考虑不确定性与相关性的退化过程建模与动态评估方法

Equipment undergoes a series of different degradation states from normal to failure. For modeling and estimation of degradation processes, we should consider the multi-source uncertainties caused by individual differences, external dynamic working conditions, unknown influencing factors, model selections and measurement errors, as well as dependences among different degradation processes due to structure and function relations. To solve this problem, first, this project proposes a comprehensive multi-source uncertainty characterization method for single degradation process, and establish the degradation model considering multi-source uncertainties based on stochastic process models, to describe the overall variation of degradation process. Secondly, in consideration of the degradation process dependences, a characterization method of degradation process dependences under different conditions is proposed, and an overall degradation model considering the multi-source uncertainty and degradation process dependence is established. Finally, based on the actual monitoring data of particular degradation process, a dynamic estimation method of particular degradation process considering uncertainty and dependence is proposed to realize the cognition of the behavioral characteristics. A case study on forming drum transmission box is conducted to demonstrate the validity and efficiency of the proposed models and methods. The purpose of this study is to improve the accuracy of the estimation results for particular degradation process under the influence of uncertainty and dependence, which is of important theoretical and practical significance for maintaining service performance and increasing service life of equipment.

设备从正常运行到失效会经历一系列不同程度的退化状态。在进行退化过程建模与评估工作时，一方面要考虑个体差异、外部多变工况、未知影响因素、模型选择与测量误差导致的多源不确定性，另一方面也要考虑由于设备结构与功能联系所导致不同退化过程间相关性。针对该问题，本项目首先针对单一退化过程，提出多源不确定性综合表征方法，基于随机过程模型建立考虑多源不确定性的总体退化模型，描述退化过程总体变化规律。其次，考虑退化过程相关性，提出不同情况下退化过程相关性表征方法，建立考虑多源不确定性与退化过程相关性的总体退化模型。最后，融合个体退化过程实际监测数据，提出考虑不确定性与相关性的个体退化过程动态评估方法，实现个体退化过程行为特征认知。以成型鼓传动箱为应用对象，对模型与方法正确性与有效性进行验证。该研究旨在提高不确定性与相关性影响下个体退化过程评估结果准确性，对于设备服役性能保持与寿命提升具有重要理论和现实意义。

设备从正常运行到失效会经历一系列不同程度的退化状态。在进行退化过程建模与评估工作时，一方面要考虑个体差异、外部多变工况、未知影响因素、模型选择与测量误差导致的多源不确定性，另一方面也要考虑由于设备结构与功能联系所导致不同退化过程间相关性。为此，本项目主要研究了考虑多源不确定性的总体退化模型的建立与综合考虑多源不确定性与不同退化过程间相关性的总体退化模型的建立，并在此基础上，研究了考虑多源不确定性与不同退化过程间相关性的个体退化过程动态评估方法。.针对以上研究内容，本项目提出了多源不确定性的综合表征方法，在常用的退化模型框架基础上，基于数理统计方法建立了描述产品退化规律的总体退化模型；并在样本量较小的情况下提出了考虑认知不确定性的退化建模方法；此外，研究了载荷共享特性所导致的退化过程相关性，建立了一种载荷共享表决系统的总体退化模型；针对模型不确定性较高的问题，提出了基于自适应核密度估计的辅助粒子滤波方法用于估计退化状态；针对个体退化数据量较少，认知不确定性影响较大的情况，提出了基于相似性的参数更新与退化状态动态评估方法；最后，针对综合考虑不确定性与相关性的个体退化过程动态评估问题，构建了状态空间模型，用于估计个体设备的退化状态与未知退化参数。.围绕以上研究，项目组在IEEE汇刊、RESS、MSSP等国内外重要期刊共发表论文16篇，项目负责人为第一作者或者通讯作者，其中SCI期刊论文15篇，且12篇发表在国际Q1区期刊上，获授权发明专利8项。此外，项目组还将模型与方法成功应用于智能制造、核电等领域。