摆式列车线路检测信号的动态自适应滤波研究

为提高摆式列车检测系统的精度,针对传统数字滤波器的不足,将非平稳随机信号时变模型参数的自适应估计与普通卡尔曼滤波算法相结合,提出一种能有效消除或削弱测量信号中高斯白噪声的卡尔曼动态自适应滤波方法及数学模型。实时建模精度是实现卡尔曼动态滤波的关键。通过对具有不同遗忘因子的递推最小二乘算法的分析和比较,结果表明,带自适应遗忘因子的递推最小二乘算法(RLSAF)由于其遗忘因子能根据信号本身的统计特性的变化自适应地进行调整,因而对非平稳随机信号具有很强的跟踪性能。采用基于RL-SAF算法的卡尔曼动态自适应滤波方法,能实现摆式列车线路检测信号(陀螺仪角速率信号)的有效滤波。

Kalman Dynamic Adaptive Filtering to Railway Line Measurement Signals of Tilting Train

To improve the accuracy of the measurement system of tilting train and in view of the disadvantages of traditional digital filters, an effective Kalman dynamic adaptive filtering method and corresponding mathematical model is put forward, which combines time-varying parameters adaptive estimation of non-stationary random signals model and normal Kalman filtering algorithm together, therefore can effectively eliminate or minimize effects of gauss white noises. Real time modeling accuracy of non-stationary random signals is key to the results of this dynamic filtering. Comparing and analyzing various RLS algorithms with different forgetting factors, it is shown that RLSAF algorithms can strongly track non-stationary random signals because of its adaptive forgetting factor changing with statistical characters of signals. In application to signals treatment of tilting train, results show that this Kalman dynamic adaptive filtering method based on RLSAF algorithm can effectively realize filtering to the railway line measurement signals, such as gyroscope angular velocity signal.