A new wind turbine health condition monitoring method based on VMD-MPE and feature-based transfer learning

Abstract

Aimed at the problem that the signal data of wind turbine faulty gearbox is difficult to obtain and the health condition is difficult to diagnose under variable working conditions, a fault diagnosis method based on variational mode decomposition (VMD) multi-scale permutation entropy (MPE) and feature-based transfer learning (FTL) is proposed. According to the vibration signal characteristics of wind turbines, a series of mode components are obtained by transforming the signals under variable conditions. The MPE of the mode components is combined with the signal time domain features as a feature vector to be input into the transfer learning algorithm. The source domain and the target domain data belong to different working conditions, so the traditional machine learning methods are not ideal for fault classification. The method adopted in this paper minimizes the covariance between the source domain and the target domain through a linear transformation matrix, and reduces the difference of data distribution between the source domain and the target domain. Then, the feature vectors of the covariance-aligned source domain and the target domain are input into the support vector machine (SVM) for training and testing. Experiment shows that the proposed covariance alignment (COVAL) of fault features has higher accuracy in rolling bearing multi-state classification under variable working conditions compared with other methods.