Abstract: In order to diagnose the fault gear under the strong noise interference, put forward the technique of synchronous time domain average (TSA) combined with AR model of gear fault diagnosis methods. The TSA technology was used to extract gear characteristic signal under strong noise interference, the order of AR model is determined by FPE criteria, the parameter vector and tolerance range of parameter of gear under the normal state is determined by AR model parameter algorithm, and then analysis of AR model parameter of the gear fault signal under the same model order, contrast to established tolerance range of parameter, so as to gear fault diagnosis. Applying this method to actual test signal analysis,the extraction of the characteristics of eight groups of normal gear signal data AR model is established,the best order for five order AR model, by the AR model parameters algorithm to get the AR model parameter vectors and parameters of the normal gear tolerance range, then in the same order number for five order of AR model analyzes the fault condition of several groups of model parameters, contrast to established tolerance range of normal AR model parameters, the result shows that it is effective to gear fault diagnosis.