The implementation of fault tree analysis

approaches in nuclear power plant

probabilistic safety assessment
Cite as: AIP Conference Proceedings 2180, 020010 (2019); https://doi.org/10.1063/1.5135519
Published Online: 10 December 2019

Julwan Hendry Purba, Damianus Toersiwi Sony Tjahyani and Deswandri

ARTICLES YOU MAY BE INTERESTED IN

The assessment of the radioactive releases from the confinement structure of AP1000 by
probabilistic safety analysis
AIP Conference Proceedings 2180, 020019 (2019); https://doi.org/10.1063/1.5135528

Preface: The 3rd International Conference on Nuclear Energy Technologies and Sciences
(ICoNETS) 2019
AIP Conference Proceedings 2180, 010001 (2019); https://doi.org/10.1063/1.5135509

Reliability program plan for field programmable gate array-based I&C system of nuclear
power plant
AIP Conference Proceedings 2180, 020035 (2019); https://doi.org/10.1063/1.5135544

AIP Conference Proceedings 2180, 020010 (2019); https://doi.org/10.1063/1.5135519 2180, 020010

© 2019 Author(s).
 The Implementation of Fault Tree Analysis Approaches in
Nuclear Power Plant Probabilistic Safety Assessment
Julwan Hendry Purbaa), Damianus Toersiwi Sony Tjahyanib), Deswandric)

Center for Nuclear Reactor Technology and Safety, National Nuclear Energy Agency of Indonesia (BATAN), Gd. 80
Kawasan Puspiptek Serpong, Tangerang Selatan – 15310, Indonesia.
a)
Corresponding author: [email protected]
b)
[email protected]
c)
[email protected]

Abstract. Probabilistic safety assessment (PSA) has been extensively implemented to assess the performance of nuclear
power plant (NPP) safety systems. One well-known modeling approach in NPP PSA is a fault tree analysis (FTA). A
fault tree is a graphical representation of possible failure scenarios of the system being evaluated. To estimate the top
event failure probability, a quantitative analysis needs to be performed based on those scenarios. Prior to performing
quantitative analysis, basic events’ failure probabilities of the system fault tree need to be provided well in advance.
Conventional FTA assumes that basic events always have precise probability distributions characterizing their lifetime to
failure. However, in practical applications, this is not the case. For example, a new system will not have sufficient
operating experiences to probabilistically estimate reliabilities of their components. To deal with this limitation, a number
of approaches has been developed and proposed. Each approach offers advantageous but also has disadvantageous. Since
the results of FTA will be used to verify NPP designs, it is necessary to select the most suitable approach. It is, therefore,
essential to clearly understand the strengths and weaknesses of each approach. The purpose of this study is to review the
implementation of various FTA approaches in NPP PSA. The strengths and weaknesses of each approach are also
discussed. To achieve research objectives, this study classified those FTA approaches into conventional FTA and fuzzy
FTA. Fuzzy FTA is further grouped into fuzzy hybrid FTA and fuzzy based FTA. This study concludes that safety
analysts need to, firstly, confirm the type of reliability data at hands. Secondly, if epistemic uncertainty is essential and
need to be considered in the study being performed, fuzzy based FTA should be applied. Otherwise, safety analysts
should apply conventional FTA or fuzzy hybrid FTA depending on how the basic events’ failure probabilities are
generated.
Keywords: Fault tree analysis, nuclear power plant, probabilistic safety assessment, fuzzy fault tree analysis.

INTRODUCTION
The main objectives of nuclear power plant (NPP) safety systems are to ensure the normal operation of the plants
without risk exposure to operators, publics, and environment; to prevent accidents when unexpected events happen;
and to mitigate the consequences of accidents when they really occur. Over the past two decades, those objectives
are achieved through probabilistic safety assessment (PSA). It is a comprehensive approach to evaluate significant
plant vulnerabilities, to construct accident scenarios, to predict the safety level of the plant, and to numerically
estimate potential risks. PSA evaluates safety and risk in NPPs by postulating potential initiating events that might
initiate accidents [1]. Those initiating events could be internal and external hazards, component failures, and/or
human errors. Designers, utilities as well as regulatory body utilize the results of PSA to verify NPP design, to
propose new design, to adjust operation procedures, to improve the reliability of safety systems, and to potentially
change regulation, or to license basic events.
One well-known modelling approach in NPP PSA is a fault tree analysis (FTA). It graphically represents parallel
and/or sequential fault events, which could lead to the system failure. The system failure is defined as a top event of

The 3rd International Conference on Nuclear Energy Technologies and Sciences (ICoNETS) 2019
AIP Conf. Proc. 2180, 020010-1–020010-7; https://doi.org/10.1063/1.5135519
Published by AIP Publishing. 978-0-7354-1927-8/$30.00

020010-1
the tree. Through a fault tree, possible failure scenarios could be logically depicted using Boolean gates [2-4]. Based
on those possible failure scenarios, a quantitative analysis is performed to estimate the failure probability of the top
event. If the top event failure probability is greater than the objective probability, system designers should redesign
the system or propose a more innovative design. These can achieved by utilizing better quality components,
modifying maintenance policy, adjusting testing activities, and suggesting component redundancies [5]. To ensure
that failure probability of the new proposed system is less than the objective probability, its reliability needs to be re-
quantified.
To perform quantitative analysis, basic events of the fault tree of the evaluated system need to have
corresponding failure probabilities [6, 7]. Conventional FTA assumes that basic events always have precise
probability distributions describing their lifetime to failure. In practical engineering applications, this is not always
the case. For example, a new system will not have sufficient operating experiences and historical failure data to
assess reliabilities of their components [8]. To deal with the limitation of generating basic events’ failure
probabilities, a number of FTA approaches has been developed and proposed. Each approach offers advantageous
but also has disadvantageous. Since the results of FTA will be used to verify NPP designs, it is necessary to select
the most suitable approach to the study of interest. Therefore, clear understanding on the strengths and weaknesses
of each approach is essential. The present paper aims to review the implementation of FTA approaches, which have
been proposed and developed for generating basic events’ failure possibilities in the quantitative analysis of FTA.
The strengths and weaknesses of each approach are also discussed to help safety analysts properly select the most
appropriate approach to their study. To achieve research objectives, the implementation of FTA approaches are
classified into two groups, i.e. conventional FTA and fuzzy FTA. Fuzzy FTA is further grouped into fuzzy hybrid
FTA and fuzzy based FTA.

METHODOLOGY
A fault tree is a graphical representation to depict logical interrelationships amongst basic events to the
predefined undesired top event that is the failure of the system being investigated. In drawing a fault tree model, the
process starts from the higher fault events to the more basic fault events. Boolean gates, then, denote the relationship
between those fault events. The higher event is the output of the gate and the lower event is the input to the gate. The
most common Boolean gates in a typical fault tree model are an OR Boolean gate and an AND Boolean gate. A
typical fault tree model is graphically shown in Figure 1.

Figure 1: A typical fault tree model [9].

The top event A0 in Figure 1(a) will be failure if one of input events Ai fails. Meanwhile, the top event A0 in
Figure 1(b) will be failure if all input events Ai fails at the same time.
The top event failure probability is a function of the basic event reliability data. Based on how basic events’
probabilities are generated to characterize their reliabilities, the implementation of FTA approaches in NPP PSA can
be organized into two groups, i.e. conventional FTA and fuzzy FTA. Fuzzy FTA can be further categorized into two
groups, i.e. fuzzy hybrid FTA and fuzzy based FTA. This classification is depicted in Figure 2.

020010-2
 Figure 2: Classification of fault tree analysis implementation in NPP PSA.
This study reviews the implementation of each approach in Figure 2 using a number of accessible scientific
publications. Strengths and weaknesses of each approach are discussed to help safety analysts decide which
approach is the most relevant to their study.

RESULTS AND DISCUSSION

The main difference between conventional FTA and fuzzy FTA is in the data used to represent the occurrence
likelihood of individual basic events constructing the fault tree of the system being interest. In conventional FTA,
basic events’ reliabilities are probabilistically estimated using their historical failure data. Meanwhile, in fuzzy FTA,
basic events’ reliabilities are characterized using fuzzy probabilities. In the sequel, each approach is elaborated and
its strengths and weaknesses are also discussed.

Conventional Fault Tree Analysis

In conventional FTA, basic event lifetime to failure is characterized by a probability distribution. It is a
mathematical representation of the occurrence probabilities of various possible outcomes within a specified event. It
characterizes a random phenomenon about the event probabilities. Basic event failure probability is represented in
the form of conventional probability of, for example, 1.8E-5.
In this FTA approach, Boolean algebras are used to quantify the output of every Boolean gate. The output of the
OR Boolean gate and the AND Boolean gate depicted in Figure 1 can be respectively quantified using (1-2) [9, 10].
𝑃 𝐴! = 1 − !!!! 1 − 𝑃 𝐴! (1)
!
𝑃 𝐴! = !!! 𝑃 𝐴! (2)
where n is the number of fault events and 𝑃 𝐴! is the probability of event 𝐴! . Various data sources have been used
by scholars to generate probability distributions of basic events 𝑃 𝐴! in (1-2). Each source offer advantageous but
also has disadvantageous.
To obtain the most actual results of FTA, basic event probability distributions should be collected from operating
experiences of the system being evaluated [11, 12]. The Living PSA paradigm is then applied to reflect changes in
the plants during their lifetime [13]. Those collected data will represent the real performance of the system being
investigated. The probability of the undesired top event calculated from basic event probabilities using (1-2) will
represent the real reliability of the system. Hence, system designers can be much confident to redesign the system or
to propose a more innovative design in order to improve system reliability. Unfortunately, basic events do not
always have probability distributions to characterize their lifetime to failure. For example, when an accident occurs
infrequently or a system design is new, sufficient operating experiences will not be available. Therefore, reasonable
historical failure data are not sufficient to statistically estimate the component reliability characteristics.
When historical failure data are still not available or insufficient, it is therefore unavoidable to apply generic data
taken from operating experiences of other nuclear facilities and non-nuclear facilities [3, 14-17]. However, the
results of FTA will not actually describe the real performance of the system being evaluated. Should system need to
be redesigned or a more innovative design needs to be proposed to improve reliability; system designers cannot
directly rely on the results. They have to evaluate uncertainties and imprecision caused by those generic data. For
this purpose, Monte Carlo simulation (MCS) is widely used. It can calculate the overall uncertainties of

020010-3
conventional fault tree analysis [18, 19]. MCS has been commonly used to solve real engineering problems in many
fields for reliability analysis. It allows to realistically modelling the behavior of complex engineering systems.
However, MCS is appropriate only for quantifying aleatory uncertainty. It is not suitable for quantifying epistemic
uncertainty [18, 20].

Fuzzy Fault Tree Analysis

In fuzzy FTA approach, fuzzy probabilities are applied to characterize basic event reliabilities. A fuzzy
probability is defined as a membership function of a fuzzy set. It is expressed in the space of probabilities to
characterize event reliability. This membership functions are usually defined in a real unit interval [0, 1].
Trapezoidal and triangular fuzzy numbers are the most common membership functions to represent fuzzy
probabilities [20]. Basic event fuzzy probability is represented in the form of membership function of triangular
fuzzy number of, for example, (0.25, 0.37, 0.49), or in the form of membership function of trapezoidal fuzzy number
of, for example, (0.15, 0.2, 0.3, 0.35).
Two different approaches have been proposed to generate basic events’ fuzzy probabilities. The first approach
utilizes limited historical failure data [21, 22]. In this approach, the membership functions describing fuzzy
probabilities are generated using the lower bound, the middle and the upper bound values, which are statistically
calculated from the available limited historical failure data [9]. The second approach utilizes experts’ judgments. In
this second approach, a group of experts are asked to subjectively justify reliabilities of basic events using
qualitative failure possibilities. These qualitative assessments already have predefined corresponding quantitative
fuzzy probabilities [23, 24]. In this second approach, it is essential to select members of the expert team who have
expertise and experiences on the system being studied to avoid bias in their judgment. It is also necessary to weight
selected experts to confirm their credibility. Kumaraningrum et al. weighted selected experts based on professional
experience, education background, technical qualifications, and the involvement in the design, construction as well
as commissioning of the reactor being evaluated [25].
The advantage offered by these two approaches is their capabilities to characterize basic event reliabilities
without being confined to probability distributions. Based on how basic event fuzzy probabilities utilized in FTA
approaches, fuzzy FTA is further organized into two sub-groups, i.e. fuzzy hybrid FTA and fuzzy based FTA. In the
sequel, each sub-group is elaborated and discussed in details.

Fuzzy Hybrid Fault Tree Analysis

In fuzzy hybrid FTA approach, fuzzy probabilities are only applied to characterize basic event reliabilities. The
reliability of intermediate events and the top event are still represented by conventional probability. Therefore, the
probabilities of the top event A0 in Figures 1(a) and 1(b) are still quantified using Boolean algebras in (1-2) [26-28].
To integrate basic events’ fuzzy probabilities into Boolean algebra quantification, they need to be, firstly converted,
into the form of conventional probability, for example, from the form of (0.25, 0.37, 0.49) into the form of 1.8E-5.
To convert basic event fuzzy probabilities into their corresponding conventional probabilities, two algorithms are
needed, i.e. a defuzzification technique and an Onisawa’s logarithmic function.
The defuzzification technique is to decode a fuzzy probability into a single scalar quantity. For example, from
the form of (0.35, 0.50, 0.65) into the form of 0.102619. Various defuzzification techniques have been proposed for
a number of different applications and purposes. Huang, Chen and Wang [29] acknowledged that no one single
defuzzification technique could be applied in any applications. An area defuzzification technique developed and
proposed by [9] has been confirmed to be suitable for NPP PSA.
The Onisawa’s logarithmic function [30] is to convert the single scalar quantity generated by the defuzzification
technique into a conventional probability. For example, from the form of 0.102619 into the form of 1.8E-5. This
generated conventional probability is then used to quantify the probabilities of intermediate events and the top event
based on Boolean algebras in (1-2).
Since fuzzy probabilities are applied to only characterize basic event reliabilities, epistemic uncertainty raised in
the basic events’ fuzzy probabilities will not be propagated into the top event failure probability. Furthermore, MCS
is also not suitable for quantifying epistemic uncertainty. However, this FTA approach can integrate two types of
basic events’ failure probabilities, i.e. fuzzy probabilities and conventional probabilities. In the case of basic events
do have their corresponding lifetime to failure; conventional probability can be statistically generated. On the other

020010-4
hand, in the case of basic events do not have their corresponding lifetime to failure; fuzzy probabilities can be
generated from expert judgments.

Fuzzy Based Fault Tree Analysis

In fuzzy based FTA, fuzzy probabilities are used to not only characterize basic event reliabilities but also
intermediate event reliabilities and the top event reliability as opposed to fuzzy hybrid FTA. While the fuzzy
probabilities of basic events are collected either from experts’ qualitative judgments or limited available data, the
fuzzy probability of the intermediate events and the top event are quantified using fuzzy combination rules instead
of Boolean algebras as in conventional FTA and fuzzy hybrid FTA.
In this approach, two fuzzy combinations rules substitute the OR and AND Boolean gates. The OR Boolean gate
is quantified using a fuzzy complementation rule. Meanwhile, the AND Boolean gate is calculated using a fuzzy
multiplication rule. Therefore, the probability of the top event A0 in Figures 1(a) and 1(b) can be quantified using (3-
4), respectively [9, 23].
!
𝜇 !! 𝑥 = 1 − !!! 1 − 𝜇 !! 𝑥 (3)
!
𝜇 !! 𝑥 = !!! 𝜇!! 𝑥 (4)
where 𝜇!! 𝑥 is the fuzzy probability of basic event 𝐴! and 𝜇!! 𝑥 is the fuzzy probability of the top event 𝐴! in
Figure 1.
This approach offers another advantage of propagating epistemic uncertainty raised in basic event fuzzy
probabilities into the top event. However, this approach cannot be partially applicable when some of basic events
come with their corresponding lifetime to failure.

From the results and discussion elaborated above, it can be summarized that it is necessary to select the most
suitable fault tree analysis for NPP PSA. It is critical for safety analysts to, firstly, know the type of the available
reliability data of the system being investigated. If basic events sufficiently have their own probability distribution of
their lifetime to failures, conventional FTA should be applied. The results will actually represent the real safety
performance of the system being investigated. When available historical failure data is not sufficient to statistically
estimate basic events’ failure probabilities or expert judgments are the only means to generate basic events’ failure
probabilities, fuzzy FTA should be used. Secondly, if epistemic uncertainty is essential and need to be considered in
the study being performed, fuzzy based FTA should be applied. Otherwise, safety analysts should apply
conventional FTA or fuzzy hybrid FTA. The decision, which one of these two approaches to be used, depends on
how basic event probabilities of the system being studied are generated or collected.

CONCLUSIONS

In conventional FTA, reliabilities of basic events, intermediate events, and the top event are characterized by
conventional probabilities. Meanwhile, in fuzzy FTA, reliabilities of basic events are characterized using fuzzy
probabilities. Different from fuzzy hybrid FTA in which reliabilities of intermediate events and the top event are still
in the form of conventional probabilities, reliabilities of intermediate events and the top events in fuzzy based FTA
are also characterized by fuzzy probabilities. Prior to the selection of the most appropriate FTA approaches in NPP
PSA, safety analysts need to, firstly, confirm the type of reliability data at hands. If basic events have their own
probability distribution of their lifetime to failures, conventional FTA should be applied. When available historical
failure data is not sufficient or expert judgments are the only means to generate basic events’ failure probabilities,
fuzzy FTA should be used. Secondly, if epistemic uncertainty is essential, fuzzy based FTA should be applied.
Otherwise, safety analysts should apply conventional FTA or fuzzy hybrid FTA depending on how basic event
probabilities are generated.

ACKNOWLEDGMENTS
The work presented in this paper was funded by the government of Indonesia through DIPA of the Center for
Nuclear Reactor Technology and Safety, National Nuclear Energy Agency of Indonesia (BATAN).

020010-5
 REFERENCES
1. Zubair, M., A. Ababneh, and A. Ishag, Station black out concurrent with PORV failure using a Generic
Pressurized Water Reactor simulator. Ann. Nucl. Energy, 2017. 110: p. 1081-1090.
2. Khakzad, N., F. Khan, and P. Amyotte, Risk-based design of process systems using discrete-time Bayesian
networks. Reliab. Eng. Syst. Saf., 2013. 109: p. 5-17.
3. Kamyab, S. and M. Nematollahi, Evaluating the core damage frequency of a TRIGA research reactor using
risk assessment tool software. Nucl. Eng. Des., 2011. 241(8): p. 2942-2947.
4. Song, H., H.Y. Zhang, and C.W. Chan, Fuzzy Fault Tree Analysis Based on T–S Model With Application to
INS/GPS Navigation System. Soft Computing - A Fusion of Foundations, Methodologies and Applications,
2009. 13(1): p. 31-40.
5. Contini, S., L. Fabbri, and V. Matuzas, A novel method to apply Importance and Sensitivity Analysis to
multiple Fault-trees. J. Loss Prev. Process Ind., 2010. 23: p. 574-584.
6. Yang, G., Potential Failure Mode Avoidance, in Life Cycle Reliability Engineering. 2007, John Wiley & Sons:
Hoboken, New Jersey. p. 194-235.
7. Verma, A.K., A. Srividya, and D.R. Karanki, eds. Reliability and Safety Engineering. 2010, Springer-Verlag:
London.
8. Kabir, S., An overview of fault tree analysis and its application in model based dependability analysis. Expert
Syst. Appl., 2017. 77: p. 114-135.
9. Purba, J.H., Fuzzy probability on reliability study of nuclear power plant probabilistic safety assessment: A
review. Prog. Nucl. Energy, 2014. 76: p. 73-80.
10. Verma, A.K., A. Srividya, and D.R. Karanki, System Reliability Modeling, in Reliability and Safety
Engineering. 2010, Springer-Verlag: London. p. 71-168.
11. Liu, T., J. Tong, and J. Zhao. Probabilistic risk assessment framework development for nuclear power plant. in
IEEE International Conference on Industrial Engineering and Engineering Management. 2008. Singapore.
12. Delaney, M.J., G.E. Apostolakis, and M.J. Driscoll, Risk-informed design guidance for future reactor systems.
Nucl. Eng. Des., 2005. 235(14): p. 1537-1556.
13. Di Maio, F., F. Antonello, and E. Zio, Condition-based probabilistic safety assessment of a spontaneous steam
generator tube rupture accident scenario. Nucl. Eng. Des., 2018. 326: p. 41-54.
14. Abdelgawad, M., A.R. Fayek, and F. Martinez. Quantitative assessment of horizontal directional drilling
project risk using fuzzy fault tree analysis in Construction Research Congress 2010: Innovation for Reshaping
Construction Practice. 2010. American Society of Civil Engineers.
15. Bickel, J.H., Risk implications of digital reactor protection system operating experience. Reliab. Eng. Syst.
Saf., 2008. 93(1): p. 107-124.
16. Kamyab, S., M. Nematollahi, and G. Shafiee, Sensitivity analysis on the effect of software-induced common
cause failure probability in the computer-based reactor trip system unavailability. Ann. Nucl. Energy, 2013.
57: p. 294-303.
17. Borges, D.S., et al., Nondeterministic method to analysis of the aging effects in PWR power plants components.
Ann. Nucl. Energy, 2015. 81: p. 249-256.
18. Hanss, M. and S. Turrin, A Fuzzy-Based Approach to Comprehensive Modeling and Analysis of Systems with
Epistemic Uncertainties. Struct. Saf., 2010. 32: p. 433-441.
19. Ferdous, R., et al., Fault and event tree analyses for process systems risk analysis: Uncertainty handling
formulations. Risk Anal, 2011. 31(1): p. 86-107.
20. Ferdous, R., et al., Fault and event tree analyses for process systems risk analysis: Uncertainty handling
formulations. Risk Anal., 2011. 31(1): p. 86-107.
21. Guimaraes, A.C.F. and C.M.F. Lapa, Parametric fuzzy study for effects analysis of age on PWR containment
cooling system. Appl. Soft Comput., 2008. 8(1): p. 1562–1571.
22. Guimaraes, A.C.F., et al., Fuzzy uncertainty modeling applied to AP1000 nuclear power plant LOCA. Ann.
Nucl. Energy, 2011. 38(8): p. 1775-1786.
23. Purba, J.H., et al., Fuzzy probability based fault tree analysis to propagate and quantify epistemic uncertainty.
Ann. Nucl. Energy, 2015. 85: p. 1189-1199.
24. Purba, J.H., et al., α-Cut method based importance measure for criticality analysis in fuzzy probability – Based
fault tree analysis. Ann. Nucl. Energy, 2017. 110: p. 234-243.

020010-6
25. Kumaraningrum, A.R., H. Hermansyah, and J.H. Purba. Experts’ selection in the application of fuzzy fault tree
analysis to evaluate an RSG – GAS primary cooling system. in The 8th Annual Basic Science International
Conference (BaSIC 2018). 2018. East Java, Indonesia: American Institute of Physics.
26. Yuhua, D. and Y. Datao, Estimation of Failure Probability of Oil and Gas Transmission Pipelines by Fuzzy
Fault Tree Analysis. J. Loss Prev. Process Ind., 2005. 18: p. 83-88.
27. Gupta, S. and J. Bhattacharya, Reliability Analysis of a Conveyor System Using Hybrid Data. Qual. Reliab.
Eng. Int., 2007. 23(7): p. 867-882.
28. Pan, N.F. and H. Wang. Assessing failure of bridge construction using fuzzy fault tree analysis. in IEEE
International Conference on Fuzzy Systems and Knowledge Discovery. 2007. Haikou.
29. Huang, D., T. Chen, and M.J.J. Wang, A Fuzzy Set Approach for Event Tree Analysis. Fuzzy Sets and Systems,
2001. 118(1): p. 153-165.
30. Onisawa, T., An approach to human reliability in man-machine systems using error possibility. Fuzzy Sets
Syst., 1988. 27(2): p. 87-103.

020010-7