Research On Remaining Useful Life Prediction Model Of Electric Spindle With Multi-parameter Related Degradation

Remaining useful life prediction is the focus and key to achieving intelligent predictive maintenance.The electric spindle is a key core component of high-end CNC machine tools.During the operation of the machine tool,as the performance of the electric spindle deteriorates,its remaining useful life gradually decreases,and the possibility of failure gradually increases.The traditional remaining useful life prediction of CNC machine tool electric spindle is based on onedimensional degradation data or sudden failure data.In fact,its performance degradation and sudden failure coexist and compete.In the era of big data,integrating multi-sensor monitoring data and considering the correlation between degradation and sudden failure to predict remaining useful life is an effective guarantee for the reliability and safety of complex electromechanical products,and is also an important prerequisite for achieving lifecycle health management.This article takes the machining center electric spindle product as the research object,integrates Miner linear cumulative damage theory,information entropy theory,random degradation process,and Copula function,conducts research work from the aspects of accelerated degradation test design,multiparameter degradation process correlation analysis,establishes a multi-parameter related degradation remaining useful life prediction model and a failure dependence remaining useful prediction model.The main research content and results of this article are as follows:(1)Step stress acceleration degradation test design for electric spindle is conducted.Historical fault mechanism analysis is used to clarify electric spindle step stress acceleration degradation test items and accelerated stress types.The improved Miner linear cumulative damage theory is applied to model acceleration factor.Based on the principle of consistency in failure mechanisms,step acceleration stress levels are determined.Two-step maximum likelihood method is applied to estimate Weibull distribution parameters under highly truncated data.According to electric spindle fault model and acceleration factor,step stress acceleration degradation test time is designed.Thus,accelerated degradation test design theory is perfected.(2)A correlation analysis method for multi-parameter degradation process of electric spindle is proposed.The correlation of electric spindle multi-parameter degradation process is the basis of information fusion.Aiming at the problem that ignoring uncertainty of degradation process information and dynamic time-varying characteristics of correlation relationships when conducting correlation analysis,based on the multi-parameter degradation information transfer entropy and parameter degradation information entropy,the net normalized transfer entropy under Markov processes is proposed for the first time.Combined information entropy difference and maximum information entropy symbolization processing,computational complexity of net normalized transfer entropy is reduced.the correlation direction of multi-parameter degradation is analyzed,the correlation strength of multi-parameter degradation processes is quantified.Thus,the theory of correlation analysis is expanded.(3)Multi-parameter related degradation remaining useful life prediction model for electric spindle is established.Aiming at the problem of determining the fusion weights of heterogeneous parameters with different physical meanings when constructing composite degradation indicators based on weighted fusion models,an importance evaluation method based on net normalized transfer entropy improved Page Rank algorithm is proposed.Using importance value as the weight,a weighted fusion model is applied to construct electric spindle composite degradation index.Under the assumption of Wiener degradation process,a dual acceleration stress degradation model for electric spindle is established,and it is generalized into a multi-parameter related degradation remaining useful life prediction model using inverse power-law model.The rationality of the proposed model is verified using mean absolute percentage error and other evaluation indicators.(4)A failure dependence remaining useful life prediction model for electric spindle is established.Aiming at the mechanism deviation of predicting the remaining useful life of electric spindles based on the independent assumption of degradation process and sudden failure,the remaining useful life prediction model for electric spindles considering the dependence between sudden failure and performance degradation is proposed.Equivalenting the step stress accelerated degradation test to constant stress acceleration tests,generating zero sudden failure data,sudden failure model related to degradation degree is constructed based on parameter estimation by Bayes and IBA-SVR.Taking the degradation failure model of composite degradation index under normal working stress and sudden failure model as the edge distribution,Copula function is introduced to establish a joint distribution,and failure dependence remaining useful life prediction model for electric spindle is derived accordingly,thus perfect remaining useful life prediction methods.