Distributed detection with data fusion for aircraft flight control computer malfunction monitoring

Timely and reliable detection of controller malfunction is a crucial task in all control systems. In flight control, it is even more crucial, since the cost of controller malfunction is potentially very high. Aircraft Flight Control Computers (FCCs) are typically implemented with redundant processing elements in order to mask random independent component failures. However, redundancy alone does not mask the effects of common-cause disturbances, such as lightning and High-Intensity Radiated Fields (HIRF), that can affect the functional integrity of all processing channels. This paper presents a distributed detection scheme with data fusion for monitoring the function of redundant processing channels of a flight critical controller during operation. Malfunctions in the system are non-Gaussian and can be detected and isolated to the control command calculation in the specific faulty channel. Three approaches for detecting errors in the control law calculations of each processor (i.e. the local detector) are presented: (1) Kalman Filter estimates are used to generate residuals which are then tested against a 3-a threshold; (2) Kalman Filter estimates are used to generate residuals which are then processed using an AR-predictive smoothing filter, and the sample variance of the raw residuals is also calculated; and (3) local detector design is based on state estimation by particle filtering. The implementation of the detection scheme using each of these approaches for the local detector in the context of distributed detection with data fusion is demonstrated using data collected during closed-loop HIRF system effects experiments on a quad-redundant FCC executing Autoland control laws under flight conditions with heavy clear air turbulence. The performance of the monitoring system using each of these local detector approaches is assessed.