Abstract: Aiming at disadvantages of wavelet neural network such as long training time and easily gets into local minimum value, the paper put forward a method that wavelet neural network fused fuzzy algorithm, and built fuzzy wavelet neural network model and its training algorithm.It gave practical sample of the model in fault prediction of frequency-conversion speed-regulation system. The application result indicated that fuzzy wavelet neural network improves network training speed and achieves good effect of function approximation.