Study On Fault Diagnosis And Degradation Identification Technology Of Rolling Bearing Under Unsteady Condition

With the development of large-scale mechanical and electrical equipment in different fields towards automation,intelligence and high reliability,the technical demand for monitoring and analyzing the operation status of mechanical and electrical equipment and its important parts and efficient operation and maintenance is growing.As an important part of rotating machinery,rolling bearing is widely used in large mechanical and electrical equipment of various industries.The bad working conditions make the bearing performance gradually degenerate and produce faults.Traditional signal analysis and fault diagnosis methods are difficult to meet the requirements.Therefore,rolling bearing fault diagnosis and degradation identification research is an important premise to improve the monitoring and diagnosis technology level of largescale mechanical and electrical equipment and ensure the equipment operation safety.However,rolling bearings are often affected by unsteady conditions including different speed,load and variable speed under actual working conditions.In addition,mining and construction machinery and other fields often appear low speed and heavy load conditions,which has distinct industry characteristics.These make the fault diagnosis and degradation identification of rolling bearing face many difficulties.In this paper,the rolling bearing under unsteady condition is taken as the research object.Through the research on local fault vibration mechanism,fault diagnosis with unknown variable speed,single pulse feature extraction,degradation identification and size measurement,the rolling bearing fault diagnosis and degradation identification technology under unsteady condition is formed,which provides theoretical basis and technical support for the safe operation of large mechanical and electrical equipment.The main research contents are as follows:(1)Based on the summary of function structure and fault characteristics of rolling bearing,a simplified rolling bearing dynamic model was established and the fault bearing vibration characteristics were analyzed.According to the research requirements,a rolling bearing fault diagnosis experimental device was built and vibration signal acquisition experiment was carried out.Vibration signals of rolling bearing with various fault types and states under different working conditions were obtained,which provides theoretical basis and basic data for the subsequent research.(2)Aiming at rolling bearing local fault vibration mechanism,a model-based measurement method of local fault size was proposed.The motion path and collision of roller were analyzed.The additional displacement and impact force formulas of inner ring and outer ring fault bearing were derived,and the rolling bearing local fault vibration model was improved.Furthermore,the influence of impact force and radial load on bearing vibration signal characteristics was analyzed.The validated simulation results were used as the reference conditions for local fault size measurement,which simplifies the mathematical model.The proposed method considers the influence of rotational speed and radial load on the dual impulse interval time,and can obtain high measurement accuracy for small local fault under heavy load condition.The research results provide theoretical basis for the following research of rolling bearing fault diagnosis and degradation identification technology.(3)For the influence of unknown variable speed conditions on bearing fault diagnosis,a fault diagnosis method based on multi curve extraction and selection,VKF and GD was proposed.Multi time frequency curves of different components are extracted iteratively from TFR of bearing signal and its envelope using curve extraction algorithm.According to the relative order,the curve attribution can be judged,and the high-precision ISRF curve and fault time-frequency curve are selected and calculated.The current bearing fault type can be determined by judging the deviation between realtime ratio of curves and FCC.For the problem that the time-frequency characteristics of early weak fault are scattered and not obvious under variable speed condition,the fault pulse components are extracted and demodulated by combining VKF and GD.The reconstructed demodulation spectrum is constructed to concentrate the scattered features and form the feature peak to realize the diagnosis of early weak fault type of rolling bearing.The proposed method can accurately diagnose bearing fault type under unknown variable speed condition,and can also identify weak bearing fault after signal demodulation.(4)For the effective extraction of rolling bearing fault feature information under unsteady and low speed condition,a monopulse feature extraction method based on phase scan was proposed.The fault phase function and key phase function are constructed using the ISRF signal.The multi fault periods are divided,and the bearing vibration envelope signal is scanned.The influence of noise and detail feature loss is eliminated by ensemble average,and the smooth monopulse waveform feature can be obtained.The peak value of monopulse waveform is used as the evaluation index to iteratively calibrate multi fault periods,and the feature amplitude is standardized to eliminate the influence of working conditions.The proposed method can effectively extract the monopulse features of inner ring and outer ring faults.The extracted monopulse feature is highly sensitive to the degradation state of local fault,and can overcome the influence of different speed,load and unsteady conditions.(5)Aiming at the identification difficulty of rolling bearing fault type and fault degree,a rolling bearing degradation identification method based on monopulse feature was studied.The feature matrix with bearing fault information is constructed using monopulse feature vectors,and the deep learning network is constructed based on CNN to accurately diagnose bearing fault types.In order to solve the problems of redundant information in sample data and misdiagnosis of small faults when a single network is used to identify bearing fault types and degradation states at the same time,a degradation identification method using multi network to process different fault types data was studied.According to the bearing fault type,the monopulse feature vector is selected,and the corresponding network is called for analysis.The proposed method can accurately identify the inner and outer ring fault types and their degradation states under different and unsteady conditions,and can also accurately measure the local fault size.This paper has 73 figures,9 tables and 167 references.