본문

서브메뉴

Data-Driven Digital Twins for Health-Aware Supervisory Control of Nuclear Reactors
Data-Driven Digital Twins for Health-Aware Supervisory Control of Nuclear Reactors
Data-Driven Digital Twins for Health-Aware Supervisory Control of Nuclear Reactors

Detailed Information

자료유형  
 학위논문 서양
최종처리일시  
20260202105245
ISBN  
9798291569658
DDC  
539.76
저자명  
Lim, Jasmin Y.
서명/저자  
Data-Driven Digital Twins for Health-Aware Supervisory Control of Nuclear Reactors
발행사항  
[Sl] : University of Michigan, 2025
발행사항  
Ann Arbor : ProQuest Dissertations & Theses, 2025
형태사항  
202 p
주기사항  
Source: Dissertations Abstracts International, Volume: 87-03, Section: A.
주기사항  
Advisor: Duraisamy, Karthik.
학위논문주기  
Thesis (Ph.D.)--University of Michigan, 2025.
초록/해제  
요약Digital twins dynamically link a physical system and its virtual counterpart through autonomous, bidirectional communication that realizes value. Reliable and real-time monitoring is empowered with the integration of data-driven methods, which facilitate whole-system analysis and uncertainty quantification. In this thesis, a digital twin framework with various data-enhanced methods is developed to reduce operational costs and increase safety awareness of a Fluoride-salt-cooled High-temperature Reactor (FHR) - a Generation-IV (Gen-IV) reactor concept. Nuclear power plants have low carbon emissions and a flexible design, making them advantageous for addressing the growing global power demands, as well as for applications such as space exploration, hydrogen production, industrial heating, and powering data centers. The conceptualized digital twin aims to support health-aware, cost-effective operations of an FHR and narrow the knowledge uncertainty gap, assisting in future design, deployment, and operational efforts.A closed-loop digital twin framework that streamlines end-to-end communication hasbeen developed for the synchronous regulation and self-adjustment of the Physical Asset - the FHR plant. The framework connects the four modules of a digital twin: the Physical Asset, the Virtual Asset, the Physical-to-Virtual Module, and the Virtual-to-Physical Module. Within the Virtual to Physical Module are two agents that autonomously drive load-following operations: a Reinforcement Learning (RL) agent that optimizes plant power and schedules maintenance; and a Reference Governor agent that enforces multiple system constraints.For the Virtual Asset, a reduced-complexity hybrid surrogate was created to model the FHR plant. The hybrid structure incorporates physics-based models to enhance application-specific accuracy and utilizes data-driven methods to reduce modeling costs. Real-time prediction is made possible with the development of statistical surrogate models, which are structured into a network to separate the high-dimensional state-space, addressing the diverse transient dynamics. In conjunction with a physics-based pump component health surrogate model, the Virtual Asset is more than 4000 times faster than physics-only simulations, with over 98% mean percent accuracy in key states. Real-time adaptation of the Virtual Asset is enabled by applying the Ensemble Kalman Filter data assimilation algorithm in the Physical-to-Virtual Module. Due to its inherent probabilistic formulation, the filtering algorithm bridges diverse information sources (e.g., simulation, data) through uncertainty quantification and propagation, promoting overall system transparency. Using a state-parameter adaptation scheme, the data assimilation corrects digital state estimations and parameters of the plant surrogate model, enabling the digital twin to recalibrate to the Physical Asset over its lifetime simultaneously.The digital twin framework is demonstrated in three cases, using physics-based simulation to emulate the Physical Asset. In the first case, a year-long supervision of the plant displays the RL component maintenance scheduling, with hourly Virtual Asset updates. The data assimilation frequency is increased in the second case to refine accuracy in the power transition regions, with no significant impact on run-time. Finally, in the last case, a system shock indicates that necessary updates are needed for the framework to adaptand capture the new dynamics. These cases demonstrate the robustness and modularity of the framework for both normal and abnormal plant environments, informing future implementations for FHRs and beyond.
일반주제명  
Nuclear engineering
일반주제명  
Aerospace engineering
일반주제명  
Computer engineering
일반주제명  
Communication
키워드  
Digital twins
키워드  
Data-driven modeling
키워드  
Surrogate modeling
키워드  
Generation-IV nuclear reactors
키워드  
Fluoride-salt-cooled High-temperature Reactor
기타저자  
University of Michigan Aerospace Engineering
기본자료저록  
Dissertations Abstracts International. 87-03A.
전자적 위치 및 접속  
로그인 후 원문을 볼 수 있습니다.

MARC

 008260126s2025        us                              c    eng  d
■001000017359980
■00520260202105245
■006m          o    d                
■007cr#unu||||||||
■020    ▼a9798291569658
■035    ▼a(MiAaPQ)AAI32272041
■035    ▼a(MiAaPQ)umichrackham006328
■040    ▼aMiAaPQ▼cMiAaPQ
■0820  ▼a539.76
■1001  ▼aLim,  Jasmin  Y.
■24510▼aData-Driven  Digital  Twins  for  Health-Aware  Supervisory  Control  of  Nuclear  Reactors
■260    ▼a[Sl]▼bUniversity  of  Michigan▼c2025
■260  1▼aAnn  Arbor▼bProQuest  Dissertations  &  Theses▼c2025
■300    ▼a202  p
■500    ▼aSource:  Dissertations  Abstracts  International,  Volume:  87-03,  Section:  A.
■500    ▼aAdvisor:  Duraisamy,  Karthik.
■5021  ▼aThesis  (Ph.D.)--University  of  Michigan,  2025.
■520    ▼aDigital  twins  dynamically  link  a  physical  system  and  its  virtual  counterpart  through  autonomous,  bidirectional  communication  that  realizes  value.  Reliable  and  real-time  monitoring  is  empowered  with  the  integration  of  data-driven  methods,  which  facilitate  whole-system  analysis  and  uncertainty  quantification.  In  this  thesis,  a  digital  twin  framework  with  various  data-enhanced  methods  is  developed  to  reduce  operational  costs  and  increase  safety  awareness  of  a  Fluoride-salt-cooled  High-temperature  Reactor  (FHR)  -  a  Generation-IV  (Gen-IV)  reactor  concept.  Nuclear  power  plants  have  low  carbon  emissions  and  a  flexible  design,  making  them  advantageous  for  addressing  the  growing  global  power  demands,  as  well  as  for  applications  such  as  space  exploration,  hydrogen  production,  industrial  heating,  and  powering  data  centers.  The  conceptualized  digital  twin  aims  to  support  health-aware,  cost-effective  operations  of  an  FHR  and  narrow  the  knowledge  uncertainty  gap,  assisting  in  future  design,  deployment,  and  operational  efforts.A  closed-loop  digital  twin  framework  that  streamlines  end-to-end  communication  hasbeen  developed  for  the  synchronous  regulation  and  self-adjustment  of  the  Physical  Asset  -  the  FHR  plant.  The  framework  connects  the  four  modules  of  a  digital  twin:  the  Physical  Asset,  the  Virtual  Asset,  the  Physical-to-Virtual  Module,  and  the  Virtual-to-Physical  Module.  Within  the  Virtual  to  Physical  Module  are  two  agents  that  autonomously  drive  load-following  operations:  a  Reinforcement  Learning  (RL)  agent  that  optimizes  plant  power  and  schedules  maintenance;  and  a  Reference  Governor  agent  that  enforces  multiple  system  constraints.For  the  Virtual  Asset,  a  reduced-complexity  hybrid  surrogate  was  created  to  model  the  FHR  plant.  The  hybrid  structure  incorporates  physics-based  models  to  enhance  application-specific  accuracy  and  utilizes  data-driven  methods  to  reduce  modeling  costs.  Real-time  prediction  is  made  possible  with  the  development  of  statistical  surrogate  models,  which  are  structured  into  a  network  to  separate  the  high-dimensional  state-space,  addressing  the  diverse  transient  dynamics.  In  conjunction  with  a  physics-based  pump  component  health  surrogate  model,  the  Virtual  Asset  is  more  than  4000  times  faster  than  physics-only  simulations,  with  over  98%  mean  percent  accuracy  in  key  states.  Real-time  adaptation  of  the  Virtual  Asset  is  enabled  by  applying  the  Ensemble  Kalman  Filter  data  assimilation  algorithm  in  the  Physical-to-Virtual  Module.  Due  to  its  inherent  probabilistic  formulation,  the  filtering  algorithm  bridges  diverse  information  sources  (e.g.,  simulation,  data)  through  uncertainty  quantification  and  propagation,  promoting  overall  system  transparency.  Using  a  state-parameter  adaptation  scheme,  the  data  assimilation  corrects  digital  state  estimations  and  parameters  of  the  plant  surrogate  model,  enabling  the  digital  twin  to  recalibrate  to  the  Physical  Asset  over  its  lifetime  simultaneously.The  digital  twin  framework  is  demonstrated  in  three  cases,  using  physics-based  simulation  to  emulate  the  Physical  Asset.  In  the  first  case,  a  year-long  supervision  of  the  plant  displays  the  RL  component  maintenance  scheduling,  with  hourly  Virtual  Asset  updates.  The  data  assimilation  frequency  is  increased  in  the  second  case  to  refine  accuracy  in  the  power  transition  regions,  with  no  significant  impact  on  run-time.  Finally,  in  the  last  case,  a  system  shock  indicates  that  necessary  updates  are  needed  for  the  framework  to  adaptand  capture  the  new  dynamics.  These  cases  demonstrate  the  robustness  and  modularity  of  the  framework  for  both  normal  and  abnormal  plant  environments,  informing  future  implementations  for  FHRs  and  beyond.
■590    ▼aSchool  code:  0127.
■650  4▼aNuclear  engineering
■650  4▼aAerospace  engineering
■650  4▼aComputer  engineering
■650  4▼aCommunication
■653    ▼aDigital  twins
■653    ▼aData-driven  modeling
■653    ▼aSurrogate  modeling
■653    ▼aGeneration-IV  nuclear  reactors
■653    ▼aFluoride-salt-cooled  High-temperature  Reactor
■690    ▼a0538
■690    ▼a0552
■690    ▼a0464
■690    ▼a0459
■71020▼aUniversity  of  Michigan▼bAerospace  Engineering.
■7730  ▼tDissertations  Abstracts  International▼g87-03A.
■790    ▼a0127
■791    ▼aPh.D.
■792    ▼a2025
■793    ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17359980▼nKERIS▼z이  자료의  원문은  한국교육학술정보원에서  제공합니다.

Preview

Export

ChatGPT Discussion

AI Recommended Related Books


    New Books MORE
    Statistics for the past 3 years. Go to brief

    Info Détail de la recherche.

    • Réservation
    • n'existe pas
    • My Folder
    • Demande Première utilisation
    • Non-Book Loan Application
    • Nighttime Book Loan Application
    Matériel
    Reg No. Call No. emplacement Status Lend Info
    TF18359 전자도서 대출가능 My Folder 부재도서신고 비도서대출신청 야간 도서대출신청

    * Les réservations sont disponibles dans le livre d'emprunt. Pour faire des réservations, S'il vous plaît cliquer sur le bouton de réservation

    Books borrowed together with this book

    Related Popular Books

    Available after logging in.