关键词: 高速列车, 悬挂参数估计, 近似贝叶斯计算, Kriging代理模型, 多体动力学仿真

Abstract: In the case of long-term service of high-speed trains, the operating parameters of the vehicle suspension are very different from the factory parameters. The identification method of vehicle physical parameters is complicated and expensive, but the parameter estimation can estimate the actual parameters of key components of vehicles only by using a few sensors to obtain the vehicle vibration state information. An approximate Bayesian calculation (ABC) method based on Kriging model is proposed to estimate the suspension parameters of high-speed trains in service by combining the measured track inspection data and multi-body dynamics simulation model. First, the authors utilize the Kriging proxy model instead of the multi-body dynamics model as the numerical model in the ABC. Second, the authors adopt single-objective optimization instead of multi-objective processing to simplify parameter estimation in ABC and obtain the posterior distributions of the suspension parameters under different weighting factors combinations. The parameter value corresponding to the maximum relative probability in the posterior distribution is the estimated value of the suspension parameter. Finally, the estimated values of the suspension parameters are introduced into the multi-body dynamics model to predict the car-body acceleration, which is compared with the real-world car-body acceleration. The results show that the Pearson correlation coefficient between the power spectral density curve of the predicted vertical acceleration value and the power spectral density curve of the ground truth is increased by 0.919 before optimization when only the vertical acceleration of the car-body is used as the optimization target compared with the simulation results based on the initial vehicle parameters. The correlation coefficient between the predicted lateral acceleration and ground truth is increased by 0.427 when only the lateral acceleration of the car-body is used as the optimization target. When the vertical and lateral accelerations of the car-body are used as the optimization targets, the correlation coefficient between the predicted vertical acceleration, lateral acceleration, and the actual values are increased by 0.861 and 0.366, respectively. This shows that the proposed method can effectively estimate the suspension parameters of high-speed trains in service, which provides technical support for further vehicle parameter optimization.

Key words: high-speed train, suspension parameter estimation, approximate Bayesian calculation, Kriging proxy model, multi-body dynamics simulation