When using traditional state equations to calculate and analyze nonlinear dynamic autonomous networks, there are some problems, such as limitations and calculation difficulties, especially the problem of calculating the time of the response crossing a boundary. To avoid these problems for the nonlinear dynamic autonomous networks, this paper presented a method based on canonical piece-wise linearization to obtain a set of canonical piece-wise-linear equations for dynamic units in nonlinear dynamic autonomous networks and the hybrid parameter equations of nonlinear dynamic autonomous networks. By resolving these equations, the simulation algorithm of fault response of nonlinear dynamic autonomous networks can be obtained. This paper used wavelet transform as a preprocessor to extract the fault features from the fault responses of nonlinear dynamic autonomous networks, and adopted Generic Algorithm to optimize the structure and parameters of BPNN. Then, the fault state features were fed to GABPNN to classify and determine the faults. Simulation results show that this fault diagnostic algorithm is an efficient analysis method.