Adaptive Observer for Fault Diagnosis in Nonlinear Discrete-Time Systems

This paper deals with the problem of fault diagnosis (FD) for a class of nonlinear systems. The scheme is based on a discrete-time diagnostic observer that computes a prediction of the system’s state. Compensation of the fault effect on the state prediction is achieved via an adaptive discrete-time approach, based on a parametric model of the faults. A stability proof is developed to prove the global exponential stability of the state estimates. A solution for fault isolation and identification is also proposed, based on a postfault analysis. The proposed FD approach is applied and experimentally tested on a conventional industrial robot manipulator.