Research On Early Warning And Fault Pattern Recognation Of Rotating Machinery

Rotating machinery is a key production tool in modern enterprises such as petroleum and chemical industry.Carrying out condition monitoring and fault diagnosis on the equipment can ensure the safety and reliable operation of these equipment,which can obtain huge economic and social benefits.Rotating machinery equipment condition monitoring and fault diagnosis are very important to carry out predictive maintenance,which can extend the service life of machinery,reduce periodic maintenance costs,and ensure the safety of equipment operation.In this paper,the early fault warning and failure mode recognition methods for rotating machinery are studied.Based on the reliability function of rotating machinery equipment and trend filtering technology,using the change of the slope absolute value of filtered vibration trend curve,an early fault warning model based on reliability function and trend filtering was constructed.Based on cluster analysis methods,a K-means clustering fault pattern recognition model was constructed,and the clustering centers for different faults were obtained based on monitoring data training.A data-driven fault diagnosis method for rotating machinery based on the Euclidean distance criterion of clustering centers was studied.Reliability Centered Maintenance(RCM)and health degree assessment were studied,and a dynamic risk assessment model was constructed.The intelligent early fault warning method and pattern recognition method were combined to design the dynamic risk assessment and maintenance decision system of rotating machinery.The main research contents are as follows:(1)Research on early fault warning method of rotating machinery.Extracting equipment vibration characteristic signals based on the reliability function of rotating machinery operation,that is,time-domain cross-correlation function,frequency domain aggregation function and spectral distance indicator function.Using l1 trend filtering technology to filter the vibration characteristic signals,according to the change of the slope absolute value of the filtered characteristic curve,an early fault warning model based on the reliability function and trend filtering technology was constructed.The model was trained using the NSFI/UCR rolling bearing No.2 full life cycle experiment.Finally,using the data of rubbing,misalignment and imbalance fault cases of centrifugal compressor to verify the engineering effectiveness of the method.(2)Research on fault pattern recognition model of rotating machinery.Based on K-means clustering analysis method,a K-means clustering fault pattern recognition model was constructed,and fault feature parameters with high fault recognition sensitivity were selected as model input parameters.The clustering centers of different faults were obtained by training based on the monitoring data of centrifugal compressors and experimental data of bearings,a data-driven fault diagnosis method for rotating machinery based on the Euclidean distance criterion of clustering centers was studied,and verified by using engineering cases.(3)Research on dynamic risk assessment and maintenance decision system design.Based on the traditional RCM risk evaluation method and health degree evaluation method of rotating machinery,a dynamic risk assessment model is constructed.Based on the level of risk,a decision tree model is used to determine targeted maintenance methods and maintenance contents.Combining intelligent early fault warning methods and failure mode identification methods,a dynamic risk assessment and maintenance decision system was established.The system's heterogeneous data collection module,sequential data storage module,sequential data preprocessing module,model training and optimization module,knowledge base module and maintenance decision module were designed.Finally,the design and development of the system were completed.