Abstract: The concept of faulty systems state observer (FSSO) is presented for the first time. The difference between FSSO and the traditional observer is analysed, and two possible structures for FSSO are pointed out. A new robust fault diagnosis method based on FSSO for a class of nonlinear systems with bounded uncertainties satisfying Lipschitz condition is proposed and its robustness and sensitivity are also analysed. This method can detect and isolate faults. It can also estimate the characteristics of the fault signal which varies with time. The result of the simulation indicates its validity.

Key words: nonlinear systems, uncertainty, fault diagnosis, state observer