Fault diagnosis in a class of nonlinear systems using identification and GLR testing

In a class of nonlinear systems, additive faults act as changes in the state-transition matrix. Each fault can be characterized by a trajectory in the parameter space. Identifying the parameters online and comparing them to pre-computed trajectories offers an approach to fault isolation. The distance between the trajectories and the observations is subjected to generalized likelihood ratio (GLR) testing. As a key element in this procedure, the best parameter estimation technique has been selected by comparing some typical estimation methods. Simulation studies are included to support the theoretical conclusions.