A theory-LSTM hybrid model of a swing-arm rubber bushing to represent the temperature,amplitude,and frequency dependencies for railway vehicle simulation

Abstract

The swing-arm rubber bushing (SARB) is a key suspension component that guarantees the vehicle’s running stability and curve safety. This paper introduces a hybrid model combining theory and Long Short-Term Memory (LSTM), designed for dynamic simulations of railway vehicles to capture the SARB’s temperature, amplitude, and frequency dependencies. The hybrid model comprised the theoretical and LSTM parts in series: the theoretical model provides mechanical predictive trends, while the LSTM enhances prediction accuracy. The experimental data verified the hybrid model’s accuracy in reflecting the three dependencies. Furthermore, the hybrid model was incorporated into a railway vehicle to assess its performance under stochastic vehicle excitation. Results indicate that the model demonstrates significant nonlinearity during dynamic simulations. The hybrid model effectively corrects the overestimation of displacement at small amplitudes and high frequencies, while also revealing that the SARB hardens at low temperatures, negatively affecting the vehicle’s curve safety.