Enhancing Train Coupling Simulation by Incorporating Speed-Dependent Energy Absorber Characteristics Through a Deep Neural Network

Recently, hydrostatic buffers have emerged as energy-absorbing components in railway vehicles. These buffers exhibit speed-dependent characteristics, with their reaction forces contingent upon compression displacement and speed. However, when dealing with a hydrostatic buffer with an unknown characteristic function in dynamic simulations, accommodating its speed-dependent attributes becomes a challenging task. In this study, we proposed a method for simulating train couplings that incorporates the speed-dependent characteristics of a hydrostatic buffer by utilizing a deep neural network (DNN). Our methodology involved the training of a DNN-based speed-dependent buffer model using empirical data obtained from dynamic buffer tests. Subsequently, this model was applied to a multibody dynamics simulation for train coupling analysis. A critical aspect of this study involved comparing speed-dependent and speed-independent models in a train coupling scenario. This comparison reveals a significant insight: neglecting speed-dependent characteristics in coupling simulations can lead to inaccurate train-coupling safety assessments. The DNN-based method demonstrated its effectiveness, even with limited test data and when the mathematical speed-dependent characteristic function of the buffer is unknown.