Ernie Kee

Bayesian nonparametric analysis of single item preventive maintenance strategies

This work addresses the problem of finding the minimal-cost preventive maintenance schedule for a single item. We develop an optimization algorithm that reduces the computational effort to find the optimal schedule. This approach relies on the item having an increasing failure rate, which is typical, and employs a Gibbs sampling algorithm to simulate from the failure rate distribution using real data. We also analyze the case when the item has a “bathtub” failure rate; we develop techniques that lead to an algorithm that finds an optimal schedule for this case as well. We then analyze the effectiveness of our approach on both artificial and real data sets from the South Texas Project nuclear power plant.Copyright

Optimizing Project Prioritization Under Budget Uncertainty

We consider a problem commonly faced in the nuclear power industry, involving annual selection of plant capital investments under the constraints of a limited and uncertain budget. When the budget is assumed known, a typical approach to such problems is built on a multi-dimensional knapsack model. This model takes as input the available budget in each year, the stream of liabilities induced by selecting each project, and the profit, i.e., net present value (NPV), of each project. The goal is to select the portfolio of projects with the highest total NPV, while observing the budget constraint for each year, as well as any additional constraints. We show that a portfolio selected in this manner can fail to hedge against uncertainties in the budget. While the budget may be known at the beginning of the planning period, external events can cause this to change as time unfolds, and hence the funds that will actually be allocated over time are typically uncertain. So, we propose a model that forms an optimal priority list of projects, incorporating multiple budget scenarios. The model is applied to example projects from the South Texas Project Nuclear Operating Company (STPNOC).

An integrated methodology for spatio-temporal incorporation of underlying failure mechanisms into fire probabilistic risk assessment of nuclear power plants

Abstract In this research, an Integrated probabilistic risk assessment (I-PRA) methodological framework for Fire PRA is developed to provide a unified multi-level probabilistic integration, beginning with spatio-temporal simulation-based models of underlying failure mechanisms (i.e., physical phenomena and human actions), connecting to component-level failures, and then linking to system-level risk scenarios in classical PRA. The simulation-based module, called the fire simulation module (FSM), includes state-of-the-art models of fire initiation, fire progression, post-fire failure damage propagation, fire brigade response, and scenario-based damage. Fire progression is simulated using a CFD code, fire dynamics simulator (FDS), which solves Navier–Stokes equations governing the turbulent flow field. Uncertainty quantification is conducted to address parameter uncertainties. The I-PRA paves the way for reducing excessive conservatisms derived from the modeling of (i) fire progression and damage and (ii) the interactions between fire progression and manual suppression. Global importance measure analysis is used to rank the risk-contributing factors. A case study demonstrates the implementation of I-PRA for a regulatory-documented fire scenario.

Project Prioritization via Optimization

We consider a problem commonly faced in industry, involving annual selection of plant capital investments. A typical approach to such a problem uses a multi-knapsack formulation, which takes as input the available budget in each year, the stream of liabilities induced by selecting each project, and the profit, i.e., net present value, of each project. The goal is to select the portfolio of projects with the highest total net present value, while observing the budget constraint for each year, as well as any additional constraints. A portfolio selected in this manner can fail to hedge against uncertainties in the budget, the liability stream and the profit. So, we propose a model that forms an optimal priority list of projects, incorporating multiple scenarios for these input parameters. Our model is not a simplistic ranking scheme. Structural and stochastic dependencies among the projects are key to our approach. We apply our methods on a set of example projects from South Texas Project Nuclear Operating Company.Copyright

Optimal Preventive Maintenance Under Decision Dependent Uncertainty

We analyze a system of N components with dependent failure times. The goal is to obtain the optimal block replacement interval (different for each component) over a finite horizon that minimizes the expected total maintenance cost. In addition, we allow each preventive maintenance action to change the future joint failure time distribution. We illustrate our methodology with an example from South Texas Project Nuclear Operating Company.Copyright

Integrated Power Recovery Using Markov Modeling

The solution to a Markov chain modeling electric power supply to critical equipment in a typical four-loop pressurized water reactor following a loss of offsite power event is compared with a convolution method. The standard “convolution integral” approach is described, and an alternative methodology based on a Markov model is illustrated.

Methodology to evaluate the monetary benefit of Probabilistic Risk Assessment by modeling the net value of Risk-Informed Applications at nuclear power plants

Abstract Probabilistic Risk Assessment (PRA) used in Nuclear Power Plants serves as a pillar of the U.S. Nuclear Regulatory Commissions Risk-Informed Regulatory framework, and is required for new reactor licenses to satisfy regulatory safety compliance. The benefits of PRA are not only experienced in terms of safety, but also from the monetary value derived from Risk-Informed Performance-Based Applications (RIPBAs), where risk estimated from PRA is utilized in decision making to expand the safe operational envelope of plants, leading to either an increase in profits or a reduction in costs. This paper introduces a methodology to evaluate this monetary value by the systematic causal modeling of the net value of RIPBAs and demonstrates the methodology for one of the RIPBAs, called Risk-Managed Technical Specifications (RMTS). The key steps of this methodology are: (i) Cost-Benefit Analysis to formulate the net value of PRA based on the net value of RIPBAs, (ii) Causal modeling to systematically model the operational scenarios leading to costs and benefits associated with RIPBAs, (iii) Uncertainty analysis, and (iv) Sensitivity analysis and validation. The results of this research could help decision makers with evaluating investment strategies in PRA that go ‘beyond-compliance’ to maximize industry profit while maintaining regulatory safety goals.

Methodological and Practical Comparison of Integrated Probabilistic Risk Assessment (I-PRA) with the Existing Fire PRA of Nuclear Power Plants

Abstract Nearly half of the U.S. nuclear power plants (NPPs) are in the process of transitioning, or have already transitioned, to a risk-informed, performance-based fire protection program. For this transition, Fire Probabilistic Risk Assessment (Fire PRA) is used as a foundation for fire risk evaluation. To increase realism in Fire PRA by reducing conservative bias, the authors have developed an Integrated Probabilistic Risk Assessment (I-PRA) methodological framework that does not require major changes to the existing plant Probabilistic Risk Assessments (PRAs). The underlying failure mechanism models associated with fire events are developed in a separate module, which can be interfaced and connected to the existing plant PRA. This paper explains the areas of methodological advancements in I-PRA, comparing them with the existing Fire PRA of NPPs. This comparison is further demonstrated in a realistic case study that applies the I-PRA framework to a critical fire-induced scenario at an NPP. The core damage frequency (CDF) for the selected scenario, computed by the I-PRA framework, is compared with the results of the Full Compartment Burn screening method and the existing Fire PRA methodology. Using the I-PRA framework, the CDF for the selected scenario has decreased by a factor of 20 compared with the Full Compartment Burn screening approach and by a factor of 2 compared to the existing Fire PRA methodology based on NUREG/CR-6850 and the subsequent NUREGs that have updated the data and methods for individual steps.

SHAKE-RoverD Framework for Nuclear Power Plants: A Streamlined Approach for Seismic Risk Assessment

Abstract A major challenge facing the nuclear energy industry is to remain competitive under current market conditions. Utility operators are searching for innovative methods to reduce nuclear power plant (NPP) operation and maintenance costs while complying with safety and reliability requirements. To support these goals, the authors suggest a streamlined approach that implements a conservative risk-informed method to reduce the costs of satisfying emergent regulatory requirements. As a streamlined approach, the Risk-informed Over Deterministic (RoverD) method was developed by some of the authors of the current paper to resolve the concerns associated with Generic Safety Issue 191 (GSI-191). The RoverD method is designed around U.S. Nuclear Regulatory Commission Regulatory Guide 1.174 (RG 1.174), which defines “risk-informed” regulation as comprising a blend of risk-based and deterministically based elements. This paper offers the Safety Hazard Analysis for earthquaKE (SHAKE)–RoverD (SHAKE-RoverD) methodology, an extension of the original RoverD methodology developed for GSI-191, to evaluate the impact of an increased seismic hazard on the performance of NPP protective systems. SHAKE-RoverD aims to reduce the cost required for developing, validating, and documenting detailed fragility curves in seismic probabilistic risk assessment by using deterministic fragility curves. The SHAKE-RoverD methodology assesses whether an increase in a seismic hazard would result in an unacceptable increase in NPP risk. If the conservative estimate of plant risk, computed by the streamlined approach, satisfies the regulatory acceptance criteria (e.g., Regulatory Guide 1.174), the plant likely would not need to make a design change (as long as defense in depth and adequate safety margin are satisfied); therefore, the use of streamlined methodology could lead to significant cost savings for the utility operator. Future work will advance SHAKE-RoverD and analyze risk management strategies based on this method.

Physics of failure, predictive modeling & data analytics for LOCA frequency

This paper presents: (a) the Data-Theoretic methodology as part of an ongoing research which integrates Physics-of-Failure (PoF) theories and data analytics to be applied in Probabilistic Risk Assessment (PRA) of complex systems and (b) the status of the application of the proposed methodology for the estimation of the frequency of the location-specific loss-of-coolant accident (LOCA), which is a critical initiating event in PRA and one of the challenges of the risk-informed resolution for Generic Safety Issue 191 (GSI-191) [1]. The proposed methodology has the following unique characteristics: (1) it uses predictive causal modeling along with sensitivity and uncertainty analysis to find the most important contributing factors in the PoF models of failure mechanisms. This model-based approach utilizes importance-ranking techniques, scientifically reduces the number of factors, and focuses on a detailed quantification strategy for critical factors rather than conducting expensive experiments and time-consuming simulations for a large number of factors. Th is adds validity and practicality to the proposed methodology. (2) Because of the evolving nature of computational power and information-sharing technologies, the Data-Theoretic method for PRA expands the classical approach of data extraction and implementation for risk analysis. It utilizes advanced data analytic techniques (e.g., data mining and text mining) to extract risk and reliability information from diverse data sources (academic literature, service data, regulatory and laboratory reports, expert opinion, maintenance logs, news, etc.) and executes them in theory-based PoF networks. (3) The Data-Theoretic approach uses comprehensive underlying PoF theory to avoid potentially misleading results from use of solely data-oriented approaches, as well as support the completeness of the contextual physical factors and the accuracy of their causal relationships. (4) When the important factors are identified, the Data-Theoretic approach applies all potential theory-based techniques (e.g., simulation and experimentation).