Research Article

Published: 2024-11-29
https://doi.org/10.20935/AcadEng7408

Multi-criteria decision-making framework for evaluating

life-cycle resilience of power-generating plants
Abba Mas’ud Alfanda1, *
Academic Editor: Kofi Agyekum

Abstract
Power generating plants, indispensable for modern societies, remain vulnerable to seismic disruptions. While efforts have been made
to mitigate seismic risks, ensuring the life-cycle resilience of these critical infrastructures remains paramount. This study introduces
a comprehensive framework, BCA-MCDM, for assessing power plant resilience, considering a multi-criteria approach encompassing
costs, benefits, downtime, payback period, and repair costs. By integrating probabilistic-based assessments and cost–benefit analyses,
the framework facilitates the identification of optimal seismic mitigation strategies. Key findings emphasize the significant influence
of business downtime on life-cycle damage, stressing the need for expanded insurance coverage. The framework quantifies total losses
by combining direct and indirect costs, providing a robust basis for decision-making. To enhance the framework's applicability, future
research should explore the integration of fuzzy TOPSIS methodology and consider additional factors. This study contributes to the
development of resilient power plants by providing a robust assessment tool for decision-makers.

Keywords: multi-criteria decision-making, life-cycle evaluation, power-generating plants, seismic resilience, sensitivity analysis

Citation: Alfanda AM. Multi-criteria decision-making framework for evaluating life-cycle resilience of power-generating plants.
Academia Engineering 2024;1. https://doi.org/10.20935/AcadEng7408

1. Introduction
Power plants are recognized as critical, interconnected infrastruc- and earthquake loads. The conventional concentrically braced
tural systems that sustain vital economic services in both advanced frames used in industrial structures have been shown to have
and developing nations. While they are essential for energy produc- poor energy dissipation capacity during earthquakes due to buck-
tion, power plants face an increasing risk of damage from natural ling behaviors and unrepairable damages [5]. While buckling-
disasters such as floods, earthquakes, and hurricanes [1]. Among restrained braces (BRBs) provide consistent hysteresis, their low
these catastrophes, earthquakes have proven to be the most de- post-yielding stiffness results in large post-earthquake residual de-
structive [1–3]. Annual losses of power plants resulting from global formation, which increases repair costs [6]. Self-centering braced
natural hazards amount to approximately 15 billion USD, which is frames have been developed to eliminate residual drift by providing
equivalent to 0.2 percent of the cost of constructing power genera- self-centering and energy dissipation capabilities [7]. However,
tion infrastructure [4] (Figure 1). When seismic disruptions occur, the practical application of self-centering systems is challenged by
industrial power plants serving municipal and industrial areas their higher costs compared to conventional seismic solutions. Hy-
suffer substantial damage not only to their structural systems but brid self-centering–damping devices could help address this issue.
also to the surrounding population, environment, and businesses. There has also been a remarkable increase in the investigation of
novel damping techniques applied within thermal power plants.
In fact, the seismic resilience of industrial power plants is of the
Pioneering studies conducted by Shu et al. [8], Peng et al. [9],
utmost importance to ensure their ability to withstand seismic
and Li et al. [10] have shed light on the utilization of coal scut-
events and swiftly recover. To fulfill these requirements, seismic
tles as unconventional tuned mass dampers (TMDs). The primary
design codes permit some structural elements such as connections,
objective behind such endeavors revolves around mitigating the
braces, damper, and base isolators to behave nonlinearly, dissi-
adverse effects of inertial forces exerted by heavy coal scuttles on
pate energy, and decrease potential damages. This motivates re-
the support structure. By employing appropriate design techniques
searchers and the engineering community to innovate various miti-
and retrofitting measures, power plants can mitigate the negative
gation strategies for extending and upgrading the useful service life
consequences of earthquakes, safeguarding not only their struc-
of power plants to facilitate their swift recovery of functionality.
tural integrity but also their operational functionality.
Over time, numerous seismic control systems have been developed
The life-cycle seismic evaluation of power-generating plants has
and applied in power plants to mitigate the adverse effects of wind
been a topic of interest in research and practice. Most power-

1
Department of Civil Engineering and Institute for Disaster Management and Reconstruction, Sichuan University,
Chengdu 610064, China
∗ email: [email protected]

ACADEMIA ENGINEERING 2024, 1 1 of 15

https://www.academia.edu/journals/academia-engineering/about https://doi.org/10.20935/AcadEng7408

consideration of multiple criteria to ensure their resilience and

sustainability. Traditional evaluation approaches usually focus on
a single criterion [10, 13] such as cost or seismic effectiveness
control, and may not provide a comprehensive and integrated view
of the evaluation.
The preceding discussions emphasize the inherent complexity of
power plant seismic mitigation and the significance of considering
both engineering and non-engineering disruptions when evalu-
ating post-earthquake business recovery in a seismically active
region. Improving the decision-making process is an urgent en-
gineering challenge that aims to guide the selection of retrofit
solutions while ensuring cost-effective business recovery. By inte-
grating these elements, the proposed framework aims to provide a
more comprehensive and effective approach for evaluating retrofit
options and guiding decision-making processes in power plant
seismic mitigation, while considering both technical and economic
aspects. The framework considers the importance of each criterion
based on its contribution to the objectives of the evaluation and
assesses the alternatives based on established evaluation models
and weights. To bridge this gap, the present study investigates
the effectiveness of commonly used seismic control techniques
within the proposed BCA-MCDM framework, which encompasses
the following aspects:

i. Integrated predictions of expected downtimes in critical com-

Figure 1 • Global overview of power plant vulnerability. (a) Po- ponents (equipment) and utility schedules, allowing for the
tential causes of downtime. (b) Top countries with corresponding appropriate sequencing of repairs and operational considera-
average annual losses. tions.

ii. Utilization of typical decision metrics based on engineer-

generating plants are designed to operate for more than 30 ing economics, enabling a comprehensive assessment of re-
years [11], and understanding the life-cycle risks associated with silience by quantifying life-cycle benefits.
unacceptable damages and their financial implications is impera-
tive. Considering the seismic vulnerability of existing power plants, iii. Incorporation of multi-criteria decision evaluation, ensuring
it is imperative to carefully weigh the life-cycle costs associated a return on investment in retrofit costs over the power plant's
with implementing such measures against the potential savings lifespan.
that would have been accrued in the absence of mitigation. Ef-
forts aimed at maintaining an acceptable performance level can
often result in exorbitant expenditures surpassing those required 2. Formalization of the proposed
for constructing entirely new facilities. It is often challenging to framework
choose an appropriate technique and determine whether to retrofit
large-scale structures such as power plants when various decision As seen in Table 1, FEMA P-58 loss predictions do not consider
variables are involved. the significant post-earthquake downtime of a building, which can
occur regardless of the extent of damage and utility availability.
Despite the advancement of performance-based design shown in REDi guidelines address these issues in FEMA-P58 [12] through
Figure 2, existing frameworks relatively neglect the fact that a a detailed downtime estimation as the sum of PACT-based direct
power plant is largely interconnected and significantly affected by repairs [12], as well as impeding factors [14] and utility disruption
recovery cost and the effectiveness of the seismic mitigation system (water, natural gas, and electrical systems) in the form of lognormal
being adopted. Notably, FEMA P-58 [12] fails to account for the cumulative distribution functions.
significant post-earthquake cost of the downtime of a building in
a damaged state and has provision neither for downtime nor for Business downtime costs play a significant role in determining
functioning utility supplies. While the REDi repair time model the annual average loss (AAL). The downtime estimation method
addresses the limitations of the FEMA P-58 downtime model, it proves to be a more reliable and accurate measure of business
has the following limitations: (i) the limit state is not explicitly con- interruption costs. Therefore, the downtime estimation method
sidered; (ii) there is a fixed repair schedule regardless of damage should be preferred for effectively assessing and addressing busi-
severity and the complexity of the building; and (iii) REDi is more ness interruption costs. To model interdependence for the estima-
suitable for residential buildings, as its repair classes depend on the tion of repair time, the BCA-MCDM (Benefit–Cost Analysis with
occupancy of a building rather than the operationality of industrial Multi-Criteria Decision-Making) framework captures not only en-
buildings. gineering (direct) but also non-engineering (indirect) factors such
as the relationship between damage and recovery activities. This
Typically, comprehensive repair costs include the cost of up- will reduce underestimation and enable effective risk-estimation
grades and other expenditure indirectly related to repair works. tools in the catastrophe insurance market.
Ideal performance evaluation loss models should, therefore,
be structure-specific, but obtaining the required data of struc-
tural/nonstructural and related earthquake consequence infor- 2. 1. Extension of conventional PEER framework
mation is challenging. Therefore, the life-cycle seismic evaluation The conventional PBEE (performance-based earthquake engineer-
of power-generating plants is a critical task that requires the ing) framework is expressed in Equation (1) as a multivariate

ACADEMIA ENGINEERING 2024, 1 2 of 15

https://www.academia.edu/journals/academia-engineering/about https://doi.org/10.20935/AcadEng7408

Figure 2 • Evolution of performance design approaches.

integral probability distribution function, capturing the conse- from steps two and three into a Multi-Criteria Decision-Making
quences of earthquake-induced damage: and Benefit–Cost Analysis (MCDM-BCA) resilience model.

y 2. 2. Damage state definition

λ (DV) = dP (DV|DS) dP (DS|EDP) dP (EDP|IM) dλ (IM) (1)
| {z } | {z } | {z } | {z } | {z }
Decision Variable 4.Loss Analysis 3.Fragility Analysis 2. Structural Analysis 1. Hazard Analysis Table 2 illustrates the severity of damage to power plant buildings
categorized as slight, moderate, extensive, or complete. Each DS is
where IM is the intensity measure (e.g., spectral acceleration),
assigned a corresponding repair class. Detailed evaluations using
EDP is the engineering demand parameter (e.g., drift), DS is the
Equation (3) and Equation (4) determine the repairability
damage state, and DV is the decision variable (e.g., financial loss
limit state and the functionality limit. Each damage state corre-
or downtime).
sponds to a set of consequences, including the repairs and related
In the performance-based earthquake engineering (PBEE) frame- costs required, as well as the repair time.
work [13], the evaluation of seismic performance for civil structures
Fragility curves (Figure 4) provide an efficient method for con-
is a multi-step process. It begins with hazard analysis, which as-
ducting probabilistic seismic evaluations of power plants. These
sesses the likelihood of seismic events occurring. This is followed by
curves enable the estimation of the likelihood of exceeding specific
structural analysis, where the building’s response to seismic forces
damage states (Table 2) for different power plant elements under
is evaluated. The next step, damage analysis, estimates the extent of
specific seismic hazard conditions.
physical damage caused by the seismic event. Finally, loss analysis
quantifies the resulting consequences in terms of economic costs, The process of obtaining seismic fragility curves involves initially
downtime (DT), and other impacts. constructing a probabilistic seismic demand model (PSDM) and
subsequently evaluating the relationship between demands and
To extend the PBEE framework further, the proposed model, as
capacities in a probabilistic manner. Among the various engineer-
shown in Figure 3, consists of four key steps: (1) the quantifi-
ing demand parameters (EDPs), the maximum inter-story drift
cation of seismic hazard scenarios based on annual probabilities
ratio (MIDR) holds significant importance in characterizing the
of exceedance for defined intensity measures, (2) the definition
seismic behavior of a power plant building on a global scale. For
of the fragility and functionality loss of the power plant under
most structural systems within power plants, as well as many
seismic events, (3) the development of recovery models addressing
nonstructural components, MIDR has been proposed as a reliable
both the restoration of plant capacity and the reinstatement of
indicator of potential damage.
component functionality, and (4) the integration of the outcomes

Table 1 • Comparison of existing recovery models.

Impending Flexible repair Utility
Reference Framework Recovery states
factors sequence disruption
ABV
Porter et al. [15] (Assembly-Based 1 No No No
Vulnerability)
PACT
(Performance
FEMA P-58 [12] 1 No No Yes
Assessment
Calculation Tool)
REDi™
(Resilience-Based
Arup et al. [16] Design Initiative 3 Yes No Yes
Seismic Downtime
Model)
BCA-MCDM
(Benefit–Cost
Proposed
Analysis with 3 Yes Yes Yes
Framework
Multi-Criteria
Decision-Making)

ACADEMIA ENGINEERING 2024, 1 3 of 15

https://www.academia.edu/journals/academia-engineering/about https://doi.org/10.20935/AcadEng7408

Figure 3 • Resilience quantification framework overview.

PSDM establishes the connection between an engineering demand The repair cost ratio (RCR) is expressed in terms of Equation (3).
parameter (EDP) such as maximum drift or residual drift and an
DS
earthquake intensity measure (IM) like spectral acceleration at the X
fundamental period Sa (T1 ) in logarithmic form represented in RCR = LRi × (DRi ) (3)
i=1
Equation (2):

Additionally, in evaluating the functional capacity of power plants

ln (edp/edp0.5 )
 
P (EDP ≥ edp|IM) = 1 − Φ √ (2) after seismic events, functional capacity refers to the percentage
β of a facility expected to remain operational post-earthquake. The
functionality can be calculated using the correlation between dif-
where Φ [·] is the standard normal cumulative distribution func- ferent damage states (DS) in Equation (4):
tion, edp0.5 is the median capacity of the structural demand for a
n
given seismic intensity, and β is the logarithmic standard deviation X
FPc = 1 − FRi × P| (DSi ) (4)
of the demand conditioned on IM.
i=1

where i is the damage state number, LRi is the weight correspond-

ing to DSi , and DRi is the ratio of units in damage state i . FPc is
the combined facility functionality, FRi is the facility restoration
percent for the ith damage state, and P(DSi) is the probability of
being in the ith damage state.
For example, with weights of 0.1 for DS1 , 0.4 for DS2 , and 0.8 for
DS3 , and 20% of the plant in DS1 , 28% in DS2 , and 2% in DS3 , the
Adjusted Damage Ratio (ADR) is (0.1 × 0.20) + (0.4 × 0.28) +
(0.8 × 0.02) = 0.148. Since this is below the FEMA P-58 threshold
of 0.5 [12], it indicates partial functionality loss.

Figure 4 • Illustration of damage state definitions. 2. 4. Life-cycle loss assessment

The use of AAL as a decision variable and performance measure

Based on Figure 4, the probability of the damage occurring within enhances the understanding of the economic impact and rational
each specified state, given a certain demand, is equal to the differ- investments in seismic retrofitting for essential facilities. In this
ence between the corresponding probabilities of that state and the framework, average annualized loss (AAL) involves combining re-
next higher state. pair costs and business interruption costs, taking into account both
direct and indirect losses weighted by the probability of exceeding
the hazard. Equation (5) can be used to express earthquake-
2. 3. Repair and retrofit cost estimation
induced losses for a power plant building with and without retrofit
Repair cost estimations are crucial in deciding whether to repair action:
n
or demolish a building. According to FEMA [12], repair costs AAL =
X
E [Li|Sa (T1 )] dλsaT1 (Sa ) (5)
exceeding 50% of the replacement cost typically justify demolition. i=0

ACADEMIA ENGINEERING 2024, 1 4 of 15

https://www.academia.edu/journals/academia-engineering/about https://doi.org/10.20935/AcadEng7408

Table 2 • Harmonized repair class and damage definitions.

Repair class Damage level Description Resilience Functionality Repair Utility Repair
states actions availability cost ratio
0 DS0 —None No damage Full function Full (100%) None Yes [0.0,0.04]
1 DS1 —Slight Slight damage Full function Several stops Minor Yes (0.04,0.10]
to a building and decrease in structural
electricity in restoration
the measures are
transmission required for
lines most
(90–60%) components
2 DS2 — Considerable Malfunction Inability to Measures may No (0.10,0.40]
Moderate damage to (full function generate be relatively
mechanical after repairs) electricity in major in some
and electrical power station cases
equipment or (60% to 40%)
loss of electric
power and
backup for 7
days
3 DS3 — A building Full loss of Inability to Replacement No (0.4,0.80]
Extensive being function generate or
extensively (unrepairable electricity in enhancement
damaged damage) power station of some
(40% to 10%) components
may be
required
4 DS4 — Building Full loss of Inability to Demolition is No [0.80,1]
Complete collapsed function generate necessary
(unrepairable electricity in
damage) power
station(<10%)

In terms of repair cost Equation (5) can be rewritten in the form for estimating business interruption costs associated with repair
of Equation (6): durations (DT1 ). The total downtime of the power plant is deter-
mined through the critical path method (CPM), which organizes
n
X repair tasks in a sequential manner. Figure 5 demonstrates the
AAL = CRep P [DSi|Sa (T1 )] LRi [DSi] dλsaT1 (Sa ) (6)
operational sequence and capacity block diagram (CBD), depicting
i=0
the interdependence of key plant components, such as boilers, coal
where CRep is the replacement value; P[DSi |Sa(T1 )] is the prob- storage, and water-cooling systems, to provide a clear view of how
disruptions in one area can affect overall plant operations.
ability of a given damage state given Sa (T1 ); dλsaT1 (Sa ) is the
absolute value of the derivative of the hazard curve with respect DT1 accounts for irrational components through a specific critical
to Sa (T1 ) ; and LRi [DSi ] denotes a global loss ratio of 0.1, 0.4, and path method (CPM) sequence, as shown in Figure 6. The repair
0.8 [17] for DS1 , DS2 , and DS3 , respectively. work breakdown structure is then transformed into a flowchart,
presenting activities in a chronological order on a network dia-
The replacement cost considers both the construction and demo- gram. In this diagram, repair tasks are represented by boxes, while
lition expenses (including clearing the site and mobilization) and arrows indicate the dependencies between them. As the repair
the costs related to recovering from potential factors. Moreover, the schedule is further developed, additional time-bound components
retrofit costs, which include the cost of materials used for retrofit are incorporated into the network diagram. Each node symbolizes
implementation, are estimated as follows [18]. a specific component, and the links between them represent both
physical and relational connections, ensuring a comprehensive
C = Q m C m + Cl (7) representation of the repair process.
where Qm is the quantity of material required, Cm is the unit cost of Critical path method-based calculations for downtime are deter-
material [19], and Cl is the cost of labor and installation. mined through backward and forward passes. The critical path
(highlighted in red) indicates the minimum time necessary to
complete the entire repair work. These calculations identify the
2. 5. Overall downtime estimations
Earliest Finish Time (EF) and Latest Start Time (LS) for each task
As part of a loss assessment, business interruptions are accounted in Equation (8) and Equation (9), respectively.
for by combined downtime due to actual repair (DT1 ), downtime
due to irrational delays (DT2 ), and downtime caused by utility in- Earliest finishtime (EF) = ES + repair duration (8)
terruption (DT3 ). CPM determines the order of component repair,
while DT3 considers the functionality of the facility.
Latest start time (LS) = LF − repair duration (9)

2. 5. 1. Repair time estimation

It should be noted that the Earliest Start Time (ES) refers to the
An accurate estimation of repair time depends on a thorough earliest point at which a task can commence, contingent upon
understanding of the overall functionality of the power plant and the completion of its preceding dependent tasks. Conversely, the
its critical components. By employing a logical repair sequence and Latest Finish Time (LF) represents the latest permissible time for
integrating both structural and nonstructural elements, as shown an activity to be completed without disrupting the overall repair
in Figure 5, the proposed framework provides a more robust tool schedule.

ACADEMIA ENGINEERING 2024, 1 5 of 15

https://www.academia.edu/journals/academia-engineering/about https://doi.org/10.20935/AcadEng7408

Table 3 • Business interruption in terms of impending factor inputs.

Impending factors Median (days) Dispersion
Repair Class Slight Moderate Extensive Slight Moderate Extensive
Inspection (essential 2 0.54
facility)
Engineering - 14 28 - 0.32 0.54
mobilization
(essential facility,
fewer than 20 stories;
engineer on contract)
Financing (insurance) 6 1.11
Permitting - 7 28 - 0.86 0.36
Water, wastewater
system, and coal
supply

2. 5. 2. Delays and utility disruption estimations in DT*i due to the ith mitigationQfactor, which may depend on the
retrofit option in question, and indicates the effect of multiple
DT2 and DT3 are calculated using the lognormal distribution curves mitigation factors considered in a multiplicative manner.
and regression functions provided in the REDi [16] and ERI [20]
guidelines, respectively. As seen in Table 3, the proposed frame-
work determines the post-earthquake inspection time by applying 2. 5. 3. Overall downtime cost
the lognormal model detailed in REDi [16].
{
Engineering + Permitting The total business downtime is estimated as the sum of all down-
DT2 = Earthquake inspection+Max Contractor mobilization (10) time due to mobilization factors for different building retrofits.
Financing
DT = DT1 + DT2 + DT3 (12)
Next, to account for utility disruptions (DT3 ), Equation (11)
presents the disruption across multiple utilities, following a where impending factors due to delays are obtained as [14] and
method similar to that used in previous calculations for such summarized in Table 3.
disruptions [20] (ERI, 2018). Utility disruptions are calculated
by combining coal outages (DT3(1)* ), sewage outages (DT3(2)* ), From Equation (12), the recovery time can be rewritten as Equa-
and water outages (DT3(3)* ). These disruptions are classified by tion (13)
duration—hours, days, weeks, and months—corresponding to the DT = DT2 + max {DT1 , DT3 } (13)
levels of slight, moderate, and very extensive outages.
n
X
DT = DTi (Days|DSi )P [DSi|Sa (T1 )] × MODDSi (14)
( nk )
(k)∗
Y
DT3 = max (1 − MFi ) × DTi (11) i=1
i=1
where DTi (Days|DSi) is the repair time obtained from CPM, and
where DT(k)∗ 3 is the unmitigated disruption due to the kth utility, MODDSi are the construction time modifiers, with 1, 0.5, and 0.5
DTi represents the mitigated form of DT*i, MFi is the reduction assigned to DS1, DS2, and DS3, respectively.

Figure 5 • Interdependence of components of power plant. (a) Operational sequence based on repair schedule. (b) Capacity block
diagram (CBD).

ACADEMIA ENGINEERING 2024, 1 6 of 15

https://www.academia.edu/journals/academia-engineering/about https://doi.org/10.20935/AcadEng7408

Figure 6 • Overall reconstruction activity network (ORAN) including 9 units.

Once the total downtime is computed, the cost in terms of financial 3. 1. Benefit quantification
loss can be estimated using Equation (15):
Benefit–Cost Analysis (BCA) is used to assess the financial viability
of seismic retrofit alternatives by weighing their costs against po-
Downtime cost (USD) = IC × CF × 24 × DT (15)
tential benefits. The process begins by identifying feasible retrofit
options, which are evaluated using criteria such as payback pe-
where IC is the installed capacity of the facility in MWh, CF is the riod, repair costs, and reduction in downtime. The total cost of
power reduction factor, and DT is the downtime in days. implementing each alternative is estimated using Equation (7).
In relation to business downtime, a measure of resilience known The net present benefit (NPB) is then calculated, which represents
as the resilience index (R-index) [10] gives an indication of how the difference between the expected benefits (reduced losses) and
quickly and efficiently a facility returns to its normal operations. the associated costs over time. NPB helps determine the long-
The R-index measures the facility’s recovery over time and is cal- term financial advantage of retrofitting. To further evaluate this,
culated using Equation (16). the annual average loss without retrofitting (AAL) and that with
retrofitting (AAL*) are computed using Equation (17), allowing a
R tO +TRE comparison of the financial impacts of each alternative.
tO
Q (t)
R-index = dt (16)
TLC
 PT
 Pt=1 (AAL − AAL∗) (1 + i) r = 0
T
NPB (T) = t=1 (AAL − AAL∗) (1 + i) (17)
where Q (t) is the functionality of the facility over time, TLC is the r>0
(1 + r)t

time for full control (typically set to 1 year), TRE is the recovery
time after disruption, and tO is the time when the earthquake event where T is the time period in years, NPB is the net present benefit
occurs. in terms of reduced losses, i is the interest rate, and r is the discount
rate.

3. BCA-MCDM framework methodology 3. 2. MCMD step-by-step procedure

MCDM (multi-criteria decision-making) has been widely applied A decision matrix is constructed, where each row corresponds to
in fields such as engineering, management, and environmental an alternative M, and each column corresponds to a criterion N, as
sciences to tackle decision-making challenges involving multiple shown in Table 4 and Equation (18).
criteria. Unlike Benefit–Cost Analysis (BCA) [21], which focuses
on a single criterion and often overlooks the combined impact of
various factors, integrating BCA with MCDM offers a more com- (Xij )M×N (18)
prehensive approach [10]. This integration allows decision-makers
to thoroughly evaluate options and estimate the costs and benefits where i = 1…..m; j = 1,…,n ; Xij is an element of the matrix.
of implementing recovery plans for critical facilities. The optimal The decision matrix is then normalized to make the values compa-
resilience alternative is identified by minimizing financial losses rable across different criteria. Equation (19) is the normalization
while considering multiple decision criteria in Table 4. TOPSIS formula:
(Technique for Order of Preference by Similarity to Ideal Solution)
is a commonly used MCDM method that breaks down complex Xij
Xij = qP where i = 1, 2, . . . , m; j = 1, 2, . . . , n (19)
decisions into a hierarchy of criteria and alternatives, determining n
Xij 2
j=1
their relative importance.
As shown in Figure 7, the BCA-MCDM (Benefit–Cost Analysis where Xij and Xij are the normalized and original decision matrix,
with Multi-Criteria Decision-Making) framework involves several respectively.
relevant equations to evaluate seismic retrofitting or disaster re-
Once the weights are assigned, the weighted decision matrix is cal-
covery strategies.
culated by multiplying the normalized values by the corresponding
Below is a detailed step-by-step explanation with all relevant equa- criterion weights:
tions included. Vij = XIJ × Wj (20)

ACADEMIA ENGINEERING 2024, 1 7 of 15

https://www.academia.edu/journals/academia-engineering/about https://doi.org/10.20935/AcadEng7408

Figure 7 • BCA-MCDM framework.

where Wj is the weight for j criterion. (SCBFs) and moment-resisting frames (MRFs) in both directions.
Each structural member is strictly evaluated for sectional strength,
Through the identification of the best and worst scenario of each
including compression, flexure, and buckling, in line with AISC
criterion from the weighted decision matrix, the separation of each
360-16 [23] requirements.
alternative from the best and worst ideal
solutions is calculated using Euclidean distance with Equation
(21) and Equation (22):
X 0.5
m
2
S+ = Vij − VJ + , j = 1, 2 . . . . . . m (21)
j=1

X 0.5
m
2
S− = Vij − VJ − , j = 1, 2 . . . . . . m (22)
j=1

where S− is the distance of alternative i from the best ideal solution

and S+ is the distance of alternative i from the worst ideal solution.
The final step in the BCA-MCDM process is to calculate the rela-
tive closeness (Pi ) to the ideal solution for each alternative using
Equation (23):

Si −
Pi = , where 0 < Pi < 1.0, i = 1, 2, 3 . . . .m (23)
Si + Si −
+
Figure 8 • Hazard curves of seismic mitigation cases.

Alternatives with Pi values closer to 1 are considered the best

options, as they are closest to the ideal solution. 4. 2. Modeling techniques
This study utilizes OpenSees for structural analysis, and the nu-
merical model considers nonlinear material properties and ele-
4. Results and discussion ments. To reach the desired precision in structural design op-
4. 1. Case example power plant timization, it is essential to repeat the process multiple times,
thus making the process quite complex. Table 5 summarizes the
To illustrate the seismic resilience of a power generation plant numerical modeling techniques developed using nonlinear mech-
using the proposed framework, an ultra-supercritical (USC) coal- anisms to capture potential failure and verified with available
fired plant with a total capacity of 2x 650 MW (1300 MW) and a net experimental data in the technical literature.
output of 832 MW is evaluated. This plant is located in a seismically
active region of China, characterized by stiff soil conditions with a
4. 3. Seismic mitigation strategies
reference shear wave velocity (Vs30) ranging from 179 m/s to 280
m/s (Figure 8). The thermal power plant is shown in Figure 9a Various design schemes involving different structural systems are
(elevation) and Figure 9b (plan view). Key components within evaluated for performance comparison, including (1) Case 1: spe-
the plant, such as the turbine hall and deaerator bay, are designed cial concentrically braced (SPCB) frame; (2) Case 2: buckling re-
as moment frames to accommodate multi-story clearance require- strained braced (BRB) frame (Figure 10); (3) Case 3: SMA-BRBF
ments. The primary lateral force-resisting systems are designed system (shape memory alloy–buckling-restrained braced frame)
according to the standards AISC 360-16 [22], AISC 341-16 [23], and (Figure 11); and (4) Case 4: equipment isolation (coal bunker
ASCE/SEI 7-16 [24], utilizing special concentrically braced frames isolation systems) (Figure 12). These alternative seismic control

ACADEMIA ENGINEERING 2024, 1 8 of 15

https://www.academia.edu/journals/academia-engineering/about https://doi.org/10.20935/AcadEng7408

Table 4 • Decision criteria related to industrial facilities.

Criteria j Weight
Reduced financial losses due to the loss of the facilities 0.3
Minimal downtime to maintain operation under extreme earthquakes 0.2
Continuance of operation within a short pay-back period at an affordable 0.2
discount rate
Cost of retrofit implementation 0.2
Annualized benefit 0.1

Figure 9 • Example power plant (a) Elevation (b) plan.

and retrofit strategies aim to reduce monetary losses in seismically For instance, the power plant in question, with a continuous
vulnerable plant buildings. production capacity of 1300 MW, operates at an efficiency rate
of 64%, which translates to a daily energy production value of
approximately (1300 × 0.64 × 24 × 112) dollars [27]. Table 7
4. 4. Selection of ground motions
summarizes the additional costs due to retrofit actions and the
A total of three sets of 15 ground motions are then scaled to associated replacement costs for Cases 2 to 4.
align with the uniform hazard spectrum corresponding to three
For each retrofit option under consideration, the expected loss
different levels: the service level earthquake (SLE), design-based
is calculated for the entire range of earthquake ground motions,
earthquake (DBE), and maximum considered earthquake (MCE).
considering the potential building damages.
The DBE hazard level, as shown in Figure 13. The scaling process
of the ground motion records aimed to achieve a mean squared A detailed comparison between Hazus [28] and the proposed BCA-
error of less than 10% between the target hazard spectrum and MCDM, shown in Figure 14, highlights the superiority of the
the average spectrum of the selected ground motions, as shown BCA-MCDM model (referred to as CPM) in estimating building
in Figure 13. For more comprehensive information regarding downtime after seismic events. BCA-MCDM offers a more compre-
the selected ground motions and their consistency with the target hensive and reliable approach by considering the repair times for
spectrum, please refer to Dai et al. [10]. individual components of a structure and integrating uncertainties
through lognormal cumulative functions. This model significantly
Nonlinear response history analyses are then performed using
improves the accuracy of functional state estimations by factor-
scaled ground motions to evaluate the structural performance of
ing in various parameters that directly affect building downtime,
the case example power plant. Using Equation (5), the average
unlike Hazus, which tends to oversimplify these factors.
annual damages are calculated by comparing the AAL of the origi-
nal building with the AAL* of the building retrofitted to a specified The analysis clearly demonstrates that the SMA-BRBF system
seismic performance level. provides significantly greater resilience compared to the bench-
mark designs. As depicted in Figure 14, the SMA-BRBF system
To estimate the total repair time, the average number of days
consistently shows quicker recovery times and a faster return to
required for completing repairs on each component is derived from
functionality, making it a superior option for reducing seismic
the research conducted by Prabhu et al. [26], along with relevant
damage and downtime in building structures.
industry standards. The duration estimates for these repairs are
presented in Table 6, providing a clear framework for assessing Furthermore, Figure 14 illustrates the comparison between BCA-
the time needed for restoring the full operational capacity of the MCDM and the Hazus approach, showing that the BCA-MCDM
affected components. method predicts significantly longer downtimes. On average, BCA-
MCDM estimations are 1.2 to 1.4 times higher than those of Hazus,
Based on building repair class tagging, the framework employs the
reflecting a more cautious and realistic approach. Specifically, the
lognormal model in Table 3 to estimate post-earthquake inspec-
analysis revealed that the predicted downtimes using BCA-MCDM
tion time. Considering factors related to mobilization for various
were 40% higher for Case 1, 28% for Case 2, 38% for Case 3,
building retrofits, the total business downtime is calculated.

ACADEMIA ENGINEERING 2024, 1 9 of 15

https://www.academia.edu/journals/academia-engineering/about https://doi.org/10.20935/AcadEng7408

Figure 10 • Calibration of single-story buckling-restrained braced frame (BRBF) against experimental test [25].

Figure 11 • Components response under cyclic loading (a) BRB (b) SMA (c) SMA-BRB.

Figure 12 • (a) Coal scuttle isolation system. (b) LRB response under cyclic loading.

ACADEMIA ENGINEERING 2024, 1 10 of 15

https://www.academia.edu/journals/academia-engineering/about https://doi.org/10.20935/AcadEng7408

Table 5 • Numerical modeling strategy.

Elements Modeling approach Features and parameters
Beam and Column Nonlinear beam–column element with a Fiber section incorporates co-rotational
layered fiber section geometric transformation to account for
geometric nonlinearities under large
displacements
Brace Fatigue fracture behaviors in braces simulated ε = 0.191 LR; M =0.671; min = −0.035; max
with low-cycle fatigue material =0.035
BRB SteelO4 Yield stress, Fy = 345 MPa; Young’s modulus,
E = 200 GPa; strain hardening rate, b = 0.1%
(kinematic hardening); R0 = 20; cR1 = 0.925;
cR2 = 0.25 (isotropic hardening parameters);
a1 = 0.4; a2 = 10; a3 = 0.4; a4 = 10
SMA-BRB Self-centering + SteelO4 material models Initial stiffness K1 = 2759 MPa; post-yield
stiffness K2 = 3882 MPa; Fy = 414MPa; β =
0.39; α = 0.58; maximum transformation
strain ( εu) = 5.5%
Lead Rubber Bearing Bi-linear model Initial stiffness; post-yield stiffness; yield
strength; characteristic strength

Figure 13 • Selection of ground motion spectra: (a) SLE; (b) DBE; (c) MCE.

and 36% for Case 4. This discrepancy underscores the limitations Table 8 presents the Pearson correlation matrix between the
of simplified models like Hazus and highlights the value of more potential input parameters and the net present benefit (NPB).
detailed, component-based performance models like BCA-MCDM, Based on this analysis, the relevance of a parameter decreases
particularly in the context of complex infrastructure such as power as its correlation value approaches zero, indicating little to no
plants. relationship. Conversely, the relevance increases as the correlation
value approaches unity, either positive (+1) or negative (-1), which
represents a perfect correlation. Parameters with high absolute
4. 5. Input parameters of probabilistic life-cycle BCA
correlation values are deemed more significant for predicting NPB.
framework
As shown in Table 9, the net present benefit (NPB) curves show
The net present benefit (NPB) fluctuates depending on the time
the total expected benefits from all potential earthquake scenar-
periods, discount rates, and the additional costs of implementing
ios. These curves provide a clear representation of the financial
retrofit measures. Alterations to any of these variables can lead to
outcomes of different seismic interventions, helping to identify
significant shifts in NPB. Including irrelevant parameters in the
the most effective strategies while maintaining computational ef-
analysis not only reduces the computational accuracy of the Monte
ficiency.
Carlo simulation but also significantly increases computational
costs and time. Furthermore, the selection of relevant parameters Figure 16 presents the net present value (NPV) generated from
is refined, capturing the uncertainties affecting decision-making. 1,500 simulations for various seismic mitigation interventions.
To identify the most influential input parameters, this study em- These interventions include isolating the coal scuttle using lead
ploys sensitivity analysis as a critical tool to eliminate irrelevant rubber bearing (LBR) isolators, as well as the application of buck-
inputs. By considering various scenarios and input variations, sen- ling restrained brace (BRB) and shape memory alloy–buckling
sitivity analysis first explores the impact of varying key seismic restrained brace (SMA-BRB) techniques.
retrofit data (Figure 15).
As observed from Table 10, when comparing various mitigation
The selection of relevant parameters from the pool of potential options, Case 3 proves to be the most effective in minimizing the
inputs is performed using Pearson correlation (PC). life-cycle loss of the power plant across all evaluated earthquake
scenarios and time intervals. This indicates that the mitigation

ACADEMIA ENGINEERING 2024, 1 11 of 15

https://www.academia.edu/journals/academia-engineering/about https://doi.org/10.20935/AcadEng7408

Table 6 • Average repair time for the case example power plant.
Description Activity Dependent Slight Moderate Extensive
Main structural frame A - 7 15 60
Coal storage B A 5 10 45
Coal handling C B 5 8 30
Ash storage D A 5 10 45
Ash handling E D 5 10 30
Boiler F D, E 5 10 45
Source/Water tank G A 5 10 45
Superheater H F 5 8 30
Economizer I F, G,H 5 10 45
Air pre-heater J G 5 8 30
Condenser K I, L 5 8 30
Turbine L H 10 25 60
Generator M L 10 25 60

Table 7 • Additional cost due to retrofit actions.

Retrofitting option Case 2 Case 3 Case 4
Cost (USD) 303,767 360,023 295,740

measures in Case 3 provide a significant reduction in poten- for complex power plants by a factor of 1.2 to 1.4. The accuracy
tial losses over the plant's operational life. In contrast, Case 4 discrepancy between the simplified PEER–Hazus and the
demonstrates the least effectiveness, suggesting that its mitigation more rigorous BCA-MCDM approaches ranges from 28% to
approach offers minimal benefits in terms of reducing life-cycle 40%, contingent upon the specific retrofit scenario.
losses when compared to the other options.
2. Sensitivity analyses indicate that the net present benefit of
retrofit solutions is highly influenced by the discount rate and
5. Conclusions the remaining life cycle of the power plant.
Seismic strategies for the design and retrofitting of power plants 3. By accounting for both direct and indirect losses, this study
are of paramount importance to engineers, stakeholders, and calculates total losses effectively. The average annual loss
decision-makers. This study emphasizes the necessity of integrat- (AAL) is identified as a critical metric, integrating repair costs
ing probabilistic time-based assessments and cost–benefit anal- and business interruption expenses, weighted by the proba-
yses to promote innovative retrofit measures and minimize the bility of exceeding hazard thresholds. The BCA-MCDM frame-
financial burdens associated with repairs. The proposed BCA- work facilitates the identification of most optimal mitigation
MCDM framework aims to address the challenges related to recov- option, achieving a balance between benefits and costs.
ery costs and downtime when selecting mitigation measures, com-
bining repair cost and downtime estimation through a system-level 4. Insurance organizations are encouraged to broaden their
performance and component-based analysis. Key findings from the risk management tools to encompass a wider array of risks,
research include the following: thereby achieving a balanced approach to energy dissipation
and self-centering capacities when facing seismic hazards.
1. Business downtime costs significantly impact the life-cycle
damage of power plants subjected to seismic loads. The sim- 5. Additionally, this framework presents opportunities for ex-
plified PEER–Hazus model tends to underestimate downtime pansion by incorporating further criteria and consequences,

Table 8 • Correlation coefficient matrix.

Input parameters Downtime Discount rate Cost of retrofit Repair cost Life cycle
Downtime 1
Discount Rate −0.972101977 1
Cost of Retrofit -1 0.972101977 1
Repair Cost 1 −0.972101977 −1 1
Life Cycle 0.93237495 −0.98462681 −0.93237495 0.93237495 1

ACADEMIA ENGINEERING 2024, 1 12 of 15

https://www.academia.edu/journals/academia-engineering/about https://doi.org/10.20935/AcadEng7408

Figure 14 • Total AAL of power plant with different mitigation cases.

Figure 15 • NPB input sensitivity plot. (a) Tornado diagram; (b) spider plot.

Table 9 • Probabilistic input parameters.

Parameters Exact Lower Upper Mean Standard Simulation
deviation
Discount rate 0.30 0.01 0.16 0.05 0.29
Case 2 Mitigation cost 303767 410085 349332 379709 10125.57 365829
Inflation 0.09 0.01 0.05 0.01 0.05
Discount rate 0.30 0.01 0.16 0.05 0.13
Case 3 Mitigation cost 360023 486031 414026 450028.75 12000.77 447640
Inflation 0.09 0.01 0.05 0.01 0.04
Discount rate 0.30 0.01 0.16 0.05 0.15
Case 4 Mitigation cost 295740 399249 340101 369675 9858 384171
Inflation 0.09 0.01 0.05 0.01 0.06

Table 10 • Decision matrix.

Retrofit option S+ S− Pi Rank
Case 2 0.14094755 0.069583741 0.330514958 2
Case 3 0.082307228 0.194730158 0.70290209 1
Case 4 0.164465454 0.035335552 0.176853725 3

ACADEMIA ENGINEERING 2024, 1 13 of 15

https://www.academia.edu/journals/academia-engineering/about https://doi.org/10.20935/AcadEng7408

Figure 16 • Net present benefit results based on 5000 Monte Carlo simulations.

and could be refined through the application of fuzzy method- Institutional review board statement
ology instead of relying solely on precise values.
Not applicable.

Acknowledgments
Informed consent statement
I would like to express our sincere gratitude to Sichuan University
for providing the necessary resources and support for this research. Not applicable.

Funding Additional information

Not applicable. Received: 2024-10-25
Accepted: 2024-10-28

Author contributions Published: 2024-11-29

A.M.A.: Conceptualization, methodology, software, validation, for- Academia Engineering papers should be cited as Academia
mal analysis, investigation, data curation, writing—original draft Engineering 2024, ISSN 2994-7065, https://doi.org/
preparation, writing—review and editing, and project administra- 10.20935/AcadEng7408. The journal’s official abbreviation
tion. is Acad. Engg.

Conflict of interest Publisher’s note

The authors declare no conflict of interest. Academia.edu Journals stays neutral with regard to jurisdictional
claims in published maps and institutional affiliations. All claims
expressed in this article are solely those of the authors and do
Data availability statement not necessarily represent those of their affiliated organizations, or
those of the publisher, the editors and the reviewers. Any product
The data that support the findings of this study are available from that may be evaluated in this article, or claim that may be made by
the corresponding author upon reasonable request. its manufacturer, is not guaranteed or endorsed by the publisher.

ACADEMIA ENGINEERING 2024, 1 14 of 15

https://www.academia.edu/journals/academia-engineering/about https://doi.org/10.20935/AcadEng7408

Copyright Redwood City (CA): Applied Technology Council; 2018.

©2024 copyright by the author. This article is an open access 13. Alfanda AM, Dai K, Wang J. Review of Seismic Fragility
article distributed under the terms and conditions of the Creative and Loss Quantification of Building-Like Industrial Fa-
Commons Attribution (CC BY) license (https://creativecommons. cilities. J Press Vessel Technol. 2022;144:060801. doi:
org/licenses/by/4.0/). 10.1115/1.4054844

14. Comerio MC. Estimating Downtime in Loss Modeling.

References Earthq Spectra. 2006;22:349–65. doi: 10.1193/1.2191017

1. Fujisaki E, Takhirov S, Xie Q, Mosalam KM. Seismic vul- 15. Porter K, Ramer K. Estimating earthquake-induced failure
nerability of power supply: Lessons learned from recent probability and downtime of critical facilities. J Bus Contin
earthquakes and future horizons of research. Proceedings of Emer Plan. 2012;5:352–64. doi: 10.69554/UNUE6649
the International Conference on Structural Dynamic, EURO-
DYN; 2014 Jun 30–Jul 2; Porto, Portugal; 2014; p. 345–50. 16. Almufti I, Willford M. The REDiTM rating system: A frame-
work to implement resilience-based earthquake design for
2. Caputo AC, Kalemi B, Paolacci F, Corritore D. Computing new buildings. Proceedings of the NCEE 2014 - 10th U.S.
resilience of process plants under Na-Tech events: Method- National Conference on Earthquake Engineering: Frontiers
ology and application to sesmic loading scenarios. Reliab Eng of Earthquake Engineering; 2014 Jul 21–25; Anchorage,
Syst Saf. 2020;195:106685. doi: 10.1016/j.ress.2019.106685 Alaska; 2014. doi: 10.4231/D3P26Q437

3. Merino Vela RJ, Brunesi E, Nascimbene R. Derivation of 17. FEMA. Multi-hazard Loss Estimation Methodology, Earth-
floor acceleration spectra for an industrial liquid tank sup- quake Model, Hazus–MH 2.1: Technical Manual. Washing-
porting structure with braced frame systems. Eng Struct. ton (DC): Federal Emergency Management Agency; 2012.
2018;171:105–22. doi: 10.1016/j.engstruct.2018.05.053
18. Babaei S, Zarfam P. Optimization of shape memory alloy
4. Nicolas C. Stronger power: Improving power sector resilience braces for concentrically braced steel braced frames. Open
to natural hazards. Washington: World Bank; 2019. Eng. 2019;9:697–708. doi: 10.1515/eng-2019-0084

5. Merino Vela RJ, Brunesi E, Nascimbene R. Seismic assess- 19. Alibaba. Alibaba Group Holding Limited. 2020. Available
ment of an industrial frame-tank s-ystem: development of from: www.alibaba.com.
fragility functions. Bull Earthq Eng. 2019;17:2569–602. doi:
10.1007/s10518-018-00548-2 20. Cremen G, Seville E, Baker JW. Modeling post-earthquake
business recovery time: An analytical framework.
6. Asgarkhani N, Yakhchalian M, Mohebi B. Evaluation Int J Disaster Risk Reduct. 2020;42:101328. doi:
of approximate methods for estimating residual drift 10.1016/j.ijdrr.2019.101328
demands in BRBFs. Eng Struct. 2020;224:110849. doi:
10.1016/j.engstruct.2020.110849 21. Zerbe RO, Falit-baiamonte A. The Use of Benefit-Cost Anal-
ysis for Evaluation of Performance-Based Earthquake Engi-
7. Christopoulos C, Tremblay R, Kim H-J, Lacerte M. Self- neering Decisions. Peer 2002. Berkeley (CA): Pacific Earth-
Centering Energy Dissipative Bracing System for the Seis- quake Engineering Research Center; 2001.
mic Resistance of Structures: Development and Validation.
J Struct Eng. 2008;134:96–107. doi: 10.1061/(ASCE)0733- 22. AISC 341-16. ANSI/AISC 341-16, Seismic Provisions for
9445(2008)134:1(96) Structural Steel Buildings. Chicago (IL): American Institute
of Steel Construction; 2016.
8. Shu Z, Li S, Zhang J, He M. Optimum seismic design
of a power plant building with pendulum tuned mass 23. AISC 360-14. Specification for Structural Steel Buildings,
damper system by its heavy suspended buckets. Eng Struct. ANSI/AISC 360-16. Chicago (IL): American Institute of Steel
2017;136:114–32. doi: 10.1016/j.engstruct.2017.01.010 Construction; 2010.

9. Kang Y, Peng L, Pan P, Wang H, Xiao G. Seismic perfor- 24. ASCE 7-16. Minimum design loads and associated criteria for
mances of a structure equipped with a large mass ratio buildings and other structures. Reston (VA): American Soci-
multiple tuned mass damper. Struct Des Tall Spec Build. ety of Civil Engineers; 2017. doi: 10.1061/9780784414248
2020;29:e1803. doi: 10.1002/tal.1803 25. Christopulos A. Improved seismic performance of buckling
10. Dai K, Alfanda AM, Wang J, Tesfamariam S, Li T, Sharbati restrained braced frames. Seattle (WA): University of Wash-
R. Comparative benefit-cost analysis for a resilient industrial ington; 2005.
power plant building with isolation system and energy dissi- 26. Prabhu S, Javanbarg M, Lehmann M, Atamturktur S. Multi-
pating devices. J Asian Archit Build Eng. 2023;22:3516–34. peril risk assessment for business downtime of industrial fa-
doi: 10.1080/13467581.2023.2193616 cilities. Nat Hazards. 2019;97:1327–56. doi: 10.1007/s11069-
11. Kumar R, Sharma AKr, Tewari PC. Cost analysis of a coal- 019-03711-3
fired power plant using the NPV method. J Ind Eng Int. 27. EIA. Updated Capital Cost Estimates for Electricity Gen-
2015;11:495–504. doi: 10.1007/s40092-015-0116-8 eration Plants. Washington (DC): US Energy Information
12. Applied Technology Council. Seismic Performance Assess- Administrations; 2018.
ment of Buildings. Volume 2 – Implementation Guide. Vol. 28. Ulmi M. Hazus-MH 2.1 Canada, User and Technical Man-
2. ual: Earthquake Module. Ottawa (ON): Natural Resources
Canada; 2014.

ACADEMIA ENGINEERING 2024, 1 15 of 15