According to the characteristics of fault vibration of rolling bearing (non-stationary signals), a method of fault diagnosis based on wavelet analysis is put forward. In the method, the signals included faulty information of rolling bearing are decomposed and reconstructed by Matlab software, and by zoom frequency spectrum analysis, the fault information composition of the rolling bearings can be extracted effectively to determine the faulty type of the rolling bearings. Based on the above, an applicable fault diagnosis system for rolling bearing with surface damage in friction pair was developed. It has been proven by practice that the diagnosis method of system is surefire and its results are stable and reliable. Through being improved and expanded,the system can be applied to collect and analyze other vibration signals.