Received: 9 October 2024 Revised: 29 May 2025 Accepted: 4 July 2025

DOI: 10.1002/suco.70247

ARTICLE

Parametric modeling for automated intelligent design

of concrete beam bridges: A global and local
optimization approach

Vinicius Nascimento de Moraes | Thomaz Eduardo Teixeira Buttignol

Department of Structures, University of

Campinas, São Paulo, Brazil Abstract
The structural design of prestressed concrete girders in beam bridges is per-
Correspondence
formed by coupling 3D parametric modeling, optimization, and statistical tech-
Thomaz Eduardo Teixeira Buttignol,
Av. Albert Einstein, 951, Cidade niques, structural analysis, and design. The objective is to explore the potential
Universit
aria, Campinas, SP 13083-852, and advantages of advanced design methodologies based on intelligent design.
Brazil.
The full 3D parametric beam bridge is modeled in the software Rhinoceros
Email: [email protected]
(computer-aided design)/Grasshooper (algorithm-aided design). An optimiza-
Funding information tion procedure using a genetic algorithm is adopted to find the minimum num-
Conselho Nacional de Desenvolvimento
Científico e Tecnol
ogico, Grant/Award
ber of supports. With the 3D bridge configuration, static and modal finite
Number: 403517/2023-0; Fundo de Apoio element analyses are performed using the Karamba3D plugin to extract the
ao Ensino, à Pesquisa e Extensão, internal forces of the prestressed girders. After that, the Ultimate and Service-
Universidade Estadual de Campinas,
Grant/Award Number: 2474/23 ability Limit States are checked by plugging the non-linear design equilibrium
equations in a Python script directly into Grasshopper. The Simplex method is
used to determine the feasible domain (space of solutions) allowing the choice
of the most appropriate pair of girder height and prestressing force.

KEYWORDS
beam bridge, genetic algorithm, intelligent design, parametric modeling, prestressed
concrete, simplex

1 | INTRODUCTION optimized girders. The analysis includes the effectiveness

of varying the compressive strength from 30 to 90 MPa.
The first part of this paper provides a brief overview of the The classical design approach can be traced back to
evolution of the structural design process, highlighting the the early 20th century, following the development of
importance of this study. The second part details the pro- structural analysis methods and the emergence of build-
posed methodology, while the third part verifies the appli- ing codes and standards. The procedure is based on a
cability of the method through case studies. Two practical deterministic approach, with manual (hand calculations)
applications are discussed: (i) analysis of the cost optimiza- or semi-automated processes (computer-aided design
tion of prestressed concrete girders described in Aydn and [CAD] and structural analysis toolkits). CAD is exten-
Ayvaz1; (ii) comparison of the structural efficiency of stan- sively used for model creation, integration with structural
dard girders of DNIT2 Caltrans: California Department of analysis, interoperability with BIM, and the representa-
Transportation,3 AASHTO,4 and Jahjouh and Erhan5 tion of structural details in design plans.

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided
the original work is properly cited.
© 2025 The Author(s). Structural Concrete published by John Wiley & Sons Ltd on behalf of International Federation for Structural Concrete.

Structural Concrete. 2025;1–20. wileyonlinelibrary.com/journal/suco 1

 17517648, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/suco.70247 by Capes, Wiley Online Library on [29/07/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2 DE MORAES and BUTTIGNOL

New softwares arose bringing more agility and pro- enables a hybrid design methodology that integrates
ductivity to the daily design routines. Building informa- data-driven optimization techniques—such as machine
tion modeling (BIM) introduced a collaborative platform learning models—with engineering domain knowledge,
that integrates architecture and engineering plans. For supported by parameterization and automated structural
example, Girardet and Boton6 adopt a parametric BIM analysis. According to Fan et al.,10 ML algorithms have
approach for bridge design aiming to develop a paramet- established new research directions in bridge engineer-
ric model that can generate all types of bridges from a ing, especially for applications including form-finding of
single parametric file. According to the authors, paramet- long-span structures, structural reinforcement, and struc-
ric modeling can improve the efficiency and interopera- tural optimization.
bility of bridge structures, although in the literature, this Figure 1 depicts the evolution of structural design,
topic is still poorly explored. Despite that, BIM capabili- from a classical approach to an intelligent design process,
ties are limited as the classical design method often with the aim of highlighting the evolution of the design
results in static BIM models that do not take advantage process. It is shifting from a deterministic and standard-
of real-time data integration and updating. BIM can be ized workflow toward an integrated, iterative methodol-
integrated with Internet of Things (IoT) systems to enable ogy that leverages both data-driven tools and structural
real-time analysis and the development of predictive engineering principles to guide decision making.
maintenance strategies for structures Tang et al.7 The methodology described in this paper consists of
The most recent structural analyses include software an integrated design system for prestressed concrete
for 3D modeling and visualization coupled with finite ele- bridge girder design, as shown in Figure 2, which
ment (FE) analysis along with optimization techniques, includes: (a) CAD (3D digital visualization of the struc-
such as shape5 and topology8,9 optimization, and ture); (b) AAD (visual scripting of the structure and para-
machine learning models.10,11 metric modeling); (c) genetic algorithm (structure
The development of new and advanced mathematical optimization); (d) FE analysis (beam internal forces);
models (for example, machine learning—ML) and soft- (e) Python programming (Simplex using prestressed con-
ware (algorithm-aided design—AAD) produced a signifi- crete beam non-linear sectional analysis).
cant change in the design methodology, leading to the The integration of multiple techniques allows for a
so-called “intelligent design” (Nawari12,13). This approach holistic view of the problem and the analysis of multiple

FIGURE 1 Classical and intelligent design processes.

 17517648, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/suco.70247 by Capes, Wiley Online Library on [29/07/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
DE MORAES and BUTTIGNOL 3

FIGURE 2 Integrated design process.

solutions. The automation of design routines can lead to the global and local optimization procedures. In this case,
increased efficiency, higher accuracy, scalability, and cost a preliminary analysis is performed to find the most suit-
reduction. Previous studies have demonstrated the able number of supports. The plugin Galapagos (genetic
advantages of enhanced design methodologies that algorithm) is used to search for the most economical
include parametric analysis and optimization tech- solution based on the minimization of a cost function. As
niques.1,5,14,15 Despite that, as stated by Girardet and input, pairs of bridge span and girder height should be
Boton,6 this topic is still poorly explored. Thus, this paper provided, either from historical data of existing bridges or
aims to contribute to the investigation of new design from a theoretical database generated by varying the
tools that enable parametric modeling, design process girder span and calculating the corresponding admissible
automation, and structural optimization. height. The admissible height is herein defined as the
minimum required girder height from a structural point
of view. This value may be affected or limited by geomet-
2 | METHOD OLOG Y ric and clearance constraints, such as vertical clearance
for roads and railways, or architectural restrictions in the
The parametric modeling and structural analysis proce- case of integrated urban infrastructure.
dures were performed using Grasshopper, a visual pro- With the final bridge configuration, FE static and
gramming environment developed by McNeel and fully modal analyses are performed, using the plugin
integrated with Rhinoceros 3D, a CAD platform widely Karamba3D,17 to obtain the internal forces and modes of
used in architectural and structural design workflows. vibration of the girders. Lastly, the design of the pre-
Grasshopper enables the creation of rule-based and para- stressed girders is performed by means of an automated
metric geometries through components connected by a design routine that is included in a Python script. The
node-based interface Tedeschi,16 allowing users to design procedures follow the Brazilian Design Code
manipulate complex shapes, control design parameters, Associação Brasileira de Normas Técnicas (ABNT)18 and
perform geometric operations, and integrate structural include the verification of Serviceability and Ultimate
and optimization analyses. In this work, it served as the Limits States, all the prestressing losses, and deflection
primary interface to build geometry-dependent algo- limits. The Simplex method is adopted to provide the fea-
rithms and automate structural analyses using the Kar- sible domain, allowing the design engineer to choose the
amba3D plugin Preisinger17 and Python scripting. most appropriate solution (for example, minimum girder
Galapagos, a native Grasshopper plugin based on a height and maximum prestressing force).
genetic algorithm, was used to execute the optimization For simpler cases, a local optimization procedure for
routines. One of the main advantages of the Rhinoceros/ prestressed concrete girders can be adopted. In this case,
Grasshopper environment is the integration of multiple the span should be predefined and no cost-based optimi-
design, analysis, and optimization routines (as shown in zation is applied.
Figure 2) into a single platform, allowing the automation The framework of the design procedure is flexible,
of the design process. allowing variations and different levels of complexity. In
The method allows for different applications accord- the methodology shown in Figure 3, the number of
ing to the complexity of the problem. Figure 3 describes girders is fixed and pre-established. Despite that, the cost
 17517648, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/suco.70247 by Capes, Wiley Online Library on [29/07/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
4 DE MORAES and BUTTIGNOL

FIGURE 3 Description of the global and local optimization procedures.

optimization can also include, together with the number concrete girders using a genetic algorithm. The vari-
of supports, the minimum number of girders to achieve ables include the span, cross-section and prestressing
the best economical solution. This can be done by simply steel area. The bridge design followed AASHTO4 stan-
including the number of girders as an additional input dards. The authors reached a reduction up to 12.6%
variable in the Galapagos plugin. The methodology can compared to the actual projects analyzed in the paper.
also be extended to include the design routines for the Rempling et al.14 describe an automatic structural
bridge deck and columns. design procedure, combining set-based design, paramet-
The main advantages of this procedure are: (i) reduction ric design and FE analysis. The applicability of the
of errors and time savings due to the automatic design rou- model was assessed on three existing Swedish bridges:
tine6; (ii) multiple solutions by using the Simplex method a concrete bridge in Örebro, a steel–concrete composite
due to the possibility of choosing the concrete properties, bridge in Nynäshamn, and a concrete frame bridge in
prestressing force, and girder height. According to Lounis Stockholm. The results demonstrated that the method
and Cohn,19 cost-effectiveness is increased by selecting the reduced by 20% to 60% the material costs and the car-
minimum prestressed and nonprestressed reinforcement bon dioxide equivalent emissions.
areas that satisfy all serviceability and ultimate limit states The methodology is detailed considering a hypotheti-
constraints; (iii) cost reduction due to the optimization of cal beam bridge, taking into account the initial input
the bridge spans1,20; (iv) holistic view due to the 3D visuali- properties described in Table 1. The bridge profile and
zation of the bridge and graphical representation of the the natural ground line are shown in Figure 4.
reactions at the base columns and the internal forces and In this example, there are no restrictions to the loca-
modes of vibration of the structural elements. tion of the columns. They are assumed to be equally
It is worth mentioning that Aydn and Ayvaz1 spaced. As a simplification, the number of girders is con-
describe a cost optimization procedure for prestressed sidered fixed.
 17517648, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/suco.70247 by Capes, Wiley Online Library on [29/07/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
DE MORAES and BUTTIGNOL 5

2.1 | 3D digital parametric model and 2.1.1 | Superstructure

preliminary analysis
The superstructure is composed of a cast-in-place
The 3D parametric digital model is constructed using the reinforced concrete (RC) slab with a thickness of 20 cm,
Grasshopper plugin, as shown in Figure 3. The paramet- precast prestressed I girders2 and rectangular RC cross-
ric model is flexible and allows the designer to simulate beams with a thickness of 20 cm and variable height,
either continuous or simply supported girders and slabs, located at the piers, as shown in Figure 6 and detailed in
depending on the design strategy. Figure 7. In this particular case, the crossbeams provide
The bridge geometry is represented by solid elements, lateral stability and support for the bearings. Intermedi-
as follows: solid bearings; continuous deck slab; I-shaped ate crossbeams, although not included in the present
DNIT standard girders2 and rectangular cross beams; model, could be incorporated to improve wheel load dis-
square column caps; square columns supported by pris- tribution among the girders.
matic pile caps; solid abutments. The natural ground line The simulation is performed considering rotational
is represented using NURBS (non-uniform rational releases at girder ends and segmentation of the deck by
B-spline). The input parameters are shown in Figure 5. the introduction of linear joints. This reflects common
The parametric model has four main parameters: precast construction practices, in which expansion joints
(i) longitudinal profile including the natural ground and and elastomeric bearings are introduced between spans
the bridge longitudinal alignment; (ii) bridge width; to accommodate thermal movements and allow rotation.
(iii) number of girders, supports, and piers per support; Figure 6 shows a rendered representation of the bridge
(iv) cantilever pier caps.
A global optimization of the 3D model of the bridge is
performed based on preliminary assessment design rules,
as described below.

TABLE 1 Input properties.

Parameter Value Unit

Characteristic compression strength 35 MPa
Tendons CP190 —
Relative humidity 60 %
Slump 8 cm
Mean temperature 22 Degree F I G U R E 5 Input parameters in grasshopper for the
preliminary assessment.
Anchored slip 4 mm
Friction coefficient 0.2 —

FIGURE 4 Geometry of the bridge: Plan and profile (units in m). FIGURE 6 3D digital model (preliminary assessment).
 17517648, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/suco.70247 by Capes, Wiley Online Library on [29/07/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
6 DE MORAES and BUTTIGNOL

FIGURE 7 Cross-section of the superstructure (units in cm).

geometry and does not display the internal joints or The diameter of the circular RC columns was parame-
boundary conditions. These conditions are defined by the terized based on the critical buckling load (Equations 1
designer within the Karamba3D environment during and 2), considering one story and neglecting shear defor-
the structural modeling. mations, and the maximum allowable stress limit
For the parameterization of the bridge section, the (Equation 3). This criterion estimates the column diameter
height of the cross-beams was defined as a function of taking into account both the material strength and buck-
the height of the girder web. DNIT2 standard design rec- ling stability. It is important to note that, for buckling sim-
ommendations were adopted to determine the geometry plified analysis, the approach described in Annex O of EN
of the girders. In the analyses, the “I” section girders have 1992-1-1 European Committee for Standardization
fixed flange (top and bottom) dimensions. The web has a (CEN)21 was adopted. For RC columns, the classical Euler
fixed thickness of 20 cm and a variable height. buckling load is not applicable, as it assumes a perfectly
A linear equation was defined to determine the most elastic and uncracked structural element, neglecting sec-
appropriate girder height as a function of the free span, ond order effects. As a result, it may significantly overesti-
based on the input properties listed in Table 1. Theoreti- mate the critical load. Although buckling is a key concern
cal discrete pairs of the girder height versus free span in slender RC columns, in the present study, the piers are
were obtained by varying the girder span and calculating short and designed to fail under axial compression.
the respective admissible height. To this end, the plugin
2 3
Karamba3D17 was used to determine the girder internal
7:8 X 4 1 kc
Ec
I c 5
forces (bending and shear) for several spans, as described Pcr ¼
  ð1Þ
2:6 1 þ 3:9
f r L2
1 þ φ
in section 2.3. The admissible heights for the specified eff ,s

spans were calculated considering the optimization script

described in section 2.4. From the discrete points, the fol- θ kc
Ec
I c
fr ¼
  ð2Þ
lowing linear equation was determined: h = 0.07
L – 0.15 M L
1þφ
eff ,s
(where: “h” is the girder height; “L” is the free span). This
equation was adopted in the preliminary design stage to
find the most cost-effective bridge configuration. Nd
σd ¼ ≤ f cd ð3Þ
Ap

2.1.2 | Substructure where: Pcr = critical buckling load; E c = modulus of elas-

ticity; I c = gross moment of inertia; θ = rotation for bend-
The superstructure is supported by RC cap beams with a ing moment M; M = bending moment at the base;
minimum height of 160 cm or equal to the diameter of L = column height; k c = 0.4 (general case); φeff ,s = 3.0
the circular RC columns plus 10 cm. The height was (first approximation), effective creep coefficient for global
established such that the neutral axis remained in the second order effects in table 5.221; N d = design load;
deformation domain 2,18 that is, failure on the steel side f cd = design concrete compressive strength; Ap = column
due to steel yielding and subsequently failure. cross-section; σ d = design compression stress.
 17517648, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/suco.70247 by Capes, Wiley Online Library on [29/07/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
DE MORAES and BUTTIGNOL 7

2.1.3 | Foundation represents an AF = 1.5, which accounts for the greater

complexity associated with the construction. A cost of $
RC walls with a thickness of 30 cm were considered for 2500/m3 is attributed for columns and foundation, which
the abutments. The foundation of the RC columns was represents an AF = 1.25 that accounts for the execution
represented by RC pile caps on six piles. In this paper, complexities (excavation and more complex foundation
the foundation and the abutments are not included in the requirements). The cost of the abutment and the
analyses. approach slab is defined as equal to $ 2800/m3, leading to
an AF = 1.4. This value includes execution features, such
as temporary containment systems.
2.2 | Evolutionary optimization The optimization procedure gives a first approximation
of the structural conceptual design of the bridge. At this
Galapagos is a native scripting component of Grasshop- stage, the dimensions of the structural elements are deter-
per that performs genetic algorithm optimization aiming mined through preliminary design routines and are
to find the best set of parameters based on an objective assumed to give a good representation of the final solution.
function. In this particular case, it is adopted to minimize In this particular case, the best solution is a six-span
a cost function. The objective is to find the number of bridge with 20 m span and two-column supports. This
supports that leads to the most economical solution. The leads to an equivalent concrete volume of 1453 m3, as
script is depicted in Figure 8. The input variable is shown in Figure 9.
the number of supports and the fitness is the cost of the
bridge. The latter should consider the costs associated
with transportation, handling, assembly, environmental 2.3 | Static analysis
impacts, and quantity of materials.
As a simplification, a cost per concrete cubic meter The structural analysis is developed using Karamba3D,17
($/m3) is defined. A volume adjustment factor (AF) was which is a Grasshopper plugin. The algorithm “Analyze”
applied to each structural element to take into account
the specific costs and different levels of construction com-
plexity of the structural elements. The most optimized T A B L E 2 Cost values per m3 and adjustment factors (AF) for
solution will be the one that presents the lowest fictitious different structural elements.
global volume.
Element $/m3 AF
Table 2 shows the costs defined in terms of the AFs
for the structural elements. The AF values represent a fic- Slab 2000 1
titious equivalent global volume of concrete, which Girder 3000 1.5
adjusts the cost of each element based on the complexity Crossbeam 2000 1
of execution and other variables. The cost per cubic meter Transverse beam 2000 1
for the slab, transverse beam and crossbeam are assumed
Columns + foundation 2500 1.25
as equal to $ 2000, with an AF equal to 1. The estimated
Abutment + approach slab 2800 1.4
cost per cubic meter for the girder is $ 3000. This

F I G U R E 8 Optimization in Galapagos: (a) Global parameters; (b) introduction of the fictitious equivalent global volume factors;
(c) Galapagos control panel.
 17517648, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/suco.70247 by Capes, Wiley Online Library on [29/07/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
8 DE MORAES and BUTTIGNOL

performs a linear FE analysis, considering small displace- considers a uniformly distributed load of 5 kN/m3 and a
ments, thus neglecting changes of length in axial or in- 450 kN load train with six wheels. The concentrated live
plane direction. loads are located at the midspan to consider the least
In the model assemblage, straight lines represent favorable case (maximum bending moment). The dead
beam, truss, and spring elements. Curved lines load considers the specific weight of the deck, pavement
represent arches. Surface elements (shells and slabs) are and barriers. The following dead loads were adopted in
based on meshes. The connection between the elements the analyses, according to ABNT NBR 7187 Associação
is carried out at the nodal points by superposition of line Brasileira de Normas Técnicas (ABNT)23: concrete slab
endpoints and mesh points. The supports and external including asphalt coating and resurfacing = 24 kN/m3 
loads are applied at the nodal points. 0.1 m + 2kN/m3 = 4.4 kN/m3; concrete rigid barriers =
In this case, it is assumed simply-supported beams 0.23  25 kN/m3 = 5.8 kN/m.
and fixed supports at the base columns. Live and dead The cross-section dimensions and the material
loads are applied on the deck, considering the recom- properties were defined during the construction of the 3D
mendations of the Brazilian National Standard ABNT digital model. In this case, a concrete with a compression
NBR 7188:2024 Associação Brasileira de Normas Técni- strength of 35 MPa, an initial modulus of elasticity of
cas (ABNT).22 Bridge load class 45 is specified, which 33 GPa, and a secant modulus of 29 GPa were specified.
The Poisson ratio is equal to 0.2.
From the analysis, the internal forces (shear and
bending) of the girders are extracted. Figure 10 displays
the visualization of the results in Rhinoceros (top, bot-
tom, right and perspective), which includes the 3D ren-
dering of the girders together with the bending moment
diagrams. The deck is 14.1 meters wide, and the five
girders are spaced 3.24 m apart.
The parametric model enables the determination of
the girders' internal forces for varying spans. The internal
forces shown in Figures 11 and 12 consider three key
design stages: (i) initial girders prestressing, without
FIGURE 9 Free span versus equivalent concrete volume. assuming the contribution of the deck; (ii) after the

FIGURE 10 Bending moment diagrams of the girders.

 17517648, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/suco.70247 by Capes, Wiley Online Library on [29/07/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
DE MORAES and BUTTIGNOL 9

material properties, loading conditions, and structural

requirements, and iterating to identify the optimal design.
The constraints account for post-tensioning and lim-
ited prestressing level, ensuring that tensile stresses
remain below the concrete characteristic tensile strength.
These conditions are widely used for the design of bridges
and viaducts. The following constraints considered are:

• Immediate stress limits in compression and tension.

• Decompression in SLS.
• Crack formation in SLS.
• Maximum allowable compression in SLS.
FIGURE 11 Maximum shear force versus free span.
• Maximum allowable tension in SLS.
• Maximum deformation in SLS.

After the verification in the SLS, the reinforcement

areas for stirrups and tendons are calculated according to
the ULS. In this particular case, the target is to find the
minimum height in the feasible (convex hull) domain.
There are seven inequality equations and only two
unknowns, which are the cross-section height and the
prestressing force. The other parameters are fixed and
predefined.
The symbology of the Equations (3)–(26) is described
below.

FIGURE 12 Maximum bending moment versus free span.

Symbology
construction of the bridge deck, without including the M gk : bending moment due to the girder self weight;
traffic load; (iii) during service life. Figure 11 shows the M gk1 : bending moment due to the RC slab self
shear forces obtained for different free spans, where: weight;
Vgk1, Vcpk, and Vcmk are, respectively, the shear force M cpk : bending moment due to the dead load;
from the self-weight of the RC slab, dead, and live loads. M cmk : bending moment due to the live load;
Figure 12 shows the bending moments obtained for dif- ψ 1 ¼ 0:6: reduction factor for accidental actions
ferent free spans, where: Mgk1, Mcpk, and Mcmk are, considering the frequent combination;
respectively, the bending moment from the self-weight of ψ 2 ¼ 0:4: reduction factor for accidental actions
the RC slab, dead, and live loads. considering the quasi permanent combination;
φ: dynamic load coefficient for static vertical
loads;
2.4 | Sectional analysis Pi : initial prestressing force;
Pinf : prestressing force after losses;
The design procedure of prestressed concrete girders fol- γ p : magnification factor for the prestressing force;
lows ABNT NBR 6118:2023 Associação Brasileira de Nor- W i0 : first moment of area of the girder bottom
mas Técnicas (ABNT).18 The analysis verifies the section without slab collaboration;
deflection limit, Serviceability (SLS) and Ultimate (ULS) W i1 : first moment of area of the girder bottom
Limit States. The Simplex method is adopted to find the section considering the slab collaboration;
feasible domain, which corresponds to a set of all possible W s0 : first moment of area of the girder top
values (solutions) that satisfy all the constraints. In a two- section without slab collaboration;
dimensional space, this region is a convex hull area. This W s1 : first moment of area of the girder top
method is adequate for limit states design, since the con- section considering the slab collaboration;
straints are predefined. The process includes defining the Ac0 : cross-sectional area of the girder without
objective functions, establishing constraints related to the slab collaboration;
 17517648, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/suco.70247 by Capes, Wiley Online Library on [29/07/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
10 DE MORAES and BUTTIGNOL

Ac1 : cross-sectional area of the girder considering For this case:

the slab collaboration; If f ck ≤ 50 MPa: f c,limit ¼ 0, 7
f ck,j . h   i
ep0 : initial eccentricity of the prestressing force; If 50 < f ck ≤ 90 MPa: f c,limit ¼ 0, 7 1  f ckj  50 =200
f ctm : mean concrete tensile strength;
f ckj .
f ctm,j : mean concrete tensile strength at age j;
f ct,limit : limit concrete tensile strength; • Inequality 3: Decompression in SLS (SLS-D)
f c,limit : limit concrete compressive strength; σ i ≤ 0 ðQuasi permanent combinationÞ ð10Þ
δCQP : midspan deflection for quasi permanent
combination; Pinf Pinf
ep0 M gk þ M gk1 M cpk þ M cmk
φ
ψ 2
δprot : counter-deflection due to prestress;   þ þ ≤0
Ac0 Wi0 Wi0 Wi1
δlimit : deflection limit; ð11Þ
E cs : Concrete Secant Modulus of Elasticity;
I x0 : gross moment of inertia without the slab Isolating Pinf :
collaboration;
I x1 : gross moment of inertia with slab    
M gk þ M gk1 M cpk þ M cmk
φ
ψ 1 Ac0
Wi0
Pinf ≥ þ

collaboration; Wi0 Wi1 Wi0 þ Ac0
ep0
L: free span; ð12Þ
qgk : load due to the girder self-weight;
qgk1 : load of the deck;
qcpk : dead load; • Inequality 4: Cracking formation in SLS (SLS-F)
qcmk : live load. σ i ≤ f ct,limit ðFrequent combinationÞ ð13Þ

Pinf Pinf
ep0 M gk þ M gk1 M cpk þ M cmk
φ
ψ 1

  þ þ ≤ f ct,limit
Ac0 Wi0 Wi0 Wi1
• Inequality 1: Immediate stress limit—max. allowable ð14Þ
tension
Isolating Pinf :
σ s ≤ f ct,limit ðtensionþÞ ð4Þ
 
M gk þ M gk1 M cpk þ M cmk
φ
ψ 1
M gk γ p
Pi γ p
Pi
ep0 P inf ≥ þ  f ct,limit
  þ ≤ f ct,limit ð5Þ  Wi0  Wi1
Ws Ac0 Ws
Ac0
Wi0

ð15Þ
Wi0 þ Ac0
ep0
Isolating Pi :
For this case: f ct,limit ¼ 0:7
α
f ctm
 
Ac0
M gk þ W s
f ct,limit
Pi ≤   ð6Þ • Inequality 5: Maximum compression in SLS (SLS-
γ p
Ac0
ep0  W s CE-FC)

For this case: f ct,limit ¼ 1, 2
f ctm,j . σ s ≤ f c,limit ðFrequent combinationÞ ð16Þ

• Inequality 2: Immediate stress limit—max. allowable Pinf Pinf
ep0 M gk þ M gk1 M cpk þ M cmk
φ
ψ 1
 þ   ≥ f c,limit
compression Ac0 Ws0 Ws0 Ws1
σ i ≥ f ct,limit ðcompressionÞ ð7Þ ð17Þ

Isolating Pinf :
M gk γ p
Pi γ p
Pi
ep
  þ ≥ f c,limit ð8Þ
Wi Ac0 Wi  
M gk þ M gk1 M cpk þ M cmk
φ
ψ 1
P inf ≥ þ  f c,limit
 Ws0  Ws1
Ac0
Ws0
Isolating Pi:
ð18Þ
Ac0
ep0  Ws0
 
Ac0
M gk þ W i
f c,limit
Pi ≤   ð9Þ For this case: f c,limit ¼ 0:6
f ck
γ p
Ac0
ep0  W i
 17517648, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/suco.70247 by Capes, Wiley Online Library on [29/07/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
DE MORAES and BUTTIGNOL 11

• Inequality 6: Maximum compression in SLS (SLS-CE-

CQP)

σ c ≤ f c,limit ðQuasi permanent combinationÞ ð19Þ

Pinf Pinf
ep0 M gk þ M gk1 M cpk þ M cmk
φ
ψ 2

 þ   ≥ f c,limit
Ac0 Ws0 Ws0 Ws1
ð20Þ

Isolating Pinf :
 
M gk þ M gk1 M cpk þ M cmk
φ
ψ 2
P inf ≥ þ  f c,limit
 Ws0  Ws1
Ac0
Ws0

ð21Þ
Ac0
ep0  Ws0

For this case: f c,limit ¼ 0:45
f ck

• Inequality 7: Maximum deformation in SLS (SLS-DEF)

δ max ≤ δlimit ðQuasi permanent combinationÞ ð22Þ

δCQP  δprot ≤ δlimit ð23Þ

δCQP  δlimit ≤ δprot ð24Þ

Isolating δprot :
FIGURE 13 Flowchart for the design analysis.
Pinf
ep0
8:5
L2
δprot ¼ ð25Þ
384
E cs
I x0
Initially, a high value is assigned to the prestressing
9:6
E cs
I x0 losses to determine the initial prestress force. Taking the
Pinf ≥
ðδCQP  δlimit Þ ð26Þ
ep0
L2 girder dimensions and the initial prestressing force,
the prestressing losses are calculated and compared to
the assigned one. If the difference is higher than 5%,
Considering the superposition of effects: the value of the assigned prestressing losses is reduced
(iterative process) until convergence is achieved.
 
5
L qgk þ qgk1 qcpk þ qcmk
φ
ψ 2
δCQP ¼
þ ð27Þ
384
E cs I x0 I x1
2.4.2 | Defining the feasible domain

The cross-section of the girders follows DNIT2 standard

2.4.1 | Estimating the prestressing losses specifications, as shown in Figure 14. The next step is the
definition of the girder web height and the prestressing
The challenge in applying the Simplex method to this force. The solution can be found in the feasible domain
problem lies in the fact that some of the objective func- constructed using the Simplex method.
tions are defined in terms of the initial prestressing force, To plot the feasible domain (convex hull area), the
while others are based on the effective prestressing web height is varied to determine the required initial pre-
force, which accounts for both immediate and long-term stressing force that meets the limit states. The solution is
losses. It is necessary to know the geometry of the cross- shown in Figure 15 and includes the seven curves repre-
section and the initial prestressing force to determine the senting the inequality equations and the feasible domain.
total prestressing losses. Hence, an iterative calculation From Figure 15, one can specify the most appropriate
should be performed, as shown in Figure 13. solution. For example, the smallest height that meets all
 17517648, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/suco.70247 by Capes, Wiley Online Library on [29/07/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
12 DE MORAES and BUTTIGNOL

F I G U R E 1 6 Summary of the verification in terms of

utilization capacity.

and 60%, indicating underutilization. The deformation

capacity (SLS-DEF) also shows an extra capacity as the
FIGURE 14 Cross-section of the girder (units in cm). degree of utilization is around 60%. The restraining condi-
tion is the solicitation on the tension side both at the time
of prestressing and during the service life (SLS-D and SLS-
F). The shear force (ULS-Shear) is 100% since the shear
reinforcement is designed for the maximum shear force.

2.5 | Modal analysis

After optimizing the structure (defining the number of sup-

ports) and the girders (defining the girder height), a modal
analysis is performed to determine the main modes of
vibration and the respective natural frequencies. This anal-
ysis should be performed as a last step, considering that the
natural frequency of the structure is a dynamic property
that depends on the stiffness and mass of the structure.
Understanding the natural frequency is crucial to
assessing whether the structure can resonate with applied
dynamic forces, such as wind, traffic, or seismic activity.
The values obtained should be compared with the refer-
ence values for each region or standard code. The ana-
lyses were conducted separately for the frame and the
bridge deck. Detailed information related to the dynamic
FIGURE 15 Simplex solution highlighting the feasible analysis can be found in Paultre.24
domain. The modal analysis was carried out using the plugin
Karamba3D. The Assemble Model component generates
the constraints is located in the intersection between the global stiffness matrix [K], the mass matrix [M] and
curves 2 and 3. In this case, a height of 122 cm and an load vector {F}. The modal analysis is performed using
initial prestressing force of 4257.32 kN are found. From the Eigenvalue Solver component, which computes the
the optimized solution, the shear and flexural reinforce- generalized eigenvalue problem shown in Equation 28.
ments in the ULS are calculated. A minimum
 
reinforcement of 0.15% of the cross-sectional concrete K  ω2
M
ϕ ¼ 0 ð28Þ
area is considered for the flexural passive reinforcement.
The utilization degree can be measured for all verifica- where: ω = natural frequency; K = stiffness matrix;
tions, as shown in Figure 16. From the graph, one can M = mass matrix; ϕ = mode shape.
identify the design limitations. In this case, the solicitation The eigenvalue solver calculates the natural frequen-
in the compression region (SLS-CE) ranges between 40% cies (ω) and the corresponding mode shapes (ϕ) of the
 17517648, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/suco.70247 by Capes, Wiley Online Library on [29/07/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
DE MORAES and BUTTIGNOL 13

FIGURE 17 Modal analysis. Graphical visualization in Rhinoceros (top, front, right, perspective).

TABLE 3 Natural frequencies—f [Hz]. Department of Transportation,3 AASHTO,4 DNIT,2 and

Jahjouh and Erhan5 optimized girders is conducted.
Mode Frame Deck Global
The study performed by Aydn and Ayvaz1 uses a
1 0.085 0.852 0.021 modified genetic algorithm to achieve the lowest cost
2 — 6.460 0.159 solution, considering design variables such as the number
3 — 8.643 0.213 of spans, cross-sectional dimensions of the prestressed
4 — 9.132 0.225
girders, and the area of the prestressing steel. In contrast,
the optimization method described in this paper uses a
simplified procedure, focusing on minimizing the cost by
structure. The natural frequencies indicate the frequencies introducing AFs that convert direct and indirect costs
at which the structure will resonate. The mode shapes into fictitious equivalent global volume factors, as
show the deformation pattern of the structure at each nat- described in section 2.2.
ural frequency, helping to identify potential issues with The main difference between the analysis presented in
dynamic loading. The results are graphically displayed this study and that performed by Aydn and Ayvaz1 lies in
using post-processing tools inside Grasshopper. the conceptual design. On the one hand, Aydn and Ayvaz1
The results of this analysis are shown in Figure 17. perform a more sophisticated cost optimization analysis to
The main values of the natural frequencies are presented determine the optimal bridge configuration. On the other
in Table 3. hand, the method described in this paper is based on two
main stages. An initial conceptual design, which includes
preliminary design procedures and a cost optimization, fol-
3 | C AS E S T UDI E S lowed by a detailed design procedure, including dynamic
analysis. The preliminary design aims to search for the
Two case studies are investigated. In the first case, the pro- optimal global bridge layout, considering the number of
cedure described in section 2 was applied to the case study supports, spans, costs, topography, and climate (tempera-
investigated by Aydn and Ayvaz.1 The objective was to opti- ture, relativity humidity), as described in sections 2.1 and
mize the total cost of the bridge by determining the optimal 2.2. Thus, this procedure ensures that all relevant design
number of spans and the optimized geometric properties of steps are addressed, which makes this method equally sig-
a multi-span bridge. In the second case, a comparative anal- nificant. The methodology allows for a local optimization
ysis of the structural efficiency of Caltrans: California of prestressed girders using the Simplex algorithm, while
 17517648, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/suco.70247 by Capes, Wiley Online Library on [29/07/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
14 DE MORAES and BUTTIGNOL

TABLE 4 Input properties for the first case study. enabling global cost optimization of the entire bridge
through a genetic algorithm approach.
Parameter Value Unit
For the comparative analysis, design parameters were
Total length of bridge 177.7 m
adopted based on technical references and commonly
Width of deck 12 m accepted structural engineering practices, as described in
Characteristic compression strength 40 MPa Table 4.
Bridge load class 450 kN The 3D parametric superstructure model (section 2.1)
Tendons CP190 — was adjusted according to the input parameters of
Table 4 to perform the static analysis and extract the
Relative humidity 40 %
internal forces corresponding to different spans. Several
Slump 8 cm
structural analyses were performed, varying the free span
Mean temperature 20 Degree between 20 and 50 m, considering increments of 5 m.
Anchored slip 4 mm Then, the Simplex method was applied to determine the
Friction coefficient 0,2 — space of solutions for different spans, as shown in
Number of girders 5 — Figure 18.
Thickness of concrete slab 20 cm From the space of solutions for different heights, the
combination of variables (h and Pi ) that result in

FIGURE 18 Space of solutions for 40 MPa considering different spans.

 17517648, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/suco.70247 by Capes, Wiley Online Library on [29/07/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
DE MORAES and BUTTIGNOL 15

F I G U R E 1 9 Summary of the
solution for 40 MPa.

the smallest concrete section for each scenario was study, the cost estimate for the alternative case with 7 sup-
extracted. These values were plotted in Figure 19 and the ports and 6 spans is $ 400/m3  6286 m3 = $ 2,514,400.00.
following equation was obtained to define the height of A difference was expected between the values due to
the beam as a function of the span: h ¼ 0:066
L  0:24 the variations in both the geometry and the applied
(where: h is the beam height; L is the free span). This methodology. It is important to note that the objective of
equation defines the height of the girders and, thus, the this study is not to determine the exact cost of the struc-
height of the web, as function of the span. ture, but rather to identify the solution that approximates
The AFs presented in Table 5 were adopted. These the most cost-effective bridge configuration.
values were determined based on the data provided by The optimization method was also applied to com-
Aydn and Ayvaz1 (Table 2—unit prices) and estimations pare the results obtained by Jahjouh and Erhan.5 The
of the reinforcement ratio for the structural elements. authors adopted a modified harmony search algorithm to
The equation that defines the height of the beam, the determine the optimal geometry of precast prestressed
factor for obtaining the global fictitious volume of concrete I-beams. The authors compared the results of
the structure, and the line of the natural terrain are intro- the shape optimization with the standard prestressed
duced in the parametric model. girders specified by AASHTO.3,4 The optimized sections
The optimal solution is a five-span bridge, with each showed 13% to 17% more structural efficiency
span measuring approximately 36 m and supported by (Equation 29) and 14% to 21% more efficiency in terms of
two-column piers, resulting in a fictitious concrete vol- the structural efficiency factor (Equation 30).
ume of 6218 m3.
Figure 20 shows the values of the equivalent r2
ρ¼ ð29Þ
fictitious concrete volume used to determine the optimized yi
ys
configuration. To estimate the total cost of each scenario,
simply multiply the corresponding fictitious concrete vol- 3:46
W i
α¼ ð30Þ
ume by the unit price of the concrete cubic meter, which Ac
h
was considered equal to $ 400. Therefore, the cost
estimation for the optimized configuration is $ where: r = radius of gyration of the girder; ys and
400/m3  6218 m3 = $ 2,487,200.00. In Aydn and Ayvaz,1 yi = distances from the centroid of the girder section to
the optimal configuration consists of 7 supports and the top fiber and bottom fiber; W i = section modulus for
6 spans of approximately 30 m. In this case, the global the bottom fiber; Ac = cross-sectional area of the girder;
value is $ 2,686,504.00. Considering the criteria of this h = depth of the girder.
 17517648, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/suco.70247 by Capes, Wiley Online Library on [29/07/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
16 DE MORAES and BUTTIGNOL

T A B L E 5 Cost breakdown and

Description Unit Qty. Unit Price Total Price AF
adjustment factors for various structural
Girders m3 1.00 2150.00 2150.00 5.38 elements, where: PC = prestressed
3
PC m 1.00 900.00 900.00 concrete; PS = steel for active
PS kg 200.00 1.50 300.00 reinforcement; F. and S. = formwork
and shoring.
Reinf. steel kg 100.00 9.50 950.00
3
Slab m 1.00 400.00 400.00 1.00
3
Concrete m 1.00 220.00 220.00
Reinf. steel kg 120.00 1.50 180.00
3
Trans. Beam m 1.00 430.00 430.00 1.08
Concrete m3 1.00 220.00 220.00
Reinf. steel kg 140.00 1.50 210.00
3
Crossbeam m 1.00 430.00 430.00 1.08
3
Concrete m 1.00 220.00 220.00
Reinf. steel kg 140.00 1.50 210.00
Column m3 1.00 420.00 420.00 1.05
3
Concrete m 1.00 220.00 220.00
Reinf. steel kg 80.00 1.50 120.00
F. and S. vb 1.00 80.00 80.00
3
Foundation m 1.00 424.00 424.00 1.06
3
Concrete m 1.00 220.00 220.00
Lean concrete m3 0.05 80.00 4.00
Reinf. steel kg 80.00 1.50 120.00
Excavation vb 1.00 80.00 80.00
3
Abutments m 1.00 424.00 424.00 1.06
3
Concrete m 1.00 220.00 220.00
Lean conc. m3 0.05 80.00 4.00
Reinf. steel kg 80.00 1.50 120.00
Excavation vb 1.00 80.00 80.00

TABLE 6 Input properties for the second case study.

Parameter Value Unit

Characteristic compression strength 35 MPa
Width of deck 14 m
Bridge load class 450 kN
Tendons CP190 —
Relative humidity 60 %
Slump 8 cm
Mean temperature 20 Degree
F I G U R E 2 0 Free span versus equivalent concrete volume for
Anchored slip 4 mm
the first case study.
Friction coefficient 0.2 —

The concept of structural efficiency for prestressed given cross-sectional area. This theoretical framework led
concrete sections was originally proposed by Guyon,25 to the development of practical indices used in bridge
who defined an efficiency factor based on maximizing engineering to evaluate the performance of precast pre-
the section moduli at the top and bottom fibers for a stressed girders. Among these, the structural efficiency
 17517648, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/suco.70247 by Capes, Wiley Online Library on [29/07/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
DE MORAES and BUTTIGNOL 17

F I G U R E 2 1 Curves of fck versus

height for different spans (L).

FIGURE 22 Curves of fck versus initial prestressing force (Pi) for different spans (L).

F I G U R E 2 3 Comparison of
structural efficiency ratio.

factor (ρ) and the structural efficiency ratio (α) are com- These concepts are not adopted in European practice.
monly used in the United States bridge engineering As stated by Nowak et al.,31 research related to bridge
community and appear frequently in studies published in design and evaluation has evolved independently in
the PCI Journal Anderson,26 Rabbat and Russell,27 Europe and North America, shaped by region-specific
Seguirant,28 Kim and Siriwardanage,29 Wang and Kim.30 conditions and priorities.
 17517648, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/suco.70247 by Capes, Wiley Online Library on [29/07/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
18 DE MORAES and BUTTIGNOL

FIGURE 24 Comparison of structural efficiency factor.

Despite that, according to Ellingwood et al.,32 modern Figure 22 depicts the relationship between the initial
national bridge design codes are based on the principles prestressing force (Pi ) and the concrete compressive
of probability-based limit states design. Although their strength for different free spans. As f ck increases, Pi fol-
implementation differs across regions, the underlying lows an upward trend, indicating the need for a higher
design philosophy is similar. At present, the growing use prestressing force to enable a more effective utilization of
of innovative materials, such as ultra high performance the concrete section.
concrete (UHPC) Foster and Bentz,33 and the adoption of The structural efficiency factor (ρ) and the structural
advanced optimization techniques for structural elements efficiency ratio (α) were calculated for the standard
(e.g., topology34) present new challenges to design engi- DNIT2 girders optimized with respect to the admissible
neers. In this context, the definition of standard perfor- minimum height and compared with the coefficients pre-
mance indicators, including sustainability, safety, and sented in Jahjouh and Erhan5 for the AASHTO,4 Cal-
reliability, is essential, yet still evolving. trans: California Department of Transportation3 standard
For the comparison with Jahjouh and Erhan5 results, girders, and the optimized girders described in Jahjouh
the analysis considers the standard recommendations of and Erhan.5
DNIT,2 as shown in Figure 7. It was not possible to iden- The structural efficiency factor (ρ) considers the
tify the environmental conditions and the strength of the radius of gyration of the beam cross-section (r2) and the
concrete adopted by Jahjouh and Erhan.5 For comparison distances from the girder centroid to the top (ys ) and bot-
purposes, the input properties shown in Table 6 were tom (yi ) fibers. The higher the value of r, the greater the
adopted. beam ability to resist bending moments due to the greater
The Simplex method was applied to explore the space distribution of the area around the centroid. Figure 23
of solutions for different span lengths and to find the shows the variation of (ρ) as a function of h. Jahjouh and
solutions that present the lowest height. Erhan5 presents better results in terms of structural effi-
The analysis considers a compressive strength of ciency, with a consistently high (ρ), demonstrating that
35 MPa and the height of the section includes a slab optimized sections are better to distribute stresses along
thickness of 20 cm. Additional calculations were per- the fibers. Caltrans: California Department of Transpor-
formed by varying the f ck from 30 to 90 MPa, maintaining tation3 shows lower values of (ρ), particularly for taller
the same design criteria. Figures 21 and 22 demonstrate, girders, suggesting a greater limitation of the structural
respectively, the variation of the girder height (h) and the efficiency in terms of stress distribution. AASHTO4 pre-
initial prestressing force (Pi ) as a function of the concrete sents intermediate values, showing a gradual increase in
characteristic strength for different free span (L) values. (ρ) for taller girders. DNIT2 shows a more significant
Figure 21 illustrates the reduction in girder height increase of (ρ) up to approximately 150 cm, followed by a
with increasing concrete compressive strength. Each convergence, for larger sections, to those of AASHTO4
curve represents a specific free span. This trend is and Caltrans: California Department of Transportation.3
expected, as a higher compressive strength allows for The structural efficiency ratio (α) relates the stiffness
shallower beams to resist the same bending forces. of the beam to its area and height. A higher value of α
 17517648, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/suco.70247 by Capes, Wiley Online Library on [29/07/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
DE MORAES and BUTTIGNOL 19

indicates that the beam can support a greater load with C O N F L I C T O F I N T E R E S T S T A TE M E N T

less material (higher efficiency). Figure 24 shows the rela- The authors declare no potential conflict of interest.
tionship between α and the height of the beam (h) for dif-
ferent cross-sections. AASHTO4 and Jahjouh and Erhan5 DA TA AVAI LA BI LI TY S T ATE ME NT
show higher values of α for larger heights. This suggests The data that support the findings of this study are avail-
that these methods favor sections that maximize stiffness, able from the corresponding author upon reasonable
particularly in taller elements. Caltrans: California request.
Department of Transportation3 shows moderate effi-
ciency values, reflecting a more conservative criterion. ORCID
DNIT2 shows an increasing curve of α with increasing h, Thomaz Eduardo Teixeira Buttignol https://orcid.org/
suggesting that sections with lower height are less effi- 0000-0002-7004-571X
cient. In contrast, for taller sections, the efficiency gain is
significant and stabilizes near 0.85. RE FER EN CES
1. Aydn Z, Ayvaz Y. Overall cost optimization of prestressed con-
crete bridge using genetic algorithm. KSCE J Civ Eng. 2013;17:
4 | C ON C L U S I ON S 769–76. https://doi.org/10.1007/s12205-013-0355-4
2. Departamento Nacional de Infraestrutura de Transportes
(DNIT). Publicação IPR-751: Álbum Projeto Tipo de Pontes
The integrated design approach described in this paper
Semipermanentes. 2022 [cited 2024 Aug 13]. Available from:
offers an efficient method for the preliminary design of https://www.gov.br/dnit/pt-br/assuntos/planejamento-e-
prestressed beam bridges. The methodology allows the pesquisa/ipr/coletanea-de-manuais/vigentes/publicacao-ipr-
optimization based on a cost analysis employing a simpli- 751-album-projeto-tipo-de-pontes-semipermanentes
fied procedure that adopts AFs that represent fictitious 3. Caltrans: California Department of Transportation. Bridge
equivalent global volumes of the concrete elements. This design practice. 2022.
approach allows for the global optimization of the struc- 4. American Association of State Highway and Transportation
ture, determining the best arrangement of spans and sup- Officials (AASHTO). Aashto lrfd bridge design specifications.
2020.
ports based on direct and indirect costs. The results are in
5. Jahjouh M, Erhan S. Optimization of prestressed concrete bridge
agreement with 23 and 24, demonstrating the effective- girder section using a modified harmony search algorithm. Struc-
ness of employing GA methods and parametric modeling ture. 2022;46:625–36. https://doi.org/10.1016/j.istruc.2022.10.093
to achieve improved cost–benefit structures. 6. Girardet A, Boton C. A parametric bim approach to foster
The Simplex method was used for the local optimi- bridge project design and analysis. Autom Constr. 2021;126:
zation of the prestressed girders after defining the 103679. https://doi.org/10.1016/j.autcon.2021.103679
global optimized bridge configuration. This approach 7. Tang S, Shelden DR, Eastman CM, Pishdad-Bozorgi P, Gao X.
A review of building information modeling (bim) and the inter-
allows the designer to select the solution that best
net of things (iot) devices integration: present status and future
meets the design requirements. Analyses considering
trends. Autom Constr. 2019;101:127–39. https://doi.org/10.
different fck have shown an increase of 50% in the pre- 1016/j.autcon.2019.01.020
stressing force by raising the fck from 30 MPa to 8. Lai Y, Li Y, Huang M, Zhao L, Chen J, Xie YM. Conceptual
90 MPa. Besides, the girder web height showed a non- design of long span steel-uhpc composite network arch bridge.
linear decrease with the increase of fck. The decrease Eng Struct. 2023;277:115434. https://doi.org/10.1016/j.
was more pronounced for conventional concretes engstruct.2022.115434
(between 30 and 50 MPa). Thus, to improve the perfor- 9. Li Y, Lai Y, Lu G, Yan F, Wei P, Xie YM. Innovative design of
long-span steel–concrete composite bridge using multi-material
mance of standard I girders, especially for high strength
topology optimization. Eng Struct. 2022;269:114838. https://
concretes, a shape optimization is recommended, as doi.org/10.1016/j.engstruct.2022.114838
proposed by Jahjouh and Erhan.5 10. Fan W, Chen Y, Li J, Sun Y, Feng J, Hassanin H, et al. Machine
learning applied to the design and inspection of reinforced con-
A C K N O WL E D G M E N T S crete bridges: resilient methods and emerging applications.
The authors gratefully acknowledge the research funding Structure. 2021;33:3954–63. https://doi.org/10.1016/j.istruc.
of both “Fundo de Apoio ao Ensino, à Pesquisa e à Exten- 2021.06.110
são” (Faepex), grant number 2474/23 and National Coun- 11. Kovačevi
c M, Antoniou F. Machine-learning-based consump-
tion estimation of prestressed steel for prestressed concrete
cil for Scientific and Technological Development (CNPq),
bridge construction. Buildings. 2023;13(5):1187. https://doi.org/
grant number 403517/2023-0. The Article Processing 10.3390/buildings13051187
Charge for the publication of this research was funded by 12. Nawari N. Intelligent design codes. In: Proceedings of the 2009
the Coordenação de Aperfeiçoamento de Pessoal de Nível Structures Congress – Don't Mess with Structural Engineers:
Superior - Brasil (CAPES) (ROR identifier: 00x0ma614). Expanding Our Role, 04 2009, pp. 1–10.
 17517648, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/suco.70247 by Capes, Wiley Online Library on [29/07/2025]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
20 DE MORAES and BUTTIGNOL

13. Salehi H, Burgueño R. Emerging artificial intelligence methods 30. Wang J, Kim YJ. A state-of-the-art review of prestressed con-
in structural engineering. Eng Struct. 2018;171:170–89. https:// crete tub girders for bridge structures. J Infrastruct
doi.org/10.1016/j.engstruct.2018.05.084, https://www. Preserv Resil. 2022;3(1):13. https://doi.org/10.1186/s43065-022-
sciencedirect.com/science/article/pii/S0141029617335526 00058-1
14. Rempling R, Mathern A, Tarazona Ramos D, Luis 31. Nowak AS, Park CH, Casas JR. Reliability analysis of pre-
Fern
andez S. Automatic structural design by a set-based para- stressed concrete bridge girders: comparison of Eurocode,
metric design method. Autom Constr. 2019;108:102936. Spanish Norma IAP and AASHTO LRFD. Struct Saf. 2001;
https://doi.org/10.1016/j.autcon.2019.102936 23(3):331–44. https://doi.org/10.1016/S0167-4730(01)00014-6
15. Ruiz R, Starchevich GB, Todisco L, Corres H. Parametric analysis 32. Ellingwood BR, Vrouwenvelder T, Gulvanessian H. Eurocodes
of variables influencing the design of i-girder bridge decks made and their implications for bridge design: background, imple-
of high-performance fiber reinforced concrete. Struct Concr. mentation, and comparison to north american practice.
2023;24(4):4606–23. https://doi.org/10.1002/suco.202200477 J Bridge Eng. 2021;19:3–4. https://doi.org/10.1061/(ASCE)BE.
16. Tedeschi A. AAD – algorithms-aided design: parametric strate- 1943-5592.0000567
gies using grasshopper. 2nd ed. Italy: Arturo Tedeschi; 2014. 33. Foster SJ, Bentz E. Design of uhpc prestressed girders for shear.
17. Preisinger C. Linking structure and parametric geometry. Archi- Struct Concr. 2024;25(2):780–95. https://doi.org/10.1002/suco.
tectural Design. 2013;83(2):110–3. https://doi.org/10.1002/ad.1564 202300738
18. Associação Brasileira de Normas Técnicas (ABNT). NBR 6118: 34. Briseghella B, Fenu L, Lan C, Mazzarolo E, Zordan T. Applica-
Projeto de estruturas de concreto – Procedimento. 2024b. Avail- tion of topological optimization to bridge design. J Bridge Eng.
able from: https://www.abntcatalogo.com.br/norma.aspx?ID= 2013;18(8):790–800. https://doi.org/10.1061/(ASCE)BE.1943-
3499 5592.0000416
19. Lounis Z, Cohn MZ. An approach to preliminary design of pre-
cast pretensioned concrete bridge girders. Comput Aided Civ
Inf Eng. 1996;11(6):381–93. https://doi.org/10.1111/j.1467- AUTHOR BIOGRAPHIES
8667.1996.tb00351.x
20. Kaveh A, Maniat M, Arab Naeini M. Cost optimum design of Vinicius Nascimento de Moraes,
post-tensioned concrete bridges using a modified colliding bod-
Department of Structures, Univer-
ies optimization algorithm. Adv Eng Softw. 2016;98:12–22.
https://doi.org/10.1016/j.advengsoft.2016.03.003
sity of Campinas, Campinas, São
21. European Committee for Standardization (CEN). EN 1992-1-1: Paulo, Brazil. Email: vinicius.
2024 – Eurocode 2: Design of concrete structures – Part 1–1: Gen- [email protected]
eral rules and rules for buildings, bridges and civil engineering
structures. Brussels: European Committee for Standardization.
CEN Standard EN 1992-1-1:2024. 2024.
22. Associação Brasileira de Normas Técnicas (ABNT). ABNT NBR
7188:2024 – Ações devido ao tr
afego de veculos rodovi
arios e de Thomaz Eduardo Teixeira But-
pedestres em pontes, viadutos e passarelas (In Portuguese). tignol, Department of Structures,
2024a. University of Campinas, Campinas,
23. Associação Brasileira de Normas Técnicas (ABNT). NBR 7187: São Paulo, Brazil. Email: thomazb@
Projeto de pontes, viadutos e passarelas de concreto. 2022.
unicamp.br
Available from: https://www.abntcatalogo.com.br/norma.aspx?
ID=3499
24. Paultre P. Dynamics of structures. 1st ed. Boca Raton, FL: CRC
Press; 2010.
25. Guyon Y. Prestressed concrete. New York: John Wiley and
Sons, Inc; 1953.
26. Anderson R. Optimized sections for precast prestressed bridge
girders. PCI J. 1982;27(4):89–101. How to cite this article: de Moraes VN,
27. Rabbat BG, Russell HG. Optimized sections for precast pre-
Buttignol TET. Parametric modeling for automated
stressed bridge girders. PCI J. 1982;27(4):88–104.
28. Seguirant SJ. New deep wsdot standard sections extend spans
intelligent design of concrete beam bridges: A
of prestressed concrete girders. PCI J. 1998;43(4):92–119. global and local optimization approach. Structural
29. Kim YJ, Siriwardanage T. New lrfd-based prestressed concrete Concrete. 2025. https://doi.org/10.1002/suco.70247
bulb-tee girders in colorado. PCI J. 2020;65(3):53–63.