Seismic Probabilistic Risk Assessment

Abstract

This chapter presents an overview of the seismic probabilistic risk assessment (PRA) for nuclear power plants. First, the basic concept of seismic PRA and its application to continuous improvement of seismic safety of nuclear power plants are explained. Then, the methodology of seismic PRA for nuclear power plants is explained. Seismic PRA process can be divided into probabilistic seismic hazard analysis, fragility analysis of structures/components, and accident sequence analysis. As for the probabilistic seismic hazard analysis, modeling of earthquake occurrence, a method for earthquake ground motion prediction, and logic tree model are introduced. Then, the fragility analysis of structures/components is described. This includes an explanation of the selection of failure modes, and the realistic strength and realistic response of buildings/components.

1 Probabilistic Risk Assessment (PRA)

The accident that took place at the Fukushima Daiichi Nuclear Power Plant as a result of the 2011 Great East Japan earthquake and consequent tsunami raised awareness of the risks associated with nuclear facilities. To achieve high level of safety for nuclear facilities, probabilistic risk assessment is considered useful because it provides insight into the weakness by identifying as many possible accident scenarios as possible.

Risk is defined as combination of consequence of an accident and its associated likelihood, i.e., occurrence probability.

When a severe accident happens at a nuclear power plant, leading to damage to the reactor core and the reactor containment vessel, large amount of radioactive material are released into the environment, exposing the public in the vicinity of the plant with respect to radiation exposure as well as environmental contamination.

As a result of probabilistic risk assessment, contributors to accident risk can be quantitatively assessed, which are potential causes of an accident. Moreover, PRA can estimate quantitatively the effectiveness of countermeasures to be taken to reduce risks.

Quantitative safety goal and subsidiary objectives, so-called performance goals, are referred to as criteria that represent the societal acceptability of the risks (Appendix 8.1).

Safety goals usually concern the radiation risk of the public as well as environmental contamination, while performance goals concern, for example, equivalent plant condition such as core damage frequency as well as large early release frequency.

2 Categorization of Initiating Events

Events that initiate accidents are classified into (a) random failure of components, (b) internal hazards such as internal flooding and an internal fire, and (c) external hazards. External hazards are classified into natural hazards such as earthquakes and tsunamis, and human-induced hazards such as an airplane attack and an intentional act.

3 Overview of Seismic PRA

In seismic PRA, attention is focused on earthquakes that may happen in the future and affect the safety of nuclear power plant. The failure probabilities of structures (e.g., buildings) and components are assessed. Uncertainty associated with the earthquake ground motion, the responses of buildings and components, and strength of structures and components are considered. Then, the occurrence probabilities (or frequencies) of various accident sequences and the magnitude of their consequences are estimated.

Furthermore, in seismic PRA, the core damage frequency is estimated on the basis of seismic hazard analysis, structures/components fragility analysis, and accident sequence analysis. The seismic hazard analysis estimates the probability of occurrence, within a certain period of time (e.g., 50 years), of an earthquake for which the ground motion, e.g., at the hypothetical free surface of bedrock, exceeds a certain level. The fragility analysis gives the failure probability of structures/components (for example, buildings, structures, and component piping systems) as a function of the seismic ground motion level. Figure 8.1 illustrates the overview of seismic PRA.

Fig. 8.1

Overview of seismic PRA

Figure 8.2 shows the seismic PRA procedure.

Fig. 8.2

Seismic PRA procedure

• (1) Collection and analysis of site/plant information, and field surveys

  To start with, the information required for seismic PRA is collected and compiled. Field surveys (site/plant walkdown surveys) are performed as required. Through field surveys, the type of information is obtained which cannot be acquired sufficiently from documents alone. They may include the information whether components and piping systems have been installed exactly as designed, the presence of any facility that may affect another facility, and factors that may complicate post-earthquake recovery actions.

• (2) Identification of accident scenarios

  According to plant information and field survey results obtained in Step (1), possible scenarios of an accident, for example, leading to core damage, are identified.

• (3) Seismic hazard analysis

  Based on information on the seismic source locations and magnitudes, occurrence frequencies of earthquakes that may affect the nuclear power plant, the probability of ground motion, e.g., at virtually free surface of bedrock, exceeding a certain level are estimated.

• (4) Ground stability analysis

  The potential effects of hazards such as sliding of foundation ground under the buildings/structures and slope failure are studied. Their potential impacts on structures and components require attention.

• (5) Structure/component fragility analysis

  On the basis of information on the dynamic response analysis and strength of structures/components, failure probabilities are estimated as a function of the earthquake ground motion intensity. The failure should not be assessed based on conservative design assumptions, but based on realistic response as well as realistic strength of structures/components.

• (6) Accident sequence analysis

  On the basis of the seismic hazard analysis, the fragility analysis and information about random failures of safety functions and the scenarios of accidents leading to core damage are analyzed (in the form of event trees and fault trees). The conditional probability and occurrence frequency of core damage are calculated. This leads to the identification of components and facilities that are likely to serve as a cause of core damage.

Figure 8.3 shows how the core damage frequency is calculated from the seismic hazard curve and the plant fragility curve (representing conditional probabilities of core damage as a function of earthquake ground motion intensity).

Fig. 8.3

The core damage frequency obtained from the seismic hazard curve and the plant fragility curve

4 Probabilistic Seismic Hazard Analysis

A relationship between the earthquake ground motion intensity and its likelihood of occurrence is obtained by the probabilistic seismic hazard analysis (PSHA). This section describes the conventional method of PSHA for a nuclear facility.

4.1 Probabilistic Seismic Hazard Analysis Flow

As shown in Fig. 8.4, a seismic hazard curve, a relationship between the earthquake ground motion intensity and its annual exceedance frequency, is obtained by taking account of various uncertainties. The uncertainties are due to the location and characteristics of seismic sources, and the characteristics of the seismic wave propagation and ground motion amplification. Typically, amplification up to reference bedrock surface (virtually free surface of bedrock) is considered in seismic hazard analysis. Amplification from reference rock to surface ground is considered in fragility analysis.

Fig. 8.4

Probabilistic seismic hazard curve

Figure 8.5 shows the seismic hazard analysis flow.

Fig. 8.5

Seismic hazard analysis flow

• (1) Identification of the seismic source model

  In seismic hazard analysis, a probabilistic model is employed to model the possibilities of earthquake occurrence that may affect the target site. The magnitudes and occurrence of such earthquakes are considered probabilistically and statistically.

• (2) Ground motion prediction model

  A ground motion prediction model, i.e., attenuation formula, is used to estimate the probability distribution of earthquake ground motion at the target site for each seismic source identified in Step (1).

• (3) Construction of a logic tree

  For the source model and ground motion prediction model, a logic tree is employed to account for epistemic uncertainty, i.e., knowledge-oriented uncertainty, in parameters that differ in assumptions concerning, for example, the fault length and the earthquake magnitude.

• (4) Seismic hazard analysis

  As a result of analysis, the seismic hazard curve, i.e., the relationship between the earthquake ground motion intensity and the annual exceedance frequency, is obtained. Uncertainties in the estimated hazard curve are analyzed using the logic tree.

• (5) Generation of artificial earthquake ground motion for fragility analysis

  A seismic hazard curve is calculated for response spectra of the ground motion for different vibration periods. A value representing a specified level of exceedance probability is selected at each vibration period for the response spectrum. Then, a uniform hazard spectrum is obtained by connecting these points. From the uniform hazard spectrum, simulated ground motion for use in dynamic response analysis is generated.

4.2 Probabilistic Seismic Hazard Analysis Method

4.2.1 Seismic Source Model

Considering all earthquakes that may affect the target site, seismic sources are modeled and the occurrence probabilities of such earthquakes are estimated. Usually, the potential seismic sources within a distance of approximately 100–150 km from the target site are considered for seismic hazard analysis.

Earthquakes of all seismogenic mechanisms, including inland crustal earthquakes, interplate earthquakes, and intra-oceanic plate earthquakes, should be addressed.

As shown in Fig. 8.6, the seismic sources are classified into two categories: earthquakes with specified faults, for which the magnitude and location can be specified in advance, and area seismic sources (i.e., earthquakes without specified faults or diffuse seismicity), where seismic activities of unspecifiable magnitudes at unspecifiable locations within a certain zone are modeled statistically.

Fig. 8.6

Earthquakes with specified faults and area seismic sources

An example of interplate earthquakes is earthquakes that are located off the Pacific coast of Tokai/Tonankai/Nankai area in Japan. The Japanese Government’s Headquarters of Earthquake Research Promotion has released estimation about the probabilities of their occurrence in the regions as shown in Fig. 8.7.

Fig. 8.7

Temporal–spatial distributions of earthquakes occurrence around the Nankai Trough [1]

The probability of occurrence of a particular type of earthquake is modeled using a time-dependent probability model to address the periodicity of earthquake occurrence, if the time and the frequency of past occurrences are known. In the case that the time of past occurrences is unknown, a homogeneous probability model such as a Poisson model is used alternatively. A homogeneous probability model assumes an unchanging probability of occurrence in time. Figure 8.8 shows three options (a–c) to estimate annual occurrence frequency or probability of earthquake.

Fig. 8.8

Three options to obtain earthquake occurrence frequency

As for the types of earthquakes that are difficult to be specified in advance in terms of magnitude, location, and their occurrence, the occurrence frequencies are modeled using the Gutenberg–Richter law as shown in Fig. 8.9. The spatial distribution of earthquakes is assumed to be uniform within the same seismotectonic zone. A seismotectonic zonation proposed for Japan is shown in Fig. 8.10.

Fig. 8.9

Gutenberg–Richter law [2]. White dots correspond to n(M) which is the number of earthquakes in which the magnitude is M. Black dots represent N(M) which is the total number of earthquakes in which the magnitude is M or above

Fig. 8.10

Seismotectonic zonation for earthquake without specified faults [2]

4.2.2 Ground Motion Prediction Model

Earthquake ground motion is estimated using a ground motion prediction equation, i.e., attenuation formula. Attenuation formula is usually a function of the magnitudes and fault-to-site distance and site amplification characteristics. The predicted earthquake ground motion is an average value. Attenuation formula is derived by means of regression analysis from ground motion records at many observation stations, and usually exhibits large variability. Referring both to the average value and the variability, the probability of ground motion exceeding a certain level is estimated. As shown in Fig. 8.11, the variability of attenuation formula, i.e., the variability of logarithm of the ground motion, is modeled as the normal distribution.

Fig. 8.11

Mean and variability of ground motion estimated using the attenuation formula

4.2.3 Logic Tree

Uncertainty in assumptions concerning the parameters is called epistemic uncertainty. In seismic hazard analysis, the epistemic uncertainty in parameters, e.g., the fault length, earthquake magnitudes, and earthquake occurrence frequency, is modeled as possible different opinions among experts. Therefore, a logic tree, like the one shown in Fig. 8.12, is constructed to account for all possible opinions. The logic tree is a tree diagram showing differences in expert opinions. With the weighing by confidence in different opinions, it contributes to the estimation of uncertainty in seismic hazard analysis.

Fig. 8.12

An example of a logic tree

4.2.4 Seismic Hazard Curve

As shown in Fig. 8.13, the seismic hazard curves for individual sources are unified to produce a seismic hazard curve. The result of seismic hazard analysis may change in the future as more seismological knowledge are accumulated. It is advisable to examine the validity of seismic hazard analysis from time to time for continuous updating.

Fig. 8.13

Seismic hazard curve

4.2.5 Generation of Artificial Earthquake Ground Motion for Fragility Analysis

A seismic hazard curve is produced for acceleration response spectrum, i.e., the maximum responses to the ground motion of one-mass-spring-damper system (Part II, Chap. 14, Sect. 14.1) with different natural vibration periods. For the same exceedance frequency, the response values are plotted at each vibration period to produce a uniform hazard spectrum as shown in Fig. 8.14. The fragility analysis is performed using the artificial time history earthquake ground motion generated so that its response spectrum fits the uniform hazard spectrum.

Fig. 8.14

Uniform hazard spectrum

5 Structure/Component Fragility Analysis

Relationships between the earthquake ground motion intensity and probability of failure of structure/component are obtained by the seismic fragility analysis. This section describes the conventional method of the seismic fragility analysis.

5.1 Fragility Analysis Flow

A fragility curve shows conditional failure probabilities for different ground motion intensity levels. On the basis of information concerning the dynamic response and strength of structures and components, a fragility curve is developed. A fragility curve is estimated for structures and components that are significant in assessing the nuclear power plant risks.

For earthquake-resistant design, a deterministic approach is employed and structural components are designed to have an enough safety margin for the deformation and stress produced by the design basis external forces. On the other hand, the actual external forces from an earthquake and the actual material strength inherently exhibit scatter. In other words, the external forces and the strength of structures and components are random variables as shown in Fig. 8.15, and the failure probability of structures and components is represented by a fragility curve like the one shown in Fig. 8.16.

Fig. 8.15

External force and strength relationship

Fig. 8.16

Fragility curve

Figure 8.17 shows the structures/components fragility analysis flow.

Fig. 8.17

Structure/component fragility analysis flow

• (1) Selection of target structures and components, and the determination of their failure modes

  After selecting target structures and components, their failure and associated response parameters to be used for determining their failure are defined.

• (2) Selection of an analysis method

  The analysis method is chosen in view of required accuracy and the objectives of the analysis.

• (3) Assessment of realistic strength

  For each of the target structures and components, the limit beyond which it will fail (i.e., the strength of the structures or components) is assessed probabilistically. The assessment method that may be employed is either an empirical method using experiments, a numerical method based on analysis, or a method that makes use of engineering judgments integrating both empirical and numerical methods.

• (4) Analysis of realistic seismic response

  For each of the target structures and components, the dynamic response is estimated stochastically, i.e., as a random variable. Conventionally, the input ground motion for this analysis is that derived from the uniform hazard spectrum (Sect. 8.4.2).

• (5) Fragility analysis

  On the basis of the result of the assessment of the probabilistically estimated strength and response, the fragility curves for the structures and components are obtained.

5.2 Building Fragility Analysis

5.2.1 Selection of the Failure Mode

First, the building’s failure mode and the elements of the building prone to failure are determined. This is followed by the selection of response parameters for assessment of failure and the ground motion intensity.

Failure modes of structure that may lead to the damage of buildings may include sliding, overturning, story collapse, failure of local elements, and the failure of nonstructural members like partitions, ceilings, and doors. The dominant failure mode is selected from among them, and the safety critical elements of the building are identified. Conventionally, a failure mode like story collapse may be assumed to lead directly to core damage.

5.2.2 Selection of the Response Parameter

A parameter of the seismic response (e.g., stress, acceleration, strain, or deformation) is chosen for assessment of failure. Note that a failure is assessed probabilistically for the chosen response parameter.

When assessing the probability of shear failure of seismic walls leading to story collapse, the shearing force or shearing strain is chosen as the response parameter.

5.2.3 Assessment of Realistic Strength

A typical process of earthquake-resistant design includes the calculation of the dynamic response of building to design input earthquake ground motion. This process is to verify that the dynamic response does not exceed the allowable limit (design strength) depending on the material property. A fragility curve is derived considering the probability distribution of the realistic dynamic response and the probability distribution of realistic strength.

The probability distribution of strength is obtained by statistically analyzed experimental data. The dominant failure mode of buildings considered for a reactor building is the shear failure of a seismic wall made of reinforced concrete.

Table 8.1 summarizes failure limit values of shearing stress and shearing strain of reinforced concrete seismic wall. Figure 8.18 shows the relationship between the shear stress and the shear strain of seismic wall. The failure limit point for seismic walls is probabilistically determined on the basis of high confidence value for experimental results. However, such values may change in the future as more data are collected in experiments. It is advisable to examine the validity of experimental data from time to time and to update the failure limit point values.

Table 8.1 Shearing stress and strain settings [3]

Fig. 8.18

Shearing stress and shearing strain relationship of seismic wall (conceptual)

5.2.4 Analysis of a Realistic Seismic Response

The dynamic response analysis is conducted probabilistically in consideration of uncertainty in the input earthquake ground motion and material property values.

In the analysis of the realistic seismic response, the factors that may significantly affect the response must be accounted for by considering the causes of the uncertainties. The probability distribution of the dynamic response can be obtained efficiently by means of sampling performed using, for example, the Monte Carlo simulation or the two-point estimate. Table 8.2 summarizes the characteristics of different sampling methods.

Table 8.2 Characteristics of different sampling methods [3]

In addition, it should be noted that there are two methods for the evaluation of realistic seismic response: a method based on response analysis and a method based on the response factors (i.e., an assessment of the probabilistic dynamic response using the response factors).

5.2.5 Determination of the Fragility Curve

Figure 8.19 shows how the failure probability F(α) for ground motion level α is calculated. As shown in the equation below, F(α) is a conditional probability that f _R (α, x), the probability density function of response to a ground motion intensity level α, exceeds f _S (x), the probability density function of realistic strength. F(α) is calculated as follows:

Fig. 8.19

Calculation of failure probability for ground motion intensity α

$$ F(\alpha ) = \int\limits_{0}^{\infty } {f_{S} (x_{R} )\left( {\int\limits_{{x_{R} }}^{\infty } {f_{R} (\alpha ,x){\text{d}}x} } \right){\text{d}}x_{R} = \int\limits_{0}^{\infty } {f_{R} (\alpha ,x_{R} )\left( {\int\limits_{0}^{{x_{R} }} {f_{s} (x){\text{d}}x} } \right){\text{d}}x_{R} } } $$

(8.1)

The large computation time is required when calculation of Eq. (8.1) is done for large number of ground motion level α. Therefore, as shown in Fig. 8.20, discretization is made appropriately and the fragility curve is obtained by interpolation. Typically, the lognormal distribution is assumed for the fragility curve.

Fig. 8.20

Interpolation of F(α) to obtain fragility curve

5.3 Components Fragility Analysis

The fragility of plant components is assessed in consideration of their required functions, paying attention to both structural failures and functional failures.

When assessing the fragility of passive, i.e., static, components such as a tank or heat exchange, the probability of losing required functions due to structural failure in the form of, for example, ductile fracture or brittle fracture is examined. As for active components such as an electrical board, pump, or valve, attention is given not only to the possibility of their structural failure, but also to the possibility of their limit for active or electrical functioning being lost.

To identify the functioning limit for active components, vibration testing is performed using a shaking table as described in Chap. 5, Sect. 5.6.

Table 8.3 summarizes component strength assessment methods. Figure 8.21 shows photos from the vibrating testing of active components.

Table 8.3 Component strength assessment [4]

Fig. 8.21

Photos from the vibration testing of active components [4]

A component fragility analysis is conventionally performed using the safety factor method. With the safety factor method, fragility is determined by referring to the result of response analysis along with the strength factors and response factors.

The fragility in relation to structural failure is determined in reference to material strength. The fragility in relation to functional failure is determined in reference to vibration test results. The median and uncertainty, i.e., logarithmic standard deviation, are estimated to obtain fragility curve.

The value of the response factor is determined from response analysis, accounting for building-related and component-related uncertainties.

Major assumptions (regarding, for example, uncertainty and the response factor) made in fragility analysis are as follows.

The fragility of a component is measured according to the maximum ground acceleration that the component may tolerate without failing. This maximum tolerable ground acceleration (referred to as the fragility acceleration) is given as a random variable as the following equation

$$ A = A_{\text{m}} \cdot \varepsilon_{R} \cdot \varepsilon_{U} . $$

(8.2)

$A_{m}$ :

Median of $A$, which is the ground acceleration beyond which the component will fail (i.e., median fragility acceleration).

$\varepsilon_{R}$ :

Random variable, i.e., variable with probability distribution, representing scatter caused by aleatory uncertainty inherent to physical randomness; the mean value is 1.0 while the logarithmic standard deviation is denoted as $\beta_{R}$.

$\varepsilon_{U}$ :

Random variable representing scatter caused produced by epistemic uncertainty; the mean value is 1.0 while the logarithmic standard deviation is denoted as $\beta_{U}$.

A fragility curve, used for the representing component fragility, is obtained as a cumulative distribution function of the fragility acceleration $A$.

Figure 8.22 shows examples of three fragility curves with 5, 50, and 95 % confidence level. In the figure, $\beta_{R}$ is represented as the gradient of the fragility curve, while $\beta_{U}$ contributes to the width between fragility curves of different confidence levels. The fragility curve considering both aleatory and epistemic uncertainties, so-called composite fragility curve, is also shown in Fig. 8.22.

Fig. 8.22

Fragility curve

Typically, the component strength can be represented according to the high confidence and low probability of failure (HCLPF) value derived from the fragility curve and then compared with the design basis ground motion. The HCLPF value is the strength at which failure probability of 5 % can be assured with 95 % confidence level.

6 Accident Sequence Analysis

6.1 Accident Sequence Analysis Flow

On the basis of the seismic hazard analysis and buildings/components fragility analysis, the probability of failure is determined for different facilities to enable the modeling of accident sequences leading to core damage. The accident sequence analysis starts with the analysis of accident scenarios and to calculate core damage frequency. Figure 8.23 shows the accident sequence analysis flow.

Fig. 8.23

Accident sequence analysis flow

• (1) Definition of accident sequences

  Accident sequences are defined using event trees and fault trees as a systematic representation of scenarios leading to core damage. The event tree shows how progressive failures of safety functions may lead to an accident including core damage, while the fault tree shows how the combinations of component failures may be a cause of a safety function failure.

• (2) Plant fragility analysis

  On the basis of the structure/component fragility analysis and the accident sequence analysis, the fragility of the entire plant is obtained.

• (3) Calculation of core damage frequency

  The core damage frequency is calculated according to the seismic hazard analysis and the plant fragility analysis. This is followed by the identification of the dominant accident sequences leading to core damage, and the components and facilities for which failure would contribute significantly to core damage are also identified.

6.2 Accident Sequence Analysis Method

Figure 8.24 shows the technical components of accident sequence analysis. Examples of the initiating events for an accident are the loss of off-site power, damage to buildings and structures (including the fracture of the reactor pressure vessel and/or the containment vessel), and piping fracture. After defining how the accident sequences progress depending on the success and failure of maintaining safety functions after the occurrence of the initiating event, the total core damage frequency is calculated as the sum of the probabilities of different accident sequences that lead to core damage.

Fig. 8.24

Technical components of accident sequence analysis

Following the calculation of core damage frequency, a cut-set analysis is performed to identify the dominant sequences leading to core damage and to identify the components and facilities for which failure contributes significantly to core damage. On the basis of the results of such analyses, introduction of measures for safety improvement (including management-oriented measures) can be decided for the reduction of the core damage frequency.

Seismic Level-2 and Level-3 PRA (Appendices 8.2 and 8.3) are considered to be a future challenge, and researches and development are being conducted toward their realization. Other important challenges include the assessment of risks due to earthquakes combined with other hazards (e.g., tsunamis, floods, and fires) and earthquake risks at multiple-unit plants and multiple sites.

Appendices

Appendix 8.1: Safety Goals of Countries/Agencies Around the World

See Table 8.4.

Table 8.4 Safety goals of countries/agencies around the world

Appendix 8.2: Levels of Probabilistic Risk Assessment (PRA)

PRA is a method for analyzing and assessing events (i.e., accidents and failures) that can happen for components and systems at a plant in a comprehensive and systematic manner, and it enables the quantitative assessment of the probability of occurrence of each event as well as the magnitude of consequence for each event.

PRA estimates three levels of risk, depending on different stages in accident progression.

Level-1 PRA develops events up to the occurrence of core damage and the core damage frequency is estimated.

Level-2 PRA further develops events leading to a release of radioactive material, and the frequency of occurrence of such release is obtained. PRA up to the assessment of the risk of containment failure is referred to as Level 1.5 PRA.

Level-3 PRA consists of assessing the societal risk including the health risks to the public in the nearby area including its occurrence frequency based on the results of Level-2 PRA.

Figures 8.25 and 8.26 show overview of Level-1, Level-2, and Level-3 PRA.

Fig. 8.25

Illustration of PRA

Fig. 8.26

Overview of Level-1, Level-2, and Level-3 PRA

A methodology for seismic Levels-2 and Level-3 PRA are considered to be a future challenge to be developed.

Appendix 8.3: Overview of Level-2 PRA for Earthquake-Initiated Events

Seismic Level-2 PRA requires the quantification of accident leading to release of radioactive materials including occurrence frequencies and the determination of the source term. But it is basically considered that seismic Level-2 PRA can be implemented by applying Level-2 PRA procedures that address internal events.

Seismic Level-2 PRA identifies and estimates accident scenarios leading to release of radioactive materials. Attention is given not only to the possibility of damage of containment vessel by the earthquake, but also to possible damage of SSCs which support the capacity of the containment vessel. The accident scenarios leading to core damage are classified into several groups according to their types. When classifying, attention is given to the particularities of and similarities among different scenarios. Then, for each group of accident sequences, representative plant damage states are assigned. When classifying accident sequences into the plant damage states, accident sequences after core damage are carefully studied in terms of how accident progresses and effects on the characteristics of source term. Due considering is given to the fact that the characteristics of accidents due to earthquake are different from those due to random failure of components. This process enables the appropriate characterization of accident sequences. To appropriately estimate the source term in seismic Level-2 PRA, it is necessary to review the condition that is simplified in conventional seismic Level-1 PRA. One of examples is an assumption that failure of reactor building directly results in core damage. An event tree is constructed for the modeling of accident sequences showing how events may develop and lead to release of radioactive materials. Accident sequences in the event tree start from the initial states (the classified plant damage states). Then, this is followed by the determination of the source term for each containment failure mode.

Seismic level-3 PRA, which is to follow the determination of source term, can be implemented basically by applying Level-3 PRA procedures that address internal events.