Gearbox Fault Identification Under Non-Gaussian Noise and Time-Varying Operating Conditions

Abstract

The Synchronous Average of the Squared Envelope (SASE) is very useful to visualise the periodicities in the instantaneous power of the machine due to damage. However, the SASE is sensitive to impulsive noise and the presence of non-synchronous damaged components and therefore provide unreliable representations of the condition of the gearbox under these conditions. Also, the instantaneous power is adversely affected by time-varying operating conditions. Impulsive noise and/or time-varying operating conditions can be encountered in the power generation (e.g. wind turbines) and mining industries (e.g. bucket wheel excavators). Hence, a method is proposed for impulsive data that were acquired under time-varying operating conditions. This method firstly estimates and removes the instantaneous power changes caused by the time-varying operating conditions, whereafter the Synchronous Geometric Average of the Squared Envelope (SGASE) is applied. A more numerically stable calculation of the SGASE is performed, which also provides further insights into its suitability for impulsive noise environments. The methodology is investigated on a bevel gearbox model that was simulated under time-varying operating conditions and an experimental dataset also acquired under time-varying conditions. The results indicate that the SGASE is to be preferred to the SASE for performing fault diagnosis in the presence of non-Gaussian noise.