Intelligent Fault Diagnosis Of Rotating Machinery Based On Signal Feature Clustering

Rotating machinery plays an important role in many fields such as aerospace,rail transportation,energy generation,petrochemical industry and automobile manufacturing.The research on condition monitoring and fault diagnosis for rotating machinery and equipment and key parts is of great practical significance to ensure the normal and stable operation of equipment,reduce maintenance costs and improve production efficiency.Clustering is an important research field in machine learning,and some clustering methods have been widely used in the field of mechanical fault diagnosis.However,intelligent mechanical fault diagnosis based on clustering analysis is still a research hotspot of fault diagnosis technology.Based on the theory of cluster analysis and signal processing technology,this paper studies the intelligent fault diagnosis based on signal feature clustering with bearings and gears as the key parts of rotating machinery.The main work of this paper is as follows:A novel feature selection method is researched to improve the accuracy of intelligent fault diagnosis model.The weight value is determined by calculating the fluctuation degree of the eigenvalue of the target feature.According to the similarity between features,some features with high similarity are removed to reduce the dimension of features.These two technologies ultimately determine the optimal feature collection of intelligent fault diagnosis.A rolling bearing fault diagnosis method with specified cluster number was researched.For affine propagation clustering algorithm under the specified cluster number,different bias degree of accuracy of clustering problems,put forward a specified clustering number affine clustering algorithm,this algorithm aims at the specified number of clustering,through iteration bias degree of parameters,find all corresponding to the result of clustering number,then compare the results of all evaluation index profile coefficient,find out the best clustering results.The validity of the proposed rolling bearing fault diagnosis method was verified by using the fault bearing data collected from the mechanical comprehensive fault simulation experimental platform.An intelligent fault diagnosis method for gears based on adaptive density peak clustering was researched.Aiming at the problem that fast searching density peak clustering can not accurately select the number of cluster center points,a method combining numerical detection and iterative optimization was proposed to select the optimal clustering results without accurately selecting the number of cluster center points.Four kinds of gear fault signals collected from the wind turbine power transmission fault diagnosis comprehensive experimental platform were used for example verification.Combined with the new feature selection method,the results show that the proposed method can be applied to gear fault diagnosis.An intelligent fault diagnosis model for rotating machinery based on adaptive clustering was researched.Adaptive clustering does not need to specify any parameters.By calculating the nearest neighbor samples of each sample,the link chain of the sample group can be found.The linked samples can be regarded as the same category.The result of each clustering can be used as the target sample set of the next clustering.This target sample is the corresponding mean value of different categories in the previous clustering.Clustering will not stop until all samples are clustered into one category.The adaptive clustering algorithm can obtain multiple clustering results and select the best clustering result by comparing the contour coefficients of all the results.Two kinds of bearing fault data sets were used to verify the intelligent fault diagnosis model of rotating machinery based on adaptive clustering.The results show that adaptive clustering can be applied to fault classification of rotating machinery.