Professional Documents
Culture Documents
Miluccio Et Al 2021
Miluccio Et Al 2021
Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct
A R T I C L E I N F O A B S T R A C T
Keywords: Even if previous studies investigated seismic fragility of existing bridges, some recent collapse cases set the need
Existing prestressed concrete bridges to investigate their vulnerability under gravity loads. In this study, the class of simply supported, beam-type,
Traffic loads prestressed concrete bridges built between 1970 and 1980 was considered to carry out a fragility analysis of
Collapse capacity
the most recurrent type of Italian bridges to traffic loads. Based on the literature and new data collected by the
Sensitivity analysis
Fragility analysis
authors from real bridges, geometric, material and load variables were defined through probability distributions
and regression models. Then, a code-based capacity modelling of bridge decks was carried out, followed by the
implementation of a simplified deck analysis method that allows extensive simulations. In a first step, a sensi
tivity analysis was performed to identify the random variables that mostly affect the structural response, to be
taken into account in the fragility analysis for the ultimate limit state. Then, a fragility analysis was carried
through a fully automatic procedure implemented in MATLAB. The main output consists of fragility models that
might be used by engineers, roadway management companies and decision-makers in large-scale risk assess
ments of existing highway bridges under code-based traffic loads, as a basis for prioritization of more refined
performance evaluations and structural retrofitting programmes.
* Corresponding author.
E-mail address: fulvio.parisi@unina.it (F. Parisi).
https://doi.org/10.1016/j.engstruct.2021.113367
Received 25 May 2021; Received in revised form 17 September 2021; Accepted 12 October 2021
Available online 23 October 2021
0141-0296/© 2021 Elsevier Ltd. All rights reserved.
G. Miluccio et al. Engineering Structures 249 (2021) 113367
load models (TLMs) that are different from those prescribed by [20] and DM 29/07/1974 [21] to include RC, PC and steel structures. As
current code provisions. a common feature, all those codes adopted the permissible stress design
(ii) Structural materials used in the past do not comply with modern (PSD) method, assuming nominal loads and a linear elastic behaviour for
building codes, hence requesting an appropriate modelling of all structural materials.
their mechanical properties. Dealing with material properties, RD 2229/1939 [19] only pre
scribed a concrete mixture with cement unit weight by volume of 300
The main goal of this study is to investigate the structural fragility of kg/m3 in order to obtain a minimum value of concrete cubic compres
existing Italian bridges subjected to traffic loads, based on code- sive strength (Rck) of 12 MPa and 16 MPa for RC and PC structures,
conforming capacity models and analysis procedures [16]. Specif respectively. Such minimum requirement was updated by DM 30/05/
ically, assuming code-conforming load patterns on the bridge deck, 1972 [20], which prescribed Rck ≥ 30 MPa for PC structures. A simpli
fragility is herein defined as the conditional probability of exceeding a fied safety check in terms of ultimate bending capacity was allowed,
damage state given the traffic load intensity. A class of simply supported, assuming a minimum global safety factor of 1.75. In this respect, design
beam-type, PC bridges is considered, because they represent more than codes did not prescribe partial safety factors for material properties and
90% of the Italian bridge stock [12]. Previous studies also reported that loads. Through the years, mild reinforcing steel was regulated according
the time frame 1960–1980 is the period when the highest number of to different classes under varying tensile properties. In PC structures, a
bridges were built due to the Italian economic growth and infrastructure minimum reinforcing steel ratio (ρs) of 0.1% was prescribed, calculated
development (see, e.g., [10,12]). Therefore, based on available data as the ratio of the total reinforcing steel area As over the tension area of
collected by the authors on real case-study bridges, this study focuses on the concrete (i.e., web area plus section enlargement at the tension side).
the traffic-load fragility of bridges built in the period 1970–1980. After According to DM 30/05/1972 [20], prestressing steel was regulated in
that traffic loads and uncertainties on both material and geometric the form of wires, strands, and bars for use in both post-tensioned and
properties of the bridge decks were modelled, a simplified yet code- pre-tensioned structures. No strength acceptance limits were defined for
conforming structural analysis was carried out to predict bending and prestressing steel because a qualification process of the manufacturer
shear strength demands to be compared with their capacity counterparts was required. This contributed to explain the reason for a limited
for the ultimate limit state (ULS). A sensitivity analysis was performed to number of available tensile tests on prestressing steel over the last
assess the influence of different geometric and mechanical properties on decades.
the structural performance, whereas a validation of the simulated design
process confirms the representativeness of adopted bridge models. The 2.2. Current code provisions
final stage of this study dealt with fragility analysis, where uncertainties
were modelled and propagated through Monte Carlo simulation and Differently from past codes, the most recent Italian regulations are
structural analysis. The computational procedure behind sensitivity and provided by DM 14/01/2018 [16] (NTC 2018 hereafter) both in terms of
fragility analyses was fully implemented in MATLAB software [17], TLMs and design provisions according to Eurocodes. NTC 2018 [16] are
allowing the impact of different modelling options to be evaluated. based on the semi-probabilistic design method (i.e., Level I reliability
method), controlling the variability in loads and material properties
2. Brief overview of code provisions for selected bridges through partial safety factors in order to meet a target failure probability
[16,22–23]. Conversely, this study makes use of a full probabilistic
To support the description of the main assumptions and methodol (Level III) reliability method [23], assuming geometric and material
ogy of this study, in this section code provisions related to the period properties as random variables (RVs) and code-based TLMs. More ac
1970–1980 and those currently in force in Italy are briefly described curate TLMs based on real traffic distributions might be alternatively
with a focus on the changes of TLMs and design provisions. considered in future studies due to an increasing number of weigh-in-
motion (WIM) systems on Italian highways [24–26]. Table 1 outlines
2.1. Code provisions at the time of construction the TLM prescribed by NTC 2018 [16], including the spacing and axles
distribution for each lane [27]. Moreover, a uniform distributed load of
The bridges under study were designed according to TLMs provided 2.5 kN/m2 is considered in the remaining part of the deck to be loaded
by the Italian regulation Circ. CSLLPP 384/1962 [18], which defined the when relevant.
following categories of bridges: Category I bridges, opened to both
military and civil vehicles; and Category II bridges, opened only to civil 3. Methodology
vehicles. Most of bridges were designed against military vehicles ac
cording to the following TLM: a military lane, considering load types This study consisted of the following steps:
(LTs) denoted as LT4, LT5 or LT6, as listed below; one or more regular
truck lanes (LT1) and pedestrians on the sidewalks (LT3). The load type (i) derivation of statistics, probability distributions and regression
LT2 was only considered for Category II bridges. LTs were associate with models for material and geometric properties, based on data
vehicle types or pedestrians as follows: available in the literature and others collected on real bridges
built in the period 1970–1980;
• LT1: multiple 2-axles truck load of 120 kN with width equal to 3.11 (ii) geometric and capacity modelling of bridge decks, according to
m; the current Italian code [16];
• LT2: single 2-axles steamroller load of 180 kN with width equal to (iii) sensitivity analysis, which was based on the generation of ca
3.11 m; pacity models using statistics of material and geometric proper
• LT3: uniformly distributed load of 4 kN/m2; ties, followed by structural analysis to identify the modelling
• LT4: multiple 6-axles military load of 615 kN with width equal to 3.5
m;
Table 1
• LT5: multiple 6-axles military load of 320 kN with width equal to 3.5
Traffic load model according to current Italian code [16].
m;
Lane Number of axles Total axle load Qk [kN] Distributed load qk [kN/m2]
• LT6: single 6-axles military load of 745 kN with width equal to 3.5 m.
1 2 300 9
At the time of bridge construction, RC bridges were designed ac 2 2 200 2.5
3 2 100 2.5
cording to RD 2229/1939 [19], which was replaced by DM 30/05/1972
2
G. Miluccio et al. Engineering Structures 249 (2021) 113367
variables that mostly influence the performance of bridge decks across those periods.
under traffic loading; As described in [28–30], uncertainties can be divided in two
(iv) fragility analysis, which was based on the random generation and different groups, namely, aleatory and epistemic uncertainties. In this
analysis of capacity models to assess the conditional probability study, epistemic uncertainties related to the bridge deck analysis
of exceeding ULS of bridge decks under varying traffic load method and capacity models are not taken into account, so they will be
intensity. implemented in future studies. By contrast, aleatory uncertainties were
duly considered as described below.
All these steps were fully implemented in MATLAB [17], delineating
the flowcharts shown in Fig. 1a and 1b for sensitivity and fragility an 3.1.1. Independent variables
alyses, respectively. In Section 3.1, all variables are defined through The database of material properties was based on laboratory test
either probability distributions or regression models. In Section 3.2, the results collected by the authors. Two different distributions were fitted
assumptions for capacity modelling of the bridge deck are illustrated, to data on concrete compressive strength fc and yield strength of rein
whereas Section 3.3 describes the structural analysis procedure to forcing steel fy, differentiating between data sets related to the periods
evaluate the effects of loads on bridge decks via engineering demand 1960–1970 and 1970–1980. Fig. 2a and 2b show three alternative
parameters (EDPs) and structural performance via demand-to-capacity probability distributions fitted to experimental data of periods
ratios (DCRs). In Sections 4 and 5, criteria and results of sensitivity 1960–1970 and 1970–1980, respectively, to model the uncertainty in fc,
and fragility analyses are respectively discussed. namely normal, lognormal and Weibull distributions. Based on best
fitting outcomes, the lognormal distribution with mean value fcm = 34.7
MPa and coefficient of variation CoV = 24.4% was assumed for the
3.1. Uncertainty modelling period 1960–1970. Those statistics changed to fcm = 38.5 MPa and CoV
= 11.4% for the period 1970–1980, thus highlighting an improved
RVs included both geometric and material properties. Based on data quality of concrete. The same procedure was applied to derive a prob
available in the literature [12] and those collected on real bridges by the ability distribution for yield strength fy of reinforcing steel (Fig. 3a and
authors of this study, a subset of RVs were modelled through probability 3b). Best fitting produced a lognormal distribution with mean value fym
distributions, whereas other RVs were assumed to be statistically = 436 MPa and CoV = 6.4% for the period 1960–1970, which changed
dependent upon the former RVs according to regression models. To that to fym = 451 MPa and CoV = 7.2% in the period 1970–1980.
aim, the authors examined the following information on PC girder A probability distribution was also fitted to experimental data on the
bridges built in the period 1970–1980: conventional yield strength of prestressing steel fp,01 for bridges built in
the period 1970–1980, fitting a lognormal distribution with a mean
• data on geometric properties; value and CoV equal to 1665 MPa and 2.5%, respectively (Fig. 4).
• experimental data on mechanical properties of concrete (100 speci Despite the limited amount of data available on such a mechanical
mens), reinforcing steel (65 specimens), and prestressing steel (21 property, these statistics agree well with different studies available in
specimens). the literature (e.g., [31–32]).
The effectiveness of the above-mentioned distributions was tested
Case-study bridges have different locations in Italy and were built by through the Kolmogorov-Smirnov (KS) test, assuming a significance
different construction companies to be representative of the selected level of 5%. The null hypothesis (i.e., the hypothesis that the data is
class of bridges. In the following sections, probability distributions and distributed as the selected probability distribution) was never rejected
regression models are described, differentiating between RVs modelled because the p-value was always found to be higher than 0.05. None
as statistically independent features and those predicted starting from theless, a lognormal distribution was assumed in line with current Ital
their knowledge. The discussion includes a comparative analysis of ian codes for existing bridges [4,8].
experimental data on mechanical properties of concrete and reinforcing The different values of material properties (especially concrete
steel related to PC bridges that were constructed in the decades strength due to different regulations that changed from RD 2229/1939
1960–1970 and 1970–1980, because of the evolution in Italian codes
3
G. Miluccio et al. Engineering Structures 249 (2021) 113367
Fig. 2. Probability distributions for concrete compressive strength: (a) 1960–1970; (b) 1970–1980.
Fig. 3. Probability distribution for reinforcing steel yield strength: (a) 1960–1970; (b) 1970–1980.
according to PSD method under the same TLMs (i.e., those provided by
Circ. CSLLPP 384/1962 [18]). This motivated the authors not to
distinguish between the two decades, hence aggregating all data for
uncertainty modelling of geometric parameters in order to obtain a
larger data set. The cross sections of the deck and longitudinal girders
are respectively shown in Fig. 5a and b, highlighting the most relevant
geometric properties considered in the analysis.
The span length L and width W of the deck were assumed as main
independent RVs. Based on data available in the literature [12] and
those collected by the authors on case-study bridges, the histogram of
the bridge span length was derived. Approximately half of the existing
bridges have a span length between 30 m and 35 m, as depicted in Fig. 6.
A lognormal distribution was assumed for L, with mean value equal to
33.2 m and CoV = 13.6%. The distribution was truncated at 15 m and
45 m to obtain realistic values of bridge length in subsequent probabi
Fig. 4. Probability distribution for conventional yield strength of prestress
listic simulation. The KS test was performed in order to demonstrate the
ing steel.
effectiveness of lognormal distribution and the null hypothesis was
rejected with a significance level of 5%. Regarding the deck width W, a
[19] to DM 30/05/1972 [20] in 1972), in addition to the lack of data on
probability mass function (PMF) with three values equal to 8.5, 12.25
prestressing steel before 1970 (no test results available before that
and 16 m having the same probability of occurrence was assumed. The
period), led the authors to assume different RVs for the two timeframes
slab thickness (denoted by s) was modelled through a uniform distri
(i.e., 1960–1970 and 1970–1980), retaining only the period 1970–1980
bution with mean value and CoV equal to 0.25 m and 12%, respectively,
for subsequent analysis. Nonetheless, particularly in the case of pre
in the range [0.2 m, 0.3 m].
stressing steel, the experimental data considered above does not include
The transverse reinforcement ratio ρsw = Asw/p (namely, the ratio
the effects of material deterioration because laboratory tests were per
between the transverse reinforcement area Asw and distance p between
formed at the time of construction of the real bridges. In this respect,
stirrups at girder ends) and prestressing ratio of the tendons σ sp/fp,01
corrosion effects should be simulated in future studies, because their
(namely, the ratio between the prestressing stress σsp and fp,01) were
impact on structural performance and fragility could be significant ac
considered as additional independent RVs. The former was modelled
cording to previous studies (e.g., [33]).
through a uniform distribution between 300 mm2/m (corresponding to
Geometric properties were defined for simply supported PC bridge
8-mm-diameter stirrups with 330 mm spacing as per minimum code
decks built during the whole timeframe 1960–1980 due to a common
requirement) and 1130 mm2/m (corresponding to 12-mm-diameter
structural system and design practice, which did not undertake relevant
stirrups with 200 mm spacing as maximum recorded value). The pre
changes. Throughout that period, beam-type decks were designed
stressing ratio was modelled in two different ways as follows: (i) uniform
4
G. Miluccio et al. Engineering Structures 249 (2021) 113367
Fig. 5. Cross sections of (a) half bridge deck and (b) longitudinal girder.
Table 2
Distributions and statistics of independent RVs for periods 1960–1970 and
1970–1980.
Category RV 1960–1970 1970–1980 Distribution
μ CoV μ CoV
[%] [%]
3.1.2. Statistically dependent variables with a coefficient of determination R2 = 0.51 and rounding the number
Data available on real case-study bridges allowed the authors to of girders to the closest integer number. The regression model was set to
observe some statistical (expected) dependence of some geometric have a number of girders between 2 and 8. After that nb is estimated, the
properties on those discussed in Section 3.1.1. Thus, regression models transverse girder-to-girder distance is calculated as ib = W/nb. It can be
were developed in order to model RVs conditioned upon independent noted that for each girder the effective width weff is equal to ib, according
RVs. The following dependent variables were considered: to geometric limitations given in Eurocode 2 [22]. A bivariate regression
was fitted to data on the girder height Hb as follows:
• number of longitudinal girders, nb; Hb = 0.28ib + 0.03L (2)
• height of longitudinal girder, Hb;
• gross area of girder cross section, Ab; with R2 = 0.81 describing the accuracy of the fitting. Eq. (2) provides
• width of girder top flange, Btop; higher values of Hb under increasing span length of the bridge and on-
5
G. Miluccio et al. Engineering Structures 249 (2021) 113367
centre distance of longitudinal girders, or equivalently the deck width observed that R2 and ρX,Y are rather close to unity, apart from the
given the number of girders. As expected, increasing nb would propor regression model defined by Eq. (1) that is affected by higher dispersion
tionally reduce Hb. This latter variable was set within the range [L/20, of data. In all cases, ρX,Y highlights a positive correlation between
3.2 m]. variables.
Geometric data also allowed the authors to develop the following The following deterministic variables were assumed:
quadratic regression model for prediction of girder gross sectional area
Ab given Hb (R2 = 0.84): 1. concrete cover equal to 30 mm;
2. girder web thickness equal to 0.2 m;
Ab = 0.43 + 0.06Hb2 (3)
3. bottom flange thickness equal to 0.3 m;
Eq. (3) allows the computation of girder self-weight per unit length 4. geometric ratio of longitudinal reinforcing steel ρs = 0.1%;
G1k as effective sectional area Ab,eff = Ab + weff ⋅s multiplied by the con 5. Young’s modulus of steel equal to 200 GPa.
crete unit weight γ c, where weff ⋅ s denotes the effective slab area of the
bridge deck. Such properties were settled as deterministic variables according to
The width of the top girder flange Btop was evaluated in order to values suggested by the code provisions during the period of construc
define the prestressing steel area Asp given ρsp. It is worth mentioning tion selected in this study. This assumption is justified by low variability
that both ρsp and ρs are defined in terms of tension area of the concrete of those properties (based on the data collected by the authors) and/or
cross section (i.e., web area plus section enlargement on the tension negligible sensitivity of the structural behaviour to such properties
side). Hence, Btop was correlated to other RVs through the following (based on a preliminary sensitivity analysis performed by the authors).
multivariate regression model: Specifically, variables 1, 3 and 5 are characterised by both low vari
ability and minor influence on the structural performance of the selected
Btop = 0.34Hb + 0.02G1k − 0.02Hb G1k + 0.002G21k (4) bridge decks. Even though variables 2 and 4 would affect the structural
response more significantly, they are characterized by a very small
with R2 = 0.71. The regression model was set to have a maximum value variability according to the collected data.
equal to 1.2 m.
The distance dsp of the prestressing steel reinforcement centroid from
the top of the girder cross section was conditioned upon Hb through the 3.2. Capacity modelling
following linear regression model (R2 = 0.97):
dsp = 0.98Hb (5) Flexural and shear capacity models are introduced in this study to
evaluate the load-bearing capacity of the edge girder at ULS, as repre
The prestressing steel ratio (ρsp) multiplied by dsp, ρsp ⋅ dsp, is obtained sentative of the bridge structural behaviour according to current codes
adopting a correlation with L: for girder bridges.
ρsp dsp = 8.16⋅10− 5 L + 1.85⋅10− 5 L2 (6) The following assumptions were made for computation of the ulti
mate bending moment resistance Mr of the girder cross section [22,16]:
with R2 = 0.88 and a minimum value of ρsp = 0.7% as provided by data.
The Pearson correlation coefficient ρX,Y between each outcome var 1) plane cross sections and perfect steel–concrete bond after flexural
iable Y and its predictor X was also computed for each regression model. deformation;
This coefficient is evaluated as the ratio between covariance of variables 2) ultimate compressive strain of concrete equal to 0.35%;
COV(X,Y) and the product of the standard deviations of two variables σX 3) uniform compressive stress distribution (i.e., stress block) and zero
and σ Y. Thus, ρX,Y turns out to be a normalised covariance that is a tensile strength of concrete;
measure of linear correlation between two data sets. The correlation 4) reinforcing and prestressing steel with elastic-perfectly plastic
coefficient falls in the range [− 1,1], with 0 and ±1 indicating no cor behaviour.
relation and perfect correlation (i.e., all data belonging to a linear trend
line in a scatter plot), respectively. Prestressing steel area Asp and reinforcing steel area As were lumped
All regression models described above are characterised by some in their respective centroids. The flexural capacity of girder cross section
variability, which is another source of uncertainty referred to as model was evaluated according to the following expression:
error (ME) and is the ratio between observed and predicted data. Each ( )
Mr = Asp fp,01 + As fy d∗ (7)
ME was considered as a normally distributed RV with zero mean and
standard deviation σME. Thus, the error of each regression model was where d* is the internal lever arm.
added to the conditional mean value predicted through one of Eqs. (1)– Due to a low value of Asp according to available data, yielding strain
(6), when randomly generating each RV conditioned upon another. In was always achieved even neglecting the initial, residual strain in pre
Table 3, the regression models used in this study are summarised stressing steel.
together with their predictors (i.e., independent variables), coefficient of As far as shear capacity Vr is concerned, three different formulations
determination, bound values and correlation coefficient(s). It can be were considered as per Eurocode 2 [22]. The first capacity model
Table 3
Regression models.
Outcome variable Predictor(s) Equation R2 Lower bound Upper bound ρX,Y
nb W 1 0.51 2 8 0.64
Hb ib, L 2 0.81 Hb ≥ L/20 3.2 m ρib ,Hb = 0.72
ρL,Hb = 0.64
Ab Hb 3 0.84 – – 0.88
Btop Hb, G1k 4 0.71 – 1.2 m ρHb ,Btop = 0.61
ρG1k ,Btop = 0.82
dsp Hb 5 0.97 – – 0.99
ρsp dsp L 6 0.88 0.007dsp – 0.94
6
G. Miluccio et al. Engineering Structures 249 (2021) 113367
7
G. Miluccio et al. Engineering Structures 249 (2021) 113367
8
G. Miluccio et al. Engineering Structures 249 (2021) 113367
case), and (2) dynamic instability [39] resulting in very large DCR were randomly generated using Monte Carlo sampling. Traffic loads
values or loss of analysis convergence (hereafter C2 case). According to according to both past and current Italian codes, i.e., Circ. CSLLPP 384/
[40] and the total probability theorem, the probability of exceeding 1962 [18] and NTC 2018 [16], were adopted to assess bending moments
collapse (i.e., event denoted by C) should be computed as follows: ME,Circ.1962 and ME,NTC2018, respectively. Bending moment capacity Mr,
DM.1972 was calculated by assuming a maximum concrete compressive
P[C|IM = im] = P[C|IM = im, C1 ]⋅(1 − P[C2 |IM = im] ) + P[C2 |IM = im]
strength σc,am = 0.55Rc and yield steel strength fy according to past code
(13) provisions.
Different effects of TLMs provided by Circ. CSLLPP 384/1962 [18]
where P[C|IM = im, C1 ] is the conditional probability calculated ac
and NTC 2018 [16] are shown in Fig. 11a in terms of ratio between ME,
cording to Eq. (12) and P[C2 |IM = im] is the probability of collapse
condition C2 due to numerical instability [40]. In this study, the meth NTC2018 and ME,Circ.1962. This ratio falls in the range [1.2,1.5] with mean
value of 1.36, demonstrating a significant increase in traffic load effects
odology adopted for bridge deck analysis produces P[C2 |IM = im] = 0,
associated with current code provisions. Fig. 11b shows a mean safety
so Eq. (13) turns out to be equal to Eq. (12).
factor SFDM.1972,m = 1.91, which is greater than minimum value of 1.75,
with a few cases resulting in a safety factor close to unity. This outcome
5.2. Validation of simulated design confirms the representativeness of the randomly generated bridge
models and their compliance with past Italian codes in force at the time
In order to establish if the proposed model variables are consistent of their construction. Additional data will be collected by the authors to
with the original design provisions, the bridges under study were further improve the accuracy of model variables.
assumed to be subjected to TLMs defined by Circ. CSLLPP 384/1962
[18] and were checked according to DM 30/05/1972 [20]. This modus
operandi was motivated by the fact that Circ. CSLLPP 384/1962 [18] 5.3. Fragility curves
and DM 30/05/1972 [20] were the Italian codes in force at the time of
construction for the selected bridges in terms of TLMs and safety check Fragility curves associated with flexural and shear failure modes
provisions, respectively. Specifically, Circ. CSLLPP 384/1962 provided were first derived separately, considering either deterministic values
the TLMs for bridge analysis, whereas DM 30/05/1972 was the Italian (Fig. 12a) and uniformly distributed values (Fig. 12b) of prestressing
code of reference for capacity modelling and safety checking of bridges. ratio. In Fig. 12, the lower x-axis indicates the selected IM (i.e., α),
Stress-based safety checks according to PSD would require a compre whereas the upper x-axis provides the corresponding first-lane tandem
hensive knowledge of the construction and prestressing stages signifi load Q1 = α Q1d, being Q1d = 600 kN the design load on first lane. Fig. 12
cantly affecting internal stress distributions, which cannot be assumed shows a higher fragility of the case-study bridges in terms of flexural
for a whole class of bridges. Alternatively, at the time of construction a failure.
simplified design check against ultimate condition could be carried out Indeed, the median value η of the flexural fragility curve is lower
in order to provide a minimum value of safety factor SFDM.1972 equal to than that of the shear fragility curve, regardless of the prestressing ratio.
1.75, calculating SFDM.1972 as the ratio between bending moment ca This means that the attainment of ULS is governed by flexural failure. As
pacity and demand (i.e., SFDM.1972 = Mr,DM.1972/ME,Circ.1962). The orig shown in Fig. 12a, shear fragility significantly reduces under increasing
inal desing process was thus validated over 103 model realizations that prestressing ratio. Table 5 outlines the median value η, dispersion β (i.e.,
Fig. 11. Simulated design: (a) increase in traffic load effects between NTC 2018 and Circ. CSLLPP 384/1962; (b) safety factor considering traffic loads provided by
Circ. CSLLPP 384/1962.
9
G. Miluccio et al. Engineering Structures 249 (2021) 113367
Fig. 12. Fragility curves associated with flexural and shear failure modes: (a) discrete distribution of prestressing ratio; (b) uniform distribution of prestressing ratio.
6. Conclusions
Fig. 13. Collapse fragility curve of case-study bridges. This study was aimed at evaluating the traffic-load fragility of
10
G. Miluccio et al. Engineering Structures 249 (2021) 113367
Fig. 14. Bending moment distributions: (a) capacity, (b) demand given α = 1, (c) overlapping of capacity and demands corresponding to multiple load intensities.
Table 6 Table 7
Mean, standard deviation and CoV of lognormal distributions representative of Mean, standard deviation and CoV of lognormal distributions representative of
flexural demand and capacity. DCR.
Bending moment μ [kNm] σ [kNm] CoV [%] α μ σ CoV [%]
Table 8
Fragility estimates derived through frequentist approach and convolution at
multiple IM levels.
α Frequentist approach Convolution
-5
0.50 4.00⋅10 4.70⋅10-4
0.75 4.10⋅10-3 8.80⋅10-3
1.00 4.40⋅10-2 5.70⋅10-2
1.25 1.63⋅10-1 1.83⋅10-1
1.50 3.50⋅10-1 3.75⋅10-1
analysis of data collected by the authors from the literature and on real
bridges was carried out. The analysis focused on girder bridge decks,
performing their geometric and capacity modelling according to current
Italian codes.
Based on statistics for model variables, a sensitivity analysis was first
conducted to identify the geometric and mechanical properties that
Fig. 15. DCR distribution under varying IM level.
mostly influence the performance of selected bridge decks under traffic
loading. According to uncertainty modelling of bridge decks, sensitivity
existing Italian, simply supported, beam-type, prestressed concrete analysis accounted for correlation between random variables, assuming
bridges built between 1970 and 1980, which was an extraordinary time lower and/or upper bounds based on engineering judgement. Sensitivity
for construction of highway bridges. After that traffic load models pre analysis results allow the following conclusions to be drawn:
scribed in past and current Italian codes were reviewed, a statistical
11
G. Miluccio et al. Engineering Structures 249 (2021) 113367
Fig. 16. Traffic-load fragility curves under varying (a) mean value and (b) CoV of fp,01.
12
G. Miluccio et al. Engineering Structures 249 (2021) 113367
Declaration of Competing Interest [17] MATLAB R2019b. Natick, Massachusetts: The MathWorks Inc.
[18] Circ. CSLLPP 14/02/1962, n. 384. Norme relative ai carichi per il calcolo dei ponti
stradali. Rome, Italy: Italian High Council of Public Works; 1962 [in Italian].
The authors declare that they have no known competing financial [19] RD 16/11/1939, n. 2229. Norme per l’esecuzione delle opere in conglomerato
interests or personal relationships that could have appeared to influence cementizio semplice od armato. Rome, Italy; 1939 [in Italian].
the work reported in this paper. [20] DM 30/05/1972. Norme tecniche alle quali devono uniformarsi le costruzioni in
conglomerato cementizio normale e precompresso ed a struttura metallica. Rome,
Italy: Italian High Council of Public Works; 1972 [in Italian].
Acknowledgements [21] DM 29/07/1974. Norme tecniche per il calcolo, l’esecuzione ed il collaudo delle
strutture in conglomerato cementizio armato normale, precompresso e per le
strutture in acciaio. Rome, Italy: Italian High Council of Public Works; 1974 [in
This study was developed in the framework of PON INSIST (Sistema Italian].
di monitoraggio INtelligente per la SIcurezza delle infraSTrutture ur [22] EN1992-1-1. Eurocode 2: Design of concrete structures – Part 1-1: General rules
bane) research project, which was funded by the Italian Ministry for and rules for buildings. Brussels, Belgium: Comité Européen de Normalisation;
2004.
Education, University and Research (Programma Operativo Nazionale [23] EN1990. Eurocode: Basis of design. Brussels, Belgium: Comité Européen de
“Ricerca e Innovazione 2014–2020”, Grant No. ARS01_00913). Normalisation; 2006.
[24] Chen S-Z, Wu G, Feng D-C, Zhang Lu. Development of a bridge weigh-in-motion
system based on long-gauge fiber Bragg grating sensors. J. Bridge Eng 2018;23(9):
References 04018063. https://doi.org/10.1061/(ASCE)BE.1943-5592.0001283.
[25] Chen SH, Wu G, Feng DC. Development of a bridge weigh-in-motion method
[1] Cook W, Barr PJ, Halling MW. Bridge failure rate. J Perform Construct Fac 2015;29 considering the presence of multiple vehicles. Eng Struct 2019;191(15):724–39.
(3):04014080. https://doi.org/10.1061/(ASCE)CF.1943-5509.0000571. [26] Chen S-Z, Wu G, Feng D-C. Damage detection of highway bridges based on long-
[2] Deng Lu, Wang W, Yu Y. State-of-the-art review on the causes and mechanisms of gauge strain response under stochastic traffic flow. Mech Syst Signal Process 2019;
bridge collapse. J Perform Construct Fac 2016;30(2):04015005. https://doi.org/ 127:551–72.
10.1061/(ASCE)CF.1943-5509.0000731. [27] EN1991-2. Eurocode 1: Actions on structures – Part 2: Traffic loads on bridges.
[3] Wardhana K, Hadipriono FC. Analysis of recent bridge failures in the United States. Brussels, Belgium: Comité Européen de Normalisation; 2003.
J Perform Construct Fac 2003;17(3):144–50. [28] Castaldo P, Gino D, Mancini G. Safety formats for non-linear finite element analysis
[4] Linee guida per la classificazione e gestione del rischio, la valutazione della of reinforced concrete structures: Discussion, comparison and proposals. Eng Struct
sicurezza ed il monitoraggio dei ponti esistenti. Rome, Italy: Italian High Council of 2019;193:136–53.
Public Works; 2020 [in Italian]. [29] Castaldo P, Gino D, Bertagnoli G, Mancini G. Partial safety factor for resistance
[5] Cosenza E, Losanno D. Assessment of existing reinforced-concrete bridges under model uncertainties in 2D non-linear finite element analysis of reinforced concrete
road-traffic loads according to the new Italian guidelines. Struct Concr 2021:1–14. structures. Eng Struct 2018;176:746–62.
[6] Di Prisco M. Critical infrastructures in Italy: State of the art, case studies, rational [30] Gino D, Castaldo P, Giordano L, Mancini G. Model uncertainty in non-linear
approaches to select the intervention priorities. In: Proceedings of fib Symposium numerical analyses of slender reinforced concrete members. Struct Concr 2021;22
2019 “Concrete: Innovations in materials, design and structures”, Krakow, Poland (2):845–70.
27–29 May 2019. [31] Wiśniewski FD, Cruz PJS, Abel A, Henriques R, Simões RAD. Probabilistic models
[7] Nuti C, Briseghella B, Chen A, Lavorato D, Iori T, Vanzi I. Relevant outcomes from for mechanical properties of concrete, reinforcing steel and pre-stressing steel.
the history of Polcevera Viaduct in Genova, from design to nowadays failure. Struct Infrastruct Eng 2012;8(2):111–23.
J Civil Struct Health Monitoring 2020;10(1):87–107. [32] Jacinto L, Pipa M, Neves LAC, Santos LO. Probabilistic models for mechanical
[8] DM 17/12/2020, n. 578. Adozione delle linee guida per la gestione del rischio dei properties of prestressing strands. Constr Build Mater 2012;36:84–9.
ponti esistenti e per la definizione di requisiti ed indicazioni relativi al sistema di [33] Belletti B, Vecchi F, Bandini C, Andrade C, Montero JS. Numerical evaluation of the
monitoraggio dinamico. Rome, Italy: Italian Ministry of Infrastructures and corrosion effects in prestressed concrete beams without shear reinforcement. Struct
Transportation; 2020 [in Italian]. Concr 2020;21(5):1794–809.
[9] Banerjee S, Shinozuka M. Mechanistic quantification of RC bridge damage states [34] Cestelli Guidi C. Cemento armato precompresso, 7th ed. Hoepli, Milan [in Italian].
under earthquake through fragility analysis. Probab Eng Mech 2008;23(1):12–22. [35] Raithel A. Costruzioni di ponti. Napoli: Liguori editore; 1977 [in Italian].
[10] Pinto PE, Franchin P. Issues in the upgrade of Italian highway structures. J Earthq [36] Raithel A. Ponti a travata. Napoli: Liguori editore; 1978 [in Italian].
Eng 2010;14(8):1221–52. [37] Parisi F, Scalvenzi M, Brunesi E. Performance limit states for progressive collapse
[11] Cardone D, Perrone G, Sofia S. A performance-based adaptive methodology for the analysis of reinforced concrete framed buildings. Struct Concr 2019;20(1):68–84.
seismic evaluation of multi-span simply supported deck bridges. Bull Earthq Eng [38] Castaldo P, Palazzo B, Alfano G, Palumbo MF. Seismic reliability-based ductility
2011;9(5):1463–98. demand for hardening and softening structures isolated by friction pendulum
[12] Borzi B, Ceresa P, Franchin P, Noto F, Calvi GM, Pinto PE. Seismic vulnerability of bearings. Struct Control Health Monitor 2018;25(11):e2256. https://doi.org/
the Italian roadway bridge stock. Earthq Spectra 2015;31(4):2137–61. 10.1002/stc.v25.1110.1002/stc.2256.
[13] Mosleh A, Jara J, Razzaghi MS, Varum H. Probabilistic seismic performance [39] Shome N, Cornell C.A. Structural seismic demand analysis: consideration of
analysis of RC bridges. J Earthq Eng 2020;24(11):1704–28. ‘Collapse’. In: Proc. of 8th ASCE specialty conference on probabilistic mechanics
[14] Gardoni P, Mosalam KM, der Kiureghian A. Probabilistic seismic demand models and structural reliability. South Bend, Indiana, USA; 2000.
and fragility estimates for RC bridges. J Earthq Eng 2003;7(sup001):79–106. [40] Jalayer F, Ebrahimian H, Miano A, Manfredi G, Sezen H. Analytical fragility
[15] Monteiro R, Zelaschi C, Silva A, Pinho R. Derivation of fragility functions for assessment using un-scaled ground motion records. Earthq Eng Struct Dyn 2017;46
seismic assessment of RC bridge portfolios using different intensity measures. (15):2639–63.
J Earthq Eng 2019;23(10):1678–94.
[16] DM 17/01/2018. Aggiornamento delle «Norme tecniche per le costruzioni». Rome,
Italy: Italian Ministry of Infrastructures and Transportation; 2018 [in Italian].
13