An auditory model,ZCPA model,was introduced to the feature extraction for mechanical fault diagnosis.The basic principles of ZCPA model were introduced.According to the characteristics of vibration signal analysis in mechanical fault diagnosis, ZCPA model was adjusted. The first adjustment was to replace logarithm

 function by S function for modeling the inner hair cells in cochlea,and the other was distributing frequency bins on frequency axis with the width of frequency bin proportionably increasing.Furthermore,the basilar membrane was modeled by Gammatone filterbank.To test the performance of ZCPA model in feature extraction,the experiments of four kinds of rotor faults were carried out and the measured vibration signals were processed by ZCPA model.The outputs of ZCPA model,the results of feature extraction,show that the ZCPA model can represent different faults only using a few numbers of value(16 values here),

and the stability and coherence for same type faults are satisfactory. Then it is quite possible that ZCPA model or some other auditory models may be applied to engineering practice.