Attribute reduction is a research hotspot in rough set theory. Traditional heuristic attribute reduction methods add the most important attribute to the decision attribute set each time, resulting in multiple redundan...Attribute reduction is a research hotspot in rough set theory. Traditional heuristic attribute reduction methods add the most important attribute to the decision attribute set each time, resulting in multiple redundant attribute calculations, high time consumption, and low reduction efficiency. In this paper, based on the idea of sequential three-branch decision classification domain, attributes are treated as objects of three-branch division, and attributes are divided into core attributes, relatively necessary attributes, and unnecessary attributes using attribute importance and thresholds. Core attributes are added to the decision attribute set, unnecessary attributes are rejected from being added, and relatively necessary attributes are repeatedly divided until the reduction result is obtained. Experiments were conducted on 8 groups of UCI datasets, and the results show that, compared to traditional reduction methods, the method proposed in this paper can effectively reduce time consumption while ensuring classification performance.展开更多
Testing is the premise and foundation of realizing equipment health management （EHM）. To address the problem that the static periodic test strategy may cause deficient test or excessive test, a dynamic sequential te...Testing is the premise and foundation of realizing equipment health management （EHM）. To address the problem that the static periodic test strategy may cause deficient test or excessive test, a dynamic sequential test strategy （DSTS） for EHM is presented. Considering the situation that equipment health state is not completely observable in reality, a DSTS optimization method based on partially observable semi-Markov decision pro- cess （POSMDP） is proposed. Firstly, an equipment health state degradation model is constructed by Markov process, and the control limit maintenance policy is also introduced. Secondly, POSMDP is formulated in great detail. And then, POSMDP is converted to completely observable belief semi-Markov decision process （BSMDP） through belief state. The optimal equation and the corresponding optimal DSTS, which minimize the long-run ex- pected average cost per unit time, are obtained with BSMDP. The results of application in complex equipment show that the proposed DSTS is feasible and effective.展开更多
近年来,随着智慧医疗的日益普及,在线医疗平台已逐步发展为满足大众基本医疗需求的重要渠道。为患者推荐合适的医生是在线问诊中的一个重要过程,优化推荐能力不仅可以提高患者的满意度,还能够推动在线医疗平台的发展。与传统推荐系统不...近年来,随着智慧医疗的日益普及,在线医疗平台已逐步发展为满足大众基本医疗需求的重要渠道。为患者推荐合适的医生是在线问诊中的一个重要过程,优化推荐能力不仅可以提高患者的满意度,还能够推动在线医疗平台的发展。与传统推荐系统不同,医生推荐领域受到隐私保护限制,无法查看患者曾经的诊疗历史,因此模型训练时仅能利用每位患者最近一次的就诊记录,面临严峻的数据稀疏问题。同样,模型预测时也仅能根据患者当前的疾病描述文本进行推荐,而由于患者对疾病描述方式的差异性,模型对不同患者的推荐能力也存在差异,这会使部分患者的需求无法得到满足,进而影响模型整体的推荐能力。基于此,本文提出了一种基于数据增强的医生推荐方法(sequential three-way decision with data augmentation,STWD-NA),通过引入不匹配的医患交互信息扩充训练数据,并利用序贯三支决策的思想训练模型。具体来说,该方法由两部分组成:一方面引入了不匹配交互信息的方法,以缓解训练冷启动问题;另一方面,提出了一种基于序贯三支决策的训练算法,以动态调整模型训练时的关注度。最后,通过好大夫平台上的真实数据集验证了本文所提STWD-NA方法的有效性。展开更多
剩余寿命(Remaining Useful Lifetime,RUL)预测和预测维修决策是预测与健康管理(Prognostics and Health Management,PHM)研究的两大核心内容。RUL预测技术已取得丰硕的成果,预测维修决策研究仍不够深入。现有预测维修决策研究更多关注...剩余寿命(Remaining Useful Lifetime,RUL)预测和预测维修决策是预测与健康管理(Prognostics and Health Management,PHM)研究的两大核心内容。RUL预测技术已取得丰硕的成果,预测维修决策研究仍不够深入。现有预测维修决策研究更多关注如何根据单次预测结果进行单次的维修决策。然而,预测是一个动态过程,相应的维修决策也是动态的,动态序贯的决策中决策的更新与停止问题是一个鲜有关注却值得研究的问题。针对以上问题,对状态可检测的系统,采用Bayesian更新和EM算法相结合的方法,实现了系统RUL和可靠性预测结果的动态更新;在此基础上制定序贯维修策略,依据自适应预测周期和RUL预测结果给出了决策停止的判断准则,在动态冗余的预测结果中选择出真正可执行的最优维修时刻;通过仿真实验验证了所提策略和方法的有效性和适用性。结果表明所提出的方法可有效降低系统的故障概率和运维成本,提升运维效率。展开更多
The optimization of condition-based maintenance (CBM) poses challenges due to the rapid advancement of monitoring technologies. Traditional CBM research has mainly relied on theory-driven approaches, which lead to the...The optimization of condition-based maintenance (CBM) poses challenges due to the rapid advancement of monitoring technologies. Traditional CBM research has mainly relied on theory-driven approaches, which lead to the development of several effective maintenance models characterized by their wide applicability and attractiveness. However, when the system reliability model becomes complex, such methods may run into intractable cost models. The Markov decision process (MDP), a classic framework for sequential decision making, has drawn increasing attention for optimization of CBM optimization due to its appealing tractability and pragmatic applicability across different problems. This paper presents a review of research that optimizes CBM policies using MDP, with a focus on mathematical modeling and optimization methods to enable action. We have organized the review around several key components that are subject to similar mathematical modeling constraints, including system complexity, the availability of system conditions, and diverse criteria of decision-makers. An increase in interest has led to the optimization of CBM for systems possessing increasing numbers of components and sensors. Then, the review focuses on joint optimization problems with CBM. Finally, as an important extension to traditional MDPs, reinforcement learning (RL) based methods are also reviewed as ways to optimize CBM policies. This paper provides significant background research for researchers and practitioners working in reliability and maintenance management, and gives discussions on possible future research directions.展开更多
文摘Attribute reduction is a research hotspot in rough set theory. Traditional heuristic attribute reduction methods add the most important attribute to the decision attribute set each time, resulting in multiple redundant attribute calculations, high time consumption, and low reduction efficiency. In this paper, based on the idea of sequential three-branch decision classification domain, attributes are treated as objects of three-branch division, and attributes are divided into core attributes, relatively necessary attributes, and unnecessary attributes using attribute importance and thresholds. Core attributes are added to the decision attribute set, unnecessary attributes are rejected from being added, and relatively necessary attributes are repeatedly divided until the reduction result is obtained. Experiments were conducted on 8 groups of UCI datasets, and the results show that, compared to traditional reduction methods, the method proposed in this paper can effectively reduce time consumption while ensuring classification performance.
基金supported by the National Natural Science Foundation of China (51175502)
文摘Testing is the premise and foundation of realizing equipment health management （EHM）. To address the problem that the static periodic test strategy may cause deficient test or excessive test, a dynamic sequential test strategy （DSTS） for EHM is presented. Considering the situation that equipment health state is not completely observable in reality, a DSTS optimization method based on partially observable semi-Markov decision pro- cess （POSMDP） is proposed. Firstly, an equipment health state degradation model is constructed by Markov process, and the control limit maintenance policy is also introduced. Secondly, POSMDP is formulated in great detail. And then, POSMDP is converted to completely observable belief semi-Markov decision process （BSMDP） through belief state. The optimal equation and the corresponding optimal DSTS, which minimize the long-run ex- pected average cost per unit time, are obtained with BSMDP. The results of application in complex equipment show that the proposed DSTS is feasible and effective.
文摘近年来,随着智慧医疗的日益普及,在线医疗平台已逐步发展为满足大众基本医疗需求的重要渠道。为患者推荐合适的医生是在线问诊中的一个重要过程,优化推荐能力不仅可以提高患者的满意度,还能够推动在线医疗平台的发展。与传统推荐系统不同,医生推荐领域受到隐私保护限制,无法查看患者曾经的诊疗历史,因此模型训练时仅能利用每位患者最近一次的就诊记录,面临严峻的数据稀疏问题。同样,模型预测时也仅能根据患者当前的疾病描述文本进行推荐,而由于患者对疾病描述方式的差异性,模型对不同患者的推荐能力也存在差异,这会使部分患者的需求无法得到满足,进而影响模型整体的推荐能力。基于此,本文提出了一种基于数据增强的医生推荐方法(sequential three-way decision with data augmentation,STWD-NA),通过引入不匹配的医患交互信息扩充训练数据,并利用序贯三支决策的思想训练模型。具体来说,该方法由两部分组成:一方面引入了不匹配交互信息的方法,以缓解训练冷启动问题;另一方面,提出了一种基于序贯三支决策的训练算法,以动态调整模型训练时的关注度。最后,通过好大夫平台上的真实数据集验证了本文所提STWD-NA方法的有效性。
文摘剩余寿命(Remaining Useful Lifetime,RUL)预测和预测维修决策是预测与健康管理(Prognostics and Health Management,PHM)研究的两大核心内容。RUL预测技术已取得丰硕的成果,预测维修决策研究仍不够深入。现有预测维修决策研究更多关注如何根据单次预测结果进行单次的维修决策。然而,预测是一个动态过程,相应的维修决策也是动态的,动态序贯的决策中决策的更新与停止问题是一个鲜有关注却值得研究的问题。针对以上问题,对状态可检测的系统,采用Bayesian更新和EM算法相结合的方法,实现了系统RUL和可靠性预测结果的动态更新;在此基础上制定序贯维修策略,依据自适应预测周期和RUL预测结果给出了决策停止的判断准则,在动态冗余的预测结果中选择出真正可执行的最优维修时刻;通过仿真实验验证了所提策略和方法的有效性和适用性。结果表明所提出的方法可有效降低系统的故障概率和运维成本,提升运维效率。
基金supported by the National Natural Science Foundation of China(Grant Nos.72401253,72371182,72002149,and 72271154)and the National Social Science Fund of China(23CGL018)+1 种基金the State Key Laboratory of Biobased Transportation Fuel Technology,China(Grant No.512302-X02301)a start-up grant from the ZJU-UIUC Institute at Zhejiang University(Grant No.130200-171207711).
文摘The optimization of condition-based maintenance (CBM) poses challenges due to the rapid advancement of monitoring technologies. Traditional CBM research has mainly relied on theory-driven approaches, which lead to the development of several effective maintenance models characterized by their wide applicability and attractiveness. However, when the system reliability model becomes complex, such methods may run into intractable cost models. The Markov decision process (MDP), a classic framework for sequential decision making, has drawn increasing attention for optimization of CBM optimization due to its appealing tractability and pragmatic applicability across different problems. This paper presents a review of research that optimizes CBM policies using MDP, with a focus on mathematical modeling and optimization methods to enable action. We have organized the review around several key components that are subject to similar mathematical modeling constraints, including system complexity, the availability of system conditions, and diverse criteria of decision-makers. An increase in interest has led to the optimization of CBM for systems possessing increasing numbers of components and sensors. Then, the review focuses on joint optimization problems with CBM. Finally, as an important extension to traditional MDPs, reinforcement learning (RL) based methods are also reviewed as ways to optimize CBM policies. This paper provides significant background research for researchers and practitioners working in reliability and maintenance management, and gives discussions on possible future research directions.
