Fault Diagnosis of Wind Turbine Gearbox based on LSGAN and VMD-MPE-KELM

Abstract

In the actual working condition，the fault samples of wind turbine gearbox are mostly unbalanced. In order to overcome the influence of sample imbalance on the diagnosis effect，a fault diagnosis method of wind turbine gearbox based on LSGAN and VMD-MPE-KELM is proposed. Firstly，LSGAN algorithm is used to generate and process a few kinds of fault samples. The generated data with original sample characteristics is expanded to make its distribution balanced. The VMD method is used to decompose the vibration signals of all kinds of faults in the sample set，and the MPE value of each modal component is calculated to extract the signal features. Then，KPCA method is used to reduce the dimension to obtain the feature vector of fault samples，which is substituted into KELM model for diagnosis. The experimental results show that LSGAN algorithm overcomes the problems of GAN gradient disappearance，unstable training and poor data quality in generating fault samples. The VMD-MPE-KPCA method can effectively extract fault features. This method effectively improves the diagnosis accuracy of unbalanced gearbox fault samples.

Keywords

• Variational mode decomposition;LSGAN;Gearbox;MPE;Fault diagnosis