Abstract: The effective separation of fault feature components is the core of rolling bearing composite fault diagnosis. Under the background of strong noise and mutual interference and coupling between various faults， the acoustic composite fault diagnosis of rolling bearing is very challenging. In this paper， a composite fault acoustic diagnosis method based on optimized singular spectrum decomposition （SSD） and parameter adaptive maximum correlated kurtosis deconvolution （MCKD） is proposed. The envelope kurtosis is used as an indicator to assist SSD to quickly determine the optimal decomposition level， so as to overcome the uncertainty of the artificial empirical determination of the decomposition level and decompose the signal into multiple singular spectral components. Combining the ratio of fault characteristic frequency energy amplitude as an index， the two singular spectral components containing the main fault characteristic information are adaptively selected. The parameter adaptive MCKD is used to filter the selected optimal component and enhance the signal feature， and the fault feature frequency is extracted by envelope spectrum analysis to realize fault diagnosis. The effectiveness of the proposed method is verified by the simulation and experimental acoustic signals of rolling bearings. The research provides a new means for the composite fault diagnosis of rotating machinery.