Research On Feature Extraction And Recognition Technology Of Fan Bearing Fault Signal

With the rapid development of the wind power industry and the requirements of high reliability and easy maintenance of the wind turbine system,the condition monitoring and fault diagnosis technology of the wind turbine has attracted wide attention of the academic and industrial circles.Bearing is the core component of the mechanical transmission system and generator system of the fan.The real-time monitoring and accurate analysis of its operation state are of great significance to the fault diagnosis and operation maintenance of the fan.In this thesis,the problem of vibration signal feature extraction in fan bearing fault diagnosis is studied.Aiming at the problem that it is difficult to accurately judge the fault of rolling bearing by using traditional transform domain method,the bearing is regarded as a generalized communication system,and a variety of information entropy characteristic indexes are constructed by combining complexity entropy with different signal analysis methods;Using the cloud model theory,the secondary feature extraction can describe the distribution characteristics of fault signal features more accurately.Aiming at the data sparsity in the process of bearing diagnosis,the optimal transport theory is deeply studied.The bearing vibration data is regarded as a low dimensional manifold of its high-dimensional physical data,in which different faults correspond to different probability distributions on the manifold,and the distance between different distributions is defined by the optimal transport theory,On the basis of this convex space,the optimal transport matrix is constructed,and then the transfer matrix is used as the fault feature matrix to realize feature extraction.According to the proposed feature extraction method,a classification optimization model based on minimum spanning tree is constructed.According to the performance difference of the base classifier for different features,a multi classifier fusion system with dynamic selection is designed,and the effectiveness of the classifier is verified by simulation experiments.Based on the above analysis,for the fault with different degrees of damage,the feature extraction algorithm based on entropy cloud theory can establish an effective feature database;for the unknown working condition environment with small training set and few sampling points,the optimal transport theory diagnosis method can quickly achieve effective feature extraction;finally,the fusion classifier algorithm is used to verify the effectiveness of the extracted features,which is the most effective Finally,the effective recognition of different fault types under different working conditions is realized,which provides an effective theoretical basis for the application of the algorithm in the field of rolling bearing fault diagnosis.