Abstract: A new approach to the sensor fault diagnosis of nonlinear systems is presented. In this method, the Takagi-Sugeno ( T-S) fuzzy model, fully-decoupled parity equations and parameter estimation are combined to detect, isolate and identify sensor faults for nonlinear systems. The fully-decoupled parity equations in linear systems are designed to generate residuals. And, the existence condition of the fully-decoupled parity vector is given. The residualwhich is sensitive to a special sensor fault is decoupled from system states, disturbance inputs and other sensor faults.The T-S model is used to extend the fully-decoupled parity equations to fuzzy parity equations for nonlinear systems.The sensor faults are represented as biases and scale factor changes. The biases and changes in the scale factor can beidentified using the information contained in the residuals by a parameter estimator based on the Kalman filter. Asimulation example of a nonlinear aircraft control system is given for illustration.

Key words: fault diagnosis, nonlinear systems, sensor, fuzzy logic, parity equation, Kalman filtering