Fault detection and isolation for the Bluebird test bed aircraft

Abstract

A Fault Detection and isolation (FDI) algorithm design is presented using the Multiple Model algorithm technique for the Bluebird aircraft being developed at the Naval Postgraduate School. The requirement to maintain high performance in the dynamic system of the aircraft necessitates the use of FDI techniques to detect and isolate malfunctions in the sensors and actuators of the aircraft without using hardware redundancy. The solution presented makes use of analytical redundancy in a bank of Kalman filters. Statistical tests using Bayesian theory are applied on the filter's innovations to perform the task of detection and isolation. The algorithm was developed using MATLAB software from The Math Works, Inc. The work presented in this thesis is related only to the task of FDI. The remaining task of the monitoring system, reconfiguration and continued operation by the observed plant after a failure detection, will not be addressed.