In the stochastic vibration analysis of train-bridge systems,the stochastic dynamic response is one of the key factors in evaluating the safety of traveling,and the existing response calculation methods have the problems of time-consuming and high cost.The ability to quickly and accurately predict the dynamic response of the train-track-bridge system is of great significance to the condition assessment and operation and maintenance of heavy-haul railroad bridges.In this paper,a stochastic vibration response prediction method for the heavy-haul train-bridge system based on Particle Swarm Optimization (PSO) Long Short-term Memory (LSTM) neural network model was proposed.The method took the random parameters of the train-bridge and the random irregularity excitation of the track as inputs and constructs an agent model with the dynamic response of the bridge as output.Firstly,the PSO-LSTM network model was constructed based on the commercial software MATLAB platform.Secondly,the stochastic dynamic response corresponding to the initial sample set was calculated by the established stochastic vibration analysis model of the train-track-bridge system,and the model was trained,while the PSO algorithm was utilized to optimize the structural parameters of the LSTM.Finally,the trained PSO-LSTM model was used to predict the bridge dynamic response.To verify the superiority and robustness of the present algorithm,the prediction efficiency of the present algorithm was compared with that of BP (Back Propagation) neural network,GRU (Gated Recurrent Unit) neural network and LSTM neural network by taking the measured data of Shuohuang heavy-haul Railway as an example.The prediction under different speeds was discussed,and the comparison analysis between the present model and the measured data and the Comparative analysis of this model with the finite element analysis data was carried out.The results are shown as follows.Under the optimization of PSO,the prediction results of the LSTM model have been improved,the correlation coefficient of the PSO-LSTM model can reach 0.97,and the other evaluation error values are smaller than those of BP neural network and GRU neural network model.The model in this paper can predict the random dynamic response of the bridge more efficiently and accurately,and provide technical support for the further development of the theory of predicting the random dynamic response of the train-track-bridge system.
关键词
随机振动/响应预测/PSO算法/LSTM神经网络/车轨桥系统
Key words
random vibration/response prediction/PSO algorithm/LSTM neural network/train-track-bridge system
维普
万方数据