In this paper, the space time-index plots were introduced and used to determine the nonstationarity of the fault diagnosis in order to handle faults in the nonstationary stage. By using the decomposition and reconstruction of wavelet transform for nonstationary signal, the author targeted on the bandwidth and select information with pertinence, and then analysed the reconstruct signal spectrum for extracting the typical character of fault. The results show that the space time-index method is applicable to judge the nonstationary and the wavelet transform method in fault diagnosis is effective for the nonstationary signal.