Research On Condition Detection And Fault Diagnosis Of Rotating Machinery System Based On Industrial Internet

As a common type of mechanical equipment in industry,rotary machinery is widely used in large rotor systems such as steam turbines and gas turbines,as well as small equipment such as motors,gear boxes and pumps.It plays an important role in industrial production.With the development of computer,sensor network,artificial intelligence and other technologies,rotating machinery equipment towards complex,intelligent direction of rapid development,coupled with complex operating environment,rotating machinery prone to failure.Parts damage or failure,may make the whole rotating machinery can not work normally,more serious will cause enterprise personnel safety accidents.The condition monitoring and fault diagnosis of rotating machinery equipment are the premise to ensure the safe and stable operation of the equipment,and it is also very important for preventive maintenance.Fault diagnosis based on deep learning can realize end-to-end diagnosis,and has been applied to some extent in the field of mechanical fault diagnosis and health monitoring.However,there exist some problems such as the contradiction between small sample fault data in actual fault diagnosis and deep network model requiring large training samples,which still need further research.At the same time,the development of new generation of information technology such as big data and industrial Internet provides new theories and new methods for equipment fault state detection and diagnosis.This thesis takes the key parts of rotating machinery prone to failure(bearings and gears)as the research object,focuses on the application of deep convolution and deep coding in fault prediction,feature extraction and fault diagnosis,and realizes fault signal acquisition,cloud storage,fault state detection and diagnosis based on industrial Internet technology.The main work and research content of this thesis are as follows:(1)In view of the contradiction between the small sample fault data in actual fault diagnosis and the deep network model requiring large training samples,a fault warning method based on massive normal monitoring data was proposed in this thesis.An early warning model based on improved convolutional self-coding was constructed,and the adaptive threshold was adopted to set the reconstruction error threshold,which was verified by the application of gearbox and bearing.The accuracy of fault prediction is above 98% and 96% respectively.(2)The monitoring big data may contain the fault mode that has not yet appeared.Based on the existing fault mode,the direct identification of faults often causes the problem of diagnosis error and low diagnostic accuracy.In order to solve this problem,the hierarchical identification method is adopted to subdivide the fault mode step by step.The data were analyzed based on adaptive hierarchical clustering,and then the Sub DAE model was established according to the clustering results.Taking the gear box as an example,the fault diagnosis accuracy was more than 99%,and the fault diagnosis rate of the equipment was improved.(3)The fault diagnosis system based on the industrial Internet is developed to realize the condition monitoring,fault warning and diagnosis of equipment,break the information island of condition monitoring,excavate hidden trouble,and provide technical support for equipment maintenance.The objective of this thesis is to complete the research on key technologies such as distributed high-dimensional feature extraction,fault warning based on massive normal monitoring data,fault diagnosis based on layers,and develop the status monitoring and fault diagnosis system of rotating machinery based on the industrial Internet.Taking gear box and bearing as an example,this thesis verifies the effectiveness of intelligent diagnosis algorithm and monitors and diagnoses the faults of rotating machinery.Improve production efficiency and equipment reliability.