Gearbox Condition Monitoring Using Sparse Filtering and Parameterized Time–Frequency Analysis

It is challenging to monitor the condition of a gearbox being operating under variable conditions. Considering that numerous variable conditions have the non-stationarity nature, a time–frequency analysis (TFA) approach is very suited to analyze the frequency components of interest. As an emerging TFA method, parameterized TFA utilizes further parameters to parameterize kernel functions, so a complex signal can be characterized more accurately. However, it is an issue to estimate the instantaneous frequency with the interference of background noise. This paper presents a method to solve this problem by using the sparse filtering method, as the sparse filtering is capable of enhancing a particular desired feature to achieve the desired effect. With the sparse filtering, the parameterized TFA is used to achieve signal decomposition and clarify the signal spectrum structure. Finally, these interferential frequencies are extracted and removed, and the processed signals are used to monitor the gearbox. The availability of this method was verified by simulated and experimental signal analysis in gearbox condition monitoring.