基于稀疏变分贝叶斯的复杂装备剩余寿命预测  

Remaining Useful Life Prediction of Complex Equipment Based on Sparse Variational Bayes

作　　者：张毅 李文博[3,4] 王浩 林文乙 刘切 ZHANG Yi;LI Wenbo;WANG Hao;LIN Wenyi;LIU Qie(School of Automation,Chongqing University,Chongqing 400030,China;Southwest China Institute of Electronic Technology,Chengdu 610036,China;Beijing Institute of Control Engineering,Beijing 100094,China;College of Automation Engineering,Nanjing University of Aeronautics and Astronautics,Nanjing 211106,China)

机构地区：[1]重庆大学自动化学院,重庆400030 [2]中国西南电子技术研究所,成都610036 [3]北京控制工程研究所,北京100094 [4]南京航空航天大学自动化学院,南京211106

出　　处：《宇航学报》2025年第2期253-261,共9页Journal of Astronautics

基　　金：国家重点研发计划(2021YFB1715000)。

摘　　要：航天器等复杂装备的剩余使用寿命(RUL)预测是装备健康管理的核心技术。针对装备存在的多个反映健康状态的变量,以及多工况和数据不确定性带来的建模挑战,提出了基于稀疏变分贝叶斯的剩余使用寿命关键变量选择及建模方法。通过回归模型描述装备健康状态与监测变量间的关系,同时考虑测量噪声和参数的不确定性,采用变分贝叶斯进行参数后验估计。在此基础上,创新性地提出了基于稀疏贝叶斯的健康状态变量确定方法。此方法首次将稀疏贝叶斯应用于剩余寿命建模,并在基准数据集上得到了验证。结果显示,相较于现有方法,所提方法在使用更少变量的同时,实现了更高的预测精度。The prediction of remaining useful life(RUL)for complex equipment,such as spacecraft,is a pivotal technique in equipment health management.Often,such equipment has multiple variables related to its health status,complicating RUL prediction.However,existing RUL prediction methods face challenges under varying working conditions and uncertainties,including low accuracy and inadequate key variable selection.To address these issues,we propose a method for key variable selection and RUL modeling based on sparse variational Bayes.This method employs a regression model to capture the relationship between equipment health state and monitoring variables.To handle uncertainties,both measurement noise and model parameters are treated as random variables,with variational Bayes used to estimate the posterior distribution of these parameters.Furthermore,we introduce a method for determining health state variables using the sparse Bayesian approach.This represents the first application of sparse Bayesian methods to RUL modeling and has been validated on a benchmark RUL prediction dataset.Our method achieves more accurate prediction results with a reduced number of variables compared to existing approaches.

关 键 词：剩余使用寿命预测 稀疏变分贝叶斯 相似性 变量选择 

分 类 号：V41[航空宇航科学与技术—航空宇航推进理论与工程]