東華大學圖書館 |

語系: 繁體中文

說明(常見問題)

回圖書館首頁

手機版館藏查詢

登入

回首頁

切換: 標籤 | MARC模式 | ISBD

FindBook

Google Book

Amazon

博客來

Value of Information and Evolution Prediction for Sequential Infrastructure Management.

紀錄類型:	書目-電子資源 : Monograph/item
正題名/作者:	Value of Information and Evolution Prediction for Sequential Infrastructure Management./
作者:	Li, Shuo.
出版者:	Ann Arbor : ProQuest Dissertations & Theses, : 2022,
面頁冊數:	150 p.
附註:	Source: Dissertations Abstracts International, Volume: 83-10, Section: B.
Contained By:	Dissertations Abstracts International83-10B.
標題:	Civil engineering. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=29062804
ISBN:	9798209941767

Value of Information and Evolution Prediction for Sequential Infrastructure Management.
Li, Shuo.

Value of Information and Evolution Prediction for Sequential Infrastructure Management. - Ann Arbor : ProQuest Dissertations & Theses, 2022 - 150 p.

Source: Dissertations Abstracts International, Volume: 83-10, Section: B.

Thesis (Ph.D.)--Carnegie Mellon University, 2022.

This item must not be sold to any third party vendors.

Information collected by sensors and monitoring systems can support the operation and maintenance (O&M) of infrastructure components, helping to reduce the uncertainty in their current condition and in the prediction of their future evolution. However, it is nontrivial to assess if it is worth implementing a specific monitoring system in an O&M process, since the impact of collecting information is a complicated function of many factors modeling the performance of monitoring systems, such as the prior information, maintenance costs, the adopted policy, etc. In this dissertation, computational framework for assessing the impact of monitoring systems and predicting the condition evolution is developed by modeling the O&M process as a Partially Observable Markov Decision Process (POMDP).Three main contributions are presented in this dissertation. First, while the evaluation of monitoring systems depends on many features in an O&M process, all these features can be embedded in the Value of Information (VoI) metric, which quantifies the utility of information gathering efforts. By modeling the O&M process as a POMDP, we investigate the relation between the VoI and key features of the component deterioration, the economic setting, and the performance of the monitoring system itself, to identify when the benefit of monitoring is high.Second, in the O&M problems, when the decision maker adopts a policy, it is not easy to predict how the condition of the system evolves in future. For example, we may ask "Can a failure event occur? And, if so, how frequently?" "What is the probability that a critical condition occurs within a specific time horizon? And that it occurs before a maintenance action is taken?" To address these questions, we illustrate how to apply Finite State Markov Chain analyses to predict relevant features of the time evolution of a system controlled by the decision maker, and presents analytical methods based on linear algebra. The analyses can be performed after a policy is selected for a Markov Decision Process (MDP) or a POMDP.Finally, although the decision maker always follows the optimal policy in the absence of external constraints and prefers to collect data, this property is not true when she acts obeying the external constraints, such as those posed by social regulations. This work illustrates how epistemic external constraints affect the VoI in sequential decision making problems, by extending POMDP models of the first part and leveraging the analysis for evaluating a fixed policy in the second part. The proposed framework can identify and understand when a decision maker avoids collecting information and we also discuss how to possibly mitigate this effect.

ISBN: 9798209941767Subjects--Topical Terms:

860360
Civil engineering.
Subjects--Index Terms:

Partially Observable Markov Decision Process

Value of Information and Evolution Prediction for Sequential Infrastructure Management.
LDR:03832nmm a2200325 4500 001 2344958
005 20220531062224.5
008 241004s2022 ||||||||||||||||| ||eng d
020 $a 9798209941767
035 $a (MiAaPQ)AAI29062804
035 $a AAI29062804
040 $a MiAaPQ $c MiAaPQ
100 1 $a Li, Shuo. $3 2057697
245 1 0 $a Value of Information and Evolution Prediction for Sequential Infrastructure Management.
260 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2022
300 $a 150 p.
500 $a Source: Dissertations Abstracts International, Volume: 83-10, Section: B.
500 $a Advisor: Pozzi, Matteo.
502 $a Thesis (Ph.D.)--Carnegie Mellon University, 2022.
506 $a This item must not be sold to any third party vendors.
520 $a Information collected by sensors and monitoring systems can support the operation and maintenance (O&M) of infrastructure components, helping to reduce the uncertainty in their current condition and in the prediction of their future evolution. However, it is nontrivial to assess if it is worth implementing a specific monitoring system in an O&M process, since the impact of collecting information is a complicated function of many factors modeling the performance of monitoring systems, such as the prior information, maintenance costs, the adopted policy, etc. In this dissertation, computational framework for assessing the impact of monitoring systems and predicting the condition evolution is developed by modeling the O&M process as a Partially Observable Markov Decision Process (POMDP).Three main contributions are presented in this dissertation. First, while the evaluation of monitoring systems depends on many features in an O&M process, all these features can be embedded in the Value of Information (VoI) metric, which quantifies the utility of information gathering efforts. By modeling the O&M process as a POMDP, we investigate the relation between the VoI and key features of the component deterioration, the economic setting, and the performance of the monitoring system itself, to identify when the benefit of monitoring is high.Second, in the O&M problems, when the decision maker adopts a policy, it is not easy to predict how the condition of the system evolves in future. For example, we may ask "Can a failure event occur? And, if so, how frequently?" "What is the probability that a critical condition occurs within a specific time horizon? And that it occurs before a maintenance action is taken?" To address these questions, we illustrate how to apply Finite State Markov Chain analyses to predict relevant features of the time evolution of a system controlled by the decision maker, and presents analytical methods based on linear algebra. The analyses can be performed after a policy is selected for a Markov Decision Process (MDP) or a POMDP.Finally, although the decision maker always follows the optimal policy in the absence of external constraints and prefers to collect data, this property is not true when she acts obeying the external constraints, such as those posed by social regulations. This work illustrates how epistemic external constraints affect the VoI in sequential decision making problems, by extending POMDP models of the first part and leveraging the analysis for evaluating a fixed policy in the second part. The proposed framework can identify and understand when a decision maker avoids collecting information and we also discuss how to possibly mitigate this effect.
590 $a School code: 0041.
650 4 $a Civil engineering. $3 860360
653 $a Partially Observable Markov Decision Process
653 $a Operation and maintenance
653 $a Value of information
690 $a 0543
710 2 $a Carnegie Mellon University. $b Civil and Environmental Engineering. $3 2097787
773 0 $t Dissertations Abstracts International $g 83-10B.
790 $a 0041
791 $a Ph.D.
792 $a 2022
793 $a English
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=29062804