Research on Hybrid Modeling Method of Yaw Damper Based on Maxwell-LSTM
The wheel tread wear of the train in the actual service environment leads to the yaw damper working at higher frequencies.The accuracy of traditional yaw damper model based on mechanics principle is significantly reduced at high frequency.It is difficult to show the dynamic characteristics of yaw damper accurately.The physical parameter model which can accurately fit the dynamic characteristics of yaw damper at high frequency is inefficient and cannot be used in multi-body dynamics simulation.To solve this problem,A hybrid yaw damper modeling method based on Maxwell equivalent parameter model coupled with LSTM neural network is proposed.Based on the traditional Maxwell model,the LSTM network captures the change characteristics of input variables and indirectly considers the frequency and amplitude of external excitation to meet the above challenges.In order to prove the feasibility of this modeling method,The hybrid model is compared with the bench test results and the nonlinear stiffness and damping piece wise Maxwell model.The results show that:the computational efficiency of LSTM hybrid model is basically the same.Compared with the segmented Maxwell model,the LSTM hybrid model reduces the model error by 22.31%on average under high frequency excitation,26.89%on average under high amplitude excitation.The dynamic stiffness error decreases by 26.35%on average,and the dynamic damping error decreases by 21.01%on average.It can be concluded that the proposed the LSTM hybrid model can more accurately represent the dynamic performance of the damper under high frequency and high amplitude.The hybrid modeling method of yaw damper based on Maxwell-LSTM can solve the contradiction between computational efficiency and accuracy of traditional dynamic model and is more suitable for studying railway vehicle system dynamics under various operating cases.
yaw damperhybrid modeling methodLSTM networkdeep learningvehicle system dynamics
万方数据
维普
