基于边缘粒子滤波的高速列车走行部关键参数估计

发布时间：2018-04-27 10:44

本文选题：故障诊断 + 参数估计　；参考：《西南交通大学》2015年硕士论文

【摘要】：本文为实现高速列车运行时转向架关键部位的参数辨识和故障诊断,对高速列车动力学模型关键参数估计方法进行研究,将基于非线性滤波器的状态估计方法应用于高速列车关键参数估计中,主要包括以下几个方面的研究内容：首先回顾了在状态参数联合估计领域以及高速列车关键参数估计领域中前人所做的有意义的工作,并讨论了运用卡尔曼滤波器、扩展卡尔曼滤波器以及粒子滤波器来解决相应参数检测问题的可能性。接下来,针对列车动力学模型,参考实际参数,建立了CRH380A列车横向动力学模型,将白噪声激扰作为模型输入,相应传感器观测结果作为模型输出。为了验证模型的有效性,将实际列车振动平台数据与高速列车模型数据进行分析与对比。在确认模型准确可靠的基础上,设定系统相应的统计学数值,采集模型输出,使用扩展卡尔曼滤波器(EKF)以及边缘粒子滤波器(Rao-Blackwellised粒子滤波器,RBPF)进行参数估计,观测并比较了两种滤波器的参数估计结果,分析了各自的性能优劣势。最后,由于参数估计体系采用线性列车模型并运用高斯白噪声模拟列车噪声输入,在实际检测中不具备良好的适应性,在Rao-Blackwellised粒子滤波器的基础上,根据状态扩展理论对Rao-Blackwellised算法进行改进,以解决原先算法中对于非线性非高斯信号适应性差的问题。运用了状态扩展理论对列车实际运行中轨道不平顺的影响进行了定量分析,并将其纳入算法中进行迭代。运用该改进算法,较好地估计出了列车在实际运行中转向架二系横向阻尼系数、抗蛇行阻尼系数和轮对踏面锥度等几个参数。估计结果较原始的Rao-Blackwellised滤波器在参数估计精度上有一定提升。接着模拟了多种可能发生的列车运行故障,使用改进后的方法估计目标参数,结果表明改进的参数估计方法对实际噪声具有良好的适应性。
[Abstract]:In order to realize the parameter identification and fault diagnosis of the key parts of the bogie when the high-speed train is running, the method of estimating the key parameters of the dynamic model of the high-speed train is studied in this paper. The state estimation method based on nonlinear filter is applied to estimate the key parameters of high-speed train. The main contents are as follows: firstly, the important work done in the field of joint estimation of state parameters and the estimation of key parameters of high-speed trains is reviewed, and the application of Kalman filter is discussed. Extend Kalman filter and particle filter to solve the problem of parameter detection. Then, according to the train dynamics model and referring to the actual parameters, the CRH380A train lateral dynamics model is established. The white noise excitation is taken as the input of the model, and the corresponding sensor observation results are taken as the model output. In order to verify the validity of the model, the actual train vibration platform data and the high-speed train model data are analyzed and compared. On the basis of confirming the accuracy and reliability of the model, the corresponding statistical values of the system are set, the output of the model is collected, and the parameters are estimated by using the extended Kalman filter (EKF) and the edge particle filter (Rao-Blackwellised particle filter (RBPF). The parameter estimation results of the two filters are observed and compared, and their performance is analyzed. Finally, because the parameter estimation system adopts the linear train model and uses Gao Si white noise to simulate the train noise input, it has no good adaptability in the actual detection. Based on the Rao-Blackwellised particle filter, According to the state expansion theory, the Rao-Blackwellised algorithm is improved to solve the problem of poor adaptability to nonlinear non- signals in the original algorithm. The influence of track irregularity in actual train operation is analyzed quantitatively by using the state expansion theory, and it is incorporated into the algorithm to iterate. By using the improved algorithm, several parameters, such as the transverse damping coefficient of the second system of the bogie, the anti-snake damping coefficient and the taper of the wheelset tread, are well estimated in the actual operation of the bogie. The estimation result is better than the original Rao-Blackwellised filter in parameter estimation accuracy. Then several possible train faults are simulated and the target parameters are estimated by using the improved method. The results show that the improved method has a good adaptability to actual noise.
【学位授予单位】：西南交通大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U270.33;TN713

【参考文献】