Weak Fault Feature Extraction And Intelligent Fault Evaluation For Rotating Machinery In Actual Complex Working Condition

Fault diagnosis of machinery is an important guarantee of safety measures.With the increasing attention of production enterprises to safety,efficiency and cost,there are growing demands for equipment health management.There are some problems in the research of mechanical fault diagnosis: emphasis on theory research but despise application,attach importance to experiment but despise industrial scene,value technology but despise experience.As the results,the commonly used mechanical fault diagnosis method cannot meet the diagnostic requirements of complex operating conditions in industrial field,the operation of the instrument is complex and high theoretical foundation is needed for the operator.From the point of view of equipment maintenance personnel,the mechanical fault diagnosis technology is "not good to be used" and "won't be used" in the industrial field.Therefore,this paper studies the fault diagnosis method of rotating machinery based on vibration signal analysis combined with the demand of the production enterprise.On one hand,considering the complex conditions of strong interference and variable speed of equipment,the vibration signal processing method of rotating machinery under complex conditions is studied.On the other hand,considering the shortage of equipment maintenance workers on the theoretical basis,this paper studies the intelligent health assessment method under complex working conditions.The application-oriented theoretical research is realized through the two aspects of the work above.Theoretical research and practical work are also verified by industrial field application.In order to extract weak fault features from vibration signals of rotating machinery in actual complex working condition,the problems of endpoint effect and modal aliasing are improved based on Hilbert-Huang transform theory,and a fast CEEMDAN algorithm is proposed to simplify the parameter adjustment of vibration signal analysis method.Then,the energy weighting method based on improved CEEMDAN adaptive non-parametric time-frequency analysis is proposed to enhance and extract the fault impulse characteristics under strong noise interference.The energy weighting method is validated by simulated vibration signal analysis and fault diagnosis experiments of train bearings.It is proved that the method has good effect on extracting the fault feature components under strong noise interference.In addition,the application scope of the energy weighting method is further expanded.The energy weighting method for variable speed conditions is proposed.The fault characteristics in actual complex working condition are enhanced by tacoless order tracking analysis and energy weighting analysis of vibration signals under variable speed and strong disturbance.A test and analysis system is developed based on the relevant theories.The validity of the energy weighting method in variable speed condition is verified by simulation signal analysis and fault diagnosis of the bearing of the hoist.Besides the study of vibration signal feature extraction method and the realization of fault diagnosis and positioning,this paper has also studied the intelligent health assessment of equipment.The intelligent fault diagnosis method based on deep learning for variable speed equipment is proposed.The vibration signal under variable condition is preprocessed into even angle signal by tacholess order tracking firstly.Then the intelligent extraction and classification of vibration signal features are utilized by using the deep convolutional model.So that,the intelligent diagnosis method is realized.Taking the blast furnace top gas recovery turbine unit and the gearbox of portal crane as the object of diagnosis,the vibration signals accumulated by long-term tracking test are analyzed,and the relevant theories are applied and verified in the industrial field.