Nonlinear degradation data modeling method based on Box-Cox transformation and random coefficient regression model
Transformation method is an important and feasible way to deal with nonlinear degradation modeling and life prediction of equipment.At present,the common transformation methods are mainly logarithmic transformation and time scale transformation,and their application scope is limited.In view of this,this paper proposes a nonlinear degradation data modeling method based on Box-Cox transformation(BCT).Firstly,the nonlinear degradation data are transformed by BCT,and the transformed degradation data are modeled by linear random coefficient regression model.Then,by constructing the probability density function of the observation data,the maximum likelihood estimation is used to identify the model parameters in BCT,and Bayesian theory is used to update the parameters online to realize the dynamic calibration of the degradation model.Finally,the actual degradation data of lithium battery and bearing after BCT are analyzed,and the linearity and correlation coefficient are increased by 69.63%and 9.19%respectively,which verifies the feasibility and potential engineering application value of this method.
NETL
NSTL
万方数据
维普
