Fault diagnosis in gears has been the subject of intensive research as gears are critical element in a variety of rotating machinery applications and an unexpected failure of the gear may cause substantial economic losses. Vibration analysis has been used as a predictive maintenance procedure and as a support for machinery maintenance decisions .In this present work monitoring of gear health using the vibration signals as well as usage of advance signal processing methods such as wavelet transform is implemented to extract the features which reveals the characteristics of the actual conditions of the gear. By meas-uring and analysing the signals, it is possible to determine both the nature and severity of the defect, and hence predict the ma-chine's failure. To reduce the dimensionality of the extracted statistical features, roughset integrated principle component analysis (RSPCA) is used after feature extraction process. Back-propagation neural network (BPNN) and probabilistic neural network (PNN) were deployed for diagnostic analysis of the sig-nal. To validate the proposed methodology, four kinds of running states are simulated with artificially created faults in gear that is accommodated in the test rig. By comparing classification accu-racy along with computation timing the best scheme is selected for diagnostic prediction. © February 2015 IJENS.