Repository logo
Log In(current)
  1. Home
  2. Colleges & Schools
  3. Graduate School
  4. Doctoral Dissertations
  5. Enhancing Nuclear Power Plant Reliability: A System-Level Framework for Prognostics and Health Management
Details

Enhancing Nuclear Power Plant Reliability: A System-Level Framework for Prognostics and Health Management

Date Issued
August 2026
Author(s)
Zanotelli, Mattia
The University of Tennessee
Advisor(s)
Jamie Coble
Additional Advisor(s)
Diego Mandelli
Xingang Zhao
Anahita Khojandi
Permanent URI
https://trace.tennessee.edu/handle/20.500.14382/61884
Abstract
Recent research in maintenance strategies has increasingly shifted from classical, static reliability techniques to dynamic approaches that leverage real-time monitoring data. Traditional reliability methods excel at providing system-level risk indications, but they rely heavily on static probabilities and historical averages, failing to capture actual asset degradation under specific operating conditions. Conversely, Prognostics and Health Management (PHM) effectively integrates real-time sensor data, expert knowledge, and historical records to perform highly accurate component-level health assessments. However, PHM traditionally struggles to translate these isolated, heterogeneous component metrics into cohesive system-level indicators. To bridge this gap, this dissertation presents a hierarchical, hybrid System-Level PHM (S-PHM) framework for complex engineered systems, with a specific focus on nuclear power plants (NPPs). By explicitly incorporating the physical and functional topology of the plant, this framework systematically aggregates diverse component-level PHM health assessments into actionable, plant-wide insights. Among the available methodologies for aggregating local health measures into higher-level indicators within the system hierarchy, three approaches are selected. The framework describes how each method is applied to specific subsystems and how the resulting subsystem-level indicators are further aggregated into a unified system-level indicator. First, an online fault-tree approach dynamically updates system risk by feeding real-time component failure probabilities into static logical structures. Second, a margin-based reliability approach provides a unified geometric metric—distance-to-failure—to integrate diverse health indicators without requiring strict probabilistic transformations. While building on established approaches, this research significantly advances these methodologies by expanding their mathematical formulations and adapting them for novel, dynamic applications. Third, advanced regression using Bayesian Random Fourier Feature (RFF) ensembles is applied to robustly map continuous component input uncertainties to subsystem physical or virtual indicators. This hybrid architecture is validated using the Asherah Pressurized Water Reactor simulator. Degradation scenarios across Balance-of-Plant systems demonstrate how these methodologies can be applied independently or structurally combined. Additionally, by leveraging the established relationships between component health indicators and overall system performance, the framework calculates actionable, risk-informed metrics, such as the Risk Reduction Measure. These results demonstrate that the proposed framework can aggregate diverse component-level health indicators into system-level measures that support maintenance prioritization under uncertainty in real-time.
Subjects

System-Level Prognost...

Nuclear Engineering

Reliability Engineeri...

Risk Assessment

Prognostics and Healt...

Degree
Doctor of Philosophy
Major
Nuclear Engineering
Embargo Date
August 1, 2027

University Libraries

1015 Volunteer Boulevard
Knoxville, TN 37996
865-974-4351

Map & Directions
Donate to the Libraries
  • About
  • John C. Hodges Society
  • Speaking Volumes magazine
  • Outreach
  • Directory
  • Employment
  • Policies
  • Library Intranet
University of Tennessee power T logo

The University of Tennessee, Knoxville
Knoxville, Tennessee 37996
865-974-1000

Events
A-Z
Apply
Privacy
Map
Directory
Give to UT
Accessibility

Built with DSpace-CRIS software - Extension maintained and optimized by 4Science