复杂条件下随机退化设备的剩余寿命预测理论与方法研究

Remaining useful life prediction plays an irreplaceable role in ensuring the safe and reliable operation of engineering equipment. Due to the complexity of engineering equipment, the associated performance degradation is difficult to physically model. Thus, the methods modeling the degradation data of equipment by stochastic process models and predicting the fisrt hitting time have become rather popular in the field of remaining useful life prediction. However, current methods pay primary attention to the equipment with single time scale, single failure model, and single degradation variable. When the equipment is operating under complex conditions including multiple time scales, multiple failure modes, and multiple degradation variables, new issues and challenges are encountered in the theories and methods for remaining useful life prediction. Therefore, the remaining useful life prediction of equipment operating in complex conditions is a frontier science problem driven by the engineering demands. ..Oriented to applications in key equipment of missile control systems, this project adopts the idea of data-driven stochastic degradation process modeling to deal with the frontier science problem and studies the novel remaining useful life prediction theories and methods of equipment operating in complex conditions. These theories and methods include: remaining useful life prediction for equipment with multiple time scales, remaining useful life prediction for equipment with multiple failure modes, remaining useful life prediction for equipment with multiple degradation variables. The key scientific problems encountered in aforementioned studies will be emphatically resolved including the degradation process characterizing and modeling problem for equipment under complex conditions and the associated solution problem of the remaining useful life for equipment under complex conditions, which aim at revealing the degradation progression process and the degradation modeling principle of equipment under complex conditions, and addressing the interaction mechanism of complex conditions and the remaining useful life. As such, the research results of this project will provide the basic theories and the key techniques for the remaining useful life for equipment under complex conditions.

剩余寿命预测是保证工程设备安全可靠运行的杀手锏。由于工程设备的复杂性，其性能退化难以机理建模，利用随机过程模型将设备退化数据建模并预测其首达时间，已成为当前非常流行的剩余寿命预测方法。然而，已有理论与方法主要针对单时间尺度、单失效模式、单变量的设备，当设备运行于复杂条件（如多时间尺度、多失效模式、多变量等）下时，遇到了新问题与新挑战。因此，复杂条件下设备剩余寿命预测是工程需求驱动的前沿科学问题。本项目以导弹控制系统关键设备为应用背景和对象，采用数据驱动随机退化过程建模思想，研究复杂条件下设备剩余寿命预测新理论与新方法，分别为多时间尺度、多失效模式、多变量等条件下设备的剩余寿命预测，重点解决复杂条件下随机退化过程表示与建模、剩余寿命分布求解等关键问题，旨在揭示复杂条件下设备退化演化过程及建模原理，阐述复杂条件与剩余寿命的影响机制，为复杂条件下设备剩余寿命预测提供有价值的基础理论与关键技术。

剩余寿命预测是保障复杂工程设备长周期安全可靠运行的关键技术。现有针对随机退化设备的剩余寿命预测研究主要针对单时间尺度、单失效模式、单变量的设备，难以应对于复杂条件下随机退化设备的剩余寿命预测问题。本项目瞄准复杂条件下设备剩余寿命预测这一前沿科学问题，研究了复杂条件下设备剩余寿命预测新理论与新方法，主要包括：（1）多时间尺度下随机退化设备剩余寿命预测；（2）多失效模式下随机退化设备剩余寿命预测；（3）多变量随机退化设备剩余寿命预测。.针对以上研究内容，项目组提出了多时间尺度下设备剩余寿命预测方法，定量刻画了不同时间尺度之间的不确定关系及对剩余寿命的影响；建立了多失效模式下设备随机退化过程模型，分别提出了考虑状态检测影响、多阶段退化、随机冲击影响、运行状态切换等因素的设备剩余寿命预测方法；针对多变量随机退化设备，建立了融合多变量数据的设备复合健康指标时变演化模型，提出了多变量随机退化设备退化模型参数和剩余寿命分布在线更新方法；在剩余寿命预测背景下提出了非失效退化轨迹、最后逃逸时间、数模联动等新概念、新原理。通过以上理论研究，揭示了复杂条件下设备退化演化过程及建模原理，阐述了复杂条件与剩余寿命的影响机制，为复杂条件下设备剩余寿命预测提供了基础理论与关键技术。部分研究成果应用于导弹设备的剩余寿命预测与延寿决策。.围绕以上研究，项目组在EJOR、IEEE汇刊、RESS、MSSP、自动化学报等国内外重要期刊发表学术论文43篇，其中上中科院二区以上期刊论文27篇，2篇入选ESI高被引论文，6篇论文被所在期刊或会议评为高影响力论文或优秀论文，获授权发明专利10项。负责人获2019年国家自然科学二等奖(3)、2018年教育部自然科学一等奖(3)、CAA自然科学二等奖(3)各1项，入选国家自然科学优秀青年基金资助(2019)、军队学科拔尖人才(2021)、火箭军导弹专家(2020)、火箭军十大砺剑尖兵(2019)。