Linear model identification for gear fault detection using higher order statistics and inverse filter criteria

Abstract

Our study in the past showed that the autoregressive (AR) modelling method could be effectively used in the detection of gear tooth cracking. In the search for further improvement, a technique of identifying linear parametric models for gear signals using higher order statistics and inverse filter criteria has been evaluated and was applied to some seeded fault gear test data. The results indicate that this approach is more effective than the AR modelling method and the conventional residual signal technique.