一种改进的残差χ^2检验方法在联邦UKF滤波器中的应用

Research on an improved residual Chi-square fault detection method for federated unscented Kalman filter

故障检测和隔离是实现联邦UKF(unscented Kalman filter)滤波器工程应用的关键。针对常规残差χ2检验法因误警导致无故障子系统被频繁隔离的问题,提出了一种改进算法。该算法采用模糊逻辑和加权平均来处理残差χ2检验法的检验阈值,可以根据突变信号的大小自动调节检测时间的长短,特别是对不易检测的小幅值突变信号,自动延长检测时间从而利用更多的信息,有效降低了系统的误警率;在此基础上,进一步设计了合理的故障检测控制规则。将改进的残差χ2检验法应用于多传感器卫星组合导航系统中,仿真结果表明,改进算法有效提高了组合导航系统的稳定性,具有实用价值。

Fault detection and isolation is the key technology for engineering application of federated unscented Kalman Filter. Because of incorrect alarm, conventional residual chi-square test will cause frequent isolation when used in local filters of the federated UKE In view of this problem, this paper puts forward an improved method, which employs fuzzy logic and weighted average to deal with the test threshold. The method is able to adjust the detection time adaptively, especially for small amplitude and transient signal, the algorithm automatically increases the detection time to make use of much information and reduces the percentage of incorrect alarm effec- tively. Based on this, the proper control rule for fault detection is designed. The new chi-square test was applied in a multi-sensor satellite integrated navigation system, and the simulation results prove that the improved algorithm increases the stability of the integrated navigation system effectively and is rather practical.