This paper presents a new fault diagnosis procedure for rotating machinery based on wavelet packets-fractal technology in combination with neural network. The main purpose is to investigate different fault mechanism in rotating machinery, such as imbalance, misalignment, or imbalance with misalignment conditions, etc. When these faults occur they usually produce nonstationary vibration signals, by using wavelet packets transform on these signals, the fractal dimension of each frequency channel is extracted and the box dimension is used to depict the failure characteristics of vibration signals, and then the failure modes can be classified by radial basis function neural network. Experiments were conducted and the results shown that the proposed method can detect and recognize different kinds of faults in rotating machinery. Therefore, it is concluded that the wavelet packets-fractal technology combined with neural network method can provide an effective way to diagnosis faults in mechanical systems.