考虑时变相关性的多退化特征设备剩余寿命预测

打开文本图片集

关键词：剩余寿命预测;退化特征；时变相关性；时变Copula函数中图分类号：TH17文献标志码：ADOI：10.7652/xjtuxb202510020 文章编号：0253-987X（2025）10-0210-11

Remaining Useful Life Prediction of Multi-Degradation Feature Equipment Considering Time-Varying Correlation

ZHANG Jialing，ZHANG Jianxun，DU Dangbo，ZHANG Zhengxin，HU Changhua （Zhijian Laboratory，Rocket Force University of Engineering，Xi'an 71oo25，China）

Abstract： To address the issue of insufficient adaptability of existing methods in remaining useful life （RUL） prediction for multi-degradation feature equipment，where static correlation models or potential assumptions about time-varying correlation forms lead to poor performance under complex operating conditions，a time-varying Copula-based RUL prediction method is proposed. First，nonlinear Wiener processes are employed to establish models for individual degradation features，while time-varying Copula functions are introduced to capture dynamic correlations among multiple degradation models. Second，parameter identification for both the nonlinear Wiener processes and the time-varying Copula functions is achieved via Bayesian Markov chain Monte Carlo Metropolis-Hastings and maximum likelihood estimation. Finally，the optimal time-varying Copula form is selected using the Akaike information criterion，and the joint distribution of predicted RUL is derived by integrating marginal distributions. Verification is performed using numerical simulation data and actual data from a blast furnace wall. Experimental results indicate that the proposed method reduces the mean square error of RUL predictions by 8.62% compared to independent modeling of three degradation characteristics， 5.57% over static Copula function， and 17.85% against dynamic Bayesian network methods. The proposed method can accurately characterizes the impact of time-varying degradation correlations of multiple degenerative features on RUL prediction，offering a more effective solution for predicting the remaining useful life of multi-degradation feature equipment.

Keywords： remaining useful life； degradation processes； time-varying correlation； time-varyingCopula function

剩余寿命（RUL）预测是预测与健康管理（PHM技术的关键核心部分，准确预测系统剩余寿命能够为系统后续的维修决策提供重要理论依据和充足的信息支撑[1-2]。（剩余16136字）