Statistical modeling of gear vibration signals and its application to detecting and diagnosing gear faults

A statistical model of gear mesh vibration signals is proposed in this paper. It is seen that the sample mean is first removed from the gear mesh vibration signal, and the remaining signal components are approximated with a set of sinusoidal functions in the signal model. The signal model parameters can be estimated by the least-square method and the optimal model order is determined based on the Akaike Information Criterion (AIC) or the modified Bayesian Information Criterion (BIC). Within the framework of the signal model, to perform the gear fault diagnosis and analysis, the residual signal between the synchronous signal average and the output of the optimal signal model is further computed and the corresponding kurtosis values are employed for efficiently detecting the gear tooth fault.