基于三支决策和强化学习的深空探测器非预期故障自主诊断与系统重构研究
结题报告
批准号:
61903015
项目类别:
青年科学基金项目
资助金额:
23.0 万元
负责人:
索明亮
依托单位:
学科分类:
F0301.控制理论与技术
结题年份:
2022
批准年份:
2019
项目状态:
已结题
项目参与者:
--
国基评审专家1V1指导 中标率高出同行96.8%
结合最新热点,提供专业选题建议
深度指导申报书撰写,确保创新可行
指导项目中标800+,快速提高中标率
客服二维码
微信扫码咨询
中文摘要
针对复杂不确定性环境下深空探测器面临的故障模式刻画不精确、非预期故障诊断自主能力不强、系统重构初始条件不确定等难点问题,借鉴人类认知决策过程原理,将三支决策和强化学习理论有机结合,依次开展基于三支决策和多属性决策的预期故障下诊断模型的自主建立与诊断决策、基于三支决策聚类和粗糙自学习的非预期状态下系统模型的自主学习与决策、基于三支决策和强化学习的非预期故障下自主系统重构方法研究。突破制约探测器自主诊断与系统重构难点问题中的关键科学问题,即非预期状态下三支决策参数自主确定及自主诊断建模问题、非预期故障下考虑探测器特殊约束的自主系统重构中强化学习机制构建问题,进而提升深空探测器非预期故障下自主诊断与系统重构的能力,并以探测器故障高发的姿控系统、电源系统和推进系统为对象进行方法和理论的验证。
英文摘要
This proposal focuses on the difficult issues of the deep space explorer under the complex and uncertain environment, which are the inaccurate description of fault mode, weak autonomy of unexpected fault diagnosis, and uncertainty of initial condition of system reconstruction. This proposal draws lessons from the principle of human cognitive decision-making process and combines the three-way decision and reinforcement learning theories organically, and then carries out the following three researches, i.e., the modeling and diagnosis decision for anticipated faults based on three-way decision and multiple attributes decision making theories, the self-learning and decision making for unexpected states based on three-way decision clustering and rough self-learning methods, and the autonomous reconstruction with respect to unexpected faults by employing three-way decision and reinforcement learning theories. This research breaks through the key scientific problems that restrict the autonomous diagnosis and system reconstruction of explorer, i.e., the autonomous diagnosis modeling and autonomous parameter determination of three-way decision under the unexpected states, the mechanism of reinforcement learning for autonomous system reconstruction after unexpected faults occurring by considering the special constraints of space explorer, so as to enhance the ability of autonomous diagnosis and system reconstruction under unexpected fault of deep space explorer, and the proposed methods and theories are verified by using the systems usually with more faults, which are the attitude control system, power system and propulsion system of space explorer.
项目针对复杂不确定性环境下深空探测器面临的故障模式刻画不精确、非预期故障诊断自主能力不强、系统重构初始条件不确定等难点问题,借鉴并深入研究了认知决策过程原理,以三支决策为基础理论支撑,开展深空探测器自主故障诊断与系统重构决策支持技术研究。(1)预期故障下诊断模型的自主建立与诊断决策。通过对数据的统计特性分析获取了一种新的风险决策矩阵,继而得到一种具有较高自主能力的决策模型,即单参数决策粗糙集,可实现诊断模型自主建立和诊断决策。(2)非预期状态下系统模型的自主学习与决策。利用最小风险融合的形式,实现决策结果的可信生成,继而可实现多故障、未知状态的判别。(3)非预期故障下自主系统重构方法研究。基于构建的决策粗糙集模型,实现对强化学习模型的修正,建立一种适用于探测器自主重构决策的被动强化学习模型,继而实现了非预期状态下的自主系统重构管控决策。通过以上研究,发表标注该基金的论文成果14篇,其中SCI期刊13篇,国际会议1篇,受理发明专利6项。培养硕士和博士研究生4人。申报获批教育部科技进步一等奖1项。
期刊论文列表
专著列表
科研奖励列表
会议论文列表
专利列表
Associated Fault Diagnosis of Power Supply Systems Based on Graph Matching: A Knowledge and Data Fusion Approach
基于图匹配的供电系统关联故障诊断:知识与数据融合方法
DOI:10.3390/math10224306
发表时间:2022-11-01
期刊:MATHEMATICS
影响因子:2.4
作者:Tao,Laifa;Liu,Haifei;Wang,Chao
通讯作者:Wang,Chao
Simultaneous-Fault Diagnosis of Satellite Power System Based on Fuzzy Neighborhood ζ-Decision-Theoretic Rough Set
基于模糊邻域γ决策理论粗糙集的卫星电力系统同步故障诊断
DOI:10.3390/math10193414
发表时间:2022-09
期刊:Mathematics
影响因子:2.4
作者:Laifa Tao;Chao Wang;Yuan Jia;Ruzhi Zhou;Tong Zhang;Yiling Chen;Chen Lu;Mingliang Suo
通讯作者:Mingliang Suo
Degradation Dynamics Cognition and Prediction of Li-Ion Battery: An Integrated Methodology for Alleviating Range Anxiety
锂离子电池的退化动力学认知和预测:缓解里程焦虑的综合方法
DOI:10.1109/access.2020.3029397
发表时间:2020
期刊:IEEE Access
影响因子:3.9
作者:Tao Laifa;Zhang Tong;Hao Jie;Wang Xiaolin;Lu Chen;Suo Mingliang;Ding Yu
通讯作者:Ding Yu
Single-parameter decision-theoretic rough set
单参数决策理论粗糙集
DOI:10.1016/j.ins.2020.05.124
发表时间:2020-10-01
期刊:INFORMATION SCIENCES
影响因子:8.1
作者:Suo, Mingliang;Tao, Laifa;Zhang, Tong
通讯作者:Zhang, Tong
DOI:10.1016/j.eswa.2022.116503
发表时间:2022-01
期刊:Expert Systems With Applications
影响因子:8.5
作者:Tong Zhang;Laifa Tao;Xiaoding Wang;Chuanfang Zhang;Shihao Li;Hao Jie;Chen Lin;Mingliang Suo
通讯作者:Mingliang Suo
国内基金
海外基金