15. Time Series Modelling of Non-stationary Vibration Signals for Gearbox Fault Diagnosis

Gearboxes often operate under variable operating conditions, which lead to non-stationary vibration. Vibration signal analysis is a widely used condition monitoring technique. Time series model-based methods have been developed for the study of non-stationary vibration signals, and subsequently, for fault diagnosis of gearboxes under variable operating conditions. This chapter presented the latest methodologies for gearbox fault diagnosis using time series model-based methods. The main contents include widely used time-variant models, parameter estimation and model structure selection methods, model validation criteria, and fault diagnostic schemes based on either model residual signals or model parameters. Illustrative examples are provided to show the applications of model residual-based fault diagnosis methods on an experimental dataset collected from a laboratory gearbox test rig. Future research topics are pointed out at the end.