Diagnosis of non-linear mixed multiple faults based on underdetermined blind source separation for wind turbine gearbox: simulation, testbed and realistic scenarios

The diagnosis of multi-fault in wind turbine gearbox based on vibration signal processing is considered challenging as the collected measurements from acceleration transducers are often a non-linear mixture of signals induced from an unknown number of sources, i.e. an underdetermined blind source separation (UBSS) problem. In this study, a novel UBSS-based algorithmic solution is proposed to address this technical challenge from two aspects: source number estimation and source signal recovery. The former is realised based on the empirical mode decomposition and singular value decomposition (SVD) joint approach; and for the latter, the observed vibration 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 are carried out to estimate the mixing matrix and recover the source signals, respectively. The performance of proposed solution is extensively assessed through experiments based on simulation, testbed and realistic wind farm measurements for a range of fault scenarios for both linear and non-linear scenarios. The numerical result clearly confirms the effectiveness of the proposed algorithmic solution for non-linear multi-fault diagnosis of wind turbines.