Multi-dimensional variational mode decomposition for bearing-crack detection in wind turbines with large driving-speed variations

Maintenance of wind turbines is recognized as one of the most important issues in wind energy sector. Bearing cracks are a dominant cause of wind-turbine gearbox failures, and gearboxes are the largest contributor to turbine downtime and costliest to repair. Because rolling bearings in wind-turbine gearboxes are often used in variable-speed and variable-load situations, early fault detection and diagnosis of bearings under these non-steady-state conditions are essential to prevent catastrophic failures and thus increase turbine availability and reduce the cost of wind energy. This work proposes a new method, namely, multi-dimensional variational decomposition (MDVD), for bearing-crack detection. In this method, variational mode decomposition (VMD) is incorporated into convolutive blind-source separation (BSS) to address the challenge of substantial driving-speed variations. One unique property of the proposed MDVD method is its ability to deal with multi-channel vibration signals with large speed/load fluctuations. Hence, MDVD does not impose any restrictions on the number of sensors and their installations (e.g., installation locations and directions) and thus overcomes the limitation of existing methods, which can only process a single sensor signal. An experimental validation of the proposed method was conducted using bearings with axial cracks in the outer race.