Está en la página 1de 14

Structure and Infrastructure Engineering

Maintenance, Management, Life-Cycle Design and Performance

ISSN: 1573-2479 (Print) 1744-8980 (Online) Journal homepage: http://www.tandfonline.com/loi/nsie20

Bridge performance indicators based on traffic


load monitoring

Ana Mandić Ivanković, Dominik Skokandić, Aleš Žnidarič & Maja Kreslin

To cite this article: Ana Mandić Ivanković, Dominik Skokandić, Aleš Žnidarič & Maja Kreslin
(2017): Bridge performance indicators based on traffic load monitoring, Structure and Infrastructure
Engineering, DOI: 10.1080/15732479.2017.1415941

To link to this article: https://doi.org/10.1080/15732479.2017.1415941

Published online: 21 Dec 2017.

Submit your article to this journal

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at


http://www.tandfonline.com/action/journalInformation?journalCode=nsie20

Download by: [University of New England] Date: 22 December 2017, At: 07:29
Structure and Infrastructure Engineering, 2017
https://doi.org/10.1080/15732479.2017.1415941

Bridge performance indicators based on traffic load monitoring


Ana Mandić Ivankovića, Dominik Skokandića, Aleš Žnidaričb and Maja Kreslinb
a
Faculty of Civil Engineering, University of Zagreb, Zagreb, Croatia; bSlovenian National Building and Civil Engineering Institute, Ljubljana, Slovenia

ABSTRACT ARTICLE HISTORY


The assessment of the existing road bridges as a part of bridge management networks has been a subject Received 11 April 2017
of several European research projects and actions in the last few decades. Bridge management includes a Revised 4 October 2017
series of activities, interconnected in order to achieve optimal balancing of required costs, potential risks Accepted 5 October 2017
and overall bridge performance. This paper discusses a valuable link between a specific indicator (traffic KEYWORDS
load information calculated using Bridge Weigh-in-Motion measurement data) and the corresponding Road bridges; performance
structural performance (reliability of existing bridge). It is shown that implementation of weigh-in-motion
Downloaded by [University of New England] at 07:29 22 December 2017

indicators; probabilistic
measurements in the probabilistic assessment of existing road bridges can reveal hidden bridge reserves assessment; traffic load
and predict bridge reliability development over the required lifetime. Consequently, such approach can monitoring; B-WIM; reliability
yield an unrestricted use of the bridge over a much longer remaining service life. Broader advantages analysis
lie in an improved road network management, road bridges in particular, and in a more sustainable
development of infrastructure network and greater satisfaction of road users and owners.

Introduction In order to evaluate the effect of each element on overall


bridge condition, it is necessary to determine its impact on the
Current standards and codes for the design of new bridges are
structural safety, traffic safety and bridge durability. This can
based on conservative assumptions regarding load and resistance
be achieved through evaluation of various performance indica-
modelling in order to be applicable on different bridge types
tors, defined as parameters that have the possibility to indicate
and to guarantee sufficient structural safety over their lifetime.
a bridge condition (Strauss, Bergmeister, & Mandić Ivanković,
Although these codes result in creation of safe and cost-effective
2016; Strauss, Mandić Ivanković, Matos, & Casas, 2016). Basic
new bridges, the use of these standards for assessment of existing
visual inspection of bridge elements provides fundamental
bridges may show that many of these bridges need to be strength-
information for evaluation of performance indicators, but addi-
ened or even replaced (Šavor & Novak, 2015; Wiśniewski, Casas,
tional site-specific data can be collected using Structural Health
& Ghosn, 2012). Recent research has revealed that the site-spe-
Monitoring (SHM) methods and tools. Bridge Weigh-in-Motion
cific bridge assessment, based on measured monitoring data, can
(B-WIM) systems, as a part of SHM, can be used to provide data
lead to reduction in maintenance costs and extension of remain-
for real-life traffic load models for bridges, along with some addi-
ing service life of specified bridges (O’Connor & Enevoldsen,
tional, measured structural parameters that can be used for an
2009; Žnidarič, Kreslin, Lavrič, & Kalin, 2012).
optimised bridge assessment (O’Brien, 2011; Skokandić, 2016).
In addition, over the past few decades, the assessment of these
The first part of this paper contains theoretical introduction
bridges as a part of bridge management networks has been a
on the definition and categorisation of performance indicators
subject of a number European research projects and actions
for road bridges. It is followed by an overview of the B-WIM
(ARCHES D08, 2009; Bien, Elfgren, & Olofsson, 2007; COST,
technology, measurements and associated traffic loading and
2014; Woodward, 2001). Furthermore, bridge performance
structural data. Then, a short introduction on the application
aspects have been analysed within the ongoing COST Action TU
of probabilistic methods in the assessment of existing bridges is
1406 (Quality specifications for roadway bridges, standardisation
presented, followed by practical application of B-WIM data in
at a European level), whose main objective is to develop a guide-
the assessment of a case study bridge, analysis of the results and
line for the establishment of quality control plans for roadway
the conclusions.
bridges (Casas, 2016; Matos, Casas, & Fernandes, 2016; Strauss,
Mandić Ivanković, Matos, & Casas, 2016). Bridge management,
as a part of the road infrastructure network, includes a series Performance indicators for optimised road bridge
of activities, interconnected to achieve optimal levels, including assessment
balancing of required costs, potential risks and overall bridge
The categorisation of performance indicators through compo-
performance. Bridge effectiveness can be analysed through dif-
nent, system and network levels (Figure 1, solid lines) accord-
ferent levels, starting from the analysis of each individual bridge
ing to appropriate technical, sustainability and socio-economic
element.

CONTACT  Ana Mandić Ivanković  mandicka@grad.hr


© 2017 Informa UK Limited, trading as Taylor & Francis Group
2   A. MANDIĆ IVANKOVIĆ ET AL.
Downloaded by [University of New England] at 07:29 22 December 2017

Figure 1. Interaction of performance indicators and performance goals through bridge management levels (solid lines) with additional input of bridge safety/reliability
assessment based on bridge WIM measurements (dashed lines).
Notes: PI: performance indicators; G(T): intermediate goals (Tasks); MIL: measured influence line; PG: performance goals; SHM: structural health monitoring tool; GDF: girder distribution factor;
DAF: dynamic amplification factor; γ: safety index; β: reliability index.

aspects is proposed (Strauss et al., 2016), in order to simplify structures over their lifetime. Nevertheless, the usefulness of
the evaluation of these indicators in compliance with certain SHM systems needs to be revealed by quantifying their value,
structural performance goal/criteria. At the bridge component to convince the infrastructure owners and operators to invest in
level, the fundamental goal to achieve is the damage assessment, these types of systems. This is the main objective of the COST
which implies detection of damage, but also its identification action TU1402 (COST, 2014). Therefore, a joint task was set
and evaluation. At this level, damage detection is usually divided between both COST actions to identify and categorise perfor-
into three bridge sub-systems: substructure, superstructure and mance indicators that may be detected, evaluated and possibly
roadway, in order to simplify characterisation of damage and its quantified with available SHM technologies, and to weigh their
consequences on the overall bridge performance. At the system importance in achieving crucial performance goals (Strauss,
level, the importance of each component is to be evaluated to Mandić Ivanković, & Soussa, 2016). The major question for
assess the impact of its condition on the overall bridge perfor- bridge owners: ‘which PI should be monitored to achieve an
mance. The evaluation is conducted according to each of the fol- optimum long-term bridge performance?’ remains partially
lowing criteria: structural safety and serviceability, traffic safety unanswered. COST actions TU1406 and TU1402 are jointly
and durability. At the final network level, bridge is assessed to seeking answers to this question.
evaluate its importance within the entire infrastructure network, The implementation of the bridge weigh-in-motion (B-WIM)
based on the data from previous levels, in order to achieve the traffic measurements, as an example of SHM tool, into the mul-
primary performance goal – priority repair ranking of selected ti-level bridge assessment might result in a higher value of bridge
bridges. safety or reliability index and, consequently, in an unrestricted
The performance indicator most widely used across Europe use of bridges over a much longer remaining service life. This
is the condition index (also known as the condition rating, dete- has been proven with the case study elaborated in this paper. To
rioration index etc., depending on the country and/or bridge enhance calculation of structural safety, it is proposed to benefit
owner), which is mainly determined through visual inspection. from the measured influence lines, girder distribution factors and
In many countries, this is the only performance indicator used dynamic amplification factors, which are the crucial performance
in practice by bridge owners and operators (Strauss, Mandic- indicators at the system level, before decisions are to be made at
Ivankovic, et al., 2016). the network level (Figure 1, dashed lines). These performance
In many cases, visual inspection will not suffice for effective indicators together with safety and reliability indexes are listed
assessment of bridge performance, and so additional data on the in Table 1, and are more thoroughly explained in the following
bridge structure will be required. Typical examples are stiffness sections.
distribution, existing material characteristics, site specific traffic
loads etc., requiring additional testing, measurements, monitor-
Bridge weigh-in-motion as a part of structural health
ing and advanced methods of processing, modelling and calcu-
monitoring
lation of bridge performance.
Structural Health Monitoring (SHM) involves a wide range of Weigh-in-motion (WIM) (Moses, 1979; Žnidarič, Kreslin, &
activities which, through different technologies and algorithms, Kalin, 2016) is a procedure that is used to measure axle load
may supply information about the performance of existing/new and gross weight of a vehicle as it drives over a measurement
STRUCTURE AND INFRASTRUCTURE ENGINEERING   3

Table 1. Performance indicators based on traffic load monitoring at the bridge – B-WIM.

Performance indicator Description Basic principle


Measured influence line MIL Difference between measured and theoretical influence lines True (measured) influence lines can result in reduced traffic
indicates changes in bridge boundary conditions load effects; their change over time can indicate degrada-
tion of bearings and/or expansion joints
Girder distribution factor – GDF Dimensionless parameter that indicates proportion of total Difference between analytical and measured GDF and
traffic load effect on a bridge, carried by an individual changes of GDF over time can indicate degradation and/or
girder structural changes in bridge girders
Dynamic Amplification factor – DAF Ratio of maximum total load (including dynamic part) effect Dynamic amplification factor decreases as static load on the
divided by maximum static load effect bridge increases
Safety index – γB-WIM Ratio of designed value of the resistance to bending MRd and Safety index may reveal reserves of the bridge based on
applied internal bending moment MEd deterministic (or semi probabilistic) approach
Reliability index – βB-WIM,T Probabilistic measure, corresponding to probability of failure Probability index may reveal reserves of the bridge based on
probabilistic approach within the required lifetime of the
bridge

site at full speed, without the need for slowing down or stop- al., 2012). Furthermore, predictions of increase in the number
ping. Two types of WIM systems exist: one with the sensors built of heavy vehicles are also taken into account.
into the pavement and the other involving the bridge weigh-in- It should be noted that the axle load and spacing of unidi-
Downloaded by [University of New England] at 07:29 22 December 2017

motion (B-WIM). The second one uses instrumented bridges rectional traffic are not entirely independent, but the degree of
as weighing scales (Moses, 1979). The main advantage of the correlated vehicles, typically up to a few per cent, varies consid-
B-WIM systems over the pavement ones is that they are portable. erably from one site to another. Furthermore, the correlation is
Furthermore, as all sensors are installed under the bridge, the difficult to model (Enright, 2010) and adds a number of addi-
installation and maintenance works do not impede the flow of tional parameters to the already complex simulation models. On
traffic (Žnidarič et al., 2012). Finally, they provide supplementary the other hand, the convolution method, which uses independent
information about bridge behaviour under traffic load, which is traffic in the adjacent lanes, belongs to the group of approximate
highly valuable for structural analysis. load modelling approaches, which has been shown to gener-
Data collected with B-WIM measurements can be used for ate results similar to the simulations (Moses & Verma, 1987;
a number of applications such as traffic analyses, pavement and Žnidarič, 2017).
bridge design and/or assessment, selection of overloaded vehi-
cles, etc. Weighing results for a two axle truck, as collected by
Static load effect for the bridge
the commercial B-WIM system called SiWIM®, are given as an
example in Table 2 This system was developed in Slovenia as an The calculation of load effect of each vehicle that crosses the
outcome of research performed in the scope of the COST Action bridge is based on the WIM measurement results (axle weights
323 (Jacob, OBrien, & Jehaes, 2002) and EC 4th Framework and spacing), which are combined with the relevant bridge influ-
Project WAVE (Jacob, 2002). ence lines, a method proposed by a number of authors (Ieng,
Site-specific traffic load models, developed using B-WIM 2015; Karoumi, O’Brien, & Quiligan, 2006; Moses, 1979; Žnidarič
measurements, are a key input for an optimised assessment of et al., 2012). Influence lines are used to transform traffic load on
existing bridges (Sivakumar, Ghosn, & Moses, 2011; Žnidarič, the bridge into load effects (moments and shear forces) and,
Kreslin et al., 2015). The main challenge in the development of in theory, they depend only on the bridge span, geometry and
these load models is the extrapolation of measured data and, boundary conditions (simply supported, continuous bridge etc.).
consequently, realistic estimation of the maximum expected The calculation is based on the formula for static load effect
load effects on a bridge (bending moments, hogging moments Qs:
and shear forces) over a certain time period/bridge lifetime. A
n
number of methods have been proposed to extrapolate the lim- ∑
ited traffic information (Getachew & Obrien, 2007; Sivakumar
Qs = Ai Ii (1)
i=1
et al., 2011; Žnidarič et al., 2012). They are either based on fitting
statistical distribution on collected data or on using simulations, where Ai is the weight of the axle i, n stands for the number of
such as the Monte Carlo method. vehicle axles and Ii is the value of the influence line due to axle i
The method used to assess the case study bridge (Žnidarič at location x. Results obtained with this method are presented in
et al., 2012) is based on the convolution method (Sivakumar terms of maximum expected moments and shear forces at critical
et al., 2011) and an assumption that the highest load effect is sections of the bridge for a specified time period. In WIM data
achieved when two trucks from independent traffic flows are post-processing, these static load effects given by Equation (1) are
placed on the bridge side by side in each traffic lane, at the place presented as histograms for each lane, creating two independent
of maximum action, which is defined as a loading event. This probability mass functions (fX and fY), based on the assumption
assumption is justified for bridges with short to medium spans, that the traffic in lanes 1 and 2 is independent. In the next step,
of up to 40 m in length where, due to typical dimensions of heavy these functions are convoluted to define the probability mass
vehicles, a critical event occurs when there is one vehicle in each function (fZ) of all possible load effects generated by pairs of
lane (Caprani, Grave, O’Brien, & O’Connor, 2002; Žnidarič et vehicles from both lanes.
4   A. MANDIĆ IVANKOVIĆ ET AL.

Table 2. Example of WIM data for single truck passage.

Number of
Time stamp Lane Speed [m/s] Class axles GSW [kN] AW1 [kN] AW2 [kN] Axle spacing [m]
2007-03-22-00- 1 17.5 41 2 123.8 37.07 86.69 6.07
39-28-955

The cumulative distribution function (CDF) FZ derived from to this load. The information about traffic load can be provided
fZ is presented in Figure 2 (curve labelled ‘CDF’). Finally, cumu- by any weigh-in-motion system that provides data about axle
lative distribution functions (convolution curves in Figure 2) are load, axle spacing and inter-vehicle distance of all vehicles that
defined for different time periods, using the extreme value theory cross the measured road section. In addition, the bridge WIM
(Ang & Tang, 1975; Castillo, 1988), by raising FZ to the power systems can, apart from being fully portable, provide additional
of expected loading events (N) in such periods. Median values structural information about the instrumented bridge behav-
for each time period, derived from CDF functions in Figure 2, iour under traffic load (Karoumi et al., 2006; O’Brien, González,
are selected as maximum static load effects for a chosen period. Dowling, & Žnidarič, 2013), including realistic values of the
Furthermore, the shape of distribution function (FZ)N converges above-mentioned indicators: influence lines, GDFs and DAF.
at high values of N to a generalised extreme value (GEV) distri- These measured values are more reliable, and in most cases, less
Downloaded by [University of New England] at 07:29 22 December 2017

bution, and it is straightforward to estimate its standard deviation conservative than the theoretical ones, and can substantially
from data in relative frequency histograms using the weighted increase the bridge resistance-to-load ratio.
standard deviation formula (Montgomery & Runger, 2014). The difference between the theoretical and measured influ-
This information enables application of traffic load effects ence lines (MIL) can reveal structural reserves or changes in the
in the assessment of existing bridges, in both the deterministic bridge behaviour due to deterioration of its bearings, expansion
and probabilistic approaches. A detailed review of the presented joints, etc. (Ieng, 2015; Karoumi et al., 2006; Žnidarič et al., 2015).
method for traffic load extrapolation, along with a short overview The true length of influence lines can be different from the the-
of other methods, is given in (Žnidarič et al., 2012). It is worth to oretical ones, due to bearings dimensions and condition, which
mention that O’Brien, Schmidt, Hajializadeh, and Zhou (2015) can affect traffic load effects at critical bridge sections. Realistic
reviewed seven methods of statistical inference for bridge safety distribution of the total load imposed on structural members of
evaluation. They have concluded that the accuracy of the results the bridge (GDF) can indicate differences in relation to bridge
is less sensitive to the method chosen but more to the quantity design, or even changes in their elements’ stiffness, which might
of data used and the assumptions made with regard to the tail. suggest certain defects or design faults. This information is very
These issues are also discussed in (Žnidarič, 2017). useful, as it can point to degradations that cannot be detected
during visual inspections.
Finally, B-WIM measurements of bridge behaviour under
Maximum load effect for a girder
traffic load can provide real-time information about the
The total predicted maximum load effects Q, used in bridge dynamic amplification factor (DAF) for each individual vehicle.
assessment, are defined by multiplying the predicted static load Comparison of recommended dynamic factors in current design
effect Qs of Equation (1) with the dynamic amplification factor codes for new bridges (Eurocode, 2003) and measured values
(DAF) and girder distribution factor (GDF). The girder distri- obtained with B-WIM measurements was conducted as a part
bution factor defines the share of the total traffic load effects of FP6 ARCHES project (ARCHES D10, 2009). The outcome of
that is carried by bridge elements and cross sections, identifying the project was that design code recommendations are in general
critical section of the bridge. In theory, the GDF depends on basic conservative, which is appreciated for new structures that will
bridge parameters, such as the cross section, geometry, material have to sustain loads over the entire lifetime, but it unnecessarily
properties etc., and on characteristics of imposed loads. increases the load of existing bridges that are assessed for the pres-
The dynamic amplification factor is defined as an increase ent traffic conditions. The main conclusions of the project were
in static load on the bridge due to the dynamic bridge–vehicle that DAF decreases with larger and heavier vehicles, and that the
interaction (ARCHES D10, 2009). It is most often described as DAF values change by even up to 50% with pavement conditions.
a ratio between the total and static load effects due to a vehicle As the critical loading event is assumed when two heavy
crossing the bridge. The DAF values are defined in design codes vehicles are placed side by side on a critical bridge section, it
for new bridges (Eurocode, 2003) based on span length, number was recommended to assess the dynamic allowances of exist-
of traffic lanes on the bridge and the required load effect (bending ing bridges with more accurate methods (ARCHES D10, 2009;
or shear). For comparison, in older design codes (DIN-1072, Caprani, 2005; Kalin, Žnidarič, & Kreslin, 2015; Kirkegaard,
1988; Furundžić, 1969), the dynamic factor depended on span Nielsen, & Enevoldsen, 1997). If B-WIM measurements are used
length only. to measure the DAF, the ARCHES project (2009) recommended
to measure at least 100 000 vehicles, or at least two months of
continuous traffic. An overview of DAF calculation methods
Measurement-based indicators
exceeds the scope of this paper but can be found in the ARCHES
An efficient bridge assessment requires information about the D10 report (2009) and in (Kalin et al., 2015). The application of
true traffic load and about behaviour of the structure subjected measured indicators (MIL, GDF and DAF) in bridge assessment
STRUCTURE AND INFRASTRUCTURE ENGINEERING   5

Figure 2. Convolution curves of expected load effects in different time periods.


Downloaded by [University of New England] at 07:29 22 December 2017

is presented in detail in the section ‘Application of additional data plans, causing big differences between the realistic and theoret-
in assessment of existing bridges’. ical results. Such deviations can be reduced by using Bayesian
methods of probability estimation (Holický, 2013). These meth-
ods provide reliable values of material characteristics, combining
Probabilistic approach to bridge assessment
prior information from literature or previous measurements with
Probabilistic approach to bridge assessment is based on calcula- new results, in order to reduce model uncertainties (Skokandić,
tion of the probability of failure Pf, which occurs when load effects Mandić Ivanković, & Džeba, 2016; Sykora, 2008).
on the bridge E exceed its structural reliability R. Traditionally, None on-site measurement data, apart from the B-WIM
one of the most frequently used reliability measures of structures results, were available for assessment of the case study bridge.
is reliability index β, quantity related to the probability of failure Therefore, the original design plans and information were used
Pf (Gulvanessian, Calgaro, & Holický, 2012). to model the bridge resistance to bending. Statistical parameters
The required reliability index for the design of new bridges (standard deviations and mean values) were defined based on
is defined in current standards (Eurocode, 2002) based on the recommendations from (JCSS, 2002).
designed lifetime of bridges. No codes have been defined so far
for existing bridges, but the method proposed by Koteš and Vičan
Application of additional data in assessment of
(2012) takes into account the age of the bridge. Unlike current
existing bridges
standards, the proposed method does not define reliability index
for the total lifetime, but for the remaining design lifetime of the The assessment of existing bridges using detailed numerical
bridge, based on its age and condition. Thus, bridges are assessed for models is a common approach to determine the bridge load
shorter time periods (difference between the bridge age and design distribution, modal shapes, deflections and other structural
lifetime), resulting in reduction of required reliability indices. parameters essential to establish its load-carrying capacity. These
The reliability index and the corresponding probability of models are mainly developed using the finite element method
failure of an existing bridge are key performance indicators in (FEM). When original design plans are available, they can be
assessment procedure, as they provide a degree of structural and developed to a very precise level, even taking into account some
traffic safety of the bridge. The calculation of these indicators is degradation over time. Nevertheless, it is hard to simulate the
based on a number of stochastic variables representing bridge realistic behaviour of an existing bridge without the on-site meas-
load and resistance characteristics, such as the geometry of cross urements and structural monitoring data, as some parameters
section and concrete and reinforcement characteristics. are hard to determine with visual inspection only, as described
The main challenge in modelling these characteristics as sto- in the first part of this paper. The inclusion of these additional
chastic variables lies in the accuracy of determination of their parameters into numerical models of the bridge, developed using
statistical parameters and correspondent type of distributions. Sofistik software for structural analysis (Sofistik, 2014), is pre-
The best approach is to directly determine the mean value and sented in the following sections (Skokandić, 2016).
standard deviation from results obtained with measurements,
and to fit those to statistical distribution. As this approach
Case study bridge analysis
requires on-site or laboratory testing and is relatively expen-
sive, statistical parameters of these variables can be calculated The bridge chosen for this analysis is a simply supported highway
theoretically from the original documentation using recommen- bridge with a single span of 24.8 m, and with the superstructure
dations from the Probabilistic Model Code (JCSS, 2002), which composed of five prefabricated I-type prestressed concrete gird-
also proposes which type of distribution should be used. A major ers connected with the monolithic concrete deck slab (Figure 3).
issue regarding this approach is that the actual bridge geometry The 3D FEM numerical model of the bridge (Figure 4) has
and built-in materials can deviate significantly from original been developed based on original design plans, along with built
6   A. MANDIĆ IVANKOVIĆ ET AL.

Figure 3. Cross section of case study bridge with built in reinforcement.

Step 1 – application of measured influence line (MIL)


At the next level, the measured influence line for bending
(ARCHES D16, 2009; Ralbovsky et al., 2014), obtained from
B-WIM measurements was compared to the theoretical influence
line for simply supported bridges.
Theoretical influence line for the simply supported bridge
is marked in Figure 5 with thick blue line, while thick green
Downloaded by [University of New England] at 07:29 22 December 2017

line is used for the measured line. The difference between the
two reveals that the bridge does not behave as an ideally simply
supported structure, as it could have been expected after the orig-
Figure 4. 3D FEM model of case study bridge. inal expansion joints were replaced with the asphalt ones. The
measured influence line (MIL) requires adjustment in numerical
in reinforcement and tendons, enabling calculation of the load model of the bridge. To simulate partially fixed behaviour, the
carrying capacity to bending and shear. Girders and monolithic model is modified with the calibrated influence line marked with
deck have been modelled using grillage method to simulate real- red line in Figure 5.
istic behaviour of bridge elements, providing theoretical trans- This modification is accounted for with additional rotational
verse load distribution on every girder (theoretical GDF). stiffness of supports, using springs on the bearings, as shown in
Figure 4. As a result of calibration, the bending moments in the
centre of the span reduced compared to the initial level, adding to
Calibration of bridge numerical model using B-WIM data the resistance/load ratio of critical cross section of the case study
A multi-level assessment for the case study bridge has been con- bridge (Table 5, step 1). Realistic influence lines can be evalu-
ducted based on the method proposed by Skokandić et al. (2016). ated from a few tens of vehicles measured with a B-WIM system
Additional data have been applied in every step to show benefits (Ralbovsky et al., 2014; Žnidarič, 2017) that was installed on the
of utilisation of bridge monitoring data. Results obtained at every bridge under inspection. The whole process lasted no more than a
level are presented in the following sections, both in determinis- few hours, which reduced the total monitoring cost considerably.
tic (as a ratio of bridge resistance to total load) and probabilistic
(comparison of calculated and targeted reliability) formats. An Step 2 – measured load distribution over girders (realistic
overview of step-by-step assessment process is explained in the GDF)
sequence. Theoretical and measured girder distributions factors (GDFs)
were compared in next step (step 2) of the assessment, so as to
Step 0 – initial reference step review possible differences that might point to degradation of
The linear elastic analysis of the bridge was conducted at the bridge elements. Furthermore, GDFs were used to determine
initial level of assessment based on current design codes for new critical cross sections of the bridge. As the case study bridge was
bridges. Only the permanent actions (Eurocode, 2004) (self- composed of identical girders with the same amount of rein-
weight, fixed equipment and road surfacing) and traffic loads, forcement and tendons, the girder subjected to highest load was
according to Traffic Load Model 1 (Eurocode, 2003) were taken selected. In the theoretical model, GDFs are based on geometry
into account. Load effects on the bridge (bending moment and and stiffness of bridge elements and on transverse position of
shear) were calculated by means of Sofistik software using the- the moving load.
oretical influence lines for simply supported bridges. Bending On the other hand, a B-WIM system stores maximum strain
resistance of girders was calculated using the built in reinforce- values obtained from all strain sensors, which allows calculation
ment. The corresponding results are presented with all results in of mean values and standard deviations of the measured GDF.
Table 5 (step 0). This initial step has resulted in conservative esti- Comparison of these two sets of values is presented in Figure 6. In
mation of the bridge load carrying capacity, and was conducted this case, the differences are small (Table 5, step 2), while the GDF
as a reference point to demonstrate benefits of monitoring in an comparison would likely show much bigger deviations in a case
optimised bridge assessment. involving one or more damaged or seriously degraded girders.
STRUCTURE AND INFRASTRUCTURE ENGINEERING   7

Figure 5. Comparison of influence lines – case study bridge.


Downloaded by [University of New England] at 07:29 22 December 2017

Figure 6. Comparison of load distribution on each girder– case study bridge.

Such results were expected for the case study bridge, as no In the calculation process, the measured influence lines and
visible signs of degradation were discovered during visual inspec- GDF values were used for the calculation of static load and load
tion. No numerical model calibration was conducted due to the distribution, respectively. For the total load effect calculation,
fact that girders are identical in geometry, stiffness and reinforce- DAF was evaluated for 288,413 vehicles with the gross vehicle
ment, and as the total load was redistributed according to meas- weight (GVW) in excess of 5 t. DAF values for every vehicle, and
ured GDF values. In cases where calibration of numerical model their comparison with design recommendations, are presented
is required, it can be done by moderately adjusting stiffness of in Figure 7 where it can clearly be seen that DAF values decrease
bridge elements in the software, without affecting the geometry. with the GVW of vehicles. For example, a DAF value for 10
This method enables detection of invisible degradations of bridge heaviest loading events amounts to 1.0067 only. The mean value
elements that would in the end result in the reduction of total for use in assessment was obtained from the relative frequency
maintenance and reparation costs. histogram presented in Figure 8. It amounts to 1.035 with the
standard deviation of 0.029.
Step 3 – site specific traffic load model To assess the load carrying capacity at the critical section (in
In the final step (step 3) of the assessment procedure, a site-spe- the middle of the span), the bending moment for traffic load-
cific traffic load model was developed from B-WIM measure- ing, calculated from the B-WIM results (Table 5, step 3), was
ments (Žnidarič et al., 2012). The model was extrapolated using combined with the permanent load effect obtained from the
the method described in the section entitled ‘Bridge weigh-in- calibrated numerical model using Sofistik. Every step of assess-
motion as a part of SHM’. The characteristic value that is being ment procedure was conducted for shear forces, too, but is not
assessed can vary. In this example, the 75-year median values presented due to restricted length of the paper. The assessment
were chosen to represent the remaining service life of the case procedure was also performed using probabilistic approach, as
study bridge. described in the following section.
8   A. MANDIĆ IVANKOVIĆ ET AL.

Figure 7. Measured DAF for vehicles with GVW above five tonnes.
Downloaded by [University of New England] at 07:29 22 December 2017

Figure 8. Relative frequency histogram for measured DAF values.

Probabilistic approach – case study bridge where MRd and MEd stand for the girder’s cross section resistance
to bending and the total bending moment on the girder, respec-
The basic limit state equation for failure probability calculation
tively. Further derivation of Equation (4) is as follows:
is defined as: [( ) ]
Φ
Z = 𝜃R ⋅ R − 𝜃E ⋅ E (2) R = MRd ⋅ 𝜃R = h − c − b ⋅ nb ⋅ As ⋅ fy + ng ⋅ d ⋅ Ap ⋅ fypk ⋅ 0.9 ⋅ 𝜃R
2
where E and R are modelled as stochastic variables, described (5)
with statistical parameters (mean value μ and standard deviation
σ) and certain type of distribution, depending on their charac- E = MG ⋅ 𝜃E,G + MQ ⋅ 𝜃E,Q
teristics. Additional functions θR and θE are defined as model
[ ]
+ ΔMg ⋅ 𝜃E,G + MQ ⋅ 𝜃E,Q (6)
( ) L2
uncertainty variables in order to reduce model deviations from = A + hd ⋅ bd ⋅ 𝛾 C ⋅
8
reality (Eurocode, 2002; JCSS, 2002). If we assume that function
Z given by Equation (2) is distributed normally, the reliability
where MQ = MT1 for traffic load according to EN 1991-2 and
index β is calculated as follows:
MQ = MT2 ⋅ DAF for traffic load according to B-WIM. All vari-
𝜇z ables included in Equations (4)–(6) for the assessment of girder
𝛽=
𝜎z (3) 3 are defined with their values and distribution types in Tables
3 and 4.
where μz and σz are statistic parameters of Z, denoting its mean
In order to solve limit state equations using probabilistic
value and standard deviation, respectively.
approach, it is convenient to use specialised software programs,
Equation (2) for the case study bridge is defined for each
as mathematical operations regarding parameters with different
girder separately as:
distributions types are complex. The first order reliability method
Z = 𝜃R ⋅ MRd − 𝜃E ⋅ MEd (4) (FORM) and corresponding Hasofer–Lind reliability index was
STRUCTURE AND INFRASTRUCTURE ENGINEERING   9

Table 3. Statistical parameters for bending resistance of girder 3.

Standard deviation
Variable Units Distribution Nominal value Mean value (μ) (σ)
Girder height h [m] Deterministic 1.66 1.66 /
Concrete cover c [m] Normal 0.02 0.02 0.025 μ
Number of bars/girder nb Deterministic 4 4 /
Number of tendons/girder ng Deterministic 2 2 /
Diameter of bar 𝛷b [m] Deterministic 0.02 0.02 /
Yield strength of reinforc. steel fy [kN/cm2] Normal 22.0 24.64 0.05 μ
Area of rebar As [cm2] Normal 3.14 3.14 0.020 μ
Diameter of tendon 𝛷t [m] Deterministic 0.053 0.053 /
Effective depth of tendons d [m] Normal 1.41 1.41 0.25 μ
Tensile strength of prestress. steel fypk [kN/cm2] Normal 135.28 151.51 0.05 μ
Area of tendon Ap [cm2] Normal 22.0 22.0 0.020 μ
Resistance uncertainty θR Lognormal / 1.2 0.15

Table 4. Statistical parameters for loading of girder 3.


Variable Units Distribution Nominal value Mean value (μ) Standard deviation (σ)
Concrete density γC [kN/m3] Normal 25.0 25.0 0.04 μ
Downloaded by [University of New England] at 07:29 22 December 2017

Bridge span L [m] Deterministic 24.8 24.8 /


Girder cross section area A [cm2] Normal 7052 7052 0.020 μ
Deck height hd [m] Deterministic 0.24 0.24 /
Deck width bd [m] Deterministic 2.80 2.80 /
Additional permanent ΔMg [kNm] Normal / 403.14 0.05 μ
load
Traffic load – EN, 1991-2 MT,1 [kNm] Normal / 2385.88 0.08 μ
Traffic load – B-WIM MT,2 [kNm] GEV / 1302.58 0.10 μ
Dynamic amplification DAF Gumbel / 1.035 0.029
factor
Permanent load uncer- θE,G Lognormal / 1.0 0.12
tainty
Traffic load uncertainty θE,Q Lognormal / 1.0 0.19

Table 5. Assessment procedure results – deterministic approach [kNm].

Girder 1 Girder 2 Girder 3 Girder 4 Girder 5


MEd [kNm] – Step 0 7553.63 7968.83 7728.16 7030.85 5686.29
MEd [kNm] – Step 1 4963.38 5367.21 5317.27 4668.96 3522.73
MEd [kNm] – Step 2 4948.28 5155.85 5421.91 4974.01 3339.48
MEd [kNm] – Step 3 3890.03 4030.52 4181.36 3879.62 2828.86
MRd [kNm] 7901.76

used for the case study bridge assessment using adequate relia- Time-variant reliability analysis
bility analysis software (Schneider, 2017).
Results of the time-variant reliability analysis over the selected
time periods, conducted for Girder 3, are presented in Figure 11.
Case study bridge – assessment results The analysis is based on the B-WIM site-specific traffic load model,
Results for bending assessment conducted with deterministic obtained in step 3 of the assessment method. The blue dashed line
approach are presented in Table 5 with bending moments calcu- illustrates how the calculated reliability index decreases with time,
lated on each girder for every step of the assessment, along with depending on the expected traffic load effects shown in the upper
the total bending resistance of girders. right corner for various time periods. This method can be used
Additionally, results are presented graphically in Figure 9 as to predict critical point in the projected bridge lifetime, when its
a ratio of girder cross section resistance to the corresponding calculated reliability drops below the minimum required level.
bending moment, defined as a safety index γ: Taking into account the reduction of resistance and increase in
traffic load over time, this type of analysis could optimise decision
MRd making in the sphere of bridge maintenance, as it provides tools
𝛾= (7) to predict bridge reliability levels over time. Time periods can be
MEd
chosen depending on the requirements, e.g. a period of 5 years can
Probabilistic analysis results are presented in Table 6 and, graph- be set as it represents the time between mandatory bridge inspec-
ically, in Figure 10, as reliability indexes in every step of the pro- tions in some countries. If possible, in order to reduce uncer-
cedure, calculated for each girder separately. Minimum required tainties, information about the bridge condition and traffic loads
reliability indexes for new and existing bridges are also presented should be adapted after main or principal bridge inspections,
in Table 6. which are conducted in Europe at every five or six-year intervals.
10   A. MANDIĆ IVANKOVIĆ ET AL.
Downloaded by [University of New England] at 07:29 22 December 2017

Figure 9. Bending moment and resistance ratio of every girder.

Figure 10. Reliability index β for every girder.

Table 6. Assessment procedure results – probabilistic approach.

Girder 1 Girder 2 Girder 3 Girder 4 Girder 5


Reliability index β – Step 0 3.82 3.49 3.80 4.25 5.58
Reliability index β – Step 1 6.02 5.55 5.67 6.37 8.16
Reliability index β – Step 2 6.04 5.80 5.48 5.88 8.48
Reliability index β – Step 3 7.76 7.53 7.27 7.72 9.47
βRequired (Eurocode, 2002) 3.80
βRequired (Koteš & Vičan, 2012) 3.70
STRUCTURE AND INFRASTRUCTURE ENGINEERING   11
Downloaded by [University of New England] at 07:29 22 December 2017

Figure 11. Time-variant analysis of bridge reliability.

Table 7. Impact on performance indicators based on B-WIM measurements in optimised bridge assessment.
Performance indicator Application Result Requirements
Measured influence line – MIL Step 1 Reveals changes in bridge boundary condition. In most cases reduces maxi- A few hours or a few 100 B-WIM
mum bending moments (in the case study bridge for around 35%) vehicle records
Girder distribution factor – GDF Step 2 Total load is redistributed on the girders identifying bridge critical element A few hours or a few 100 B-WIM
(cross section). Can reveal invisible degradation on structural elements. (In vehicle records
the case study bridge, changes the load effect from the step 1 for around
+/–6%)
Dynamic amplification factor – Step 3 Static traffic calculated from B-WIM data is multiplied with DAF to define If available, at least 100 000,
DAF the total site specific load effect. Can be used as an indicator for pavement or minimum two months of
condition. (In the case study bridge, reduces bending moment from step continuous vehicle records
2 for around 23%)

Conclusions, recommendations and future research at a European level and COST Action TU 1402 – Quantifying the Value
of Structural Health Monitoring, both supported by COST (European
The implementation of B-WIM measurements in the probabilis- Cooperation in Science and Technology) and on results of the Croatian
tic assessment of existing road bridges can reveal hidden bridge national project Performance indicators for assessment of existing bridges.
reserves and predict bridge reliability development over a speci-
fied lifetime. Consequently, such measurements can permit unre- Disclosure statement
stricted use of a bridge over a much longer remaining service life.
Therefore, it is recommended to recognise parameters and No potential conflict of interest was reported by the authors.
indexes calculated from the B-WIM measurement results as key
performance indicators for the assessment of bridge functionality Funding
at the system level, before the issue is raised to the network level.
This work was supported by COST (European Cooperation in Science and
The summary of these parameters and their contribution to an Technology) and Zagreb University.
optimised bridge assessment is presented in Table 7.
Further research will require evaluation of economic aspects
by quantifying the value of initial investments in B-WIM meas- References
urements, followed by more detailed bridge assessment, and Ang, A. H. & Tang, W. H. (1975). Probability concepts in engineering,
comparison with the reduction of bridge maintenance costs planning and design. New York, Toronto: Wiley.
within the management of a set of bridges at the network level. ARCHES D08. (2009). Report D08: Recommendations on bridge traffic load
monitoring. Brussels: European Commission.
ARCHES D10. (2009). Recommendations on dynamic amplification
Acknowledgments allowance. Brussels: European Commission.
This paper is based on the work carried out in the scope of the COST Action ARCHES D16. (2009). Recommendations on the use of soft, diagnostic and
TU 1406 – Quality specifications for roadway bridges, standardization proof load testing. Brussels: European Commission.
12   A. MANDIĆ IVANKOVIĆ ET AL.

Bien, J., Elfgren, L., & Olofsson, J. (Eds.). (2007). Sustainable bridges Montgomery, D. C. & Runger, G. C. (2014). Applied statistics and probability
– Assessment for future traffic demands and longer lives. Wrocław: for engineers (6th ed.). Hoboken, NJ: Wiley.
Dolnoslaskie Wydawnictwo Edukacyjne. Moses, F. (1979). Weigh-in-motion system using instrumented bridges.
Caprani, C. C.. (2005). Probabilistic analysis of highway bridge traffic ASCE Transportation Engineering Journal, 105, 233–249.
loading (PhD thesis. University College Dublin, Dublin, Ireland. Moses, F. & Verma, P. (1987). Load capacity evaluation of existing bridges
Caprani, C., Grave, S., O’Brien, E. J., & O’Connor, A. J. (2002). Critical (Report No. 301). Washington, DC: National Cooperative Highway
loading events for the assessment of medium span bridges. International Research Program (NCHRP).
Conference on Computational Structures Technology, 147, 1–11. O’Brien, E. J. (2011). Aplicações de pesagem em movimento na engenharia
Casas, J. R. (2016). European standardization of quality specifications for de pontes [Applications of weigh-in-motion in bridge engineering] In
roadway bridges: an overview. In Proceedings of the 8th international Proceedings of the 1st International Seminar of Weight in Motion,
conference on Bridge Maintenance, Safety and Management, Foz do Florianopolis, Santa Catarina, Brazil.
Iguaçu (Brazil), Taylor and Francis. O’Brien, E. J., González, A., Dowling, J., & Žnidarič, A. (2013). Direct
Castillo, E. (1988). Extreme value theory in engineering. Cambridge, MA: measurement of dynamics in road bridges using a bridge weigh-in-motion
Academic Press. system. The Baltic Journal of Road and Bridge Engineering, 8(4), 263–270.
COST. (2014). Memorandum of understanding – COST 1402. Brussels: O’Connor, A. & Enevoldsen, I. (2009). Probability-based assessment of
European Cooperation in the Field of Scientific and Technical Research. highway bridges according to the new Danish guideline. Structure and
COST 345. (2001). COST 345 – Procedures required for assessing higway Infrastructure Engineering, 5(2), 157–168.
structures. Brussels: European Cooperation in the Field of Scientific and O’Brien, E., Schmidt, F., Hajializadeh, D., & Zhou, X.-Y. (2015). A review of
Technical Research. probabilistic methods of assessment of load effects in bridges. Structural
DIN-1072. (1988). Straßen und Wegbrücken, Lastannahmen Erläuterungen. Safety, 53, 44–56.
Downloaded by [University of New England] at 07:29 22 December 2017

Berlin, Wien, Zurich: Beuth. Ralbovsky, M., McRobbie, S., Šajna, A., Leban, M. B., Sekulič, D., Žnidarič,
Enright, B. (2010). Simulation of traffic loading on highway bridges (PhD A. (2014). Final report of advanced bridge monitoring techniques –
thesis). University College Dublin, Dublin, Ireland. TRIMM Report D32. Vienna: AIT.
Eurocode. (2002). Eurocode 0 – Basis of structural design. Brussels: Šavor, Z. & Novak, M. Š. (2015). Procedures for reliability assessment of
European Committee for Standardization CEN. existing bridges. Journal of the Croatian Association of Civil Engineers,
Eurocode. (2003). Eurocode 1: Actions on structures – Part 2: Traffic loads 67(6), 557–572.
on bridges. Brussels: European Committee for Standardization CEN. Schneider, J. & Vrouwenvelder, T. (2017). Introduction to safety and
Eurocode. (2004). Eurocode 2: Design of concrete structures – Part 1–1: reliability of structures, structural engineering documents. Zurich: IABSE.
General rules and rules for buildings. Brussels: European Committee for Sivakumar, B., Ghosn, M., & Moses, F. (2011). Protocols for collecting and
Standardization CEN. using traffic data in bridge design (NCHRP Report No.683). Washington,
Eurocode. (2005). Eurocode 2: Design of concrete structures –- Part 2: DC: Transport Research Board.
Concrete bridges – Design and detailing rules. Brussels: European Skokandić, D. (2016). STSM Report – Application of B-WIM measurements
Committee for Standardization CEN. in assessment of existing road bridges. Zagreb.
Furundžić, B. (1969). Privremeni tehnički propisi za opterećenje mostova Skokandić, D., Mandić Ivanković, A., & Džeba, I. (2016). Multi – level road
na putovima PTP-5 (1949), Zbirka tehničkih propisa u građevinarstvu. bridge assessment. In Proceedings of 19th IABSE Congress, Stockholm,
Beograd: Građevinska knjiga. IABSE.
Getachew, A. & Obrien, E. J. (2007). Simplified site-specific traffic load Sofistik. (2014). Sofistik software for structural analysis. Germany:
models for bridge assessment. Structure and Infrastructure Engineering, SOFISTIK AG.
3(4), 303–311. Strauss, A., Bergmeister, K., & Mandić-Ivanković, A. (2016). Applied and
Gulvanessian, H., Calgaro, J. A., & Holický, M. (2012). Designers’ guide to research based performance indicator database for highway bridges
EN 1990 Eurocode: Basis of structural design. London: Thomas Telford. across Europe. In IALCCE – Fifth International Symposium on Life-Cycle
Holický, M. (2013). Introduction to probability and statistics for engineers. Civil Engineering. Delft, The Netherlands.
Verlag Berlin Heidelberg: Springer Science & Business Media. Strauss, A., & Mandić-Ivanković, A. (2016). Performance indicators for
Ieng, S.-S. (2015). Bridge influence line estimation for bridge weigh- road bridges – Categorization overview. In Paper presented at the WG
in-motion system. Journal of Computing in Civil Engineering, 29(1), meeting of COST Action TU1406 Quality Specifications for Roadway
06014006. Bridges, Standardization at a European Level. Belgrade, Serbia.
Jacob, B. (2002). Weigh-in-Motion of axles and vehicles for Europe (WAVE), Strauss, A., Mandić-Ivanković, A., Matos, J. C., & Casas, J. R. (2016).
General Report. Paris. Performance Indicators for roadway bridges, technical report of the
Jacob, B., OBrien, E. J., & Jehaes, S. (2002). Final Report of the COST 323 working group 1: Performance indicators.
action. Paris. Strauss, A., Mandić-Ivanković, A., & Soussa, H. (2016). Performance
JCSS (2002). Probabilistic model code (12th ed.). Lyngby: Author. indicators for road bridges. In Factsheet presented at the 3rd workshop of
Kalin, J., Žnidarič, A., & Kreslin, M. (2015). Using weigh-in-motion data the COST Action TU 1402. Barcelona, Spain.
to determine bridge dynamic amplification factor. In EVACES’15, 6th Sykora, M. (2008). Assessment of existing bridges using Bayesian updating.
International Confererence on Experimental Vibration Analysis for Civil In Proceedings of the 3rd International Conference ‘Reliability, safety
Engineering Structures. Duebendorf, Switzerland. and diagnostics of transport structures and means 2008’, University of
Karoumi, R., O’Brien, E. J., & Quiligan, M. (2006). Calculating an influence Pardubice, Czech Republic.
line from direct measurements. Bridge Engineering, 159(1–2), 31–34. Wiśniewski, D. F., Casas, J. R., & Ghosn, M. (2012). Codes for safety
Kirkegaard, P. H., Nielsen, S. R., & Enevoldsen, I. (1997). Heavy vehicles assessment of existing bridges – Current state and further development.
on minor highway bridges: calculation of dynamic impact factors from IABSE Structural Engineering International, 22(4), 552–561.
selected crossing scenarios. Structural Reliability Theory, R9722(172), Woodward, R. J. (2001). Guidelines for assessing load carrying capacity –
1–44. Final report of BRIME project. Brussels.
Koteš, P. & Vičan, J. (2012). Reliability levels for existing bridges evaluation Žnidarič, A. (2017). Influence of number and quality of weigh-in-motion
according to eurocodes. Procedia Engineering, 40, 211–216. data on evaluation of load effects on bridges (Doctoral dissertation).
Matos, J. C., Casas, J. R., & Fernandes, S. (2016). Quality specifications for University of Ljubljana, Ljubljana.
roadway bridges (Bridge Spec). In Proceedings of the 8th international Žnidarič, A., & Kreslin, M. (2012). Structures performance assessment,
conference on Bridge Maintenance, Safety and Management. Taylor and HeROAD Report 2.1. Ljubljana: ERANET Road II. Retrieved from www.
Francis. eranetroad.org
STRUCTURE AND INFRASTRUCTURE ENGINEERING   13

Žnidarič, A., Kreslin, M., Lavrič, I., & Kalin, J. (2012). Simplified approach Žnidarič, A., Kreslin, M., Leahy, C., O’Brien, E. J., Schmidt, F., & Pedersen,
to modelling traffic loads on bridges. Procedia – Social and Behavioural C. (2015). Guidelines on collecting WIM data and forecasting of traffic
Sciences, 48, 2887–2896. load effects on bridges. Re-GEN D3.1 Report.
Žnidarič, A., Herga, L., Pirman, B., Willenpart, T., Hevka, P., Močnik, C., Žnidarič, A., Kreslin, M., & Kalin, J. (2016). Weigh-in-motion and traffic
et al. (2015). Management of bridges in Slovenia, national report. Seoul: load monitoring. Fact Sheet WG, 2–9. COST Action TU 1402.
PIARC – World Road Congress.
Downloaded by [University of New England] at 07:29 22 December 2017

También podría gustarte