Research On Remaining Life Prediction Considering Dynamic Transition Of Degradation State

Prognostics and health management(PHM)identifies and manages the occurrence of failures by predicting the remaining life of the system,and provides guarantees for the stability and safe use of the system.Therefore,PHM has become a hot research direction of multi-discipline,such as electronics,reliability,and information,and provides new research ideas and methods for the optimal maintenance of equipment.In the entire PHM system,the remaining useful life(RUL)prediction research is the key direction,which can save the maintenance cost of the system to a greater extent and increase the stability of the system.This has become the research of system failure prediction and maintenance decisionmaking.New hotspots and new trends.This paper uses the received real-time monitoring data that can characterize the degradation characteristics of the gearbox to study the remaining life prediction method based on the kernel density estimation,and verifies the reliability of the prediction model with the prediction accuracy,which provides for the establishment of a systematic prediction and health management system Theoretical support and application cases.The main contents are as follows:(1)For the degraded state of a complex system,a sudden change will occur in the actual operating environment,which will change the trend and law of the degraded state of the system.The mutation point hides a large amount of remaining life information,which plays a key role in identifying changes in health status.Even if a single prediction model has the ability to adjust parameters adaptively,it is difficult to monitor the dynamic transition of the degraded state under the limitations of the model itself.An adaptive multi-stage degradation model based on real-time clustering is proposed,which accurately clusters the data representing the different degradation states of the system,obtains the dynamic transition mode of the degradation state,and then studies the data distribution characteristics,considering the high computational complexity of real-time estimation The problem is to establish a recursive model of adaptive smoothing parameter kernel density estimation for dynamic transition of degradation state to realize real-time remaining life prediction.(2)Aiming at the existing kernel density estimation models are assumed to be independent and identically distributed time-invariant systems,this paper constructs a real-time residual life prediction model for hybrid time-varying dynamic kernel density estimation,and adaptive smoothing parameters based on real-time monitoring data to solve different For the problem of sample contribution,a time-varying adaptive kernel density estimation model for predicting the remaining life is established considering the dynamic transition of the component degradation state.