Multi-hazard integrated probabilistic risk assessment framework for seismically induced internal flooding in nuclear power plants

Komaki, Jumpei

Permalink

https://hdl.handle.net/2142/122272 Copy

Description

Title

Multi-hazard integrated probabilistic risk assessment framework for seismically induced internal flooding in nuclear power plants

Author(s)

Komaki, Jumpei

Issue Date

2023-12-08

Director of Research (if dissertation) or Advisor (if thesis)

Mohaghegh, Zahra

Committee Member(s)

Sakurahara, Tatsuya

Department of Study

Nuclear, Plasma, & Rad Engr

Discipline

Nuclear, Plasma, Radiolgc Engr

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

M.S.

Degree Level

Thesis

Keyword(s)

• probabilistic risk assessment
• PRA
• multi-hazard
• earthquake
• internal flooding, nuclear

Abstract

The Fukushima-Daiichi NPP Accident in 2011 highlighted that to avoid future severe accidents, it is vital to advance probabilistic risk assessment (PRA) techniques to treat complex spatiotemporal phenomena due to multiple hazards. The current industry and regulatory practices in the U.S. treat multi-hazard risk scenarios, such as seismically induced floodings and fires, through qualitative consideration (e.g., plant walkdowns and vulnerability identification) and quantitative screening using conservative, bounding assumptions to demonstrate the risk contribution by multi-hazard scenarios is insignificant. This fact, however, does not negate the need for multi-hazard PRAs at NPPs, considering that (i) vulnerability of NPPs to multi-hazard events is design- and site-specific, and some of the non-U.S. countries are concerned about multi-hazard scenarios as significant risk contributors; and (ii) for future advanced reactors, due to highly reliable passive systems, the total risk profile is likely to be dominated by extreme external hazards that could cause not only simultaneous damage to multiple safety systems but also the secondary hazard initiation at the plant site (e.g., fires, floodings). In the literature, a few academic studies have developed a dynamic PRA model for seismically induced flooding scenarios at NPPs. The construction of a dynamic event tree for plant-level risk scenarios was guided by a simulation model of flooding progression initiated by an earthquake. Dynamic PRA is capable of capturing time-dependent evolution of plant response driven by dynamic interactions between system performance and operator’s action more realistically than the state-of-practice static PRA; however, there are inherent challenges in dynamic PRA that kept it from being widely adopted by the nuclear industry as of now, including significantly high computational cost due to scenario explosion, visualization of a large volume of complex risk information, and extra time and resource burdens for the compliance with regulatory requirements and standards (e.g., industry peer review). There is another group of academic research that developed probabilistic graph models to analyze multi-hazard scenarios, for instance, the DQFM (Direct Quantification of the Fault tree using the Monte Carlo simulation) method and the Bayesian network method. These studies, however, relied on engineering judgment to estimate probabilistic inputs that characterize the dependency induced by multiple hazards, such as the correlation coefficients for seismic and flooding fragility curves in the DQFM method and the conditional probability tables for equipment failure events in the Bayesian network method. Considering rapid developments and advancements in modeling and simulation capabilities for various hazards, it is desirable for multi-hazard PRAs to be able to utilize state-of-the-art simulations fully. In summary, there is a need for a new multi-hazard PRA approach that can satisfy the following two requirements: (i) the existing plant PRA model (using static event trees and fault trees) can be utilized without significant modifications to its structure, and (ii) simulation data for hazard initiation, progression, and plant impact can be incorporated into PRA scenarios. As a new approach that meets these two requirements, this thesis develops an integrated PRA (I-PRA) methodology for multi-hazard PRA of NPPs. The I-PRA methodology was initially developed for a single hazard PRA of NPPs in the previous research by the Socio-Technical Risk Analysis (SoTeRiA) Research Laboratory. The I-PRA methodology enhances the realism of risk estimates by integrating spatiotemporal simulations with existing plant PRA models through a probabilistic interface. The interface is equipped with uncertainty quantification, probability estimations with dependency treatment, and a hazard scenario model so that the PRA inputs (i.e., initiating event frequencies and failure event probabilities) can be estimated from simulation data. This thesis advances the I-PRA methodology to analyze multi-hazard risk scenarios at NPPs, specifically focusing on those caused by an earthquake combined with seismically induced internal flooding. The methodological advancements in this thesis include: 1) Advance the probabilistic interface between PRA and underlying simulation models to integrate a coupled multi-hazard simulation with a plant PRA model, focusing on two aspects: o Advance the hazard scenario model: In the initial I-PRA methodology for a single hazard, the hazard scenario model was introduced to compute the probabilistic inputs to PRA (i.e., initiating event frequencies and failure probabilities of PRA equipment) from a simulation model of underlying physical phenomena that typically generates data at a lower level in accident causation, such as smaller pieces and parts of PRA equipment. The hazard scenario model constructs scenarios with respect to how the lower level physical damage can propagate to the PRA events, considering damage propagation and defense measures in place. For multi-hazard scenarios, this thesis advances the hazard scenario model to account for interrelationships between two hazards, including (i) one hazard can initiate another hazard and (ii) the first hazard can impact the defense measures against the second hazard and make PRA equipment more vulnerable to the second hazard. o Advance the dependent failure analysis: In the initial I-PRA methodology developed for a single hazard, a simulation-informed approach was developed to treat the dependent failure of multiple PRA components induced by a shared failure mechanism; for instance, both components A and B are damaged by a single failure mechanism. In this thesis, the simulation-informed dependency analysis in I-PRA is advanced to deal with the dependent failure of multiple PRA components induced by two interrelated hazards; for instance, component A is damaged by an earthquake, while component B is damaged by seismically induced internal flooding. This is achieved by formulating the probability of dependent failure of multiple components as a multiplication of conditional probabilities using the chain rule of probabilities and directly estimating each condition probability from the coupled multi-hazard simulation model. 2) Extend importance measure analyses for the multi-hazard I-PRA methodology. The initial I-PRA methodology for a single hazard was equipped with two types of importance measure analyses: (i) classical importance measures to rank the PRA events at the system and component levels and (ii) global importance measures to rank the risk contributing factors at the underlying phenomenon level, for instance, physical design parameters. In multi-hazard PRA, global importance measures could play a more crucial role for two reasons: (a) classical importance measures for the PRA events, which are typically computed by varying each event probability at a time, are difficult to interpret due to a higher degree of dependency induced by the underlying hazard interactions; and (b) the interdependencies induced by multiple hazards can result in complex system interactions that are hard to predict based on expert’s intuition or risk insights from a single-hazard PRA. Under this situation, ranking the underlying risk contributing factors using the global importance measure can provide valuable risk insights, for instance, how a design parameter should be modified (e.g., increase or decrease) to reduce the overall plant risk effectively. In this thesis, the global importance measure analysis in I-PRA is extended by (i) developing an algorithm to generate the ranking of design options (i.e., alternatives of design parameter values) in addition to the ranking of input parameters based on their contributions to the uncertainty for the plant risk metrics; and (ii) creating a quick-running surrogate PRA model to manage the high computational cost for multi-hazard PRA due to the increased size of the PRA model and relatively large component failure probabilities that would significantly increase the number of cut sets quantified in PRA after screening and truncation. These methodological advancements are demonstrated by an NPP case study, where multi-hazard risk due to seismically induced internal flooding caused by rupture of fire protection piping is analyzed.

Graduation Semester

2023-12

Type of Resource

Thesis

Copyright and License Information

Copyright 2023 Jumpei Komaki

Owning Collections