自适应滤波算法在导弹动力系统故障诊断中的应用

Application of Kalman Filter Adaptive Algorithm to the Fault Detection of a Liquid Missile Power System

基于时间序列的AR模型方法在用于液体导弹动力系统稳态工作段故障检测时遇到了两难题。其一,由于系统差异,在以试车时的动力系统状态的均值作为实际运行时动力系统状态均值时,得到的时间序列必定不再是零均值;其二,导弹动力系统工作环境不同时刻不尽相同,表现在AR模型中其观测噪声是时变的。在实践中,我们采用了基于扩展Kalm an滤波的自适应算法以及带有时变噪声统计的自适应滤波算法很好地解决了相关问题。

The AR model based on time series when applied to the fault detection of the steady working process of a liquid missile power system（LMPS） has two problems. One is that the time series is no longer zero mean value when the mean value of the power system status at trial operation is used as the mean value at actual operation. The other is time-varying measurement noise, namely the working environment of the LMPS varies as time varies. The two problems are solved by using extended Kalman filter adaptive algorithm and the adaptive filter algorithm with time-varying noise statistics.