基于自适应滤波辅助残差χ2法的导航系统故障诊断方法

针对残差χ2法对软故障检测不敏感，且当量测噪声阵不准确时，也不能很好地检测突变故障的问题，提出了一种限制自适应滤波辅助残差法的诊断方法。基于贝叶斯理论推导出一种限制自适应滤波，利用其自适应调节性在线估计量测噪声阵，抑制滤波预报值跟踪软故障，以辅助残差χ2法进行故障检测。为了克服直接隔离故障法导致滤波精度降低，误检率增加的问题，提出了基于一步预测的故障处理和重构方法，即用前一时刻解算的结果预测当前时刻的导航信息。仿真结果表明，所提出的方法能实时有效地诊断出突变和软故障，保证系统在故障阶段的精度，并使系统在故障消失后及时恢复正常，提高了系统的可靠性。

Fault diagnosis method based on residual chi square assisted by adaptive filter for navigation system

Residual chi square test is not sensitive to soft faults and also unable to detect suddenly changing faults well in inaccurate observation noise. Therefore, a diagnosis method of residual chi square test assisted by restricted adaptive filter was proposed to solve the problem. A restricted adaptive filtering algorithm was derived based on Bayesian theory to estimate the observation noise matrix automatically and on line. In doing so, the proposed algorithm could prevent the filtering prediction from tracking soft faults and then assist the residual chi square test during fault detection. To overcome the problems of filtering precision decline and the increase of the error detection rate caused by immediately isolating fault method,a fault disposal and reconstruction method was developed based on one step predicting information, which is generated by forecasting the current navigation information using the resolved results in the frontal time. Simulation results show that the proposed method can effectively diagnose suddenly changing faults and soft faults online, and ensure system precision during the failure phase. Moreover, the system can be recovered timely from the fault state when the fault signals disappear, and thus the system reliability can be improved.