[en] This study presents and efficient methodology that derives design alternatives and performance criteria of safety functions/systems in commercial nuclear power plants. Determination of design alternatives and intermediate-level performance criteria is posed as a reliability allocation problem. The reliability allocation is performed for determination of reliabilities of safety functions/systems from top-level performance criteria. The reliability allocation is a very difficult multi objective optimization problem (MOP) as well as a global optimization problem with many local minima. The weighted Chebyshev norm (WCN) approach in combination with an improved Metropolis algorithm of simulated annealing is developed and applied to the reliability allocation problem. The hierarchy of probabilistic safety criteria (PSC) may consist of three levels, which ranges from the overall top level (e.g., core damage frequency, acute fatality and latent cancer fatality) through the interlnediate level (e.g., unavailiability of safety system/function) to the low level (e.g., unavailability of components, component specifications or human error). In order to determine design alternatives of safety functions/systems and the intermediate-level PSC, the reliability allocation is performed from the top-level PSC. The intermediated level corresponds to an objective space and the top level is related to a risk space. The reliability allocation is performed by means of a concept of two-tier noninferior solutions in the objective and risk spaces within the top-level PSC. In this study, two kinds of towtier noninferior solutions are defined: intolerable intermediate-level PSC and desirable design alternatives of safety functions/systems that are determined from Sets 1 and 2, respectively. Set 1 is obtained by maximizing simultaneously not only safety function/system unavailabilities but also risks. Set 1 reflects safety function/system unavailabilities in the worst case. Hence, the intolerable intermediate-level PSC determined by Set 1 should not be violated. On the other hand, Set 2 is obtained by maximizing safety function/system unavailabilities and at the same time minimizing risks. Thus, Set 2 reflects the flexibility in designing and operating safety functions/systems with the lowest possible risk to the public. The global optimization results from strong nonlinearity of the probabilistic safety assessment (PSA) model and nonconvexity of the problem. The Boolean algebra for the safety function/system unavailabilities and the risks are strongly nonlinear equations. Furthermore, the reliability allocation MOP has an infinite number of two-tier noninferior solutions. The transformation of the MOP to a single objective optimization problem is performed to find the two-tier noninferior solutions. After the scalarization of the reliability allocation MOP to a single objective optimization problem by the weighted Chebyshev norm (WCN) approach, the global optimization is performed by the improved Metropolis algorithm. The global minimum corresponds to one of the infinite two-tier noninferior solutions of the reliability allocation MOP. The methodology is applied to a realistic streamlined PSA model that is developed based on the PSA results of Surry Unit 1 nuclear power plant. The WCN approach with the improved Metropolis algorithm results in drastically reduced calculational efforts. The results of this study show that risks to the public are drastically changed according to the combination of safety system unavailabilities. When design and modification of a nuclear power plant or regulatory actions are performed, it is necessary that particular attention be paid to auxiliary feedwater system (AFW system) and onsite electric power system (OEP system) than to the other safety functions/systems. If we improve the unavailabilities of AFW and OEP systems, the induced risks are drastically reduced. On the other hand, the other safety function/system unavailabilities could be relaxed, e.g., the allowed outage times (AOTs) and surveillance testing intervals (STIs) of those systems could be relaxed. The intermediate-level PSC and desirable design alternatives based on the two-tier noninferior solutions obtained by the methodology in this study are balanced and consistent with the top-level PSC. Hence, safety of the plant can be improved by utilizing the self-consistent intermediate-level PSC or desirable design alternatives. Furthermore, the methodology developed in this study can be used as an efficient design tool for desirable safety function/system alternatives and for determination of the intermediate-level performance criteria

[en] In order to calculate the more accurate top event probability from cutsets or minimal cut sets (MCSs) than the conventional method that adopts the rare event approximation (REA) or min cut upper bound (MCUB) calculation, advanced cutset upper bound estimator (ACUBE) software had been developed several years ago and shortly became a vital tool for calculating the accurate core damage frequency of nuclear power plants in probabilistic safety assessment (PSA). Usually, the whole cutsets in the industry PSA models cannot be converted into a Binary decision diagram (BDD) due to the limited computational memory. So, the ACUBE selects the major cutsets whose probabilities are larger than the others, and then converts the major cutsets into a BDD in order to calculate more accurate top event probability from cutsets. This study (1) suggests when and where the ACUBE should be employed by predicting the amount of overestimation of the top event probability depending on the cutset structure, (2) explains the details of the ACUBE algorithm, and (3) demonstrates the efficiency of the ACUBE by calculating the top event probability of some PSA cutsets. - Highlights: • EPRI report [32] introduces many successful events in the seismic PSA cutsets. • This results in drastically overestimated top event probability. • In order to overcome this problem, the author developed ACUBE software. • ACUBE calculation can be determined according to the cutset structure (Section 4). • ACUBE calculation removes unnecessary conservatism in the top event probability

[en] In order to verify FTREX functions and to confirm the correctness of FTREX 1.5, various tests were performed 1.fault trees with negates 2.fault trees with house events 3.fault trees with multiple tops 4.fault trees with logical loops 5.fault trees with initiators, house events, negates, logical loops, and flag events By using the automated cutest propagation test, the FTREX 1.5 functions are verified. FTREX version 1.3 and later versions have capability to perform bottom-up cutset-propagation test in order check cutest status. FTREX 1.5 always generates the proper minimal cut sets. All the output cutsets of the tested problems are MCSs (Minimal Cut Sets) and have no non-minimal cutsets and improper cutsets. The improper cutsets are those that have no effect to top, have multiple initiators, or have disjoint events A * -A

[en] The fault tree quantification uncertainty from the truncation error has been of great concern for the reliability evaluation of large fault trees in the Probabilistic Safety Analysis (PSA) of nuclear plants. The truncation limit is used to truncate cut sets of the gates when quantifying the fault trees. This report presents measures to estimate the probability of the truncated cut sets, that is, the amount of truncation error. The functions to calculate the measures are programmed into the new fault tree quantifier FTREX (Fault Tree Reliability Evaluation eXpert) and a Benchmark test was performed to demonstrate the efficiency of the measures. The measures presented in this study are calculated by a single quantification of the fault tree with the assigned truncation limit. As demonstrated in the Benchmark test, Lower Bound of Truncated Probability (LBTP) and Approximate Truncation Probability (ATP) are efficient estimators of the truncated probability. The truncation limit could be determined or validated by suppressing the measures to be less than the assigned upper limit. The truncation limit should be lowered until the truncation error is less than the assigned upper limit. Thus, the measures could be used as an acceptability of the fault tree quantification results. Furthermore, the developed measures are easily implemented into the existing fault tree solvers by adding a few subroutines to the source code

[en] In order to evaluate accurately a Station BlackOut (SBO) event frequency of a multi-unit nuclear power plant that has a shared Alternate AC (AAC) power source, an approach has been developed which accommodates the complex inter-unit behavior of the shared AAC power source under multi-unit Loss Of Offsite Power (LOOP) conditions. The approach is illustrated for two cases, 2 units and 4 units at a single site, and generalized for a multi-unit site. Furthermore, the SBO frequency of the first unit of the 2-unit site is quantified. The SBO frequency at a target unit of Probabilistic Safety Assessment (PSA) could be underestimated if the inter-unit dependency of the shared AAC power source is not properly modeled. The effect of the inter-unit behavior of the shared AAC power source on the SBO frequency is not negligible depending on the Common Cause Failure (CCF) characteristics among AC power sources. The methodology suggested in the present report is believed to be very useful in evaluating the SBO frequency and the core damage frequency resulting from the SBO event. This approach is also applicable to the probabilistic evaluation of the other shared systems in a multi-unit nuclear power plant

[en] Since the attack on the World Trade Center on September 11, 2001, US Nuclear Regulatory Commission (NRC) has enforced electric power utilities to evaluate the level of physical security when designing a new nuclear power plant. In addition, International Atomic Energy Agency (IAEA) established security related department and has been strengthening security measures against potential sabotage. Accordingly, there has been growing worldwide interest in developing Vital Area Identification (VAI) method as one of the possible measures against sabotage. The United States is most actively taking advantage of Vital Areas by determining and protecting them

[en] In order to evaluate accurately a station blackout (SBO) event frequency of a multi-unit nuclear power plant that has a shared alternate AC (AAC) power source, an approach has been developed which accommodates the complex inter-unit behavior of the shared AAC power source under multi-unit loss of offsite power conditions. The SBO frequency at a target unit of probabilistic safety assessment could be underestimated if the inter-unit dependency of the shared AAC power source is not properly modeled. The approach is illustrated for two cases, 2 units and 4 units at a single site, and generalized for a multi-unit site. Furthermore, the SBO frequency of the first unit of the 2-unit site is quantified. The methodology suggested in the present paper is believed to be very useful in evaluating the SBO frequency and the core damage frequency resulting from the SBO event. This approach is also applicable to the probabilistic evaluation of the other shared systems in a multi-unit nuclear power plant

[en] A new Vital Area Identification (VAI) method was developed in this study for minimizing the burden of VAI procedure. It was accomplished by performing simplification of sabotage event trees or Probabilistic Safety Assessment (PSA) event trees at the very first stage of VAI procedure. Target sets and prevention sets are calculated from the sabotage fault tree. The rooms in the shortest (most economical) prevention set are selected and protected as vital areas. All physical protection is emphasized to protect these vital areas. All rooms in the protected area, the sabotage of which could lead to core damage, should be incorporated into sabotage fault tree. So, sabotage fault tree development is a very difficult task that requires high engineering costs. IAEA published INFCIRC/225/Rev.5 in 2011 which includes principal international guidelines for the physical protection of nuclear material and nuclear installations. A new efficient VAI method was developed and demonstrated in this study. Since this method drastically reduces VAI problem size, it provides very quick and economical VAI procedure. A consistent and integrated VAI procedure had been developed by taking advantage of PSA results, and more efficient VAI method was further developed in this study by inserting PSA event tree simplification at the initial stage of VAI procedure.

[en] In Probabilistic Safety Assessment (PSA) of nuclear power plants, (1) event trees and fault trees are modelled for the accident scenarios, (2) Minimal Cut Sets (MCSs) are calculated from the integrated fault trees, (3) MCS post-processing is performed to delete impossible MCSs and manipulate human errors, and (4) core damage frequency is finally calculated from these post-processed MCSs. Regular fault tree solvers restructure a fault tree, convert it into a modularized one, generate modularized MCSs from the modularized fault tree, and expands modules in MCSs. These modularized MCSs should be expanded for performing MCS post-processing. It is well known that post-processed MCSs cannot be easily modularized. This paper proposes a new method to factorize MCSs. Importance measures are employed for the acceleration of this factorization. The benchmarks tests showed the effectiveness of this algorithm. The algorithm in this paper minimizes computational memory and quickly detects modules in MCSs

[en] The purpose of this paper is to show an effect of the Initiating Event (IE) modeling method on the major the importance measures of components in a Probabilistic Safety Assessment (PSA). We have presumed that the importance for the components relevant to the IE would be changed if we directly use an IE frequency Fault Tree (FT) instead of an IE frequency value in a quantification process. In general, there are two approaches to estimate the frequencies for initiating events. They are a FT modeling method and a Bayesian analysis method for the experienced data. However, the IE frequency has been usually handled as a value in a quantification process even though it was obtained through an FT analysis. Accordingly, we have questioned that if we use an IE frequency value in a quantification process, we would not obtain the exact importance for the components relevant to an IE. It means that the importance for the mitigating system's components related to an initiating event would be changed if we reflect the effect of a failure inducing an initiating event. Therefore, in this paper, we revised the Loss of Component Cooling Water (LOCCW) FT to use in a quantification process directly and evaluated the major importance measures such as Fussel-Vesely (F-V), Risk Achievement Worth (RAW), Risk Reduction Worth (RRW) importance for the Component Cooling Water (CCW) system and the Essential Chilled Water (ECW) system components