Sparse component analysis-based under-determined blind source separation for bearing fault feature extraction in wind turbine gearbox

The signal processing-based bearing fault diagnosis in wind turbine gearbox is considered challenging as the vibration signals collected from acceleration transducers are, in general, a mixture of signals originating from an unknown number of sources. Even worse, the source number is often larger than the number of installed sensors, and hence the fault characterisation is effectively an under-determined blind source separation problem. In this study, a novel sparse component analysis-based algorithmic solution is proposed to address this technical challenge from two aspects: source number estimation and source signal recovery, to enable accurate and efficient bearing fault diagnosis. The source number estimation is implemented based on the empirical mode decomposition and singular value decomposition joint approach. The observed signals are transformed to the time-frequency domain using short-time Fourier transform to obtain the sparse representation of the signals. The fuzzy C-means clustering and l(1) norm decomposition methods are used to estimate the mixing matrix and recover the source signals, respectively. The proposed solution is assessed through simulation experiments for scenarios of linearly and non-linearly mixed bearing vibration signals, and the numerical result confirms the effectiveness of the proposed algorithmic solution