面向产品可靠性评估的退化和寿命数据分步融合方法  

Phased Fusion Method of Degradation and Lifetime Data for Product Reliability Evaluation

作　　者：李博文 贾祥 赵骞 郭波[1] LI Bowen;JIA Xiang;ZHAO Qian;GUO Bo(College of Systems Engineering,National University of Defense Technology,Changsha 410073;College of Information Communication,National University of Defense Technology,Xi’an 710106)

机构地区：[1]国防科技大学系统工程学院,长沙410073 [2]国防科技大学信息通信学院,西安710106

出　　处：《机械工程学报》2022年第16期430-440,共11页Journal of Mechanical Engineering

基　　金：国家自然科学基金(71801219,72071208);湖南省科技创新团队(2020RC4046);湖南省优秀青年基金(2021JJ20050)资助项目

摘　　要：产品的退化数据和寿命数据是两类重要的可靠性信息,通过融合利用两类数据可显著提高产品可靠性评估结果的精度。现有融合方法常利用Bayes理论,但计算过程繁琐复杂,需要通过提高计算量来保证结果精度,且两类样本数据量的不均衡性会导致出现“数据淹没”的问题。考虑这一问题,提出了基于退化和寿命数据分步融合的可靠性评估方法,首先利用产品的退化数据,建立随机过程退化模型,对参数进行初始估计,随后结合寿命数据对寿命样本的失效概率进行更新,分两步将两类信息进行折合计算,从而达到数据融合的目的。最后通过寿命分布曲线拟合计算退化模型参数,并给出产品的可靠性评估结果。仿真试验和算例分析表明,与传统Bayes融合方法相比,在新的数据融合思路下,该方法能够简化可靠性评估工作实施步骤,同时提高评估精度和运算效率,还有效避免了“数据淹没”问题。The degradation data and lifetime data of products are important reliability information.The accuracy of reliability evaluation usually can be improved by the fusion of these data.However,the existing methods are mainly based on Bayes theory where the calculation is complicated and required high computational cost to satisfy accuracy requirement.Meanwhile,the unbalanced data sample size can easily result in“data cover”problem.A phased fusion approach of degradation and lifetime data for product reliability assessment is proposed,which contained modelling the stochastic process degradation model on basis of degradation data,calculating the point estimations of parameters and updating the failure probability on basis of lifetime data.Data fusion is achieved by folding the two types of information in two phases.The estimations of parameters in degradation model and product assessment are finished by fitting the lifetime distribution finally.Fact is proved by simulation study and illustrative example that,compared with Bayes fusion approach,the proposed phased fusion method efficiently improves the accuracy and computational cost under new idea for reliability data fusion.More importantly,it also avoids the“data cover”problem effectively.

关 键 词：可靠性评估 Bayes理论 随机过程 失效概率 

分 类 号：TB114.3[理学—概率论与数理统计]