列车动力学模型时变环境参数自适应辨识

考虑列车制动性能与制动距离对列车安全的重要影响, 分析了列车运行的动力学特性, 构建了列车离散化制动模型,并针对影响列车制动性能的关键参数 — 钢轨粘着系数难以直接观测、随钢轨环境变化的特点, 提出基于滑动窗口与最大期望理论的轮轨粘着系数在线辨识算法. 首先, 依据数据特征确定滑动窗口位置与窗口尺寸; 然后, 构造列车动力学模型参数的条件数学期望, 并结合粒子滤波与粒子平滑算法以及贝叶斯理论, 估计预设模型参数下的列车运行状态; 在此基础上, 分析粘着系数的后验概率, 并极大化条件数学期望对模型参数预设进行优化更新, 进而实现模型真实参数的逐步逼近. 最后, 考虑雪地、隧道等场景下的粘着系数变化, 对本文方法进行了仿真验证, 并数值分析了粘着系数对制动距离的影响. 仿真结果表明本文算法可快速、准确地对粘着系数进行实时辨识, 掌握轮轨间实时粘着状态.

Adaptive Identification of Time-varying Environmental Parameters in Train Dynamics Model

Considering the braking performance and distance on train safety, the dynamic characteristics of train are analyzed, and the discrete braking model is constructed. Aiming at the pivotal rail-wheel adhesion coefficient, which is difficult to be observed directly and varies with rail environment, an online identification of adhesion coefficient based on sliding window and expectation maximization is proposed. Firstly, the position and size of window are set up. Then, the conditional expectation about the parametric train dynamic model is constructed, and the train running state under the preset parameters is estimated by the particle filter, particle smoothing and Bayesian theory. On this basis, the posterior probability of the adhesion coefficient is analyzed, and the conditional expectation is maximized to optimize and update the model parameters, so that the real parameters can be approached step by step. Finally, considering the change of adhesion coefficient in snow, tunnel and other scenes, the method is simulated and validated, and the influence of adhesion coefficient on braking distance is analyzed numerically. The simulation results show that the adhesion coefficient can be identified quickly and accurately, and the real-time adhesion state can be grasped in real time.