According to modulation characteristics of rotating machinery fault vibration signals and limitation of the traditional envelope spectrum analysis needs to rely on experience for determining the parameters of band- pass filter,a fault diagnosis method for rotating machinery based on maximum correlated kurtosis deconvolution（ MCKD） and envelope spectrum is proposed. Firstly,by using MCKD,the noise in the rotating machinery vibration signal is reduced,then the envelope spectrum is obtained by Hilbert transform to identify the fault,simulation signal analysis and engineering applications demonstrated the effectiveness of the method.