See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/312043117

Reliability of Critical Infrastructure Networks: Challenges

Article · January 2017

CITATION READS
1 411

2 authors:

Konstantin Zuev Michael Beer

California Institute of Technology Leibniz Universität Hannover
45 PUBLICATIONS   528 CITATIONS    301 PUBLICATIONS   3,082 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Modeling and dealing with vague and sparse information using machine learning View project

Finding Hidden Structures in Financial Networks View project

All content following this page was uploaded by Konstantin Zuev on 06 October 2017.

The user has requested enhancement of the downloaded file.

 Reliability of Critical Infrastructure Networks: Challenges

Konstantin M. Zuev, Ph.D.1 and Michael Beer, Prof. Dr.-Ing.2,3,4

1
Department of Computing and Mathematical Sciences, California Institute of Technology,
1200 E. California Blvd., Pasadena, CA, USA; e-mail: [email protected]
2
Institute for Risk and Reliability, Leibniz Universität Hannover, Callinstraße 34, 30167
Hannover, Germany; e-mail: [email protected]
3
Institute for Risk and Uncertainty, University of Liverpool, UK
4
International Joint Research Center for Engineering Reliability and Stochastic Mechanics,
Tongji University, China

ABSTRACT

Critical infrastructures form a technological skeleton of our world by providing us with water,
food, electricity, gas, transportation, communication, banking, and finance. Moreover, as urban
population increases, the role of infrastructures become more vital. In this paper, we adopt a
network perspective and discuss the ever growing need for fundamental interdisciplinary study
of critical infrastructure networks, efficient methods for estimating their reliability, and cost-
effective strategies for enhancing their resiliency. We also highlight some of the main challenges
arising on this way, including cascading failures, feedback loops, and cross-sector
interdependencies.

1. INTRODUCTION

Critical infrastructure networks, such as transportation systems (roads, rails, and airlines),
electric power grids, natural gas and petroleum networks, water distribution networks, cellular
grids, and the internet, are tightly interwoven into the fabric of the modern world. These complex
distributed systems ensure the functioning of our society by providing us with services critical
for everyday life, such as water, food, energy, banking, and finance. Moreover, they facilitate
transport-dependent economic activities, and make communication and access to information
fast and efficient. In a sense, critical infrastructures form a technological skeleton of our
civilization.
About ten years ago, we have reached a “tipping” point when, for the first time in human
history, more than half of the world’s population was living in cities. For more developed
regions, such as North America and Europe, this happened even earlier. Figure 1 shows the
dynamics of the percentage of total population living in urban areas for different regions between

1
1967 and 2015. Furthermore, it is projected that by 2050, 66% of the world’s population will be
urban (United Nations 2014). As urban conglomerations are growing, the dependence of our
society on critical infrastructure networks, spanning cities, countries, and even continents, is
constantly increasing. This makes the study of critical infrastructure networks, the enhancement
of their resiliency, and the quantification of the associated uncertainties one of the most
challenging and important problems of modern engineering.

Figure 1. Percentage of population living in urban areas by region, 1967-2015.

Data from database: World Development Indicators, <http://databank.worldbank.org/>.

Improving the resilience of (especially aging) infrastructures is absolutely necessary for

avoiding failures such as the one happened in 2010 in San Bruno (a suburb of San Francisco,
CA), when a 30-inch diameter natural gas pipeline (installed in 1956) exploded, killing 8 people,
injuring 58, and destroying 38 homes (Lagos et al. 2010; NTSB 2011). Because of the strong
shaking caused by the explosion, some of the local residents, first responders, and news media
thought initially that it was an earthquake. It took nearly an hour to find out that it was actually a
natural gas pipeline explosion. Although this failure was truly devastating, it was still a “local”
failure in the sense that it did not propagate and did not cause any other failures. But in principle
the consequences of this local failure could have been even worse.
It is well known that the network of natural gas pipelines and electric power grid are
strongly interconnected, since pipelines provide fuel for electrical generators and the grid
provides power for compressors, storage, and control systems in the gas network. This coupling
creates a wealth of potential scenarios for failure propagation, one of which is schematically
illustrated in Figure 2: the pipeline explosion (denoted by the fire sign) causes an electrical
generator failure in the grid, which triggers a sequences of cascading power outages (power grids
are prone to cascading failures), which, in turn, leads to substantial losses of natural gas
production. Electric power and other infrastructure problems in California provide many
examples of this kind of interdependencies and feedback loops (Rinaldi et al. 2001).

2
Figure 2. Failure propagation in coupled infrastructure networks: U.S. natural gas pipeline
network (on the left, source: Energy Information Administration) and U.S. power grid (on
the right, source: <https://commons.wikimedia.org/wiki/File:UnitedStatesPowerGrid.jpg>).

In this paper, we adopt a network view on critical infrastructures and discuss how to
estimate their reliability and what challenges need to be overcome to make this accurately and
efficiently. The rest of the paper is organized as follows. In Section 2, we formulate a general
network reliability problem and explain why it is computationally very challenging. In Section 3,
we briefly describe a recently introduced network reliability method, and, taking this method as
an example, we focus on the challenges in Section 4. Section 5 concludes with a brief summary.

2. NETWORK RELIABILITY PROBLEM

Suppose that an infrastructure network is modeled as a graph with 𝑛𝑛 nodes and 𝑚𝑚 links. Assume
that all nodes are absolutely reliable and don’t fail, but links may fail. The network state is then
represented by a vector 𝑠𝑠 = (𝑠𝑠1 , … , 𝑠𝑠𝑚𝑚 ), where 𝑠𝑠𝑖𝑖 = 1 if the 𝑖𝑖 𝑡𝑡ℎ link is fully operational, 𝑠𝑠𝑖𝑖 = 0 if
it is fully failed, and 0 < 𝑠𝑠𝑖𝑖 < 1 if it is partially operational. Geometrically, the network state
space, 𝑆𝑆 = {𝑠𝑠 = (𝑠𝑠1 , … , 𝑠𝑠𝑚𝑚 ), 0 ≤ 𝑠𝑠𝑖𝑖 ≤ 1} is thus an 𝑚𝑚-dimensional hypercube, where vertices
(1, … ,1) and (0, … ,0) correspond to an ideal state and complete failure state, respectively.
Next, let 𝑓𝑓(𝑠𝑠) be a probability distribution function (PDF) on the network state space 𝑆𝑆
which provides a probability model for occurrence of different states. In applications, the PDF
𝑓𝑓(𝑠𝑠) is constructed based on expert knowledge and available data. Also, let 𝑢𝑢(𝑠𝑠) be a
performance function that quantifies the degree to which the network provides the required
service (Ghosn 2016). Here, 𝑢𝑢(𝑠𝑠) is interpreted as a utility function, meaning that higher values
of 𝑢𝑢(𝑠𝑠) correspond to better performance. Finally, let’s define a failure domain 𝐹𝐹 ⊂ 𝑆𝑆 as a
collection of network states with performance below a certain critical threshold 𝑢𝑢∗ ,

𝐹𝐹 = {𝑠𝑠 ∶ 𝑢𝑢(𝑠𝑠) < 𝑢𝑢∗ }. (1)

The network reliability problem is then to estimate the probability of failure 𝑝𝑝𝐹𝐹 which is
given by the following integral:

3
 𝑝𝑝𝐹𝐹 = � 𝑓𝑓(𝑠𝑠)𝑑𝑑𝑑𝑑 = � 𝑓𝑓(𝑠𝑠)𝐼𝐼𝐹𝐹 (𝑠𝑠)𝑑𝑑𝑑𝑑, (2)
𝐹𝐹 𝑅𝑅 𝑚𝑚

where 𝐼𝐼𝐹𝐹 (𝑠𝑠) is the indicator function of the failure domain 𝐹𝐹, i.e. 𝐼𝐼𝐹𝐹 (𝑠𝑠) = 1 if 𝑠𝑠 ∈ 𝐹𝐹 and 𝐼𝐼𝐹𝐹 (𝑠𝑠) =
0 if 𝑠𝑠 ∉ 𝐹𝐹. This problem is of fundamental importance for both risk assessment of existing
technological networks and design of future infrastructures.
There exist a few factors which make estimating the failure probability (2) challenging.
First, the number of links 𝑚𝑚 is very large. For example, the U.S. Western States power grid has
more than 6,000 links (Watts and Strogatz 1998) and the California road network has more than
2.5 million links, i.e. roads connecting intersections and endpoints (Leskovec et al. 2009).
Therefore, the failure domain (1) is high-dimensional and numerical integration methods are
simply computationally infeasible. Next, the probability of failure 𝑝𝑝𝐹𝐹 is very small, since the
network failure is a relatively rare event. This means that if we want to employ the Monte Carlo
method (Robert and Casella 2004; Zio 2013), which is a standard general method for high-
dimensional integration, then, to get an accurate estimate of 𝑝𝑝𝐹𝐹 , we need to generate a large
number of samples. This stems from the fact that the coefficient of variation (COV) of the Monte
Carlo estimate is approximately 𝛿𝛿 ≈ 1/�𝑁𝑁𝑝𝑝𝐹𝐹 , where 𝑁𝑁 the total number of samples (e.g. Beck
and Zuev 2017). Finally, computing the performance function 𝑢𝑢(𝑠𝑠) is typically time-consuming
due to the complexity and size of the network. As a result, the computational effort for
evaluating the integrand 𝑓𝑓(𝑠𝑠)𝐼𝐼𝐹𝐹 (𝑠𝑠) in (2), involving checking whether 𝑠𝑠 is a failure state, is
significant, which makes the use of a large number of Monte Carlo samples computationally
prohibitive.
Estimating network reliability is thus indeed a very challenging problem, and we are still
at the very beginning of developing efficient practical methods for solving this problem. In
recent years, several promising methods have been developed for network reliability, resilience,
and risk analysis. Liu and Li (2012) introduced an analytical method, based on the minimal cut
recursive decomposition algorithm, for evaluating the network connectivity reliability. Fang and
Zio (2013) proposed a general framework for multi-scaled representation of critical
infrastructure networks, which is based on a recursive unsupervised spectral clustering method,
and showed how this representation can be used for network reliability analysis at different
scales. Gómez et al. (2014) used agent-based modeling for developing an efficient integrative
methodology for risk assessment and management of infrastructure networks.
In what follows, we focus our attention on a recently proposed stochastic simulation
method for network reliability estimation (Zuev et al. 2015), which is based on the Subset
Simulation method, originally introduced by Au and Beck (2001) for estimating small failure
probabilities of complex structural systems, such as buildings and bridges, subject to earthquake
excitations. This focus on simulation is driven the desire for a high general applicability of the
approaches for estimating network reliability. In the next two sections, we briefly describe this
method and highlight some of the main challenges that need to be overcome to make it and other
proposed network reliability methods accurate and efficient in real-world applications.

4
3. SUBSET SIMULATION FOR NETWORK RELIABILITY ESTIMATION

The key idea behind Subset Simulation is to decompose a very small failure probability into a
product of larger conditional probabilities, each of which can be efficiently estimated by a
Monte-Carlo-like method. This is achieved by considering a sequence of nested subsets, called
intermediate failure domains, that starts from the entire network state space 𝑆𝑆 and finishes at the
target failure domain 𝐹𝐹,
𝑆𝑆 = 𝐹𝐹0 ⊃ 𝐹𝐹1 ⊃ ⋯ ⊃ 𝐹𝐹𝐿𝐿−1 ⊃ 𝐹𝐹𝐿𝐿 = 𝐹𝐹. (3)

Each intermediate failure domain 𝐹𝐹𝑖𝑖 in (3) is defined similarly to the target failure domain in (1)
by relaxing the value of the critical threshold 𝑢𝑢∗ :

𝐹𝐹𝑖𝑖 = {𝑠𝑠 ∶ 𝑢𝑢(𝑠𝑠) < 𝑢𝑢𝑖𝑖∗ }, where 𝑢𝑢∗ = 𝑢𝑢𝐿𝐿∗ < 𝑢𝑢𝐿𝐿−1
∗
< ⋯ < 𝑢𝑢1∗ . (4)

Given subsets (3), the failure probability can be written as a product of conditional probabilities:

𝑝𝑝𝐹𝐹 = 𝑃𝑃(𝐹𝐹) = 𝑃𝑃(𝐹𝐹1 )𝑃𝑃(𝐹𝐹2 |𝐹𝐹1 ) … 𝑃𝑃(𝐹𝐹𝐿𝐿 |𝐹𝐹𝐿𝐿−1). (5)

By choosing the intermediate thresholds 𝑢𝑢𝑖𝑖 appropriately, one can make all conditional
probabilities 𝑃𝑃(𝐹𝐹𝑖𝑖 |𝐹𝐹𝑖𝑖−1 ) in (5) large enough, so that the corresponding conditional events 𝐹𝐹𝑖𝑖 |𝐹𝐹𝑖𝑖−1
are not rare and can be estimated efficiently by sampling.
The first factor, 𝑃𝑃(𝐹𝐹1 ), can be in fact efficiently estimated by direct Monte Carlo,
provided that 𝑢𝑢1∗ is sufficiently large. To estimate 𝑃𝑃(𝐹𝐹𝑖𝑖 |𝐹𝐹𝑖𝑖−1 ) for 𝑖𝑖 ≥ 2, one should be able to
sample from the conditional distribution 𝑓𝑓(𝑠𝑠|𝐹𝐹𝑖𝑖−1 ), which is, in general, a nontrivial task. In
Subset Simulation, this is achieved by using the Modified Metropolis algorithm (MMA) (Au and
Beck 2001; Zuev and Katafygiotis 2011), which is a Markov chain Monte Carlo (MCMC)
algorithm (Liu 2001; Robert and Casella 2004) specifically tailored for sampling from
conditional distributions in high dimensions. The main difference between MMA and the
original Metropolis algorithm (MA) (Metropolis et al. 1953) is in the way the candidate states of
Markov chains are generated. Unlike MA, where a high-dimensional proposal distribution is
used to generate a candidate state, MMA generates a candidate state component-wise using a
sequence of univariate proposal distributions. It was shown in (Au and Beck 2001) that MMA
outperforms MA in high dimensions, provided that random variables 𝑠𝑠1 , … , 𝑠𝑠𝑚𝑚 are independent.
A geometric explanation as to why this happens, based on the properties of multivariate normal
distribution in high-dimensions, is provided in (Katafygiotis and Zuev 2008). For more details on
MMA, the reader is referred to (Zuev et al. 2012), where both the algorithm and its practical
implementation are discussed in detail.
The efficiency of Subset Simulation strongly depends on the choice of intermediate
critical thresholds 𝑢𝑢1∗ , … , 𝑢𝑢𝐿𝐿−1
∗
, since they control the values of conditional probabilities
𝑃𝑃(𝐹𝐹𝑖𝑖 𝑖𝑖−1 ) in (5). It is difficult to find the optimal values of 𝑢𝑢𝑖𝑖∗ in advance, and so they are
|𝐹𝐹

5
defined adaptively and sequentially. First, a moderate number 𝑛𝑛 of Monte Carlo samples
distributed according to 𝑓𝑓(𝑠𝑠) is generated. Based on these samples, 𝑢𝑢1∗ is defined such that
𝑃𝑃(𝐹𝐹1 ) = 𝑝𝑝0, where 𝑝𝑝0 ∈ (0,1) is some fixed value (typically, 𝑝𝑝0 = 0.1). Next, MCMC samples
distributed according to the conditional PDF 𝑓𝑓(𝑠𝑠|𝐹𝐹1 ) are generated using MMA. Based on these
samples, 𝑢𝑢2∗ is defined such that 𝑃𝑃(𝐹𝐹2 |𝐹𝐹1 ) = 𝑝𝑝0. The algorithm proceeds in this way until the
target failure domain 𝐹𝐹 has been sampled sufficiently. The Subset Simulation estimate of the
network failure probability 𝑝𝑝𝐹𝐹 is then given by

𝑛𝑛𝐹𝐹
𝑝𝑝̂ 𝐹𝐹𝑆𝑆𝑆𝑆 = 𝑝𝑝0𝐿𝐿−1 , (6)
𝑛𝑛

where 𝐿𝐿 is number of levels, i.e. the number of intermediate failure domains in (3), 𝑛𝑛 is the
number of generated samples at each level, and 𝑛𝑛𝐹𝐹 is the number of samples in 𝐹𝐹 at the last level.
The efficiency of Subset Simulation for network reliability estimation is demonstrated in
(Zuev et al. 2015), where two small-world network models are compared in terms of the
maximum-flow reliability of the networks they generate. For more details on and general
introduction to Subset Simulation and its applications to rare event estimation, the reader is
referred to the original paper, where the method was developed (Au and Beck 2001), a detailed
exposition at an introductory level with implementation in MATLAB (Zuev 2015), and a
fundamental monograph (Au and Wang 2014).

4. CHALLENGES

Estimating reliability of critical infrastructure networks is one of the central problems of modern
civil engineering, operations research and applied statistics. Yet we are at the very beginning of
developing efficient solutions, since there exist several fundamental challenges that make this
problem both conceptually and computationally very difficult. In this section, we highlight some
of these challenges and propose ways to overcome them.
In the IEEE report “Reliability analysis of complex network systems: research and
practice in need” (Zio 2007), Prof. Zio observed that “…the classical methods of reliability and
risk analysis fail to provide the proper instruments of analysis.” A moment of thought shows that
this observation is directly applicable to Subset Simulation, which is a very efficient classical
reliability method, originally developed for structures (Au and Beck 2001), not networks. As
explained in Section 3, to extend Subset Simulation to networks, one needs to assume that
components 𝑠𝑠1 , … , 𝑠𝑠𝑚𝑚 of the network state 𝑠𝑠 are independent, which is, of course, an
oversimplification, since in real infrastructure networks 𝑠𝑠1 , … , 𝑠𝑠𝑚𝑚 are correlated: a link failure
often increases the probability of failure of neighboring links. Moreover, many infrastructures
are prone to cascading failures, where a local failure propagates through the network and leads to
a global failure (Dueñas-Osorio and Vemuru 2009; Hines et al. 2009). This shows that Subset
Simulation solves the network reliability problem only approximately, and to make it useful for

6
estimating the reliability of real infrastructure networks, realistic models of link correlations and
cascading failures must be incorporated into the framework.
Over the last two decades, many different models of cascading failures in complex
networks as well as methods for assessing the criticality of network components with respect to
cascading failures have been proposed by researchers from different research communities, e.g.
sociology (Watts 2002), physics (Crucitti et al. 2004), applied mathematics (Swift 2008),
computer science (Iyear et al. 2009), and engineering (Zio and Sansavini 2011; Hernandez-
Fajardo and Dueñas-Osorio 2013; Song et al. 2016), to mention but a few. A qualitatively
realistic modeling of the effects of cascading failures on an infrastructure network is challenging
because of 1) the complexity and variety of the physical and engineering mechanisms involved
and 2) the interactions between physics of cascading failures and the decisions made by humans
who operate the network. Therefore, a thorough analysis and domain-specific synthesis of
developed models of cascading failures as well as modeling of “soft” factors (such as human
errors, impact of online social networks, etc.) is required for accurate estimation of critical
infrastructure reliability.
Another challenge in assessing reliability and resilience of critical infrastructure networks
comes from the fact that they are strongly interconnected, mutually dependent, and subject to
complex feedback loops. In Section 1, we discussed that the network of natural gas pipelines and
the electric power grid are coupled (see Fig. 2). This cross-sector interdependence is not an
exception, it is rather a rule: essentially all critical infrastructures are interconnected (Rinaldi
2001). Examples of infrastructure interdependencies are demonstrated in Figure 3. As a result,
failures or poor performance of one infrastructure network can have a negative effect on other
infrastructures, and this may result into a big shock to the national or even global economy.
Moreover, these multi-sector interdependencies make infrastructure networks more fragile and
vulnerable to cascading failures triggered by both random failures and intentional attacks, than
the single networks taken in isolation (Buldyrev, 2010). These mechanisms of cascading failures
are critical for a realistic risk assessment. Consequences of initial failure scenarios evolve
through cascades between networks finally resulting in an economic loss or loss of human life.
Since these cascades cannot be captured properly with the current network analysis approaches
risk assessment is generally lacking a substantiated quantification of failure consequences.
Taking cross-sector interdependencies into account is thus absolutely necessary for both
estimation of the reliability and enhancement of the resilience of critical infrastructures as well as
for optimal design of interdependent networks (González et al. 2016). A survey of different
approaches to modeling critical infrastructure interdependency is provided in (Pederson et al.
2006). More recently, a new methodology, based on the so-called multi-layer networks
(Boccaletti et al. 2014; Kivelä et al. 2014), has been developed for modeling dependence and
interactions between different complex networks and various processes on the corresponding
“networks of networks” (D’Agostino and Scala 2014). Multi-layer networks have been
successfully used for analyzing complex networks and extracting useful insights about the
underlying complex systems in different fields of sciences and engineering.

7
 Figure 3. Multi-sector interdependencies of critical infrastructure networks.

This brings us to the last point of this paper: interdisciplinary collaboration is the key for
understanding how to assess the reliability of infrastructure and how to make it more resilient.
Complex networks are intrinsically multidisciplinary objects and they are studied by very diverse
and heterogeneous research communities. For example, networks are used to analyze the spread
of epidemics in human networks (Newman 2002), for predicting a financial crisis (Elliott et al.
2014), and for developing a theory of quantum gravity (Boguñá et al. 2014). Also networks may
help us to understand how a brain works (Krioukov 2014) and how to treat cancer (Barabási et
al. 2011). Moreover, insights obtained in network studies are often truly interdisciplinary and
results obtained, for instance, for biological networks, could be extended and applied to
infrastructure networks, which could be very important for making critical infrastructures more
reliable and resilient. Despite the fact that the value of interdisciplinary research has been long
recognized and promoted (NSF 2016), interdisciplinary collaboration in the area of critical
infrastructure networks is still in its infancy. This is mainly because the area itself is relatively
new: for example, prior to the 1990s, little attention was given to infrastructure
interdependencies (Visarraga 2011). So it is important to organize more interdisciplinary
workshops and conferences, where researchers from different fields can discuss the progress they
have made, exchange ideas, describe the open problems they are facing, and, most importantly,
form interdisciplinary groups for attacking these problems. Another issue concerns the
conservatism of our publication media, which are typically focused on single-discipline
developments. Cross-disciplinary developments are often considered as not fitting into the scope
of a journal because of elements from a second discipline. We need to break off this silo

8
structure in order to facilitate interdisciplinary discussions on this new generation of
developments to make substantial progress. Similarly, research foundations and stakeholders
need to translate their initial opening for interdisciplinary research into a comprehensive
implementation of a interdisciplinary assessment and funding structure with interdisciplinary
units, panels, etc.
It is important to realize however that this interdisciplinary “dialog” will not be smooth
and easy due to substantial differences in educational backgrounds and “vocabularies” used by
researcher from different fields. As illustration, Adam at al. (2015) developed a theoretical
approach for studying cascading effects, which can be potentially very useful for understanding
cascading failures in infrastructure networks. Understanding this paper, however, requires
familiarity with notions of topological closer, isomorphism, with elements of universal algebra,
and other notions which are far beyond the standard mathematical education of civil engineers.
To use the results of this paper, engineers have to either educate themselves, or convince the
authors to explain their results in the language engineers can understand. Neither of these two
tasks is easy, but we believe both are necessary.

5. CONCLUSION

This paper adopts a network perspective to highlight some of the main challenges researchers
face when estimating reliability of critical infrastructures. In particular, we discuss the extension
of Subset Simulation, an efficient classical reliability method for estimating small failure
probabilities of structures, to networks. We show that for accurate network reliability estimation,
several important challenges must be overcome, including realistic domain-specific models for
link correlations and cascading failures, large-scale modeling of cross-sector interdependencies
and feedback loops, modeling of “soft” factors, such as human errors and interaction of
infrastructure with online social networks. Domain knowledge and expertise from multiple
branches of engineering, system analysis, applied mathematics, statistical physics, computer
science, and computational statistics are required to successfully address the outlined challenges.
To succeed in these tasks, interdisciplinary collaboration is thus a must.

ACKNOWLEDGMENTS

The first author would like to thank the organizers of the ASCE Workshop “Resiliency of Urban
Tunnels and Pipelines” for inviting him to give a keynote talk. Both authors thank the participants
of the workshop for fruitful discussions and valuable suggestions.

REFERENCES

Adam, E.M., Dahleh, M.A., and Ozdaglar, A. (2015) “Towards an algebra for cascade effects.”
Available online at < https://arxiv.org/abs/1506.06394>.

9
Au, S.K. and Beck, J.L. (2001). “Estimation of small failure probabilities in high dimensions by
subset simulation.” Probabilistic Engineering Mechanics, 16(4), 263-277.
Au, S.K. and Wang, Y. (2014). Engineering Risk Assessment and Design with Subset Simulation.
John Wiley & Sons, Singapore.
Barabási, A.-L., Gulbahce, N., and Loscalzo, J. (2011). “Network medicine: a network-based
approach to human disease.” Nature Reviews Genetics, 12, 56-68.
Beck, J.L. and Zuev, K.M. (2017). “Rare event simulation.” In Handbook of Uncertainty
Quantification, R. Ghanem, D. David, and H. Owhadi, eds. Springer.
Boccaletti, S., Bianconi, G., Criado, R., del Genio, C.I., Gómez-Gardeñes, J., Romance, M.,
Sendiña-Nadal, I., Wang, Z., and Zanin, M. (2014). “The structure and dynamics of
multilayer networks.” Physics Reports, 544(1), 1-122.
Boguñá, M., Kitsak, M., and Krioukov, D. (2014). “Cosmological networks.” New Journal of
Physics, 16, 093031.
Buldyrev, S.V., Parshani, R., Paul, G., Stanley, H.E., and Havlin, S. (2010), “Catastrophic
cascade of failures in interdependent networks.” Nature, 464, 1025-1028.
Crucitti, P., Latora, V., and Marchiori, M. (2004). “Model for cascading failures in complex
networks.” Physical Review E, 69, 045104(R).
D’Agostino, G. and Scala, A. (2014). Networks of Networks: the Last Frontier of Complexity.
Berlin, Springer.
Dueñas-Osorio, L. and Vemuru, S.M. (2009). “Cascading failures in complex infrastructure
systems.” Structural Safety, 31(2), 157-167.
Elliott, M., Golub, B., and Jackson, M.O. (2014). “Financial networks and contagion.” American
Economic Review, 104(10): 3115-3153.
Fang, Y.-P., and Zio, E. (2013). “Hierarchical modeling by recursive unsupervised spectral
clustering and network extended importance measures to analyze the reliability
characteristics of complex network systems.” American Journal of Operations Research,
3, 101-112.
Ghosn, M., Dueñas-Osorio, L., Frangopol, D., McAllister, T., Bocchini, P. Manuel, L.,
Ellingwood, B., et al., (2016). “Performance indicators for structural systems and
infrastructure networks.” Journal of Structural Engineering, 142(9): F4016003.
Gómez, C., Sánchez-Silva, M., and Dueñas-Osorio, L. (2014). “An applied complex systems
framework for risk-based decision-making in infrastructure engineering.” Structural
Safety, 50, 66-77.
González, A. D., Dueñas-Osorio, L., Sánchez-Silva, M., and Medaglia, A.L. (2016). “The
interdependent network design problem for optimal infrastructure system restoration.”
Computer-Aided Civil and Infrastructure Engineering, 31(5): 334–50.
Hernandez-Fajardo, I. and Dueñas-Osorio, L. (2013). “Probabilistic study of cascading failures
in complex interdependent lifeline systems.” Reliability Engineering & System Safety,
111, 260-272.

10
Hines, P, Balasubramaniam, K., and Cotilla-Sanchez E. (2009). “Cascading failures in power
grids.” IEEE Potentials, 28(5), 24-30.
Iyear, S.M., Nakayama, M.K., and Gerbessiotis, A.V. (2009). “A Markovian dependability
model with cascading failures.” IEEE Transactions on Computers, 58(9), 1238-1249.
Katafygiotis, L.S. and Zuev, K.M. (2008). “Geometric insight into the challenges of solving
high-dimensional reliability problems.” Probabilistic Engineering Mechanics, 23, 208-
218.
Kivelä, M., Arenas, A ., Barthelemy, M . Gleeson, J.P., Moreno, Y., and Porter, M.A. (2014).
“Multilayer networks.” Journal of Complex Networks, 2 (3), 203-271.
Krioukov, D. (2014), “Brain theory.” Frontiers in Computational Neuroscience, 8, 114.
Lagos, M., Fagan, K., Cabanatuan, M., and Berton, J. (2010). “San Bruno fire levels
neighborhood – gas explosion.” San Francisco Chronicle, September 10, 2010.
Leskovec, J., Lang K., Dasgupta, A., Mahoney, M. (2009). “Community structure in large
networks: natural cluster sizes and the absence of large well-defined clusters.” Internet
Mathematics, 6(1), 29-123.
Liu, J.S. (2001). Monte Carlo Strategies in Scientific Computing. New York: Springer Verlag.
Liu, W. and Li, J. (2012). “An improved cut-based recursive decomposition algorithm for
reliability analysis of networks.” Earthq Eng & Eng Vib., 11, 1-10.
Metropolis, N., Rosenbluth, A.W., Rosenbluth, M.N., Teller, A.H., and Teller, E. (1953).
“Equation of state calculations by fast computing machines.” Journal of Chemical
Physics, 21(6), 1087-1092.
Newman, M.E.J. (2002) “Spread of epidemic disease on networks.” Physical Review E, 66,
016128.
Pederson, P., Dudenhoeffer, D., Hartley, S., and Permann, M., (2006). “Critical infrastructure
interdependency modeling: a survey of U.S. and international research.” Idaho National
Laboratory report INL/EXT-06-11464.
Rinaldi, S.M., Peerenboom, J.P., and Kelly, T.K. (2001). “Identifying, understanding, and
analyzing critical infrastructure interdependencies.” IEEE Control Systems Magazine, 11-
25.
Robert, C.P. and Casella, G. (2004). Monte Carlo Statistical Methods. New York, Springer
Verlag.
Song, J., Cotilla-Sanchez, E., Ghanavati, G., and Hines P. (2016). “Dynamic modeling of
cascading failure in power systems.” IEEE Transactions on Power Systems, 31(2), 1360-
1368.
Swift, A.W. (2008). “Stochastic models of cascading failures.” Journal of Applied Probability,
45, 907-921.
NSF. (2016). <https://www.nsf.gov/od/oia/additional_resources/interdisciplinary_research/>
(Dec. 12, 2016).
NTSB (National Transportation Safety Board). (2011). Pacific Gas and Electric Company
Natural Gas Transmission Pipeline Rupture and Fire, San Bruno, California, September

11
 9, 2010, Accident Report NTSB/PAR-11/01, PB2011-916501. Available from
<http://www.ntsb.gov/investigations/AccidentReports/Reports/PAR1101.pdf>
United Nations. (2014). World Urbanization Prospects: The 2014 Revision, New York: United
Nations DoEaSA, Population Division. Available from <https://esa.un.org/unpd/wup/>.
Visarraga, D.B. (2011). “Understanding complex systems: infrastructure impacts.” In Unraveling
Complex Systems, Mathematics Awareness Month. Available from
<http://www.mathaware.org/mam/2011/essays/complexsystemsVisarraga.pdf>
Watts, D.J. and Strogatz, S.H. (1998). “Collective dynamics of small-world networks.” Nature,
393, 440-442.
Watts, D.J. (2002). “A simple model of global cascades on random networks.” Proc. Natl. Acad.
Sci. 99, 5766–5771.
Zio, E. (2007). “Reliability analysis of complex network systems: research and practice in need,
IEEE Reliability Society Annual Technology Report.
Zio, E. and Sansavini, G. (2011) “Component criticality in failure cascade processes of network
systems.” Risk Analysis, 31(8), 1196-1210.
Zio, E. (2013). The Monte Carlo Simulation Method for System Reliability and Risk Analysis.
Springer Series in Reliability Engineering, Springer-Verlag London.
Zuev, K.M. and Katafygiotis, L.S. (2011). “Modified Metropolis-Hastings algorithm with
delayed rejection.” Probabilistic Engineering Mechanics, 26, 405-412.
Zuev, K.M., Beck, J.L., and Katafygiotis, L.S. (2011). “On the optimal scaling of the Modified
Metropolis-Hastings algorithm.” Proc., 11th International Conference on Applications of
Statistics and Probability ICASP-2011, Zurich, Switzerland.
Zuev, K.M., Beck, J.L., Au, S.K., and Katafygiotis, L.S. (2012). “Bayesian post-processor and
other enhancements of Subset Simulation for estimating failure probabilities in high
dimensions.” Computers & Structures, 92-93, 283-296.
Zuev, K.M. (2015). “Subset simulation method for rare event estimation: an introduction.” In
Encyclopedia of Earthquake Engineering, M. Beer, I.A. Kougioumtzoglou, E. Patelli,
S.K. Au, eds. Springer Verlag Berlin Heidelberg, 3671-3691.
Zuev, K.M., Wu, S., and Beck, J.L. (2015). “General network reliability problem and its efficient
solution by Subset Simulation.” Probabilistic Engineering Mechanics, 40, 25-35.

12

View publication stats