based Residual Lstm Neural Network with Temporal Pattern Attention Mechanism

As urbanization advances, metro vehicles are navigating an increasing number of curves, bringing challenges to both vehicle safety and passenger comfort. There is no doubt that reliable acquisition of wheel-rail force is critical, since it has great significance for the safety and stability of vehicle operation. However, conventional wheel-rail force measurement methods are costly and difficult to measure high-frequency forces accurately. A data-driven approach to inverting the wheel-rail force will overcome the above problems. In this work, a transfer learning-based residual long short-term memory neural network with temporal pattern attention mechanism (TPA-ResLSTM) is proposed to realize real-time monitoring of wheel-rail force even when the dataset lacks adequate features. Firstly, according to the physical relationship between the wheel-rail force and acceleration, the learnable wheel-rail force inversion network model is established. Subsequently, a dynamic model for a B-type metro vehicle is adopted to simulate diverse cases as a virtual source and feed the dataset to the neural network. Afterward, the performance of the model is synthetically validated by the ablation study and field experimental data. Finally, the deep learning model is further improved by the transfer learning network, whose performance is comprehensively evaluated using limited data in small radius curve cases. The results show that the inversion model still has remarkable accuracy, in which the coefficient of determination reaches 0.949, under the case of limited training data. It means the proposed method reduces data demands for the network and provides real-time monitoring and feedback of wheel-rail force, possessing a more realistic sense for the operational safety of trains.