Decentralized fault detection and diagnosis using combined parity space and filter innovations based methods

Summary form only given, as follows. A decentralized system architecture is utilized through an Information Filter implementation of the Kalman filter to estimate states pertinent in the operation of an unmanned aerial vehicle. This paper looks at the decentralized data fusion of an Inertial Measurement Unit (IMU) with data from the Global Positioning System (GPS) and an Air Data System (ADS) in order to perform fault detection and diagnosis. For this integrated GPS/IMU/ADS system model we investigate the Fault Detection and Diagnosis (FDD) methodologies born out of observing the information filter innovations as well as the residuals from Parity Space Methods. The viability and the apparent benefits of a joint implementation of these methods is presented, The advantages of both FDD methods become apparent at various stages of operation and the usefulness of applying the methods in conjunction is demonstrated. The Parity Space Methods with their superior isolability and robustness characteristics when combined with the temporal effect properties of the filter innovations provide very promising results. The effectiveness of a decentralized system from a robustness and integrity point of view is also exposed