Research On Bearing Health Management Based On Time Series Prediction

Rolling bearings are widely used in many mechanical devices.As a key bearing part,it is long-term used in complex and poor environment.Thus it has high failure rate,and easy to damage,which is the main source of mechanical failure.So the rolling bearing degradation assessment and remaining life prediction technology has become the urgent needs in today’s industry.Traditional reliability analysis methods and maintenance strategies can no longer meet the equipment requirements.Therefore,fault prediction and life prediction analysis methods based on online monitoring have emerged.It can provide basis and decision for maintenance before the occurrence of fault,and will greatly reduce the occurrence of catastrophic accidents.Taking rolling bearing as the research object,this paper has carried out in-depth research on the degradation mechanism analysis,vibration feature extraction,performance degradation evaluation and residual service life prediction of bearing.The main research contents of this paper are as follows:This paper first analyzes the bearing Failure mechanism,and carries out a Failure Tree Analysis qualitative Analysis on the bearing system,so as to point out the fault modes with the highest influence degree and several Failure causes with the greatest structural importance,providing a basis for the optimization design of bearing reliability.Then,this paper elaborates the testability of spindle bearing fault,establishes the relationship between its vibration mechanism and fault characteristics,and turns the key problem of spindle bearing fault prediction into fault feature extraction based on vibration signal.After introducing the conventional method of vibration signal analysis and multidomain feature extraction,a feature extraction method based on and Hilbert transform is adopted in this paper to accurately extract and identify the early fault characteristics of spindle bearing.By EMD decomposition of vibration signals in different states,IMF components containing fault characteristic information of each order are obtained,and the common points of modal components in different bearing faults are analyzed,so as to realize numerical identification of the health status of spindle bearing in different states.On this basis,the root mean square value mode decomposition of the energy return one becomes Health Index,implemented to the distance between the test sample to the trouble-free sample to describe the performance degradation of bearing state,and take advantage of the time series prediction,based on historical time series Autoregressive Moving Average model,set up the spindle bearing performance degradation prediction curve,fitting degree can reach 96%.Finally,this paper develops a rolling bearing performance degradation and life prediction system by using hidden Markov model and combining the above methods.The test results show that this method can accurately and stably extract and identify the early fault characteristics of the spindle bearing,further improve the accuracy of fault diagnosis,and provide a basis for the maintenance of the bearing of the same type.