Engineering Structures 285 (2023) 116031

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Superstructure bridge selection based on Life-Cycle cost analysis for short

and medium span bridges
Stefan L. Leiva-Maldonado a, Mark D. Bowman b
a
School of Civil Engineering, Universidad de la Salle, Cra 2 #10-70, Bogota, 111711, Bogota, Colombia
b
Lyles School of Civil Engineering, Purdue University, 550 Stadium Mall Drive, West Lafayette, 47907-2051, Indiana, USA

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

Keywords: Bridge life cycle cost analysis (BLCCA) is a method that evaluates the total cost of a structure. Historically,
Bridge deterministic approaches of the method have been used for the cost-effective selection of maintenance/pre­
Life-Cycle Cost Analysis ventive alternatives for different projects. However, decision makers are inherently risk averse and deterministic
Monte Carlo Simulations
approaches are not strong enough to do a comprehensive informed based selection. Therefore, implementation of
Stochastic Dominance, Superstructure
Selection
stochastic methods could be beneficial for decision makers and structural engineers in the superstructure se­
lection process. Based upon common practice in Indiana, three different span length ranges were selected and
eight different superstructure types both in steel and concrete were evaluated. Substructures were neglected in
the analysis. The analysis was conducted for span configurations that included simple spans, two and three span
continuous girders, for each span range. Cost-effective life-cycle profiles for each superstructure type were used
based on Indiana Department of Transportation experience. Both deterministic and stochastic analyses were
performed. In general, the inclusion of long term-costs using BLCCA reduces the differences in cost-effectiveness
between alternatives for the same span length. Galvanized steel options were found to be cost-effective for
different span lengths, especially for shorter spans, while prestressed bulb tees were generally the best option for
longer spans.

1. Introduction different superstructure options and different span configurations.

Therefore, there is a need to include the LCCA in the decision process to
In everyday professional practice, when a bridge project is idealized, improve the superstructure selection and optimize budget allocations.
the designer must face two different challenges: first, the solution has to The LCCA process applied to bridges is called bridge life-cycle cost
be structurally efficient, and second, it has to be cost-effective. The analysis (BLCCA). The BLCCA method is used to determine the true cost
balance between those two elements results in an optimized solution. In of a bridge structure, which involves the cost to build, inspect and
general terms, structural efficiency has been well covered, giving a wide maintain the structure over the entire lifespan of the bridge. There are
variety of solutions for the same structural problem. However, cost- very few data or prior published studies regarding the life-cycle cost of
effective solutions have often been ignored in this continuing evolu­ entire bridge structures that utilize different materials. A study to
tion. As a rule of thumb, bridge type selection is based primarily upon a evaluate these costs would be useful for efficient and cost-effective
comparison of initial costs, and long-term costs are often neglected in future planning.
this planning stage. Historically, little effort has been made to examine BLCCA has been studied for the past decades with increasing interest
the bridge management issue on a comprehensive basis [1]. Addition­ in the academic and professional community. Therefore, continuous
ally, different authors have focused their efforts on the use of life-cycle advances have been made not only in the understanding of the technique
cost analysis (LCCA) in the analysis of different retrofitting/rehabilita­ itself but also in new approaches to optimize its implementation.
tion/maintenance work actions for specific projects rather than on the Comprehensive literature reviews presented by Frangopol & Soliman
superstructure selection at any stage of planning or life-cycle [2–6]. On [11] and Frangopol et al. [12] not only summarizes the more modern
the other hand, different researchers have addressed the need for the use advances done but also points out gaps that the practitioners and re­
of the LCCA process at the planning stages or in the designing process searches need to address in order to fully implement this method in the
[7–10]. However, no comprehensive analysis has been done comparing decision-making process. Specifically, Frangopol & Solimnan [11]

E-mail address: [email protected] (S.L. Leiva-Maldonado).

https://doi.org/10.1016/j.engstruct.2023.116031
Received 26 May 2022; Received in revised form 19 January 2023; Accepted 18 March 2023
Available online 4 April 2023
0141-0296/© 2023 Elsevier Ltd. All rights reserved.
S.L. Leiva-Maldonado and M.D. Bowman Engineering Structures 285 (2023) 116031

stated that future research efforts should focus on formulating the superstructure types are considered for different span lengths. For con­
necessary tools to implement BLCCA in the hands of decision-makers crete structures, the most common superstructure types within the state
and infrastructure managers. In this sense, this research objective not are slab bridges, prestressed concrete box beams, prestressed concrete
only focus on that gap, but also intend to give a procedure to designers to AASHTO beams, and prestressed concrete bulb tees. For steel structures,
select cost-effective superstructure types during the planning process Indiana typically uses steel rolled beams or steel plate girders, depend­
that could result in agency budget savings, helping the decision-makers ing upon span length. In addition to those, two different systems are
along the way. considered: folded plate girders (FPG) and simply supported steel beams
The inclusion of stochastic processes into the BLCCA is also a topic for dead load and continuous for live load (SDCL). FPGs utilize U-type
that have been studied by different researchers. For example, Kim et al. shapes built by cold-bending flat steel plates into tub sections using a
[13] used probabilistic approaches to optimize Bridge Maintenance press-brake. Folds are uniform but thicknesses and dimensions vary
Management Systems for deteriorating component including service-life depending on project conditions. Concrete is typically cast in the shop to
extensions due to maintenance working actions for concrete bridges connect the folded plates to the deck as part of a prefabricated section.
under corrosion. Additionally, Han & Frangopol [14], uses stochastic The original idea was proposed by [19], and later modifications have
analysis for risk-based life-cycle maintenance strategy for steel struc­ been proposed by different researchers [20–23]. Design recommenda­
tures considering different structural steels. Finally, stochastic and tions by Barth et al. [23] were used in this research. SDCL on the other
probabilistic approaches have been also used for the forecasting of hand, are simple span steel members utilized at the early construction
future condition in deteriorating structures, for the bridge case different stages (dead load only), and then modified by adding the required
authors have proposed different approaches, like probabilistic long-term continuity tension and compression details during construction to create
resilience (Qiang et al. [15]) or Two-step Cluster Analysis (Moscoso et al. a continuous structural system. This structural system eliminates field
[16]). As it can be seen, current approaches are focused on a variety of splices when spans are shorter than transportation limitations. The SDCL
topics related to bridges, however, superstructure bridge selection for method is considered as a special construction process rather than an
planning purposes is still a research topic that needs development. This application of special bridge elements. This option was used in Indiana
research aim is focused in helping to close that gap. in a project structure built in 2016 (on SR 246 over the Eel River in Clay
This research, as described in detail in Leiva-Maldonado [17], is County) with positive opinions not only from the agency but also from
focused on short to medium span bridges (between 9.0 and 40.0 m) the construction contractors as well as addressed by different re­
which represents 65\% of the NBI Indiana bridge inventory. Bridges are searchers [24–27].
categorized in three different groups of span ranges. Range 1 included According to the NBI database, bridges with maximum spans below
bridges with spans within 9.0 m and 18.0 m, range 2 with span lengths 40.0 m represent 98 % of the total Indiana bridge inventory. However,
between 18.0 m and 27.0 m, and range 3 spans with lengths in the range county, and minor road bridges are not considered in the characteriza­
from 27.0 m to 40.0 m. Design types were selected depending on their tion of the bridge designs. Structures with spans shorter than 7.0 m (4.4
cost-effectiveness potential for each of the span ranges. Different su­ %) are considered “culverts” and are out of the scope of this research.
perstructure types are considered for both concrete and steel options. Likewise, bridges between 7.0 m and 9.0 m (14.6 %) use predominantly
Types considered include bulb tees, AASHTO prestressed concrete slab and culvert superstructure types. Consequently, bridges between
beams, slab bridges, prestressed concrete box beams, steel beams, steel 9.0 m and 40.0 m were selected as the span range of interest of this
girders, folded plate girders and simply supported steel beams for dead research.
load and continuous for live load (SDCL). A design plan composed of With the aim of evaluating different design options depending on the
simply supported bridges and continuous spans arrangements was car­ maximum span length, 3 different span ranges were used. Span range 1
ried out. Deterministic analysis for short and medium span bridges in includes bridges with spans within 9.0 m and 18.0 m, span range 2 spans
Indiana based on BLCCA is presented for different span ranges and span between 18.0 m and 27.0 m, and finally span range 3 with spans be­
configurations. In addition to the deterministic analysis, Monte Carlo tween 27.0 m and 40.0 m. Design types were considered depending on
(MC) simulations were used to include the stochastic nature of the their cost-effectiveness potential for each of the span ranges. In addition,
variables considered. Stochastic dominance rules (SD) were used to the NBI database shows that one, two and three span configurations
categorize the most cost-effective superstructure types depending on the comprised 83 % of the concrete and steel bridges used in Indiana.
span length. Findings will help designers to consider the most cost- Moreover, for span ranges 1 and 2, one and three spans arrangements
effective bridge solution for new projects, resulting in cost savings for are the most frequently used span configuration, while for span range 3 a
agencies involved. two-span configuration is the most common alternative. Furthermore,
the span ratios for three span bridges were idealized as exterior spans
2. Superstructure types and design equivalent to 80 % the mid-span length, while two span configurations
were modelled as equal span bridges. Using this trend, the design plan
The first step in a BLCCA is to determine which alternatives to utilized simple and three-span structures for span ranges 1 and 2, and
compare. Different superstructure solutions are available depending not simple and continuous two-span bridges for span range 3. The summary
only on the span length but also on the span configuration and material of the design matrix can be seen in Table 1.
type. In this study, features such as material type, span length, span To simplify the design process, a single cross section was assumed
configuration, and common superstructure types in the state of Indiana common for all the superstructure types. Two 3.65 m lanes in opposite
were obtained from the National Bridge Inventory (NBI) database and a directions are considered along with 2.45 m shoulders on each side of
contractor database provided by the Indiana Department of Trans­ the bridge, giving a total bridge width equal to 13.10 m including the
portation (INDOT). Due to privacy of the information used internally by barrier railings. Fig. 1 presents the typical cross section of all bridges
INDOT the contractors database can not be accessed freely. However, designed; this particular view depicts the configuration for five pre­
the summary of it can be accessed in Leiva-Maldonado [17] and stressed concrete bulb tee beams.
Bowman & Leiva-Maldonado [18]. To summarize, the databases showed Designs were then developed for each of the superstructure types and
that more than 90 % of bridges within the state are either concrete or span length values shown in Table 1 for the cross section depicted in
structural steel. Among them, approximately-two-thirds of the bridges Fig. 1. A total of 60 structural designs were developed that, including all
were concrete or prestressed concrete (distributed almost evenly), while different variations, totals 109 different alternatives. The designs were
27 % were structural steel. developed to satisfy the requirements in the AASHTO LFRD Bridge
Superstructure types designed were selected according to their span Design Specification [28] and the Indiana Design Manual [29]. Design
range applicability and possible cost-effectiveness. Eight different combination used includes strength I, service I and fatigue I for the

2
S.L. Leiva-Maldonado and M.D. Bowman Engineering Structures 285 (2023) 116031

Table 1 Table 2
Bridge Design Matrix. Design standard values.
Superstructure Type Span Length* (m) Item Unit Value

Simply Continuous (1) (2) (3)

Supported Bridge Railing – Concrete Railing1
Skew degree 0.02
3 Spans 2 Spans
AADT – Moderate3
(1) (2) (3) (4) Deck Thick. m 0.2
Slab Bridge 9.0, 13.5 9.0, 13.5 – Min. Reinf. mm 15.9
Prestressed Concrete Box 9.0, 13.5, 18.0 9.0, 13.5, – Max. Reb. Space. m 0.2
18.0 Structural Steel – ASTM A709 gr.345
Prestressed Concrete AASTHO 9.0, 13.5, 18.0 13.5, 18.0, – Reinf. Steel – AASHTO A615 gr.420
Beam** 22.5 PS Strands. – Low relax. strands
22.5, 27.0 27 Slab Concrete MPa 27.5
Structural Steel Beam (5 9.0, 13.5, 18.0 13.5, 18.0 – PS Concrete Beams MPa 48
beams)*** 1
22.5, 27.0
INDOT Standard Drawing No. E 706-BRSF-01.
2
Structural Steel Beam (4 9.0, 13.5, 18.0 13.5, 18.0 – Average skew 20\textdegree.
3
beams)*** Average truck traffic values below 3500 trucks per day that are represen­
22.5, 27.0 tative of the majority of bridges in Indiana.
Folded Steel Plate 9.0, 13.5, 18.0 – –
Structural Steel SDCL Beams*** 18.0, 22.5,
as traffic, weather, location, etc. Deterministic and stochastic models
– –
27.0
Prestressed Concrete Bulb Tee 18.0, 22.5, 27.0 18.0, 22.5, 27.0, 33.5, have been used to asses the deterioration of different structures. Three
Beams** 27.0 40.0 different deterioration models made specifically for the Indiana bridge
33.5, 40.0 inventory were considered, Sinha et al. [31] (deterministic model),
Structural Steel Plate Girders*** 27.0, 33.5, 40.0 27.0 27.0, 33.5,
Moomen et al. [32] (deterministic and probabilistic model) and Cha
40.0
et al. [33] (constant rate model). Deterioration rates vary depending on
*
For continuous arrangements corresponds to the maximum span length. the database and method used to compute it. Nonetheless, it is clear that
**
Design used for options with and without support diaphragms. deterioration rate is time dependent. For steel superstructure deterio­
***
Design used for painted and galvanized options.
ration curves proposed by Sinha et al. [31] appear to fit better the his­
torical data, Equation 1 shows the model used.
structural steel alternatives, and strength I, service I and service III
(prestressed concrete only) for concrete alternatives. To verify the val­ ER = 1.70219 +
51.82974
(1)
idity of the designs process utilized, some of the bridge designs were 7.03877 + 0.00859*year1.70602
compared with actual designs developed by independent consultants for For deck behavior Moomen et al. [32] indicates that superstructure
INDOT. Table 2 lists the properties used for the concrete, prestressed material type is not a factor that affects the deterioration behavior. The
concrete and structural steel sections. The table also highlights other service life proposed by this approach is close to 37 years. The likelihood
pertinent properties used in the design. The suitability and efficiency of of programming a deck replacement at a much greater service life is low
the designs were bench marked by comparing the designs obtained with according to actual data and INDOT experience, and it is often scheduled
other recent designs on INDOT projects with comparable spans and between 30 and 40 years. This means the deterministic method can be
design features. used reliably. Equation 2 shows the model used, where SERVICEUNDER
corresponds to 0 for crossings different than waterways, SPANNO could
3. Deterioration models be 1 to 3 spans (do not affect greatly the outcome for service-life),
DECKPROT as 0 (no deck protection considered), FRZINDX as 1 (for
After obtaining the design matrix, the second step needed to develop northern Indiana regions) and finally, ADTT for a heavy traffic roadway
the BLCCA involves the definition of the deterioration models for each with 10.000 trucks per day.
superstructure type. Historically, the NBI condition rating database is
commonly used to determine the effects of aging in bridge structures. ER =8.55637 − 0.24129*AGE + 0.0096*AGE2 − 0.0001667*AGE3
Condition ratings are described as a numerical representation of the − 0.04301*SERVICEUNDER − 0.01218*SPANNO (2)
current physical state of the asset that ranges from 9 to 0 (from excellent + 0.51375*DECKPROT − 0.05182*FRZINDX − 0.01872*ADTT
to failed condition); more detailed information can be found in the
documentation provided by the Federal Highway Administration [30]. Threshold rating age for different concrete superstructure types lies
There are multiple approaches that have been used to determine the between 55 and 65 years not only for the curve fitting approach but also
deterioration rate of a given structure due to many different factors such for the constant deterioration rate method. INDOT experience indicates

Fig. 1. Typical bridge cross section.

3
S.L. Leiva-Maldonado and M.D. Bowman Engineering Structures 285 (2023) 116031

that is unlikely to have a concrete superstructure older than 70 years comprehensive life cycle cost profiles can be proposed and later the
without any rehabilitation or repair. Deterioration models proposed by BLCCA can be obtained.
Moomen et al. [32], appear to better reflect the common practices in Additionally, options with integral abutments for prestressed con­
Indiana for concrete superstructures. Equations 3 to 5 shows the models crete bridges were also considered. Deterioration of prestressed mem­
used, from concrete slabs, PS concrete box and PS concrete Beams bers at the bridge beam ends were commonly observed by INDOT
respectively. Where in addition to the variables described for decks, INT engineers and is believed to be the reason for shortened service lives;
corresponds to 1 for highway bridges, SKEW is 0 degrees, NRFTC as slightly higher service lives were observed for steel beam ends. The use
number of Freeze-Thaw Cycles (1), and LENGTH depended on the total of integral abutments eliminates the joint at the bridge ends and is
length and span configuration. believed to be effective in increasing the overall bridge service life. It
was assumed that inclusion of those abutments will increase the service
ER =9.5820 − 0.27195*AGE + 0.00874*AGE2 − 0.0000933*AGE3
life of the bridge to be at least comparable with the service life of steel
− 0.1991*INT − 0.17981*SERVICEUNDER − 0.71169*FRZINDX structures. In other words, an increase of service life of 15 years. Like­
(3) wise, SDCL system service life is also extended in the same proportion
since the system itself is based on the same principle of integral abut­
ER = EXP(2.52216 − 0.01574*AGE − 0.21057*INT − 0.00628*NRFTC) ments and intermediate pier diaphragms, extending its service life up to
(4) 95 years.
The use of corrosion protection systems other than paint-based
ER =9.67048 − 0.03574*AGE − 0.00076366*AGE2 − 0.12316*INT coatings could also increase the service life of steel elements signifi­
− 0.00089223*LENGTH − 0.66583*FRZINDX − 0.0178*NRFTC cantly. According to the American Galvanizers Association [35], for
(5) suburban environments a zinc average thickness of 4.0 mils or more
could extend the service life of the initial coating up to 100 years or
Then, the deterioration models from two different research studies
more, representing an extension of the service life of 20 years compared
for Indiana bridges were selected. For steel superstructures the model
with the painted elements. Moreover, such a system would eliminate the
developed by Sinha et al. [31] was selected (deterministic approach),
need for periodic re-coating. Variation of galvanization thicknesses and
while for concrete structures and concrete decks the models by Moomen
processes could significantly impact the service life and working action
et al. [32] were chosen (deterministic approach used to be comparable
distribution. These changes and additional considerations can be
with the steel deterioration model). These models are presented in
addressed in future works focused on this problem. Accordingly, an
Fig. 2. It is important to mention that among all the different variables
equivalent extension in the service life is considered for the SDCL
that could impact the deterioration rates, the environmental conditions
galvanized option with integral end abutments, improving its service life
can be impactful, and its watchful consideration should be addressed as
to 115 years.
appropriate.
Based on the deterioration curves presented, the service life of each
4. Life-Cycle cost profiles
superstructure type can be obtained. The service life is computed as the
expected bridge age at which the structure reaches the lower limiting
The life-cycle cost profile (LCCP) is a graphical representation of all
condition of serviceability. This threshold could vary depending on the
costs involved in the service life of a given structure. Major working
budget allocation and availability. For the State of Indiana,”poor con­
actions such as the reconstruction of elements, overlays, deck re­
dition”, or condition rating 4 in the NBI database, is considered as the
placements among others, along with routine or regular working actions
minimum allowable rating for a bridge to be considered as safe
due to the aging of the structure, are captured in those profiles. A
(assumption made by different studies including one made by Nasrol­
complete BLCCA is composed of the unique set of working actions
lahhi and Washer [34]. Consequently, the service-life for different
needed for each of the major components that configures a bridge,
concrete superstructure types are given as follows: 58 years for concrete
which includes the deck, superstructure, and substructure. However,
slabs, 60 years for prestressed concrete box beams, and 65 years for “I”
this research is focused on the superstructure only. Generalization of soil
shaped prestressed concrete beams. In contrast, the projected service-
and substructures in general throughout Indiana is considered an over­
life for steel structures is 80 years. Based on these services lives,
simplification and is not within the scope of this research.
Multiple factors are involved in the process of creating a LCCP.
Features such as superstructure type, scheduling related to the service
life of the working actions considered, possible extensions of the service
life due to preventive procedures, among others, impact the BLCCA
result to some extent. Each one of these factors and its variations creates
different alternatives and therefore different LCCPs to consider in
BLCCA.
Working actions related to the superstructure often involve the full
or partial intervention of the deck. Therefore, LCCP proposed herein are
a combination of preventive / maintenance / repair / rehabilitation
strategies of both elements. As a consequence, preventive and mainte­
nance activities for decks must be considered in the life-cycle of the
superstructure. Working actions included in the LCCP are cleaning and
washing of the deck surface (WC), deck crack sealing (SC), deck overlays
(O), bridge demolition (BD), deck replacement (DR), bearing replace­
ment (BR), spot painting of steel structures − 10 % of exposed surface-
(SP), and salvage recycling value (SR). Table 3 presents the LCCP’s used
in this study depending on the superstructure type and material. It is
important to point out that aspects such as the increasing the concrete
cover in concrete structures can improve the service life and possibly the
cost effectiveness of different working actions. However, such changes
Fig. 2. Deterioration curves. are not considered in current INDOT practices. Future works supported

4
S.L. Leiva-Maldonado and M.D. Bowman Engineering Structures 285 (2023) 116031

Table 3
Cost-Effective LCCP.
Group I.D. SL Working action (years of occurrence)

(yrs) WCa SCb O BDd DR BR SPc SRd

Conc. Slab SB 58 Y 5 40 SL – – – –
PS Concrete ABe or BTf or CBg 65 Y 5 – SL 40 40 – –
PSC Diap. ABDe or BTDf 80 Y 5 – SL 40 – – –
Steel Paint SBXPh,i or SPGXPh,j 80 Y 5 – SL 40 40 10 SL
Steel Galv. SBXGh,i or SPGXGh,j or FPGXh,k 100 Y 5 – SL 50 50 – SL
SDCL Pnt. SDCLXPh,l 95 Y 5 – SL 50 – 10 SL
SDCL Galv. SDCLXGh,l 115 Y 5 – SL 40,80 – – SL
a
Working action performed on a yearly basis.
b
Working action performed every 5 years until the end of the service life.
c
Working action performed every 10 years until the end of the service life.
d
Performed once at the end of the service life.
e
For Prestressed Concrete AASHTO Beams use I.D. “AB” or “ABD” if diaphragms are used.
f
For Prestressed Concrete Bulb Tee Beams use I.D. “BT” or “BTD” if diaphragms are used.
g
For Prestressed Concrete Box Beams use I.D. “CB”.
h
“X” denotes the number of structural beams used.
i
For Structural Steel Beams use I.D. “SBXP” or “SBXG” if galvanized.
j
For Structural Steel Plate Girders use I.D. “SPGXP” or “SPGXG” if galvanized.
k
For Structural Steel Folded Plate Girders use I.D. “FPGX”.
l
For Structural Steel SDCL Beams use I.D. “SDCLXP” or “SDCLXG” if galvanized.

by field and lab testing can be done to capture the impact of those Transportation (INDOT) has been collecting information on highway
recommendations. construction projects from 2011 to 2017 and this database was used to
It should be noted that the LCCP options shown in Table 3 are compute the current value of different pay items involved in bridge
believed to be the optimal LCCP for each bridge type. Multiple LCCP construction. Inflation rates were calculated using the current consumer
options were considered for each of the different structure types. Sto­ price index (CPI) published monthly by the Bureau of Labor Statistics
chastic analysis done for the LCCP selection can be found in Leiva (BLS).
Maldonado [17]. The introduction of regular deck crack sealing (SC) One of the most relevant steps when unit costs and bills of quantities
was believed to extend the deck life and eliminated the need for inter­ are used is the proper identification of outliers. Inclusion of those values
mediate bridge deck overlays (O) prior to a full deck replacement (DR). could lead to wrong interpretation of results or misleading analyses. The
Refer to Leiva Maldonado & Bowman [18] for more information on outlier identification was done using interquartile ranges (IQR) and the
various LCCP options evaluated. Tukey’s fence approximation. In order to take into account, the eco­
nomics of size of the projects, a weighted average and standard devia­
5. Cost allocation tion are chosen to use as an input in the deterministic approach of
BLCCA. Values not only for construction costs but also maintenance
As mentioned before, BLCCA involves the sum of the costs of all working actions can be consulted in the work done by Leiva Maldonado
different work actions needed during the service-life of a given bridge. [17].
The total cost, then, is the sum of the agency costs - which includes On the other hand, the stochastic nature of the unit costs can be
design costs (DC), construction costs (CC), maintenance costs (MC), considered by using a probability distribution of such events using the
rehabilitation costs (RC) salvage costs (SV), and user costs (UC), as historical data. The Kolmogorov-Smirnov (K-S) and Anderson-Darling
described in Equation 6. Unless there is a reason to do otherwise, agency (A-D) tests for probability Density Function (PDF) selection were used
costs are typically assumed to be incurred at the end of the period in for the unit costs, as well as an adequate visual inspection. In order to
which expenditures actually will occur [36]. The BLLCA is dependent on properly select the adequate distribution, both tests are considered.
the discount rate (i) and the service life (SL) of each option, while each of Then, a weight is given to each distribution depending on its ranked
the cost components are dependent on the discount rate (i) and the time position for each, the K-S and the A-D tests. Numeric values from 1 to 6
of execution of each working action (t). are given in descending order from the best to the worst option. In case
that the logical hypothesis result rejects the null hypothesis, a weight
BLCCA(i, SL) = CD(i, t) + CC(i, t) + MC(i, t) + RC(i, t) + UC(i, t) + SV(i, t)
value of 0 is assigned. Then, an ordered weight is computed as the in­
(6) verse of the sum of both weights. The distribution with the lowest or­
To estimate the user costs, one needs to obtain detailed site-specific dered weight is then selected as the distribution that best fits the pay
information such as traffic data and clearance limitations. However, all item historical data, detailed procedure and results can be consulted in
bridge designs in this study are hypothetical and have no specific loca­ Leiva-Maldonado [17]. In addition, PDF selection for different pay items
tion along any specific road. Assumptions regarding traffic, velocity and is included as supplemental information to this document.
detours are considered an oversimplification of the problem and could According to the literature (Sun [37]), if no historical data is avail­
negatively impact the outcome of the BLCCA comparison. As a recom­ able the use of triangular distribution or uniform distribution is rec­
mendation to practitioners, it is important to underline that user costs ommended. On the other hand, Peters [38], estates that uniform,
need to be included if specifics on the project location are known. triangular, PERT and Beta distributions can be used for the same case.
Additionally, and as shown later in the study, inclusion of UC will not However, the use of PERT distribution is a good solution because of its
make any change in the superstructure selection process. The most ease-of-use and greater flexibility compared with the uniform distribu­
widely used basis to estimate construction, rehabilitation and mainte­ tion. Additionally, recognizing that the Beta distribution is more flexible
nance costs are the utilization of unit costs and bills of quantities. In the than the PERT, the major challenges of the first one is the practicality of
absence of this information, parametric cost estimating models may be use. Then, the PERT distribution is advised for cases with non-historical
used as a best-guess estimate (Hawk, 2002). The Indiana Department of data available. Pay items without historical records (e.g. deck surface

5
S.L. Leiva-Maldonado and M.D. Bowman Engineering Structures 285 (2023) 116031

washing) are modelled as a PERT distribution with most likely value BLCCA works and how it can be implemented. The consensus among
(ML) equal to the unit price obtained using a weighted average, a DOT decision makers is that BLCCA can be fully implemented only using
maximum value (max) of 1.25 times the ML, and a minimum value (min) deterministic approaches, stochastic analysis are often neglected. The
equivalent to 0.75 times the ML. Table \ref{Table 4.} shows the corre­ use of PDF selection, Monte Carlo Simulations and the usage to help the
sponding PDF’s selected for each pay item used according to the LCCP decision-making process needs to be introduced in order to have more
matrix presented, with their respective distribution parameters. informed decision that could led to budget savings.
Finally, some PDF’s have heavier tails than others, which means that It is important to clarify that all analyses and alternative cost con­
when a simulation uses a random number generator those values could siderations are made in constant dollars as is commonly done for eco­
be easily chosen for a number of iterations. However, according to the nomic analysis. Inflation rates will not be considered on the assumption
historical data, extreme values are unlikely to happen. Consequently, for that all costs and benefits are affected equally by inflation [39]. How­
simulation purposes only, a range between 0.75 times the minimum ever, if it is considered that the inflation will affect the future costs
value and 1.25 times the maximum value in the unit coat data set was differently from a given alternative, such adjustment needs to be made
used. Values outside these boundaries were taken as the 0.75 times the accordingly [40].
minimum and 1.25 times the maximum values, respectively.

6.2. Deterministic analysis

6. Bridge Life-Cycle cost analysis

One of the most important aspects, when conducting a BLCCA, is to

6.1. Life-Cycle cost profiles
establish a comparable service life for all alternatives. If two alternatives
with different service lives are to be compared, the least common mul­
Deterioration curves will be used to predict the maintenance, reha­
tiple of the two estimated service lives of the two alternatives must be
bilitation, and reconstruction scheduling for each of the design options
used [41]. Consequently, analyses include numerous working action
considered. Once an element reaches the threshold for each condition, a
repetitions. In lieu of this, it is assumed that it is likely to replace the
jump in the condition rating will be assumed and the deterioration af­
structure indefinitely or in perpetuity. This assumption is based on the
terward will follow the correspondent curve. Final deterioration profiles
fact that a long service-life asset like bridges is most likely to be replaced
will be used to allocate costs during the BLCCA process.
in the same location over and over again, rather than in a different
There are different challenges during implementation of the BLCCA
location. This factor implies that the life cycle is recurrent, independent
process. Two examples are the inability to set a common analysis period
of the structure type used. Consequently, the life cycle of a given bridge
for all the superstructure types and the reliability of the prediction of
can be considered as a cost in perpetuity and its present cost can be
future costs. For this reason, the BLCCA must not be considered as a
computed under this premise.
precise tool to allocate budgets but as a helpful instrument to provide
Fwa [42] and Ford et al. [43] present different methods to compute
better information to decision-makers. However, the most important
the present worth of life cycle cost in perpetuity based on if the life-cycle
issue is attributed to the lack of transparency and full knowledge of how
is considered at the beginning or at the end of the service life

Table 4
Probability distribution functions for different pay items.
Item Unit Data Min Max Distr. Dist. Parameters

Parameter Value

Concrete Type C /m3 354 270.84 687.15 Gamma Shape (κ) 4.81
Scale (θ) 40.62
Shift (ω) 255.4
P/S Concrete Bulb Tee /m3 132 0.75 1.91 Logistic Location (μ) 1.22
Scale (σ) 0.11
3
P/S Concrete Box /m 13 4.06E-07 8.51E-07 Lognormal Mean (λ) − 14.35
STD (ζ) 0.22
3
P/S Concrete Beam /m 55 1.14 3.55 Lognormal Mean (λ) 0.58
STD (ζ) 0.2
Structural Steel /kg 63 0.29 1.36 Normal Mean (μ) 0.89
STD (σ) 0.24
Reinforcing Steel /kg 150 0.3 0.61 Gamma Shape (κ) 35.28
Scale (θ) 0.01
Overlay
Overlay /m2 226 5.05 13.42 Gamma Shape (κ) 21.6
Scale (θ) 0.41
Overlay Removal /m2 121 0.15 1.59 Weibull Shape (κ) 0.98
Scale (λ) 2.9
Hydrodemolition /m2 212 1.66 13.02 Weibull Shape (κ) 2.79
Scale (λ) 6.72
2
Overlay Additional /m 263 1.53 2.55 PERT Mean (ML) 2.04
Deck Cleaning /m2 – 0.15 0.25 PERT Mean (ML) 0.2
Sealing of cracks /m2 – 0.09 0.21 PERT Mean (ML) 0.12
Bridge Removal /m2 – 0.77 1.29 PERT Mean (ML) 1.03
Deck Reconstruction
Deck Reconstruction /m3 65 290.22 818.04 Inv. Gauss Mean (μ) 534.78
Shape (λ) 10,860
2
Structure Removal /m 144 1.05 4.55 Lognormal Mean (λ) 0.67
STD (ζ) 0.37
Spot Painting /m2 – 0.15 0.25 PERT Mean (ML) 0.2
Recycle Struc. Steel /kg – 0.03 0.05 PERT Mean (ML) 0.04
Galvanizing /m2 – 0.02 0.034 PERT Mean (ML) 0.027
Discount rate % – 2 6 PERT Mean (ML) 4

6
S.L. Leiva-Maldonado and M.D. Bowman Engineering Structures 285 (2023) 116031

Fig. 3. Cost-effectiveness for simply supported beams - Deterministic Approach (i = 4 %).

7
S.L. Leiva-Maldonado and M.D. Bowman Engineering Structures 285 (2023) 116031

Fig. 4. Cost-effectiveness for continuous beams - Deterministic Approach (i = 4 %).

8
S.L. Leiva-Maldonado and M.D. Bowman Engineering Structures 285 (2023) 116031

respectively. Equation 7 shows Ford’s alternative, where Pp is the pre­ costs are considered, bulb tees would be the optimal solution for the
sent worth of BLCCA in perpetuity, P is the life cycle cost of a single entire span range.
service life at the beginning of the service life, i is the discount rate used Fig. 4 shows the results for continuous beams. For span range 1, slab
and SL is the service life in years of each option. Using this equation, it is bridges and galvanized steel continuous beams are the most cost-
possible to compare different alternatives with different service lives in effective solutions for the two halves of the span range, respectively.
terms of life-cycle costs. However, prestressed concrete AASHTO beams are also a competitive
option for spans between 13.5 and 18 m. Additionally, for span range 2,
P(1 + i)SL 4 galvanized rolled beams are still cost-effective for spans shorter than
Pp = (7)
(1 + i)SL − 1 21 m. For longer spans including the entire span range 3, no variance in
Historically, discount rates for highway assets have fluctuated from the cost-effectiveness of the bulb tee option is noticed between the initial
4 % to 6 % [44]. The Indiana Design Manual [29] recommends a 4 % cost comparison and the inclusion of long-term costs, although the cost
discount rate be assumed for several purposes. This is the value used differential is notably reduced.
herein for the deterministic comparisons made. It is important to underline the fact that results shown are not a
Equally important is the evaluation of the initial cost for each precise measurement of cost-effectiveness. Rather, they are an approx­
alternative. As stated before, only superstructure elements are consid­ imation and the first approach to designers at the moment of bridge
ered. Depending on the superstructure type, different elements must be planning. This tool could clarify which super-structure options could be
considered in the determination of the initial cost of each alternative. cost-effective during the planning process. However, final site condi­
However, as a generalization, these are the main elements considered in tions and project level cost estimations should represent accurately the
this computation: concrete deck material, reinforcement steel, bearing best option for construction [17].
pads, concrete diaphragms, concrete beam elements, and structural The FPG option could be considered as a cost-effective solution
steel. Railing, as well as other miscellaneous elements, were not depending on the span length of the structure. Regardless of the results
considered on the basis that is common for all the alternatives and they shown in this document, a more accurate cost estimation of construction
will not impact the final comparison. cost, not only for steel elements but also for prefabricated composite
Initial cost comparisons, as well as BLCCA, were made for every modules is needed. Such analysis needs to be done separately to
superstructure type considered. Results shown correspond to the ratio demonstrate the cost-effectiveness viability of this system. It is impor­
(Effectiveness Ratio -ER-) between the cost of the option analysed, Calt , tant to remark that the conclusions shown regarding this alternative do
and most cost-effective among all the alternatives for a given span length not disregard its structural usage, such behavior is cover in detail by
as shown in Equation 8 proposed by the authors. ER values shown other authors.
correspond to the ratio between the option analyzed and the lowest price
among all the alternatives for a given span length, it means that values 6.3. Stochastic analysis
equal to 1 for a given span length are considered as the cost-effective
option among all the superstructures considered. Deterministic analysis could be useful as a first approximation to
assess the superstructure selection. However, as seen in Fig. 3 and Fig. 4
ER =
Calt
(8) the ER ratio for BLCCA difference between different alternatives is often
min(CAlt1 , CAlt2 , ⋯, CAlti ) minimal and therefore a preference on superstructure type may not be
Fig. 3 and Fig. 4 illustrate the superstructure categorization for clear. Additionally, only a single case is examined since average values
different span ranges. X-axis shows the maximum bridge span length alone are used to compute the BLCCA. Nonetheless, there are multiple
and the Y-axis the effectiveness ratio. In general, the superstructure other combinations that can be analyzed depending on the different
option for a given span length that has an ER equal to one, is considered working actions considered, variation in costs among others. Hence­
as the most cost-effective option according to the deterministic analysis. forth, a Monte Carlo Simulation (MCS) is used to assess the probabilistic
Fig. 3 shows a comparison of the initial cost and BLCCA for simply nature of all the variables involved in the BLCCA of each alternative.
supported beams for all span ranges using the deterministic approach. As defined by Ang and Tang [45], “simulation is the process of
As a conclusion drawn from in the figures, the use of BLCCA reduces the replicating the real world based on a set of assumptions and conceived
difference between all the alternatives due to the inclusion of long-term models of reality”. When the problem is based on random variables with
costs for the same span length. In the case of span range 1, the slab assumed or known probability distributions, the Monte Carlo simulation
bridge alternative is the most cost-effective solution considering or not (MCS) method is very useful. Generally speaking, the MCS method is a
considering long-term costs for spans less than 11.5 m. On the other repetition of many different simulations, using a different set of values of
hand, for spans longer than 11.5 m, the inclusion the four (4) beam the random variable, based on the corresponding probability distribu­
galvanized configuration option is the most cost-effective. Conversely, tion. Deterministic analysis on the other hand, is then considered as a
when only initial costs are considered, painted rolled beams and pre­ single simulation process.
stressed concrete AASHTO beams would be the preferable options. The MCS was performed using the Microsoft excel complementary
Additionally, the FPG option is not the most cost-effective however, is software called @RISK, developed by Palisade software company. The
among the optimal alternatives for second half of the span range. simulations used in this document use a random number generator
In addition, for spans shorter than 20 m (span range 2), 4 galvanized based on Mersenne Twister pseudo random numbers, Latin hypercube
rolled beams are still cost-effective, however, prestressed concrete bulb sampling and 10,000 iterations per simulation. As a result, a cumulative
tees are the optimal solution for longer spans. Again, if an analysis with density function (CDF) can be obtained using the results of each set of
only initial cost is made, prestressed concrete bulb tees would be iterations.
selected for the entire span range. This trend is attributed to the lower Using those CDF’s, it is possible to determine the preference for a
material and fabrication costs and, resistance optimization achieved by specific alternative using the concept of Stochastic Dominance (SD).
the bulb tee system. First and second stochastic dominance (FSD - ≻1 - and SSD - ≻2 -
Span range 3 results when long-term costs are considered can be respectively) along with Almost first and Almost second stochastic
divided in two different trends. Firstly, for spans up to 33.5 m two dominance AFSD - ≻almost
1 - and ASSD - ≻almost
2 - respectively) rules were
optimal options can be identified (i.e., prestressed concrete bulb tees used to rank the alternatives based on the BLCCA.
and galvanized steel plate girders). Secondly, bulb tees are the most cost- In general terms, SD rules are based on comparisons between CDF’s.
effective solution for longer spans. Again, as in span range 2, if only first FSD makes a direct comparison between them, while SSD uses an indi­
rect comparison based on the areas under the CDF’s compared. Both

9
S.L. Leiva-Maldonado and M.D. Bowman Engineering Structures 285 (2023) 116031

will be the alternative that is dominated by all the other alternatives,

and the inefficient set the complementary options. Additionally, if an
alternative (A) is dominated by another (B), no matter the degree of
dominance, it is considered that A is preferred to B, A≻B. However, if the
result does not conclude dominance, it is considered that there is no
preference between the alternatives and the decision maker is indif­
ferent between the options, A ~ B.
Due to the multiple comparisons needed for each of the span lengths,
a graphical representation for each case is insufficient to determine
accurately the preference of any alternative among the others, see ex­
amples in Fig. 5. Fig. 5a presents the results of continuous beams with a
maximum span of 9.0 m. As it can be seen, slab bridges (SB) are clearly
dominated by the prestressed concrete box beam (CB) alternatives. In
other words, the SB alternatives provides a lower cost (LCCAP) at any
given density value of the CDF than the CB alternative. On the other
hand, Fig. 5b shows the results of simply supported beams with spans of
33.5 m, where it can be seen that no clear dominance of any alternative
(prestressed bulb tee (BT), prestressed bulb tee with diaphragm (BTD), 5
beams painted steel girder (SPG5P) and 5 beams galvanized steel girder
(SPG5G)) can be observed using the graphical representation. Sensi­
tivity analysis can be found in Leiva Maldonado [17].
The comparison in Fig. 5b is now further examined using the Sto­
chastic Dominance (SD) noted before. A summary table showing the four
SD values for each alternative is given in Table 5. In the table, a sto­
chastic matrix selection is presented. Each cell shows a set of logical
values composed of 4 figures, namely, first, second, almost first and
almost second stochastic dominance. As a convention, 0 indicates not
dominance of the row option to the column option contrasted, while 1
means dominance in any degree of the row alternative to the column
option. For example, for row 4 of Table 5, galvanized 5 steel plate girders
(SPG5G), and column 1, prestressed concrete bulb tee beams (BT), the
logical output is ’’0–0-1–1′′ , meaning that SPG5G almost dominates in
first degree BT, BT≻almost
1 SPG5G. In other words, BT is more cost-
effective than SPG5G and is preferable for that specific span length.
The final objective is to find a column that is dominated by every row
in the stochastic dominance matrix. In the previous case, since the
prestressed bulb tee alternative (BT) is dominated by all the other op­
tions, then it is the most cost-effective option for simply supported
beams with a 33.5-m longer span (i.e. BT ≻almost1 SPG5G ≻almost
1 SPG5P
almost
≻2 BTD). As a result of the analysis of the simulation results and the
summarization of the stochastic dominance matrices, the superstructure
selection for each case is summarized as follows:

Fig. 5. Simulation results, CDFs. • Simply Supported Beams:

• Span = 9.0 m: SB ≻almost 1 SB4G ≻almost
1 FPG4 ≻2 SB5G ≻1
methods assumed that the rules are true by all the values of the range. In SB4P ≻almost
1 FPG6 ≻ 2 SB5P ≻1 AB ≻2 ABD ≻1 CB
contrast, AFSD and ASSD are special cases of the first two rules assuming • Span = 13.5 m: FPG4 ≻1 SB4G ≻almost 1 SB5G ≻almost
1
almost
that the rule itself is true for “most” of the CDF range. Examples of each SB4P ≻1 SB5P ≻2 AB ≻2 SB ≻2 ABD ≻1 CB
rule made for this research can be found in Leiva Maldonado [17]. • Span = 18.0 m: SB4G ≻almost 1 FPG4 ≻almost
1 SB5G ≻almost
1
almost almost
Mathematical derivations and statistical justification can be found in SB4P ≻1 SB5P ≻2 BT ≻2 BTD ≻1 CB ≻almost
1 AB ≻2 ABD
Levy [46]. • Span = 22.5 m: BT ≻1 BTD ≻almost 1 AB ≻almost
1 SB4G ≻almost
1 SB4P ≻2
When decision makers only have partial information available, and ABD ≻almost
1 SB5G ~ SB5P
need to select an alternative, the process is known as partial ordering, • Span = 27.0 m: BT ~ SPG5G ≻almost 1 SPG5P ≻almost
2 BTD ≻almost
1 SB4P
which is the case of the Stochastic Dominance (SD). For this study, the ≻almost
1 SB4G ≻ almost
1 AB ≻ almost
1 SB5P ≻ 2 SB5G ≻ 2 ABD
objective is to find the alternative with the lower price, i.e., the alter­ • Span = 33.5 m: BT ≻almost 1 SPG5G ≻almost
1 SPG5P ≻2 BTD
almost
native with a higher probability to have a lower price. Henceforth, the • Span = 40.0 m: BT ≻1 BTD ≻almost
1 SPG5P ≻2 SPG5G
decision rule used to divide the feasible set is such that the efficient set • Continuous Beams
• Three spans, longer span = 9.0 m: SB ≻1 CB
• Three spans, longer span = 13.5 m: SB4G ≻almost 1 ABD ≻almost
1 AB
Table 5 ≻almost SB5G ≻almost SB ≻almost SB4P ≻almost SB5P ≻1 CB
1 1 1 1
Stochastic dominance matrix - Simply supported beams, span = 33.5 m.
• Three spans, longer span = 18.0 m: SB4G ≻almost 1
Alternative BT BTD SPG5P SPG5G SDCL5P ≻almost
1 SDCL4P ~ AB ≻almost 1 ABD ≻almost
1 SB4P ≻almost
1 SB5G
almost almost almost almost
BT – 0–0–0–0 0–0–0–0 0–0–0–0 ≻1 SB5P ≻1 BT ≻1 BTD ≻1 SDCL5G ≻almost
1 SDCL4G
BTD 0–1–1–1 – 0–1–1–1 1–1–1–1 ≻1 CB
SPG5P 0–0-1–1 0–0–0–0 – 0–0-1–1
SPG5G 0–0-1–1 0–0–0–0 0–0–0–0 –

10
S.L. Leiva-Maldonado and M.D. Bowman Engineering Structures 285 (2023) 116031

• Three spans, longer span = 22.5 m: BT ≻almost1 BTD ≻almost

1 structural steel galvanized structures dominate the span lengths be­
AB ≻almost
1 SDCL4P ≻ almost
1 ABD ≻ almost
1 SDCL5P ≻1 tween 11 m and 22.5 m for both span distributions.
SDCL4G ≻almost
1 SDCL5G Based on the results obtained through stochastic dominance, it is also
• Three spans, longer span = 27.0 m: BT ≻almost1 BTD ≻almost
1 clear that the most cost-effective options implicitly consider the least
AB ≻almost
1 ABD ≻almost
1 SPG5G ≻almost
1 SPG5P ≻almost
1 number of working actions during their service lives. Specifically, in all
SDCL4P ≻almost
1 SDCL5P ≻1 SDCL4G ≻almost1 SDCL5G the span lengths where structural steel options are the most cost effec­
• Two spans, equal spans = 27.0 m: BT ≻almost 1 BTD ≻almost
1 SPG5G tive, galvanized options are preferred among the painted options. This
≻almost
1 SPG5P finding corroborates indirectly the assumptions of not considering user
• Two spans, equal spans = 33.5 m: BT ≻almost1 BTD ~ costs in the whole analysis.
SPG5G ≻almost
1 SPG5P Generally speaking, it can be inferred that the inclusion of user costs
• Two spans, equal spans = 40.0 m: BT ≻almost 1 BTD ≻almost
1 SPG5P associated with the different working actions, will increase the working
≻almost
1 SPG5G actions final cost proportionally with the duration and degree of inter­
vention of the original structure. It is expected that specifically for minor
As can be seen, prestressed concrete bulb tee beams are preferable for deck interventions, such as washing and cleaning and deck sealing, the
spans longer than 22.5 m for both continuous and simply supported user costs will be low compared with the cost associated with major deck
spans. Additionally, slab concrete bridges are also preferable for span interventions such as the construction of overlays or reconstruction of
lengths shorter than 11 m for both span configurations. However, the deck itself. In this sense, if user costs are considered, and an

Fig. 6. Superstructure Selection Flowchart 1.

11
S.L. Leiva-Maldonado and M.D. Bowman Engineering Structures 285 (2023) 116031

Fig. 7. Superstructure Selection Flowchart 2.

alternative considering overlays are more expensive than one without 7. Conclusions
them, then the difference will increase, and the preference of the latter
option will be reaffirmed. The same logical inference can be applied Summarizing, conclusions can be divided into those for deterministic
with other significant working actions. Hence, inclusion of minor and stochastic analysis, respectively.
working action user costs will not change the superstructure The deterministic analysis, explicitly for simply supported beams,
categorization. showed that for Span Range 1 the slab bridge is the most cost-effective
As a complementary information and clarity, Fig. 6 and Fig. 7 present solution either considering or not considering long-term costs for
the workflow for the superstructure selection explained in this spans less than 11.5 m. For longer simple spans, 4 beam galvanized steel
document. structures are the preferable alternative up to 19.5 m. In contrast, for
Span Range 2, the prestressed concrete bulb tees sections became the
optimal solution for longer spans, as well as the upper part of Span
Range 3, structures longer than 33.5 m. Additionally, Span Range 3

12
S.L. Leiva-Maldonado and M.D. Bowman Engineering Structures 285 (2023) 116031

results show that including long-term costs suggests multiple cost- could be implemented in future works in order to enrich the outcome of
effective design solutions for spans up to 33.5 m, with prestressed con­ this research. Even though the findings presented are the result of
crete bulb tees and galvanized steel plate girders being the two optimal multiple cases and stochastic processes, those are recommendations and
solutions. cannot be interpreted as design or cost absolute advise.
For continuous beams, again for the deterministic analysis, it is
shown for Span Range 1 that slab bridges and galvanized steel contin­ CRediT authorship contribution statement
uous beams are again the most cost-effective solutions for the lower and
upper parts of the span range, respectively. However, prestressed con­ Stefan Leiva-Maldonado: Conceptualization, Methodology,
crete AASHTO beams are also a competitive option for spans between Writing – original draft, Visualization, Investigation. Mark D. Bowman:
13.5 m and 18 m. Span Range 2 suggests that 4 beams galvanized rolled Supervision, Writing – review & editing.
beams are still cost-effective for spans shorter than 21 m. Additionally, it
is noticeable that prestressed bulb tees and AASHTO beams become
Declaration of Competing Interest
more attractive for longer spans. Finally, for Span Range 3, no variance
in the cost-effectiveness of the bulb tee option is noticed between the
The authors declare that they have no known competing financial
initial cost comparison and the inclusion of long-term costs.
interests or personal relationships that could have appeared to influence
The stochastic analysis showed that the most cost-effective options
the work reported in this paper.
implicitly consider the least number of working actions during their
service lives. For example, prestressed concrete bulb tees are preferable
Data availability
for spans longer than 21 m for both continuous and simply supported
spans. Additionally, slab concrete bridges are preferable for span lengths
Data will be made available on request.
shorter than 12 m for both span configurations. However, structural
steel galvanized structures dominate the span lengths between 12 m and
21 m for both span distributions. If the results are analyzed carefully, it Acknowledgements
can be shown that the compared alternatives that were not selected by
the analysis often include a higher number of working actions actively This study was supported by the Joint Transportation Research
considered during their lifespan. Program administrated by the Indiana Department of Transportation
Even though comparing the results from the deterministic and the and Purdue University. The content of this paper reflects the views of the
stochastic analyses may appear to suggest the same superstructure se­ authors, who are responsible for the facts and accuracy of the data
lection, it is important to differentiate the extent of both analyses. The presented herein, and do not necessarily reflect the official views or
consideration of only average values will be enough for risk-neutral policies of the sponsoring organization. These contents do not constitute
assessments. However, decision makers are often identified as risk- a standard, specification, or regulation. The authors also would like to
averse and therefore, superstructure selection need to be based on present a posthumous tribute to Purdue University Professor Wallace
more informed analysis considering different scenarios. Risk-behavior of Tyner who gave his uncountable help in the process of understanding
policy and decision-makers depends on many different factors such as and implementation of stochastic processes and Monte Carlo Simula­
economic sectors or geographic locations among others. For reference, tions. Additionally, we would like to thank Universidad del Valle’s
studies made by Sabatino et al. [47], Cha & Ellingwood [48], Boyle et al. professor Daniel Gomez for his help during the coding process.
[49], McDougal [50] among others investigates the risk-preference of
decision-makers and its causes and consequences. The risk aversion Data Availability Statement
preference is even more relevant for cases in which the deterministic
analysis for some span lengths shows cost-effectiveness ratios close to Some or all data, models, or code that support the findings of this
each other for different superstructure types. As it can be seen, the study are available from the corresponding author upon reasonable
deterministic analysis results are vaguer in terms of superstructure se­ request (PDF selection code, unit costs database, condition rating
lection, giving a wide variety to the decision maker, while the stochastic database, life-cycle cost analysis @Risk file and code, superstructure
analysis results narrow the possibilities. selection code and SD analysis code).
Generally speaking, it can be inferred that the inclusion of user costs
associated with the different working actions, will increase the working Appendix A. Supplementary material
actions final cost proportionally with the duration and degree of inter­
vention of the original structure. It is expected that specifically for minor Supplementary data to this article can be found online at https://doi.
deck interventions, such as washing and cleaning and deck sealing, the org/10.1016/j.engstruct.2023.116031.
user costs will be low compared with the cost associated with major deck
interventions such as the construction of overlays or reconstruction of References
the deck itself. In this sense, if user costs are considered, and an alter­
native considering overlays are more expensive than one without them, [1] Sinha KC, Saito M, Jiang Y, Murthy S, Tee A-B, Bowman MD. The development of
then the difference will increase, and the preference of the latter option optimal strategies for maintenance, rehabilitation and replacement of highway
bridges. IN: West Lafayette; 1991.
will be reaffirmed. The same logical inference can be applied with other
[2] V. Mara, R. Haghani, A. Sagemo, L. Storck, and D. Nilsson, “Comparative Study of
significant working actions. Hence, inclusion of minor working action Different Bridge Concepts Based on Life- Cycle Cost Analyses and Life-Cycle
user costs will not change the superstructure categorization concluded Assessment,” pp. 1–6, 2011.
[3] Zayed TM, Chang L-M, Fricker JD. Life-cycle cost based maintenance plan for steel
in this chapter.
bridge protection systems. J Perform Constr Facil 2002;16:55–62. https://doi.org/
Results presented in this document correspond to unit cost data 10.1061/(ASCE)0887-3828(2002)16:2(55).
specifics of the state of Indiana. However, the methodology used can be [4] So KKL, Cheung MMS, Zhang EXQ. Life-cycle management strategy on steel girders
implemented elsewhere using local data available as needed. It is in bridges. Adv Civ Eng 2012. https://doi.org/10.1155/2012/643543.
[5] L. R. Feldman, M. Boulfiza, J. Zacaruk, P. N. Christensen, and G. Sparks, “Life cycle
important to note that this research was conducted following only the deterioration and cost comparison of bridge deck designs including rehabilitation
assumptions described in the paper. Therefore, limitations such as the strategies,” pp. 143–148, 2008.
cost data generalization for one particular state (Indiana) have to be [6] A. Kendall, G. a. Keoleian, and G. E. Helfand, “Integrated Life-Cycle Assessment
and Life-Cycle Cost Analysis Model for Concrete Bridge Deck Applications,”
considered before interpreting the information presented. Topics such as J. Infrastruct. Syst., vol. 14, no. 3, pp. 214–222, 2008, doi: 10.1061/(ASCE)1076-
uncertainties of deterioration, working action timing among others 0342(2008)14:3(214).

13
S.L. Leiva-Maldonado and M.D. Bowman Engineering Structures 285 (2023) 116031

[7] M. G. Barker, “Historical Life Cycle Cost of Steel Concrete Girder bridges,” Aisi [27] Zanon R, Ochojski W, Hechler O, Klimaszewski P, Lorenc W. “Road bridges with
Smdi, 2016, [Online]. Available: http://www.shortspansteelbridges. prefabricated composite beams in high-strength steel - Realization Example”, in
org/~/media/Files/SSSBA/Learning/Historical LC Costs of Steel and Concrete Steel Bridges: Innovation & New. Challenges 2015:283–92.
Girder Bridges Final Report.pdf?la=en. [28] American Association of State Highway and Transportation Officials. AASHTO
[8] Okasha NM, Frangopol DM, Fletcher FB, Wilson AD. Life-cycle cost analyses of a LRFD Bridge Design Specification. 7th ed. Washington D.C.: American Association
new steel for bridges. J Bridg Eng 2012;17:168–72. https://doi.org/10.1061/ of State Highway Transportation Officials; 2015.
(ASCE)BE. [29] Indiana Departement of Transportation. Indiana Design Manual. Indianapolis, IN:
[9] K. Ozbay, N. a. Parker, and D. Jawad, “Guidelines for Life Cycle Cost Analysis,” Indiana Department of Transportation; 2013.
Security, no. July, 2003, [Online]. Available: http://trid.trb.org/view.aspx? [30] Federal Highway Administration, “Recording and Coding Guide for the Structure
id=798636. Inventory and Apprisal of the Nations’s Bridges,” Washington D.C., 1995.
[10] Y. Zhang, D. a. Novick, A. Hadavi, and R. J. Krizek, “Whole Life Cycle Cost for [31] Sinha KC, Labi SA, McCullouch BG, Bhargava A, Bai Q. “Updating and Enhancing
Chicago Type Bascule Bridges,” Cost Eng., vol. 50, no. 4, pp. 28–32, 2008, [Online]. the Indiana Bridge Management System (IBMS)”, West Lafayette. Indiana 2009.
Available: http://cat.inist.fr/?aModele=afficheN&cpsidt=20308007. https://doi.org/10.5703/1288284314306.This.
[11] Frangopol DM, Soliman M. Life-cycle of structural systems: recent achievements [32] Moomen M, Qiao Y, Agbelie BR, Labi S, Sinha KC. Bridge deterioration models to
and future directions. Struct Infrastruct Eng 2016;12(1):1–20. https://doi.org/ support Indiana’s bridge management system (Joint Transportation Research
10.1080/15732479.2014.999794. Program Publication No. FHWA/IN/JTRP-2016/03). West Lafayette, IN: Purdue
[12] Frangopol DM, Dong Y, Sabatino S. Bridge life-cycle performance and cost: University; 2016. p. 144.
analysis, prediction, optimisation and decision-making. Struct Infrastruct Eng [33] Cha H, Liu B, Prakash A, Varma AH. “Efficient load rating and quantification of
2017;13(10):1239–57. https://doi.org/10.1080/15732479.2016.1267772. life-cycle damage of Indiana bridges due to overweight loads (Joint Transportation
[13] S. Kim, G. Baixue, and D. M. Frangopol, “Probabilistic Optimum Bridge System Research Program Publication No. FHWA/IN/JTRP-2016/06)”. West Lafayette, IN:
Maintenance Management Considering Correlations of Deteriorating Components Purdue University; 2016. p. 94.
and Service Life Extensions,” ASCE-ASME J. Risk Uncertain. Eng. Syst., vol. 8, no. 3 [34] Nasrollahi M, Washer G. Estimating inspection intervals for bridges based on
(September), 2022, doi: 10.1061/AJRUA6.0001235. statistical analysis of national bridge inventory data. J Bridg Eng 2015;20(9):
[14] Han X, Frangopol DM. “Risk-Based Optimal Life-Cycle Maintenance Strategy for 04014104. https://doi.org/10.1061/(asce)be.1943-5592.0000710.
Bridge Networks Considering Stochastic User Equilibrium”, ASCE-ASME. J Risk [35] American Galvanizers Association. Costs Less, Lasts Longer. Centenial, CO:
Uncertain Eng Syst Part A Civ Eng 2022;8(2):4022011. https://doi.org/10.1061/ American Galvanizers Association; 2015.
AJRUA6.0001222. [36] H. Hawk, “Bridge life-cycle cost Analysis, NCHRP Report 483.” Transportation
[15] Qian J, Dong Y, Frangopol DM. Probabilistic long-term resilience of bridges under Research Board, Washington D.C., p. 138, 2002.
seismic and deterioration processes. Proc Inst Civ Eng - Bridg Eng 2022:1–12. [37] Y. Sun and D. G. Carmichael, “A simplified approach to address uncertainty in Life
https://doi.org/10.1680/jbren.21.00049. Cycle Costing (LCC) analysis,” in Life-Cycle Analysis and Assessment in Civil
[16] Y. F. M. Moscoso, L. F. Rincón, S. L. Leiva-Maldonado, and J. A. S. C. Campos e Engineering: Towards an Integrated Vision, 2019, pp. 1417–1424.
Matos, “Bridge deterioration models for different superstructure types using [38] L. Peters, “Impact of Probability Distributions on Real Options Valuation,”
Markov chains and two-step cluster analysis,” Struct. Infrastruct. Eng., vol. 0, no. 0, J. Infrastruct. Syst., vol. 22, no. 3, 2016, doi: 10.1061/(asce)is.1943-555x.0000289.
pp. 1–11, 2022, doi: 10.1080/15732479.2022.2119583. [39] Sinha KC, Labi SA. Transportation decision making principles of project evaluation
[17] Leiva Maldonado SL. Superstructure bridge selection based on bridge life-cycle cost and programming. New York, NY: John Wiley and Sons; 2011.
analysis. Purdue University; 2019. [40] American Association of State Highway Transportation Officials. A manual on user
[18] Leiva Maldonado SL, Bowman MD. Life-cycle cost analysis for short- and medium- benefit analysis of highway and bus-transit improvements. Washington D.C.:
span bridges. West Lafayette, IN 2019. https://doi.org/10.5703/1288284316919. American Association of State Highway Transportation Officials; 1978.
[19] Taly N, Gangaro HVS. Prefabricated Pre-Formed Steel T-box Girder bridge System. [41] Grant EL, Grant-Ireson W. Principles of engineering economy. 4th ed. New York,
Eng J AISC 1979;16(3):75–82. NY: Ronald Press; 1960.
[20] Nakamura S. Bending Behavior of Composite Girders with Cold Formed Steel U [42] Fwa TF. The handbook of highway engineering. 6th ed. Boca Raton, FL: CRC Press;
Section. J Struct Eng 2002;128(9):1169–76. https://doi.org/10.1061/(ASCE)0733- 2006.
9445. [43] Ford K, et al. NCHRP Report 713: estimating life expectancies of highway assets.
[21] B. S. Pavlich and R. Burgueño, “Evaluation of prefabricated composite steel box Final R 2012;vol. 2.
girder systems for rapid bridge construction.” Michigan State University and [44] Craig J, O’Conner D, Ahlskog J. Economics of bridge deck protection methods.
Michigan Depatment of Transportation, East Lansing, MI, 2008, [Online]. Mater Perform 1982;22(11):32–4.
Available: https://www.michigan.gov/documents/mdot/MDOT_Research_Report [45] Ang A-H-S, Tang WH. Probability concepts in engineering : emphasis on
_RC-1507_238838_7.pdf. applications in civil & environmental engineering. 2nd ed. New York: Wiley; 2007.
[22] Burner KA. Experimental investigation of folded plate girders and slab joints used [46] Levy H. Stochastic dominance investment decision making under uncertainty. 2nd
in modular construction. University of Nebraska; 2010. ed. US: Springer; 2006.
[23] K. E. Barth, G. K. Michaelson, and M. G. Barker, “Development and Experimental [47] Sabatino S, Frangopol DM, Dong Y. Sustainability-informed maintenance
Validation of Composite Press Brake – Formed Modular Steel Tub Girders for Short- optimization of highway bridges considering multi-attribute utility and risk
Span Bridges,” J. Bridg. Eng., vol. 20, no. 11, 2015, doi: 10.1061/(ASCE)BE.1943- attitude. Eng Struct 2015;102:310–21. https://doi.org/10.1016/j.
5592.0000770. engstruct.2015.07.030.
[24] A. Azizinamini, A. Yakel, and M. Farimani, “Development of a steel bridge system [48] Cha EJ, Ellingwood BR. Risk-averse decision-making for civil infrastructure
simple for dead load and continuous for live load Volume 1 Analysis and exposed to low-probability, high-consequence events. Reliab Eng Syst Saf 2012;
Recommendations NDOR Research Project Number P542.” University of Nebraska, 104:27–35. https://doi.org/10.1016/j.ress.2012.04.002.
Lincoln, NE, p. 486, 2005. [49] Boyle PA, Yu L, Buchman AS, Bennett DA. Risk aversion is associated with decision
[25] Azizinamini A. Simple for dead load – continuous for live load steel bridge systems. making among community-based older persons. Front Psychol 2012;vol. 3:1–6.
Eng J AISC 2014;51:59–82. https://doi.org/10.3389/fpsyg.2012.00205.
[26] Hoorpah W, Zanon R, Dabee V, Muhomud SR. “Colville deverell bridge, Mauritius [50] McDougal YB. Decision making under risk: risk preference, monetary goals and
Island composite deck with steel main girders and concrete crossbeams: economic information search. Pers Individ Dif 1995;18(6):771–82. https://doi.org/10.1016/
and fast construction technology for developing countries”, in Steel Bridges: 0191-8869(94)00207-9.
Innovation & New. Challenges 2015:277–82.

14