Files in this item



application/pdfSAKURAHARA-DISSERTATION-2018.pdf (10MB)Restricted Access
(no description provided)PDF


Title:Integrated probabilistic risk assessment (I-PRA) methodology and computational platform for fire PRA of nuclear power plants
Author(s):Sakurahara, Tatsuya
Director of Research:Mohaghegh, Zahra
Doctoral Committee Chair(s):Mohaghegh, Zahra
Doctoral Committee Member(s):Aluru, Narayana R; Stubbins, James F; Uddin, Rizwan
Department / Program:Nuclear, Plasma, & Rad Engr
Discipline:Nuclear, Plasma, Radiolgc Engr
Degree Granting Institution:University of Illinois at Urbana-Champaign
Subject(s):Fire Probabilistic Risk Assessment (PRA)
Integrated Probabilistic Risk Assessment (I-PRA)
Nuclear Power Plant
Uncertainty Analysis
Monte Carlo
Fire Dynamics Simulator (FDS)
Fire Brigade
Human Reliability Analysis (HRA)
Common Cause Failure
Global Importance Measure
Probabilistic Validation
Bayesian Approach
Abstract:Probabilistic Risk Assessment (PRA) is a systematic methodology to estimate risk in complex technological systems. PRA has been utilized by both the nuclear industry and the U.S. Nuclear Regulatory Commission (NRC) to enhance Nuclear Power Plant (NPP) safety. In resolving emergent regulatory requirements (e.g., the ones created in the aftermath of the Fukushima accident), the existing classical PRAs of NPPs were found to have limitations in generating the required realism for plant risk estimations. This research develops an Integrated PRA (I-PRA) methodology that explicitly incorporates the underlying science of accident causation into plant risk scenarios and provides a feasible solution for adding realism to the plant risk estimations. In I-PRA, the underlying failure mechanisms associated with the areas of safety concern in NPPs are simulated in separate modules and are integrated with classical PRA through a probabilistic interface methodology. Although the I-PRA is applicable for various NPP safety challenges of concern, this research focuses on the risk assessment for internal fires and makes the following methodological and practical contributions: 1. Develops an I-PRA methodological framework for Fire PRA: a. Provides a unified, multi-level probabilistic integration starting with the underlying failure mechanisms, connecting them to the component-level failures, and then linking them to the system-level risk scenarios in classical PRA; b. Integrates a Computational Fluid Dynamics (CFD)-based fire model, the Fire Dynamics Simulator (FDS), with classical PRA through a probabilistic interface that is equipped with uncertainty quantification, dependency modeling, and Bayesian updating. To alleviate the challenges associated with the high computational cost and large volumes of data, a parallel simulation using the MPI, along with (i) convergence studies on mesh resolution and (ii) convergence studies on uncertainty quantification, is utilized; c. Explicitly models the interactions between fire progression and manual fire suppression. This explicit interface is created by modifying the Heat Release Rate (HRR) curve of the fire source, which is one of the inputs to the FDS code, based on the key timings associated with the performance of manual suppression. d. Applies the I-PRA to a realistic fire scenario at NPPs. The case study shows that, as compared to the current Fire PRA methodology, the I-PRA methodology reduces the core damage frequency estimate by 50%. Improving the realism of PRA can contribute to more efficient risk applications in commercial NPPs by: (i) generating a more accurate and stable categorization of Structures, Systems, and Components (SSC) (e.g., for the 10 CFR 50.69 process) and (ii) providing NPPs with more design alternatives that can satisfy the risk acceptance criteria. 2. Develops a Global Importance Measure (Global IM) method for the I-PRA framework to generate the ranking of the underlying risk-contributing factors. This Global IM method is capable of accounting for (i) uncertainty of the input parameters, (ii) nonlinearity and complex interactions in the model, and (iii) uncertainty of the model output. The credibility of the Global IM method is demonstrated by conducting several case studies, including reduced-order I-PRA frameworks developed for Generic Safety Issue 191 and Fire PRA. The Global IM analysis helps identify the critical risk-contributing factors that are closely related to the controllable design parameters at the level of failure mechanisms; thus making a contribution toward more effective accident prevention. 3. Develops the Spatio-Temporal Probabilistic methodology for Common Cause Failure (CCF) modeling. The objectives of this new CCF methodology are twofold: (i) to create a quantitative relationship between the results of simulation modules in I-PRA and the existing data-driven CCF methods of NPPs and (ii) to advance the CCF modeling and quantification in the existing PRA of NPPs. This research develops a computational algorithm to operationalize the new CCF methodology in the Interface Module of I-PRA. In this computational algorithm, Bayesian updating is used to consider all available sources of data related to the results of simulations. The epistemic uncertainty associated with the estimated probabilities are also analyzed in order to execute a probabilistic validation. The new CCF methodology is applied for the emergency service water pumps at an NPP, and the results demonstrate that, by updating the existing plant-specific CCF parameters with the simulation-based estimates, the degree of uncertainty associated with the resultant core damage frequency can be reduced significantly. 4. Develops the Human Reliability Analysis (HRA)-based method for human performance modeling in manual fire suppression module of the I-PRA framework. Compared to the previous studies, this research advances the HRA-based manual suppression analysis by (i) using a CFD-based fire progression model to define the success criteria; (ii) addressing the HRR reduction during the manual suppression phase using the empirical water suppression model; and (iii) quantifying the times to complete individual tasks. The HRA-based manual suppression model is applied for the fire brigade data from two NPPs, and the results indicate that the performance of manual suppression can be strongly influenced by the plant-specific contextual factors (e.g., plant geometry, fire location, and fire brigade procedure); therefore, relying on the industry-wide historical fire event records to derive the industry-average fire brigade performance, as done in the existing Fire PRA, can mislead NPP risk estimation.
Issue Date:2018-04-11
Rights Information:Copyright 2018 Tatsuya Sakurahara
Date Available in IDEALS:2018-09-04
Date Deposited:2018-05

This item appears in the following Collection(s)

Item Statistics