Reliability Engineering and System Safety 237 (2023) 109409

Contents lists available at ScienceDirect

Reliability Engineering and System Safety

journal homepage: www.elsevier.com/locate/ress

Resilience evaluation and optimal design for weapon system of systems

with dynamic reconfiguration
Zhiwei Chen a, Dongpao Hong b, *, Weiwei Cui b, c, *, Weikang Xue b, Yao Wang b, Jilong Zhong d
a
Unmanned System Research Institute, Northwestern Polytechnical University, Xi’an 710109, China
b
China Academy of Launch Vehicle Technology, Beijing 100076, China
c
School of Intelligent Science, National University of Defense Technology, Changsha 410073, China
d
National Institute of Defense Technology Innovation, PLA Academy of Military Science, Beijing 100091, China

A R T I C L E I N F O A B S T R A C T

Keywords: The weapon system of systems (WSoS) is a collection of various weapon systems used to achieve a common goal
Dynamic reconfiguration and mission. Stakeholders are committed to lowering design costs while achieving greater resilience. In this
Evaluation paper, we establish a resilience evaluation and optimal design method for WSoS under different attack and
Optimal design
reconfiguration strategies that can balance its resilience and cost. Firstly, a WSoS-oriented three-phase resilience
Resilience
System of systems
model and evaluation algorithm are developed, taking into account resistance, adaptability, and recovery factors.
Then, a resilience-based WSoS cost model is presented to assess the cost of new equipment, improved equipment,
and resilience-associated costs. Additionally, a resilience-based optimization model is proposed that considers
reduced costs and increased resilience. Non-dominated sorting in genetic algorithms: II (NSGA-II) is selected to
achieve the Pareto optimal solution. Finally, a typical 5-node formation air defense WSoS is utilized to illustrate
the resilience evaluation and optimization design, which can serve as a reference for WSoS design.

1. Introduction reconfiguration [6–8], (3) resilience evaluation, and (4) resilience

optimal design [9–11].
The widespread application of information technology has ensured In constituent system reliability improvement, Yılmaz et al. [12]
that the linkages and interactions between systems have become more proposed an approach for the modeling of uncertainty propagation in
frequent and direct. The weapon system of systems (WSoS) is a collec­ fault trees to improve system reliability. Zhang et al. [13] proposed a
tion of various weapon systems to achieve a common goal and mission. strength-redundancy allocation approach by considering the optimal
It is widely used in the fields of defense and military, focusing on solving component strength and redundancy strategies to improve multi-state
multi-system integration, multi-dimensionality, multivariable elements, system reliability. Focusing on the aviation multiplex networks, Fan
complex emergence, and evolution problems. Resilience, defined as the et al. [14] proposed a dynamic strategy of system design to improve the
capability of maintaining satisfactory performance levels by adopting reliability and resilience of system from a holistic perspective. Zhao et al.
dynamic reconfiguration in the face of disruptions [1–3], provides a [15] proposed a time-variant reliability-based design optimization
novel approach for solving disruptions. Although resilience has been the method by improving extreme value moment. In literature [16], an
focus of research in many fields, dealing with resilience is still chal­ autonomous maintenance method is explored to increase the mission
lenging and exciting in the context of WSoS. WSoS often works in reliability of the control system. However, these traditional system
harsher environments and suffers from frequent shocks. WSoS resilience reliability and maintainability methods suffer great challenges. Because
enhancement can improve its resistance, adaptability, and recover­ WSoS usually encounters unpredictable and unavoidable external ma­
ability. Researchers have mainly considered the following four aspects licious attacks, it is futility to improve system reliability and maintain­
in strengthening the resilience of WSoS, which mainly include (1) ability before the disruption (Table 1).
improving constituent system/equipment reliability [4,5], (2) dynamic Dynamic reconfiguration is a typical resilient recovery strategy and a

Abbreviations: WSoS, Weapon system of systems; SoS, System of systems; OODA, Observation, Orientation, Decision, Action; CMDA, Current max degree attack;
NSGA-II, Non-dominated sorting in genetic algorithms: II; MREA, WSoS-oriented resilience evaluation algorithm; FAD-WSoS, Formation air defense WSoS.
* Corresponding authors at: China Academy of Launch Vehicle Technology, Beijing 100076, China.
E-mail addresses: [email protected] (D. Hong), [email protected] (W. Cui).

https://doi.org/10.1016/j.ress.2023.109409
Received 30 July 2022; Received in revised form 1 April 2023; Accepted 25 May 2023
Available online 27 May 2023
0951-8320/© 2023 Elsevier Ltd. All rights reserved.
Z. Chen et al. Reliability Engineering and System Safety 237 (2023) 109409

Table 1 the killing chains. Feng et al. [24] proposed resilience design and
The definitions of the primary notations. evaluation methods based on a meta-structure under internal deterio­
Notation Definition Notation Definition ration and external shocks. Chen et al. [7] put forward a resilience model
based on performance threshold, which is considered the mission
R SoS WSoS resilience g(t) WSoS performance at time t
k(t) mission baseline di resistance factor baseline. Yang et al. [4] developed a hybrid system resilience metric by
si adaptability factor hi recovery factor employing reliability engineering parameters. Ed-daoui [25] proposed
Φ the structure function for φi the structure function for an SoS safety and resilience evaluation method, which considered risk
WSoS WSoS element i monitoring design and structural analysis in economic infrastructure. Xu
the resilience of WSoS the weight of WSoS element i
et al. [26] discussed the dynamic resilience of complex networks when a
Ri ϖi
element i
si detection system of WSoS ci C2 system WSoS element i fraction of nodes is attacked, which is helpful in understanding the dy­
element i namic mechanism under critical conditions in real-world networks.
wi weapon system of WSoS ti communication system of However, the existing resilience evaluation metrics and methods do not
element i WSoS element i
sufficiently consider dynamic reconfiguration under different shock
λs ,μs the failure and repair rate λc ,μc the failure and repair rate of
of detection system C2 system strategies. Meanwhile, these resilience evaluation methods cannot be
λw ,μw the failure and repair rate λt ,μt the failure and repair rate of directly applied to WSoS owing to the difficulty in obtaining and col­
of weapon system communication system lecting input data.
λe1 The failure rate of λe2 The failure rate of wired LAN The essence of SoS resilience optimization is a decision-making
wireless data link within platforms
between platforms
problem of SoS resilience design, which is oriented to specific objec­
cy new equipment cost cr resilience-associated cost tives and missions. Thanks to the mechanism of sharing WSoS resources
cg improved equipment cost m the number of WSoS elements and capabilities, the WSoS uses fewer resources to achieve more
n the number of equipment mij the number of similar or powerful functions. The WSoS optimization design methods mainly
types in WSoS element i backup equipment of type j in
include three-tier comprehensive optimization, SoS countermeasure
WSoS element i
cuij the design and production uij the performance of type j simulation, executable model, multi-objective, and mathematical pro­
cost of type j equipment equipment in WSoS element i. gramming methods [8,10,27]. Feng et al. [11] investigated a resilience
gj optimization model for an unmanned aerial vehicle SoS to achieve rapid
the feasibility of αij the ratio of improved and accurate reconfiguration under random attacks. Wang et al. [8]
increasing performance equipment cost to original
presented a mixed integer program model to provide resilience-oriented
of type j equipment in equipment design and
WSoS element i production cost optimal reconfiguration solutions for coupled traffic-power systems.
ni the number of similar or fid the feasibility of increasing Tran et al. [28] defined the SoS networks based on a scale-free topology
backup systems in WSoS resistance of element i and found that a resilient network design is more cost-effective than a
element i,
robust one. Ren et al. [29] established a multi-objective optimization
fis the feasibility of fih the feasibility of increasing
increasing adaptability of recovery of element i model for the complex integrated system by considering the dimensions
element i of survival probability, reactive timeliness, and budget cost. Daven­
cdi the cost of maintaining CSoS WSoS cost dralingam [30] presented a novel robust optimization framework for
the resilience of element i architecting an SoS by considering inter-nodal performance and con­
the minimum resilience ∼
the upper limit of the total
straints associated with connectivity under performance uncertainty.
̃ sos
R Csos
required probability cost for WSoS
η the penalty factor Nsim the number of iterations Kuznetsova et al. [31] established an integrated framework of
agent-based modeling and robust optimization for the microgrid. Adlera
and Dagli [32] integrated a simple interdependent failure model of
significant expression of resilience for WSoS. Reconfiguration technol­ networked SoS into an optimization objective function to select SoS
ogies are increasingly used in industrial automation and control systems, architecture and improve resilience. The dynamic reconfiguration of
cyber-physical systems, and system of systems (SoS) [17]. Si et al. [18] system resources among the various WSoS platforms improves robust­
studied the component importance of reconfigurable systems and ness and resilience during the mission process. Most resilience optimi­
analyzed the effect of system structure changes on component impor­ zation design problems are a trade-off between system resilience and
tance. Wang et al. [8] considered several reconfiguration strategies, such cost [13,33,34]. The resilience optimization design of WSoS points to a
as the links reversing in road networks and line switching in power multi-objective optimization problem that maximizes resilience and
networks, to optimize the resilience of coupled traffic-power systems. minimizes cost. However, there are few studies on the optimal design of
Zhao et al. [19] optimized the design of reconfigurable systems WSoS resilience, and the cost of enhancing WSoS resilience is not
considering component degradation by two dynamic reconfiguration considered. Meanwhile, the optimization design needs to consider the
strategies. However, these studies mainly focus on system-level recon­ random and deliberate attacks of WSoS as well as the real-time re­
figuration through fault tolerance, fault diagnosis, preventive mainte­ quirements of the reconfiguration strategy.
nance and lack research on SoS-level reconfiguration methods. The WSoS resilience specifically refers to its ability to continuously
identical systems or equipment in WSoS can replace each other when a accomplish the mission through dynamic reconfiguration under internal
system or equipment fails to perform. Therefore, collaborative reconfi­ or external shocks [7]. Traditional system resilience methods are not
guration strategies of constituent systems should be considered. suitable for solving WSoS resilience issues effectively. To fill the
Resilience quantitative assessment methods can visually present research gaps described before, this study establishes a WSoS-oriented
evaluation results and facilitate the comparison of different system resilience evaluation and optimal design by considering dynamic
resilience. Ahmadian et al. [20] presented a quantitative model of reconfiguration, which would be compatible with WSoS. The main
network resilience, which can reflect the effect of component-level contributions of this paper are summarized as follows:
disruption on the network. Zhang et al. [21] defined traffic resilience
based on the spatiotemporal evolution of jammed clusters, which im­ • To the best of our knowledge, this is the first work that investigates
proves the adaptation and recovery of traffic SoS from various forms of the three-phase resilience evaluation problem for WSoS by consid­
interference. Uday et al. [22] established the resilience-based system ering its resistance, adaptability, and recovery factors. Under
importance measures based on the effect of constituent systems on SoS. different attack and reconfiguration strategies, the WSoS-oriented
Li et al. [23] proposed a combat network modeling and structural resilience evaluation algorithm is proposed to analyze resilience
robustness evaluation method for the combat network of WSoS based on

2
Z. Chen et al. Reliability Engineering and System Safety 237 (2023) 109409

and its factors based on OODA (Observation, Orientation, Decision, • Internal shock: the software and hardware failure of systems or
Action) loops. equipment in the WSoS, such as the natural degradation failure of the
• Most of the existing work studies on resilience evaluation consider system and equipment. The effect of changes in the relative position
system performance changes after a disruptive event, where the or framework between the constituent systems in the WSoS, such as
WSoS performance is complex and unavailable. This work treats the the influence of the relative distance between platforms on the radar
number of constituent systems as input data, which can be obtained detection range of the WSoS.
more easily. • External shock: the human-made and intentional malicious attack on
• A new integrated WSoS cost model is presented to calculate new the WSoS, such as a virus attack on the control system, an electro­
equipment, improved equipment, and resilience-associated cost. The magnetic pulse shock to the reconnaissance system, and a fire attack
resilience-associated cost is allocated to developing defensiveness, on various platforms.
adaptability, and recovery.
• The resilience design for WSoS is presented as an optimization WSoS resistance can be improved by resisting and dealing with in­
problem, aiming to generate a Pareto-optimal set of design plans by ternal and external shocks, such as: improving the constituent system
maximizing three resilience factors and minimizing total WSoS reliability or increasing the number of backups to avoid system failure;
integration cost. By solving the WSoS optimal design model based on enhancing firewall and cryptographic protocols to improve the software
NSGA-II, a group of optimal design schemes instead of a single so­ reliability of the control system; improving the air defense capability of
lution is obtained for stakeholders. WSoS to deal with enemy firepower; and so on.
As far as the WSoS is concerned, reliability and survivability can also
The rest of the paper is organized as follows. The WSoS-oriented indicate its resistance ability. Therefore, the resistance factor can be
resilience evaluation is explained in Section 2, the WSoS cost model is expressed as the probability parameter di, which represents the resis­
presented in Section 3, the optimization model is presented in Section 4, tance rate to internal and external shocks. Furthermore, di is given by:
a case is examined in Section 5, and finally, the conclusions drawn from
N − Nf
this study are summarized in Section 6. di = , (1)
N
2. WSoS-oriented resilience evaluation where N is the number of WSoS constituent systems, and Nf is the
number of system failures under the WSoS performance above the
In this section, a WSoS-oriented resilience model is extended. Sub­ mission baseline k(t) during the resistance stage.
sequently, a resilience evaluation algorithm is proposed to analyze the
resistance, adaptation, recovery, and resilience metrics when under 2.1.2. Adaptability factor
external attack. According to the resilience triangle model [35,36], the The adaptability factor mainly represents the ability of WSoS to
performance-threshold-based resilience model [7], and the resilience buffer the effect of subsequent shocks and disturbances after the failure
definition of WSoS, the resilience factors of WSoS can be divided into the of WSoS resistance. This ability can keep WSoS in a degraded operation
resistance factor, the adaptation factor, and the recovery factor, which state, and it can only complete part of the mission. The time interval
represent the characteristics of resilience. WSoS resilience factors are from tf to tm represents the absorption and adaptive stage after a
shown in Fig. 1, where g(t) is the WSoS performance at time t, g0 is the disruptive event when the performance is less than the mission baseline,
initial WSoS performance, and k(t) is the mission baseline of WSoS. as shown in Fig. 1, In this process, WSoS can reduce the adverse effects of
disruptive events and enhance performance, which is similar to
2.1. Resilience factors analysis robustness. WSoS may not meet the minimum performance re­
quirements and represents the limit of WSoS resilience recovery under
2.1.1. Resistance factor the influence of internal and external shocks.
The resistance factor mainly represents the ability of the WSoS to As far as the WSoS is concerned, robustness and tolerance can be
prevent or resist internal and external shocks, which can keep the WSoS used to indicate its adaptive ability. Therefore, the adaptability factor
performance above the mission baseline k(t), and this is the fundamental can be expressed as the probability parameter si, which represents the
requirement of the WSoS to complete the mission in a military adaptive ability to internal and external shocks; si is given by:
confrontation. As shown in Fig. 1, the time interval from t0 to tf repre­
N − Nf − Ns
sents the resistance stage, and the disruptive events in the mission si = , (2)
process mainly come from two aspects: N − Nf

where Ns is the number of system failures during the adaptive stage,

which means the WSoS performance under the mission baseline k(t).

2.1.3. Recovery factor

The recovery factor mainly represents the ability of the WSoS to
recover performance after adapting to internal and external shocks. It is
the main factor that distinguishes resilience from other system proper­
ties in engineering fields. As shown in Fig. 1, the time interval from tm to
ts represents the recovery stage after adapting to the shocks. WSoS
resilience recovery is primarily achieved through dynamic reconfigu­
ration. WSoS is reconfigured by adding or repairing the systems in this
situation. At the same time, the identical constituent systems on a
platform or each platform can replace each other when a system fails to
perform. Therefore, it is necessary to reconfigure the WSoS by elimi­
nating the failed platform nodes or adding new and repaired platform
nodes.
As far as the WSoS is concerned, maintainability and supportability
can be used to indicate its recovery ability. Therefore, the recovery
Fig. 1. WSoS resilience factors.

3
Z. Chen et al. Reliability Engineering and System Safety 237 (2023) 109409

factor can be expressed as the probability parameter hi, which represents ∑n k

where j=1 uij = 1, ukij ≥ 0 denotes the weight of constituent system j in
the recovery ability after internal and external shocks. Furthermore, hi is
WSoS element i determined using the kth subjective assignment method.
given by:
(2) Objective weight method
hi =
Nr
, (3) The objective weight method was used to determine the element
Nf + Ns weights of WSoS, and q − p subjective assignment methods are selected
to assign weights. The weights of each WSoS element can be obtained as
where Nr is the number of added or repaired constituent systems. follows.
( )
2.2. Three-phase resilience model vkij = vki1 , vki2 , vki3 , ⋯, vkin , k = p + 1, p + 2, ⋯, q, (8)
∑n k
In this study, the resilience model is viewed within the framework of where j=1 vij = 1, vkij ≥ 0 denotes the weight of constituent system j in
physical resources. The WSoS resilience quantitative model can measure WSoS element i determined using the kth subjective weight method.
the probability that WSoS could finish the mission under internal and (3) ADE combination weight method
external shocks. According to the resilience factors and three-stage Decision makers give the importance ak of various weight methods,
∑p ∑q
resilience model [37], the quantification of WSoS resilience is a where ak (k = 1, 2, ⋯, p, p +1, ⋯, q) and 1 ak = 1, p+1 ak = 1. Let
comprehensive measurement of resistance, adaptability, and recovery. γ(0 ≤ γ ≤ 1) be the priority factor for the weight method, the weights of
As the resistance, adaptability, and recovery processes are independent
WSoS element i are given by ϖ i = {ϖ 1 , ϖ 2 , ϖ 3 , ⋯, ϖ m }T , where
[29], a quantitative model of resilience is presented by measuring the ∑n
j=1 ϖ ij = 1, ϖ ij ≥ 0, the weights of each constituent system j in WSoS
probability of the resistance, adaptability, and recovery factors. Ac­
cording to killing chains and OODA loops [6], constituent systems with element i are given by:
⎧
the same or similar functions are formed as elements. A WSoS combines ⎪ ∑p ∑q
⎪
multiple elements, such as detection, command and control (C2), ⎪
⎪
⎪
⎪
ϖ i1 = γ ak uki1 + (1 − γ) ak vki1
⎪
weapon, and communication. The resilience of WSoS can be calculated ⎪
⎪
⎪
k=1 k=p+1
⎪
as follows: ⎪ ∑ p ∑ q
⎪
⎪
⎨ ϖ i2 = γ ak uki2 + (1 − γ) ak vki2
R SoS = Φ(di , si , hi ), (4) k=1 k=p+1 . (9)
⎪
⎪
⎪
⎪
⎪
⎪ ⋮
where R SoS is the resilience of WSoS, Φ is the structure function for ⎪
⎪
⎪
⎪ p
∑ q
∑
⎪
WSoS, and the calculation approach is determined according to the ⎪
⎪ ak ukin + (1 − γ) ak vkin
⎩ ϖ in = γ
⎪
structure of WSoS. As per the WSoS architecture framework and failure k=1 k=p+1

conditions, the resilience quantification is given by

∑
m 2.3. Resilience evaluation algorithm
R SoS = ϖi R i , (5)
i=1 The WSoS is constructed by combining different platforms and sys­
tems based on the mission. Similar systems on each platform can replace
R i = φi (di , si , hi ). (6) each other. WSoS pays greater attention to the coordination capacity
Due to the different contributions of each WSoS element to resil­ between platforms or systems. Dynamic reconfiguration can adjust the
ience, R i is the resilience of WSoS element i(i = 1, 2, ..., m), φi is the WSoS architecture when systems or platforms fail. In this section, we
structure function for WSoS element i, ϖ i is the weight of WSoS element propose a WSoS-oriented resilience evaluation algorithm (MREA) based
i, which can be achieved by combination weight method. on OODA loops and complex network theory. The algorithm can analyze
There are two main approaches for determining the weight of ele­ the resistance factor, adaptation factor, recovery factor, and resilience of
ments: subjective weighting and objective weighting. In this paper, we the WSoS. We consider the instance of a formation air defense WSoS
propose an ADE weight method that combines the analytic hierarchy (FAD-WSoS) as a typical WSoS in this study. The FAD-WSoS comprises
process (AHP) [38], decision-making trial and evaluation laboratory diverse platforms: early warning airplanes, frigates, and destroyers, and
(DEMATEL) [39], and entropy weight method (EWM) [40] to determine the different systems embodied in each platform Vi . The set of four
the weight of WSoS elements. While different subjective and objective systems is denoted by Vi = {si , ci , wi , ti }, where si , ci , wi , and ti represent
weighting methods have their respective advantages and limitations, detection system si , C2 system ci , weapon system wi and communication
AHP mainly ranks the relative weights of same-level indicators without system ti with the failure rate λs , λc , λw , λt and repair rate μs , μc , μw , μt
considering their interrelationships, while DEMATEL focuses on the respectively. The failure rate of wireless data link between platforms is
logical relationships between same-level indicators to determine their λe1 . The failure rate of wired LAN within platforms is λe2 , which can be
weights. EWM can improve the objectivity of weight coefficient deter­ ignored. WSoS is reconfigured by adding or repairing the systems or
mination but may not give accurate weights or may deviate significantly platforms. The structure of the FAD-WSoS is shown in Fig. 2.
from the actual importance of indicators when objective data is scarce. The failure conditions for a platform are given as follows:
To overcome these limitations, we propose the ADE combination
weighting method that leverages the strengths of AHP, DEMATEL, and • Platform Vi fails on suffering an external attack.
EWM to comprehensively evaluate different factors and elements in a • Communication node ti fails on suffering an external attack.
decision-making process. This approach has been proven effective in • Detection system si , C2 system ci , weapon system wi fail
various contexts and can be used to determine the weights of various simultaneously.
levels of indicators in the WSoS.
(1) Subjective weight method The cooperative capability serves as the critical element of the WSoS,
The subjective weight method was used to determine the element which indicates the presence of at least ks OODA loops in a WSoS.
weights of WSoS, and p subjective assignment methods are selected to Otherwise, there is WSoS failure. The cooperative conditions of the
assign weights. The weights of each WSoS element can be obtained as WSoS are given as follows:
follows.
( ) • At least ks platforms are in a perfect state.
ukij = uki1 , uki2 , uki3 , ⋯, ukin , k = 1, 2, ⋯, p, (7)

4
Z. Chen et al. Reliability Engineering and System Safety 237 (2023) 109409

wi , which can form an OODA loop. This paper uses three reconfiguration
strategies for WSoS, as shown in Fig. 3.
The identical systems (element) in each platform can replace each
other when a system node fails to perform. Reconfiguration strategy I,
which is reflected in the cooperative conditions of WSoS, degrades the
performance of WSoS but keeps it working above the mission baseline
and preserves the basic capabilities of WSoS. A system and platform lose
cooperative capacity and exit the WSoS structure when the communi­
cation node (ti ) or platform (Vi ) are destroyed. In such an event, the
WSoS is reconfigured by adding a new one or repairing the failure sys­
tem node. Reconfiguration strategy II repairs the failure systems, and
reconfiguration strategy III is adding a new platform. Reconfiguration
strategies II&III can restore WSoS performance back to a perfect state,
but it demands additional resources and costs. In addition, the priority of
reconfiguration strategy II&III is lower than reconfiguration strategy I.
The reconfiguration strategy algorithm of WSoS is shown in Table 3.
The resilience evaluation algorithm of the WSoS with dynamic
reconfiguration analyzes and shows the changes and effects of resis­
tance, adaptation, recovery, and resilience metrics during the birth and
Fig. 2. Structure of the FAD-WSoS. death under different attack strategies. The resilience evaluation algo­
rithm of WSoS is shown in Fig. 4.
• At least ks detection systems, C2 systems, weapon systems, and their
communication systems are in perfect state. 3. WSoS cost model

The WSoS appears quite vulnerable to various disturbance modes, As far as WSoS is concerned, the different WSoS elements are derived
such as trojan horses, electromagnetic strikes, and firepower attacks. from resource interaction and sharing of the systems or equipment. This
The random attack strategy removes systems or equipment from the study establishes the WSoS cost model using WSoS elements and
WSoS in a random mode. The current max degree attack (CMDA) resilience-associated costs [41,42]. Mettas [42] and Li [43] proposed the
strategy removes network nodes in a descending order of degree. In this system cost model by considering component performance and reli­
study, we mainly review the effect of the random attack and CMDA on ability. Li and Zuo [43] developed a cost model for calculating
the resilience of WSoS. The platform and system node failure algorithm performance-associated cost and availability-associated cost. Youn [41]
under random attack and CMDA has been presented as Pseudocodes in proposed a resilience optimization design framework for complex en­
Table 2. gineering systems and created a component cost function considering
WSoS can suitably respond to different types of disturbances by the initial price, prognostics and health management, preventive
employing different reconfiguration strategies. The WSoS combines maintenance, and life cycle cost (LCC).
multiple constituent systems, primarily including four elements: The investigation shows that there are two main types of problems in
communication node ti , detection systemsi , C2 system ci , weapon system the field of reliability/resilience optimization design:

• Selecting the system when the quality requirements (such as cost,

Table 2
performance, reliability, and resilience) are known.
Platform and system failure algorithm under random attack and CMDA.
• The performance and quality (reliability, maintainability, and resil­
Algorithm 1: platform and system failure algorithm under random attack and
ience) of the system are regarded as design variables, and the cost is a
CMDA:
1: Input: Initial WSoS list Vi = {si ,ci ,wi ,ti }; λs , λc , λw , λt , μs , μc , μw , μt ,λe1 ,λe2 and their
pre-defined function of the performance and quality characteristics.
distribution function.
2: Output: remove list, new network topology of WSoS V (t).
′
The second optimization design problem has been more widely
3: Determine the failure time and number of platforms and systems by Monte Carlo studied. According to the characteristics of WSoS architecture, resilience
simulation factors, and LCC, we assume that the WSoS cost mainly includes new
Determine the attack modes, including random attack and CMDA
Switch (attack mode)
equipment cost cy, improved equipment cost cg, and resilience-
case 0: random attack; randomly remove the failure systems and their edges associated cost cr. cy and cg are primarily composed of performance-
according to the sampling results; associated costs. Resilience-associated cost cr is additional cost to
default: CMDA; the systems in the current network are arranged in descending improve the WSoS resilience, including cost related to resistance,
order according to
adaptability, and recovery factors.
the current node degree, and the corresponding number of failed systems are
removed (1) According to the system cost model proposed by Mettas [42] and
Switch (failure node type) Li [43], new equipment cost cy refers to the cost of design and produc­
case 0: platform Vi ; remove Vi and its edges; tion of new equipment of WSoS, and cy is given by:
case 1: system ti ; remove Vi and its edges;
case 2: system si ; remove si and its edges; m ∑
∑ n

case 3: system wi ; remove wi and its edges; cy = mij cuij (i = 1, 2, …, m; j = 1, 2, …, n), (10)
default: system ci ; remove ci and its edges; i j

for each in platform list Vi

if failure si , wi and ci in a platform Vi then where m is the number of WSoS elements, ni is the number of equipment
remove platform Vi and its edges types in WSoS element i, mij is the number of similar or backup equip­
end if
ment of type j in WSoS element i, and cuij is the design and production cost
end for
4: Determine and remove the failure edges by Monte Carlo simulation of type j equipment, which is determined by performance-associated
5: Remove the isolated systems. cost [43,44]. The cost function is extended to make it suitable for
6: Set remove list platform and system state to failure. WSoS, cuij is given by:
7: Return remove list and a new V (t).
′

5
Z. Chen et al. Reliability Engineering and System Safety 237 (2023) 109409

Fig. 3. Three types of WSoS reconfiguration strategies.

maintain and improve the resistance, adaptability, and recovery abilities

Table 3
of the WSoS. cr is given by:
Reconfiguration strategy algorithm.
Algorithm 2: Reconfiguration strategy algorithm cr = cd + cs + ch (13)
1: Input: V (t); Remove list; μs , μc , μw , μt and their distribution function.
′

It is often assumed that there is an inverse power relationship or

2: Output: New network topology of WSoS V′′ (t) after reconfiguration
3: for each in Remove list power exponent between cost and failure rate [42]. Furthermore, we
Add new one or repair the failure systems by reconfiguration strategies and Monte derive the cost model by modifying an existing cost model presented in
Carlo Ren et al. and Frangopol et al. [29,45]. According to the characteristics
Switch (added or repair node type) of WSoS, the resistance-associated cost cd is given by:
case 0: platform Vi ; set Vi to 1 and connect to other platforms;
case 1: ti ; set ti to state 1 and connect to Vi and other ti ; ∑
m (n )
(14)
i
case 2: nodesi ; set si to state 1 and connect to Vi ; cd = cdi × exp ,
case 3: node wi ; set node wi to state 1 and connect to Vi ; i=1
4
default: node ci ; set node ci to state 1 and connect to Vi ;
[ ]
end for ( ) di − dimin
if generated or repaired platform and systems are isolated nodes then cdi = exp 1 − fid , (15)
dimax − di
remove the isolated systems
end if
4: for each in platform list [V1 , V2 , V3 , ..., Vn ] where ni is the number of similar or backup systems in WSoS element i,cdi
if si ,wi and ci nodes in a platform Vi without ti then is the cost of maintaining the resilience of element i, fid is the feasibility
remove si ,wi , ci and their edges
of increasing the resistance of element i. dimin is the minimum acceptable
end if
end for resistance value of element i, dimax is the maximum resistance that
( )
5: Return V′′ (t) element i can achieve, exp n4i is the cost of resource sharing for a con­
stituent system, mainly including the cost of additional communication
[ ] and dynamic reconfiguration.
cuij = gij exp uij − uijmin , (11) Adaptability-associated cost cs is given by:
[ ] (n )
where uij is the performance of type j equipment in WSoS element i. uijmin ∑ m
( ) si − simin
(16)
i
cs = exp 1 − fis × exp .
is the minimum performance of equipment, gij is the feasibility of i=1
si max − s i 4
increasing performance of type j equipment in WSoS element i, which is
usually a constant, indicating the level of difficulty to improve the Recovery-associated cost ch is given by:
equipment performance relative to others in WSoS. It mainly depends on ∑m [
( ) hi − himin
] (n )
(17)
i
the complexity and technical limitations. Improving some equipment in ch = exp 1 − fih
h − h
× exp
4
,
imax i
WSoS is usually challenging.
i=1

(2) Improved equipment cost cg is the cost of integrating equipment

where fis and fih are the feasibility of increasing adaptability and recovery
into the WSoS by improving the original equipment. Improved equip­
of element i, simin and himin are the minimum acceptable adaptability and
ment cost cg can be calculated in proportion to the design and production
recovery value of element i, simax and himax are the maximum adaptability
costs of the original equipment through analogy estimation and expert
and recovery value that element i can reach.
judgment estimation methods, and cg is given by:
According to the above analysis, Eqs. (10) and (11) ignore the effects
m ∑
∑ n
of WSoS resilience on cost, while Eq. (12) discounts the effects of system
cg = mij (αij cuij ) (12) performance on cost. According to the above analysis, it is a reasonable
i j
method to establish a WSoS cost model by accumulating various costs.
where αij is the ratio of improved equipment cost to original equipment This study gives the WSoS cost model by considering new equipment,
design and production cost, and 0 < αij < 1. Its value can be obtained by improved equipment, and resilience- associated costs. WSoS cost CSoS
engineering estimation, analogy estimation, and expert estimation can be calculated as follows:
methods. CSoS = cy + cg + cr . (18)
(3) Resilience-associated cost cr is the additional cost required to

6
Z. Chen et al. Reliability Engineering and System Safety 237 (2023) 109409

Fig. 4. Resilience evaluation algorithm for WSoS with dynamic reconfiguration.

4. Optimization design model ∼

probability, Csos is the upper limit of the total cost for WSoS.
WSoS involves a massive number of systems or nodes with great
Resilience-driven WSoS optimal design is the optimization of uncertainty and high computational complexity, it is necessary to
resilience-related multi-objective problems. The main influencing fac­ choose efficient optimal algorithms. In this paper, NSGA-II is used to
tors of resilience design are resistance, adaptability, and recovery. The solve multi-objective optimization problems, which significantly im­
cost of improving WSoS resilience should also be considered. Therefore, proves computational efficiency compared to the traditional genetic
the optimal design of the WSoS based on resilience is the allocation and algorithm (GA) [46]. Moreover, it incorporates unique dominated
reconfiguration of physical resources of the resilience factor parameters sorting and elitism without sharing parameters [29]. It can also prevent
(di,si,hi) and the system performance parameters uij under the constraint the loss of reasonable solutions once they are determined.
of cost. This section establishes a multi-objective optimization model of
WSoS based on resilience through the WSoS resilience quantification
5. Illustrative experiments
and cost models. The multi-objective optimization model is given by:
Min : CSoS = cy + cg + cr A typical 5-node FAD-WSoS with 5 platforms, shown in Fig. 5, is
. (19)
Max : R SoS = Φ(di , si , hi ) taken as a case study to verify the feasibility and effectiveness of the
proposed method. FAD-WSoS includes warning airplanes, frigates, and
According to the Eqs. (4) and (5), the quantitative resilience model of
destroyer platforms, and each platform includes detection, C2, weapon,
WSoS with m elements is given by:
and communication systems. A wireless data link connects each plat­
∑
m
form, and a wired LAN connects each system. The constituent systems
R SoS = ϖ i [φi (di , si , hi )], (20)
i=1
and critical equipment in the 5-node FAD-WSoS form an organic whole
through the communication network. The constituent systems and
Subject to: platform are destroyed when the WSoS suffers from external distur­
∼
R SoS ≥ R sos bances. Similar systems of each platform can replace each other. When
∼
the communication systems or platforms are ruined, the platform loses
CSoS ≤ Csos
its cooperative operation capacity and exits the WSoS.
dimin ≤ di ≤ dimax
simin ≤ si ≤ simax , where R
∼
sos is the minimum resilience required
himin ≤ hi ≤ himax 5.1. Resilience evaluation
0 ≤ Ri ≤ 1
m
∑ In this section, we analyze the resilience of the WSoS, considering the
ϖi = 1
i=0
7
Z. Chen et al. Reliability Engineering and System Safety 237 (2023) 109409

Fig. 5. A typical 5-node FAD-WSoS.

effects of resistance, adaptation, recovery, and dynamic reconfiguration CMDA, the WSoS resilience is smaller than under random attack, and the
on the WSoS. The WSoS is reconfigured by adding to or repairing the resilience fluctuations are more dramatic as critical systems or equip­
systems or platforms under external attacks. We assume that the systems ment are attacked more frequently. It is shown that the deliberate at­
must be repaired only when they are in a failure state, where the failure tacks are more destructive to the WSoS.
and repair rates of the systems are assumed to change exponentially with
time. The mission baseline of FAD-WSoS is ks . In this MREA program, the
simulation time is 500-unit time, and the number of iterations is 1000. 5.2. Optimization model
The other parameters of the different platforms and systems are listed in
Table 4. In this section, we considered a 5-node FAD-WSoS as an example for
According to Eqs.(1)–(6) and Monte-Carlo simulation, the resistance, optimizing design, and there are four platforms in FAD-WSoS. We
adaptability, and recovery factors of FAD-WSoS with dynamic reconfi­ evaluate different resilience design schemes and find the best scheme
guration under random attack and CMDA are shown in Fig. 6. The fig­ when the resilience value of each WSoS element is unknown. According
ures show that resistance and adaptability factors fluctuate and show a to the 5-node FAD-WSoS, the multi-objective optimization model based
declining trend. A sharp decrease in the resistance factor can be on resilience with design parameters di,si,hi,uij is given, and NSGA-II is
observed at the initial stage. Under random attack, the resistance factor selected to solve this optimization problem.
is projected to remain steady after t = 360. The recovery factor increases Problem:
sharply to 0.88 during the initial stage (0 < t < 68) and reaches a Min : CSoS = cy + cg + cr
.
Max : R SoS = Φ(di , si , hi )
steady-state value of 0.98 at about t = 360, which reveals that the
Subject to:
reconfiguration strategy can enhance the WSoS recovery capacity and ∼
resilience against adverse external disturbances. Under CMDA, the R SoS ≥ R sos
resistance and adaptability factors fluctuate more dramatically, and the
∼
CSoS ≤ Csos
lowest value of adaptability factor reaches 0, which means that the FAD- 0.8 ≤ di ≤ 0.9999
WSoS fails at sometimes and cannot achieve the mission baseline. The
0.7 ≤ si ≤ 0.9999 .
recovery factor increases sharply to 0.88 during the initial stage (0 < t
0.6 ≤ hi ≤ 0.9999
< 18) and reaches a steady-state value of 0.98 at about t = 300. It means
0 ≤ Ri ≤ 1
that greater recovery capacity (additional resources and costs) is ∑m
required to enhance WSoS resilience under CMDA. ϖi = 1
The resilience value indicates that the WSoS can still keep the ca­ i=0

pacity to fulfill the mission. The resilience of FAD-WSoS over time under The FAD-WSoS needs to develop new equipment and improve orig­
random attack and CMDA is shown in Fig. 7, which indicates that inal equipment to share resources and information among shipboard
resilience fluctuates and shows a trend of decreasing first and then platforms. FAD-WSoS is divided into four types of elements: detection,
increasing. It is shown that FAD-WSoS has sufficient resistance, adapt­ C2, weapon, and communication, as shown in Fig. 8. New equipment
ability, and recovery capabilities to withstand disturbances. Under cost cy, improved equipment cost cg and resilience-associated cost cr are
independent of each other for the cost of different equipment. The new
equipment includes radar cooperative control equipment u11 in the
Table 4 detection system, formation cooperative control equipment u22 in the C2
The parameters of MREA. system, collaborative positioning equipment u41, and data link u42 in the
Parameters Values Parameters Values
communication system. Moreover, each platform has two data links. The
improved equipment includes the original formation control equipment
N(0) 19 ks 4
u21, ship control equipment u23 in the control system, and weapon
Nsim 1000 μs 0.3
T 500 μc 0.08 control equipment u31 in the weapon system. The cost function of the
λs 0.0016 μw 0.2 equipment system is given by Eqs. (10)–(18). In addition, the parameters
λc 0.004 μt 0.1 need to be set, uij is the performance index of equipment, which is a
λw 0.0004 λe1 0.0006 dimensionless quantity and represents the relative value of equipment
0.01 0
performance, and uijmin =[1 1 2 1 1 1 2], ni =[4 4 4 8], i = 1, 2, 3, 4, and g4j
λt λe2

8
Z. Chen et al. Reliability Engineering and System Safety 237 (2023) 109409

Fig. 6. Resilience factors of FAD-WSoS with dynamic reconfiguration under random attack and CMDA.

solve complex system reliability or resilience and cost optimization

problems [47]. GA is the most widely used heuristic algorithm, and is
commonly used in solving optimization problems. In this paper,
NSGA-II, which has better computing efficiency, is used to solve the
Pareto optimal solution of the optimization problem. NSGA-II combines
the advantages of dominated sorting, elitism, and no shared parameters
when compared to a traditional GA [46]. Therefore, NSGA-II is selected
to solve the optimization design model based on resilience. The opti­
mization model is expressed as a penalty function.
Minimize:
( ∼ ) ( ∼ )
Csos + max Csos − Csos , 0 ∗ η + max R sos − R sos , 0 ∗ η
( ∼ ) ( ∼ ) , (21)
− R sos + max R sos − R sos , 0 ∗ η + max Csos − Csos , 0 ∗ η

where η is the penalty factor, which is a larger constant and η = 1000 in

this case study. The solving process can be divided into the following six
Fig. 7. Resilience of FAD-WSoS with dynamic reconfiguration under random steps:
attack and CMDA.

is 0.85. The other parameter values of the objective function are shown Table 5
in Table 5. According to the ADE weight method in Section 2, γ = a1 = Parameter values of the optimization model.
a2 = 0.5, m = 4, p = 2 and q = 3, the weights of the WSoS element are ∼
dimin simin himin mij fid fih
0.1796, 0.3510, 0.1684, and 0.3010. Csos

200 0.8 0.7 0.6 4 0.8 0.99

5.3. Optimal solution and result analysis ∼
R sos dimax simax himax αij fis gij
0.8 0.9999 0.9999 0.9999 0.5 0.9 1
A heuristic algorithm is considered the most effective method to

Fig. 8. Structure of FAD-WSoS elements.

9
Z. Chen et al. Reliability Engineering and System Safety 237 (2023) 109409

Step 1: Population initialization

Genes are decision variables for the component design: resistance

factor, adaptability factor, recovery factor, and relative equipment
performance. In this study, the population size is 60, each chromosome
has 19 coding positions, and the maximum genetic algebra is 200. Then,
a feasible initial population is randomly generated according to the
range of objective functions and constraints, and the first-generation
chromosome is shown in Fig. 9.

Step 2: Non-dominated sorting

The initial population is sorted based on a non-dominated algorithm

[48].

Step 3: Crowding distance

Fig. 10. Pareto solutions.
The select sorting of chromosomes in the same non-dominated
interface is achieved by calculating the Euclidean distance of chromo­ The resilience factors and performance are optimized by choosing
somes in the same non-dominated interface. NSGA-II, which offers high accuracy and speed. A case study is presented
to illustrate how to evaluate WSoS resilience and optimize WSoS design.
Step 4: Selection The main conclusions are summarized as follows: (i) WSoS resilience
and its factors are achieved, with mission baseline and dynamic recon­
After sorting the chromosomes in step 3, the crowded comparison figuration taken into account. (ii) Based on a comparison of resilience,
operator (≺) selects chromosomes, improving the algorithm’s the experiments indicate that dynamic reconfiguration can enhance the
convergence. WSoS recovery capacity and resilience against adverse disturbances. (iii)
WSoS resilience is more vulnerable to CMDA than random attack, which
Step 5: Genetic operators means greater recovery capacity (additional resources and costs) is
required to enhance WSoS resilience under CMDA. (iv) Resilience fac­
The proportion of crossover is 0.8, the mutation ratio is 0.3, and the tors and performance of constituent systems are optimized, and this can
probability of mutation is 0.5. provide guidance to design and establish a new WSoS to maintain the
resilience.
Step 6: Recombination and selection Since this paper mainly focuses on optimizing WSoS design and of­
fers some direction, the optimization problem of fuzzy constraints and
This step combines the original population and the current popula­ uncertainties needs to be further studied in the future.
tion to ensure the elitism of the population. Step 2 is used to continue
non-dominated sorting, and the population size generates a new gen­
CRediT authorship contribution statement
eration. The NSGA-II algorithm has fast convergence speed, and can
quickly and conveniently solve the problem.
Zhiwei Chen: Conceptualization, Methodology, Writing – review &
In this study, we ran the NSGA-II program in MATLAB R2014a using
editing, Investigation. Dongpao Hong: Methodology, Funding acquisi­
a computer with a 1.60 GHz Intel Core (TM) I5 CPU and 8 GB of RAM
tion, Project administration. Weiwei Cui: Supervision, Writing – review
with the Windows 10 Professional operating system. The Pareto-optimal
& editing, Project administration. Weikang Xue: Writing – review &
solution was obtained based on NSGA-II in approximately 69.09s. The
editing. Yao Wang: Methodology, Validation, Formal analysis. Jilong
Pareto solutions of the optimal design of WSoS are obtained as shown in
Zhong: Supervision, Resources, Validation.
Fig. 10. Then, the Pareto optimal solution set, which is shown in Table 6,
is received after the results with the same resilience and higher cost are
excluded. Optimal design often finds some "satisfactory" solutions, and
Declaration of Competing Interest
WSoS optimization results are neither unique nor the only effective
design scheme. Therefore, decision-makers and stakeholders can make
We would like to submit the enclosed manuscript entitled “Resilience
further choices based on their preferences.
Evaluation and Optimal Design for Weapon System of Systems with
Dynamic Reconfiguration”, co-authored by Zhiwei Chen, Dongpao
6. Conclusions
Hong, Weiwei Cui, Weikang Xue, Yao Wang, and Jilong Zhong for
possible publication in the Reliability Engineering and System Safety
In this paper, we establish a new resilience model and evaluation
journal. No conflict of interest exists in the submission of this manu­
algorithm for WSoS with dynamic configuration and present a new cost
script, and it is approved by all authors for publication. I would like to
model considering new equipment, improved equipment, and resilience-
declare on behalf of my co-authors that the work described was original
associated costs. Combining the two models, an optimization method is
research that has not been published previously, and not under
proposed by considering resilience maximization and cost minimization.
consideration for publication elsewhere, in whole or in part.

Fig. 9. First generation of chromosomes.

10
 Z. Chen et al.
Table 6
Pareto-optimal set of FAD-WSoS.
d1 d2 d3 d4 s1 s2 s3 s4 h1 h2 h3 h4 u11 u21 u22 u23 u31 u41 u42 Csos R sos

1 0.95 0.96 0.96 0.90 0.86 0.98 0.90 0.91 0.96 0.83 0.87 0.96 1.38 1.05 3.13 1.40 3.05 2.66 3.18 179.54 0.999697
2 0.95 0.96 0.96 0.90 0.86 0.97 0.90 0.91 0.96 0.83 0.87 0.96 1.38 1.10 3.13 1.40 3.05 2.63 3.16 171.43 0.999685
3 0.95 0.96 0.96 0.90 0.86 0.97 0.90 0.91 0.96 0.83 0.87 0.96 1.38 1.10 3.13 1.40 3.04 2.62 3.15 170.65 0.999683
4 0.95 0.96 0.96 0.90 0.86 0.97 0.90 0.91 0.96 0.83 0.87 0.96 1.38 1.13 3.13 1.39 3.03 2.60 3.13 167.39 0.999674
5 0.95 0.96 0.96 0.91 0.86 0.96 0.90 0.90 0.96 0.84 0.87 0.96 1.38 1.19 3.12 1.39 3.02 2.55 3.09 162.85 0.999657
6 0.93 0.96 0.96 0.92 0.93 0.83 0.90 0.91 0.99 0.94 0.87 0.96 1.38 2.19 3.13 1.24 3.05 1.87 2.45 152.18 0.999651
7 0.93 0.96 0.96 0.92 0.93 0.83 0.90 0.91 0.99 0.94 0.87 0.96 1.39 2.18 3.13 1.25 3.04 1.87 2.45 150.79 0.999639
8 0.93 0.96 0.96 0.92 0.93 0.83 0.90 0.90 0.99 0.94 0.87 0.96 1.39 2.16 3.11 1.26 2.99 1.87 2.45 147.83 0.999627
9 0.93 0.95 0.96 0.92 0.92 0.84 0.90 0.90 0.99 0.93 0.87 0.96 1.40 2.13 3.10 1.28 2.92 1.87 2.44 144.00 0.999605
10 0.93 0.95 0.96 0.92 0.92 0.84 0.90 0.89 0.99 0.93 0.87 0.96 1.41 2.11 3.08 1.29 2.84 1.88 2.44 141.38 0.999597
11 0.94 0.95 0.95 0.92 0.92 0.85 0.89 0.88 0.98 0.93 0.87 0.95 1.42 2.07 3.06 1.31 2.74 1.88 2.42 137.00 0.999565
12 0.94 0.95 0.95 0.92 0.92 0.86 0.89 0.88 0.98 0.93 0.87 0.95 1.43 2.04 3.04 1.32 2.68 1.88 2.42 135.03 0.99955
11

13 0.94 0.95 0.95 0.92 0.91 0.86 0.89 0.87 0.98 0.93 0.87 0.95 1.44 2.01 3.02 1.34 2.59 1.89 2.41 132.58 0.999528
14 0.95 0.95 0.95 0.92 0.91 0.87 0.89 0.86 0.98 0.92 0.88 0.95 1.46 1.96 2.99 1.36 2.47 1.90 2.40 129.84 0.999501
15 0.95 0.94 0.94 0.91 0.90 0.89 0.88 0.84 0.98 0.92 0.88 0.95 1.48 1.89 2.95 1.40 2.30 1.91 2.39 127.00 0.999458
16 0.95 0.94 0.94 0.91 0.90 0.88 0.88 0.85 0.97 0.91 0.88 0.95 1.48 1.91 2.93 1.40 2.36 1.89 2.39 126.19 0.999431
17 0.95 0.94 0.94 0.91 0.90 0.89 0.88 0.84 0.96 0.90 0.88 0.95 1.50 1.88 2.87 1.42 2.29 1.89 2.39 124.13 0.999378
18 0.95 0.94 0.94 0.91 0.90 0.89 0.88 0.84 0.95 0.90 0.88 0.95 1.50 1.87 2.83 1.43 2.29 1.89 2.39 123.20 0.999344
19 0.95 0.94 0.94 0.92 0.90 0.88 0.88 0.83 0.93 0.88 0.88 0.95 1.52 1.87 2.72 1.44 2.35 1.87 2.40 121.17 0.999246
20 0.94 0.94 0.94 0.92 0.89 0.88 0.88 0.83 0.91 0.87 0.88 0.95 1.52 1.87 2.66 1.44 2.38 1.86 2.41 120.40 0.999192
21 0.94 0.94 0.94 0.92 0.89 0.88 0.88 0.84 0.90 0.84 0.88 0.94 1.52 1.87 2.58 1.46 2.29 1.86 2.41 119.00 0.999068
22 0.94 0.94 0.94 0.92 0.89 0.88 0.88 0.84 0.90 0.82 0.88 0.94 1.52 1.87 2.56 1.47 2.24 1.86 2.41 118.52 0.999027
23 0.94 0.94 0.94 0.92 0.89 0.88 0.88 0.84 0.89 0.80 0.88 0.93 1.52 1.87 2.49 1.49 2.12 1.86 2.41 117.43 0.998925

Reliability Engineering and System Safety 237 (2023) 109409

24 0.94 0.94 0.94 0.92 0.89 0.88 0.88 0.84 0.87 0.78 0.88 0.93 1.52 1.87 2.42 1.50 2.08 1.86 2.42 116.43 0.99883
25 0.95 0.94 0.94 0.92 0.89 0.88 0.89 0.84 0.87 0.74 0.88 0.92 1.51 1.87 2.35 1.53 1.85 1.87 2.40 115.62 0.998717
26 0.94 0.94 0.94 0.92 0.89 0.87 0.89 0.84 0.84 0.71 0.88 0.91 1.52 1.87 2.22 1.55 1.80 1.86 2.41 113.93 0.998534
27 0.94 0.94 0.94 0.92 0.89 0.87 0.89 0.84 0.81 0.62 0.88 0.89 1.52 1.87 1.98 1.61 1.47 1.86 2.41 111.76 0.998169
Z. Chen et al. Reliability Engineering and System Safety 237 (2023) 109409

Data availability [22] Uday P, Marais KB. Resilience-based system importance measures for system-of-
systems. Proced. Comput. Sci. 2014;28:257–64. https://doi.org/10.1016/j.
procs.2014.03.033.
No data was used for the research described in the article. [23] Li J, Tan Y, Yang K, Zhang X, Ge B. Structural robustness of combat networks of
weapon system-of-systems based on the operation loop. Int. J. Syst. Sci. 2017;48:
659–74. https://doi.org/10.1080/00207721.2016.1212429.
[24] Feng Q, jiezhao X, Fan D, Cai B, Liu Y, Ren. Resilience design method based on
Acknowledgments meta-structure: a case study of offshore wind farm. Reliab. Eng. Syst. Saf. 2019;
186:232–44. https://doi.org/10.1016/j.ress.2019.02.024.
This work was supported by the National Natural Science Foundation [25] Ed-daoui I, El Hami A, Itmi M, Hmina N, Mazri T. Resilience assessment as a
foundation for systems-of-systems safety evaluation: application to an economic
of China under Grant 72101270, 72001213. infrastructure. Saf. Sci. 2019;115:446–56. https://doi.org/10.1016/j.
ssci.2019.02.030.
References [26] Xu F, Si S, Duan D, Lv C, Xie J. Dynamical resilience of networks against targeted
attack. Phys. A 2019;528:121329. https://doi.org/10.1016/j.physa.2019.121329.
[27] Luo L, Xing L, Levitin G. Optimizing dynamic survivability and security of
[1] Bruneau M, Chang SE, Eguchi RT, Lee GC, O’Rourke TD, Reinhorn AM, et al.
replicated data in cloud systems under co-residence attacks. Reliab. Eng. Syst. Saf.
A framework to quantitatively assess and enhance the seismic resilience of
2019;192:106265. https://doi.org/10.1016/j.ress.2018.09.014.
communities. Earthquake Spectra 2003;19:733–52. https://doi.org/10.1193/
[28] Tran HT, Domerçant JC, Mavris DN. A network-based cost comparison of resilient
1.1623497.
and robust system-of-systems. Proced. Comput. Sci. 2016;95:126–33. https://doi.
[2] Paes CEdB, Neto VVG, Moreira T, Nakagawa EY. Conceptualization of a system-of-
org/10.1016/j.procs.2016.09.302.
systems in the defense domain: an experience report in the Brazilian scenario. IEEE
[29] Ren F, Zhao T, Jiao J, Hu Y. Resilience optimization for complex engineered
Syst. J. 2019;13:2098–107. https://doi.org/10.1109/JSYST.2018.2876836.
systems based on the multi-dimensional resilience concept. IEEE Access 2017;5:
[3] Zhiwei C, Tingdi Z, Jian J, Yaqiu L. System of systems architecture modeling and
19352–62. https://doi.org/10.1109/ACCESS.2017.2755043.
mission reliability analysis based on DoDAF and Petri Net. In: Proceedings of the
[30] Davendralingam N, DeLaurentis D. A robust optimization framework to
Annual Reliability and Maintainability Symposium (RAMS); 2019. p. 1–6. https://
architecting system of systems. Proced. Comput. Sci. 2013;16:255–64. https://doi.
doi.org/10.1109/RAMS.2019.8769028.
org/10.1016/j.procs.2013.01.027.
[4] Yang B, Zhang L, Zhang B, Wang W, Zhang M. Resilience metric of equipment
[31] Kuznetsova E, Li Y-F, Ruiz C, Zio E. An integrated framework of agent-based
system: theory, measurement and sensitivity analysis. Reliab. Eng. Syst. Saf. 2021;
modelling and robust optimization for microgrid energy management. Appl.
215:107889. https://doi.org/10.1016/j.ress.2021.107889.
Energy 2014;129:70–88. https://doi.org/10.1016/j.apenergy.2014.04.024.
[5] Feng Q, Liu M, Dui H, Ren Y, Sun B, Yang D, et al. Importance measure-based
[32] Adler CO, Dagli CH. Study of the use of a genetic algorithm to improve networked
phased mission reliability and UAV number optimization for swarm. Reliab. Eng.
system-of-systems resilience. Proced. Comput. Sci. 2014;36:49–56. https://doi.
Syst. Saf. 2022;223:108478. https://doi.org/10.1016/j.ress.2022.108478.
org/10.1016/j.procs.2014.09.036.
[6] Sun Q, Li H, Wang Y, Zhang Y. Multi-swarm-based cooperative reconfiguration
[33] Chen Z, Zhao T, Jiao J, Ren F. Availability analysis and optimal design of multistate
model for resilient unmanned weapon system-of-systems. Reliab. Eng. Syst. Saf.
weighted K-out-of-N systems with component performance requirements. IEEE
2022:108426. https://doi.org/10.1016/j.ress.2022.108426.
Access 2018;6:51547–55. https://doi.org/10.1109/ACCESS.2018.2865933.
[7] Chen Z, Zhao T, Jiao J, Chu J. Performance-threshold-based resilience analysis of
[34] Yousefi N, Coit DW, Song S, Feng Q. Optimization of on-condition thresholds for a
system of systems by considering dynamic reconfiguration. In: Proceedings of the
system of degrading components with competing dependent failure processes.
Institution of Mechanical Engineers, Part B: Journal of Engineering Manufacture;
Reliab. Eng. Syst. Saf. 2019;192:106547. https://doi.org/10.1016/j.
2020. p. 1–11. https://doi.org/10.1177/0954405420937528.
ress.2019.106547.
[8] Wang H, Fang Y-P, Zio E. Resilience-oriented optimal post-disruption
[35] Hosseini S, Barker K, Ramirez-Marquez JE. A review of definitions and measures of
reconfiguration for coupled traffic-power systems. Reliab. Eng. Syst. Saf. 2022:
system resilience. Reliab. Eng. Syst. Saf. 2016;145:47–61. https://doi.org/
108408. https://doi.org/10.1016/j.ress.2022.108408.
10.1016/j.ress.2015.08.006.
[9] Alkhaleel BA, Liao H, Sullivan KM. Risk and resilience-based optimal post-
[36] Zobel CW, Khansa L. Characterizing multi-event disaster resilience. Comput. Oper.
disruption restoration for critical infrastructures under uncertainty. Eur. J. Oper.
Res. 2014;42:83–94. https://doi.org/10.1016/j.cor.2011.09.024.
Res. 2022;296:174–202. https://doi.org/10.1016/j.ejor.2021.04.025.
[37] Henry D, Emmanuel Ramirez-Marquez J. Generic metrics and quantitative
[10] Hussain A, Bui V-H, Kim H-M. Resilience-oriented optimal operation of networked
approaches for system resilience as a function of time. Reliab. Eng. Syst. Saf. 2012;
hybrid microgrids. IEEE Trans. Smart Grid 2019;10:204–15. https://doi.org/
99:114–22. https://doi.org/10.1016/j.ress.2011.09.002.
10.1109/TSG.2017.2737024.
[38] Mussi S. Facilitating the use of multi-attribute utility theory in expert systems:an
[11] Feng Q, Hai X, Sun B, Ren Y, Wang Z, Yang D, et al. Resilience optimization for
aid for identifying the right relative importance weights of attributes. Expert Syst.
multi-UAV formation reconfiguration via enhanced pigeon-inspired optimization.
1999;16:87–1102. https://doi.org/10.1111/1468-0394.00098.
Chin. J. Aeronaut. 2022;35:110–23. https://doi.org/10.1016/j.cja.2020.10.029.
[39] E. Fontela, A. Gabus. The DEMATEL Obsever. Switzerland, Geneva: Battelle
[12] Yılmaz E, German BJ, Pritchett AR. Optimizing resource allocations to improve
Geneva Research Center; 1976.
system reliability via the propagation of statistical moments through fault trees.
[40] Zhang XQ, Feng WH, Li NN. Attribute recognition model based on entropy weight
Reliab. Eng. Syst. Saf. 2023;230:108873. https://doi.org/10.1016/j.
and its application to evaluation of groundwater quality. AMM 2010;(29–32):
ress.2022.108873.
2698–702. https://doi.org/10.4028/www.scientific.net/AMM.29-32.2698.
[13] Zhang J, Li L, Chen Z. Strength–redundancy allocation problem using artificial bee
[41] Youn BD, Hu C, Wang P. Resilience-driven system design of complex engineered
colony algorithm for multi-state systems. Reliab. Eng. Syst. Saf. 2021;209:107494.
systems. J. Mech. Des. 2011;133:101011. https://doi.org/10.1115/1.4004981.
https://doi.org/10.1016/j.ress.2021.107494.
[42] Mettas A. Reliability allocation and optimization for complex systems. In:
[14] Fan D, Sun B, Dui H, Zhong J, Wang Z, Ren Y, et al. A modified connectivity link
Proceedings of the Reliability and Maintainability Symposium; 2000. p. 216–21.
addition strategy to improve the resilience of multiplex networks against attacks.
https://doi.org/10.1109/RAMS.2000.816310.
Reliab. Eng. Syst. Saf. 2022:108294. https://doi.org/10.1016/j.ress.2021.108294.
[43] Li W, Zuo MJ. Optimal design of multi-state weighted k-out-of-n systems based on
[15] Zhao Z, Zhao Y-G, Li P-P. A novel decoupled time-variant reliability-based design
component design. Reliab. Eng. Syst. Saf. 2008;93:1673–81. https://doi.org/
optimization approach by improved extreme value moment method. Reliab. Eng.
10.1016/j.ress.2008.01.009.
Syst. Saf. 2023;229:108825. https://doi.org/10.1016/j.ress.2022.108825.
[44] Faghih-Roohi S, Xie M, Ng KM, Yam RCM. Dynamic availability assessment and
[16] Zhang J, Liu T, Qiao J. Solving a reliability-performance balancing problem for
optimal component design of multi-state weighted K-out-of-N systems. Reliab. Eng.
control systems with degrading actuators under model predictive control
Syst. Saf. 2014;123:57–62. https://doi.org/10.1016/j.ress.2013.10.002.
framework. J. Frankl. Inst. 2022;359:4260–87. https://doi.org/10.1016/j.
[45] Frangopol DM, Lin K-Y, Estes AC. Life-cycle cost design of deteriorating structures.
jfranklin.2022.04.007.
J. Struct. Eng. 1997;123:1390–401. https://doi.org/10.1061/(ASCE)0733-9445
[17] Gunes V, Peter S, Givargis T, Vahid F. A survey on concepts, applications, and
(1997)123:10(1390).
challenges in cyber-physical systems. KSII Trans. Internet Inform. Syst. 2014;8:
[46] Zitzler E, Deb K, Thiele L. Comparison of multiobjective evolutionary algorithms:
4242–68. https://doi.org/10.3837/tiis.2014.12.001.
empirical results. Evol. Comput. 2000;8:173–95. https://doi.org/10.1162/
[18] Si S, Levitin G, Dui H, Sun S. Importance analysis for reconfigurable systems.
106365600568202.
Reliab. Eng. Syst. Saf. 2014;126:72–80. https://doi.org/10.1016/j.
[47] Khorshidi HA, Gunawan I, Ibrahim MY. A value-driven approach for optimizing
ress.2014.01.012.
reliability-redundancy allocation problem in multi-state weighted k-out-of-n
[19] Zhao J, Si S, Cai Z. A multi-objective reliability optimization for reconfigurable
system. J. Manuf. Syst. 2016;40:54–62. https://doi.org/10.1016/j.
systems considering components degradation. Reliab. Eng. Syst. Saf. 2019;183:
jmsy.2016.06.002.
104–15. https://doi.org/10.1016/j.ress.2018.11.001.
[48] Deb K, Agrawal S, Pratap A, Meyarivan T. A Fast Elitist Non-Dominated Sorting
[20] N. Ahmadian, G.J. Lim, J. Cho, S. Bora A quantitative approach for assessment and
Genetic Algorithm for Multi-Objective Optimization: NSGA-II. Parallel Problem
improvement of network resilience. Reliab. Eng. Syst. Saf. 2020:106977. 10.1016
Solving from Nature PPSN VI. Berlin, Heidelberg: Springer; 2000. p. 849–58.
/j.ress.2020.106977.
https://doi.org/10.1007/3-540-45356-3_83.
[21] Zhang L, Zeng G, Li D, Huang H-J, Stanley HE, Havlin S. Scale-free resilience of real
traffic jams. Proc. Natl Acad. Sci. 2019;116:8673–8. https://doi.org/10.1073/
pnas.1814982116.

12