Abstract: Radial Gaussian function networks based fuzzy systems with adaptive capability,are applied to the state estimation and fault detection of nonlinear time-varying systems .In order to extract fuzzy IF-THEN rules from input and output sample data through learning,the Gaussian function is employed to represent the membership functions of the premise part of fuzzy rules ,and then a hierarchically structural self-organizing learning method is given.Based on this method of state estimation and fault detect for nonliear systems is proposed .The structure and characteristics of the observer are discussed. The results of simulation show that the proposed nonliear state observer can observe real system state and detect system fault satisfactorily.