See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/341194565

Bridge Load Testing for Identifying Live Load Distribution, Load Rating,
Serviceability and Dynamic Response

Article  in  Frontiers in Built Environment · May 2020

DOI: 10.3389/fbuil.2020.00046

CITATIONS READS

4 508

5 authors, including:

Chuan-Zhi Dong Selcuk Bas

Palo Alto Research Center University of Central Florida
45 PUBLICATIONS   623 CITATIONS    48 PUBLICATIONS   147 CITATIONS   

SEE PROFILE SEE PROFILE

Marwan Debees Ninel Alver

University of Central Florida Ege University
1 PUBLICATION   4 CITATIONS    65 PUBLICATIONS   361 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

FBG sensing technology for structural health monitoring (SHM) View project

High-speed camera based experimental modal analysis View project

All content following this page was uploaded by Selcuk Bas on 07 May 2020.

The user has requested enhancement of the downloaded file.

 ORIGINAL RESEARCH
published: 06 May 2020
doi: 10.3389/fbuil.2020.00046

Bridge Load Testing for Identifying

Live Load Distribution, Load Rating,
Serviceability and Dynamic
Response
Chuanzhi Dong 1 , Selcuk Bas 1,2 , Marwan Debees 1 , Ninel Alver 1,3 and F. Necati Catbas 1*
1
Civil, Environmental and Construction Engineering Department, University of Central Florida, Orlando, FL, United States,
2
Department of Civil Engineering, Bartin University, Bartin, Turkey, 3 Department of Civil Engineering, Ege University, İzmir,
Turkey

In this article, dynamic and static load tests of a concrete highway bridge, which is a
deteriorated and repaired, are presented depending on displacement and strain data
for engineering decision making about the operation of a critical bridge. Static load
test was carried out to determine the live load distribution factor (DF) and load-rating
factor (RF) as well as serviceability by means of deflection limits. Modal characteristics
in terms of structural frequencies and mode shapes and impact factor (IM) were
identified from the dynamic load test for different truck-load and speed cases, and
finite element (FE) model. The DF and rating factor (RF) were also compared with
those calculated according to AASHTO standard and FE model. The results showed
that the DF calculated by American Association of State Highway and Transportation
Edited by: Officials (AASHTO) standard gave more conservative results when compared with the
Xin Ruan,
Tongji University, China experimental and FEM approaches. Similarly, the load-rating factor (RF) calculated by
Reviewed by: AASHTO standard yielded to more conservative results comparing with the experimental
Osman Eser Ozbulut, FEM approaches using practical DFs. Maximum deflections in static cases and dynamic
University of Virginia, United States
Wanshui Han,
cases were found to be within the limit calculated by (L/800) given in the AASHTO
Chang’an University, China code. Impact factors among all the cases were obtained much smaller than the one
*Correspondence: recommended by AASHTO standard (33%). The modal properties were obtained to
F. Necati Catbas track changes in dynamic behavior due to stiffness and boundary effects as well as for
[email protected]
finite element model calibration. The calibrated FE model of the bridge also indicated
Specialty section: that the load carrying capacity of the bridge is adequate after repair. Finally, the results
This article was submitted to
from the current study reveal that use of experimental data can be utilized to obtain
Bridge Engineering,
a section of the journal load rating with minimum interruption to bridge operations through computer vision
Frontiers in Built Environment technology and methods.
Received: 06 December 2019
Keywords: concrete bridge, load testing, load rating (RF), distribution factor (DF), impact factor (IM), modal
Accepted: 24 March 2020
characteristics
Published: 06 May 2020
Citation:
Dong C, Bas S, Debees M,
Alver N and Catbas FN (2020) Bridge
INTRODUCTION
Load Testing for Identifying Live Load
Distribution, Load Rating,
Bridge load testing is commonly employed to determine issues that cannot be easily
Serviceability and Dynamic Response. resolved by visual inspection or simple analysis. Visual inspection, load-testing, structural
Front. Built Environ. 6:46. health monitoring (SHM), non-destructive testing (NDT) and finite element (FE)-based
doi: 10.3389/fbuil.2020.00046 structural modeling are commonly utilized to address issues related to a bridge

Frontiers in Built Environment | www.frontiersin.org 1 May 2020 | Volume 6 | Article 46

Dong et al. Bridge Testing for Rating

or a population of bridges. For example, AASHTO-MBE (2018) effective complementary tool. More recently, civil infrastructure
recommends load testing for structural condition rating of technologies were grouped to be utilized for commonly seen
highway concrete deck bridges. The load test objectives may bridge failures (Bas and Catbas, 2019).
vary from case by case, and such a test may be needed on The main objective of this paper is to present a bridge load
particular bridge or a population of similar bridges in question test with particular engineering objectives regarding a multi-span
to make decisions such as bridge closure, bridge load posting, bridge with several spans of the same geometry and material
replacement, and retrofit. properties. The span under consideration is the worst condition
A general structural identification framework that also span and acceptable performance from this span will be favorably
encompasses bridge testing was presented in detail along with extrapolated to the entire bridge. The specifics goals of the load
utilization of field experimental and analytical studies for test are as follows: (i) obtain impact factors (IM) under different
decision making (Catbas et al., 2013). A particular bridge can loads and speeds, (ii) obtain dynamic responses in terms of
be tested to understand critical issues, and sometimes a sample structural frequencies and mode shapes, and (iii) obtain load
representative bridge can be tested to address issues related to distribution of the bridge and evaluation of the load carrying
the similar bridge population (Gokce et al., 2011). Similarly, a capacity of the bridge. During the bridge test, the strain/stress
representative bridge population sample can be tested to be able responses, displacements and accelerations under different truck
to make decisions on the entire bridge population (Catbas et al., loads were collected by using proper sensors and data acquisition
2005). In order to conduct rapid experimental test on a reinforced systems. For this aim, instrumentation plan and truck load
concrete bridge population, the researchers proposed a method configuration for the bridge were given. From static load-test
to determine the moment DFs for single-span-T-beam bridges with different truck loads, distribution (DF) and load rating (RF)
(Catbas et al., 2012). They presented that the new approach could factors, and deflection check of the bridge were determined.
be predicted live load reasonably well when compared to standard Dynamic load-testing with different truck loads and speeds were
girder analysis given in the (AASHTO, 2017) code. Based on carried out to calculate the IM and dynamic characteristics of the
the load and resistant factor rating (LRFR) approach, load rating bridge. The DF results from the experimental field test and FEM
factors were obtained for a fully instrumented bridge for three were compared with the standard formulation given in AASHTO
different methods (standard, experimental strain data, and FEM) (2017).
(Sanayei et al., 2016). Standard approach resulted lower rating
factors than the others. Static and dynamic testing was also
carried out by Catbas et al. (2006) for a concrete T-Beam bridge GENERAL FEATURES OF THE BRIDGE
taking into account before and after retrofit of the bridge through
As shown in Figure 1A, the bridge considered in the study is a
carbon fiber–reinforced polymer (CFRP) material. According
multi-span concrete bridge with a bascule section in the middle,
to the results of experimental data for both cases, they clearly
located in Florida, United States. The bridge was constructed
showed that the CFRP retrofit had an ability to improve structural
in 1964 and has a total length of 912 m. Each span consists of
response of the concrete bridge. In order to quantify the effect of
five pre-stressed I-beam spans, two flanking spans, and a steel
deterioration on live-load response of an existing concrete bridge,
double leaf bascule main span, which is 39.5 m between trunnion
Torres et al. (2019) performed experimental and numerical study.
centers. AASHTO Type II Girders are spaced at 2.4 m with
Based on the results from load-test as well as visual inspection,
an 18 cm cast in place deck and 5.1 cm wearing surface. All
they found heavily deteriorated deck, undamaged girders and
approach spans are 15.9 m each. The substructure is composed
moderate connection problems at the longitudinal joints. By
of two cast in place reinforced concrete end bent caps founded
performing a parametric study on the calibrated the FE model
on 61 cm square pre-stressed concrete piles with rubble riprap
of the bridge, the moment and shear girder distribution factor
slope protection retained by a seawall system, 53 intermediate
(DF) equations were developed in that study. Tawadrous et al.
reinforced concrete bent caps founded on 61 cm square pre-
(2019) carried out live load-testing on two concrete bridges
stressed concrete piles. The aged bridge underwent a retrofit of
with different deck systems: (i) newly developed precast concrete
the deteriorated girders, including removal of all spalled and
deck and (ii) standard cast-in-place (CIS) deck to compare their
delaminated concrete, cleaning the corroded steel and rebar,
performances on the basis of strain and deflection. The influence
installing special splice when a strand was severed or has more
of foundation movements and geohydraulic hazards on load
than 50% sectional loss, repairing hairline cracks. The load test
rating of a highway bridge was also investigated by Davis et al.
here would explore if the load carrying capacity is adequate.
(2018). The proposed load rating approach was obtained to give
more conservative RF values than those of the standard method
if foundation movements were considered. More recent load- INSTRUMENTATION PLAN
test implementations to concrete bridge can be found in the
study of Omar and Nehdi (2018). On the other hand, some As shown in Figure 1B, three types of sensors were installed on
researchers also recently demonstrated that more effective bridge the bridge. Totally, 15 accelerometers were installed at the 1/4
condition assessments could be done through other technologies span, mid span and 3/4 span of five girders (G1–G5) to test the
(computer vision, image, thermal camera, etc.) (Agdas et al., 2016; dynamic responses during load test.
Zaurin et al., 2016; Hiasa et al., 2018; Dong and Catbas, 2019). Five displacement sensors (i.e., potentiometers) were installed
In these studies, these technologies were determined to be an at the mid span of each girder to measure the displacement.

Frontiers in Built Environment | www.frontiersin.org 2 May 2020 | Volume 6 | Article 46

Dong et al. Bridge Testing for Rating

FIGURE 1 | (A) The concrete highway bridge (B) Sensor instrumentation plan.

Three cameras were employed to measure the displacements at the static loading cases. In this study, to point out the loading
the same location. The camera-based monitoring and computer location or step, Si (i = 1, 2, 3, 4) is added at the end of
vision implementation will be presented separately in other the case name. For example, the load location in Figure 3A is
publications. Five strain gauges were installed at the 1/4 span represented by T1L1S1.
of each girder. One camera was employed to record the traffic In the dynamic test, the load configuration was similar to the
footage. Details of the sensors and cameras are shown in Table 1. static test. The difference is instead of putting the trucks statically
All the sensors were installed at the bottom of the girders as in all four locations, the trucks moved in the lane with different
shown in Figure 2A. speeds. According to the types of the truck, moving speeds and
the lanes, Table 3 summarizes the dynamic loading cases.

LOADING PLAN
STATIC LOAD-TEST
The truck loading test plan consisted of static and dynamic loads
with two different trucks separately. Two types of trucks, Truck General
1 (T1) and Truck 2 (T2) were operated to conduct the load test. Figure 4A shows the displacement results of T2L1. This
The trucks and loading plan are shown in Figures 2B, 3A,B. figure shows both the displacement results from cameras
In the static test, the truck (T1 or T2) was stopped and and potentiometers. The results from the cameras and the
remained at four different locations of each lane (Lane 1 and potentiometers are very consistent with each other and the
Lane 2), and it took four steps for one test round. According maximum difference is within 2.5% range. Only small motions
to the types of the truck and the lanes, Table 2 summarizes of the potentiometers installed at the two exterior girders (P6 and

Frontiers in Built Environment | www.frontiersin.org 3 May 2020 | Volume 6 | Article 46

Dong et al. Bridge Testing for Rating

TABLE 1 | Specifications of sensors and cameras.

Sensor Specifications

Accelerotneter ModelNo.:PCB603C01
Weight: 1.8 gz,(51 g)
Sensitivity: (=10%) 100 mV/g (10.2 mV/(m/s2 )
Frequency Range: (±3dB) 30–600,000 cpm (0.5–10,000 Hz)
Sensing Element: Ceramic
Measurement Range: ±50 s (=490 m/s2 $2 )

Strain gauge Model No.: KYOWA-kc-120-120-Al-ll

Gauge factor: 2.13±1.0%
Gauge length: 120 mm
Gauge resistance (24◦ C, 50%RH): 119.8±0.2 

Displacement sensor Model No.: BEI9615 potentiometer

Linearity: ±0.35%; Full scale: 38 mm

Camera (1) Camera for displacement: Z Camera El

Resolution: 4K (3840 × 2160 pixels)
Speed: 30 frame per second
Lens: Olympus 75–300 mm zoom lens
(2) Camera for traffic: Canon VDOA HF R42
Resolution: lOSOp (1920 × 1080 pixels)
Speed: 60 frame per second
Lens: built-in 32× zoom 2.8–89.6 mm lens

FIGURE 2 | (A) Sensors installed on the bottom of the girders (B) Loading trucks.

P10) were observed during the load test. Therefore, only results when the truck was loaded on L1, the girder under the truck
from cameras are shown in Figure 4A and only these results were has the largest displacement response. For example, here P9 is
used for assessment. Details of the test and use of computer vision the mid span of Girder 4 and it gives the largest displacement,
based implementation are given by Catbas et al. (2019) and will 2.76 mm at step 3 (S3) among all the girder measurement points.
be presented in other presentations. From Figure 4A, it can be The displacement of P6, P7, P8, and P10 at step 4 is 0.41, 1.48,
seen that the displacement gives a flat level at each test step and 2.69, and 1.49 mm, respectively.

Frontiers in Built Environment | www.frontiersin.org 4 May 2020 | Volume 6 | Article 46

Dong et al. Bridge Testing for Rating

TABLE 2 | Static loading cases. FE Model of the Bridge

Case no. Case name Truck Lane As shown in Figure 5, FE model of the bridge was established
using beam elements and shell elements for the reinforced
1 T1L1 Tl LI concrete prestress girders and deck, respectively. Besides, tendon
2 T1L2 Tl L2 elements was utilized to consider the prestress effect in the
3 T2L1 T2 LI analyses. All considerations for FE model of the bridge were
4 T2L2 T2 L2 obtained according to its calculation report and project drawings.
For this aim, structural analysis software, SAP2000 (CSI, 2019),

FIGURE 3 | (A) Static loading plan of Truck 1 (T1) (B) Truck 2 (T2).

Frontiers in Built Environment | www.frontiersin.org 5 May 2020 | Volume 6 | Article 46

Dong et al. Bridge Testing for Rating

TABLE 3 | Dynamic loading cases. at the same section depending on the data being used. The
DFs calculated from the experimental displacement results are
Case no. Case name Truck Speed (mph) Lane
shown in Figure 4B for T2L1 load case. It should be noted that
5 T1L1-35 Tl 35 LI Figure 4B has two vertical axes and the left vertical axis represents
6 T1L2-35 Tl 35 L2 the displacement and the right one is for DF. The DF can also
7 T2L1-35 T2 35 LI be calculated from the strain data. Figure 7A shows the strain
8 T2L2-35 T2 35 L2 result of each girder at 1/4 span. Due to the signal noise and
9 TlLl-55 Tl 55 LI the small strain measurements, the raw data (in red line) was
10 T1L2-55 Tl 55 L2 filtered and the filtered data was shown in blue line. The DF
11 T2L1 -55 T2 55 LI results are very similar to those obtained from displacements.
12 T2L2-55 T2 55 L2 The strain of Girder 4 (G4) gives the largest value due to the
loading also shown in Figure 7A. Figure 7B illustrates the DFs
obtained from the load test with Truck 2 (loaded truck T2) on
was utilized. The bridge was then updated as per the modal
Lane 1 (L1). It is observed that the max DF is 0.4 right under the
characteristics and displacement results obtained from the
truck load. This DF of 0.4 is indicating that there is good load
experimental field test. The live load DF and load-rating (RF) of
distribution across the bridge. This fully loaded truck DF results
the updated FE model are also taken into account for the sake
can be predicted due to T2 on L2, and this will allow addition
of the comparison of experimental and AASHTO (AASHTO,
of the DFs to obtain multi-load case, which is more conservative
2017) calculation.
than the multi-presence factor given in AASHTO (2017). Similar
calculations for DF were carried out based on the results from the
Linearity Check FE analysis as given in Figure 7A and almost similar DF values to
The linearity check of the bridge was carried out mainly for
the experimental test were yielded.
two reasons. First, it is to determine the load rating with
AASHTO (2017) utilizes a multi-parameter formulation for
larger load levels and secondly to be able to calculate multiple
the load distribution. The DF was employed for the load rating
presence factors by combining separate test results. Under regular
of the bridge by means of girder line analysis. The detailed
operational loads, the bridge response should not reach ultimate
formulation for DF calculation can be seen in Table 4. It should
response levels. As such, it would be possible to observe linear
be noted that the DF calculations by using displacement or strain
behavior under given increasing load conditions. To check the
only consider one single truck in one of the two lanes. In real
linearity of the bridge, the girders with the largest responses were
cases, there is still a chance that multiple vehicles are present in
selected and here it is for the case the truck loaded on L1, the
multiple lanes at the same time. AASHTO (2017) code considers
girder is G4. Two loading cases T1L1 and T2L1 were taken as
this scenario and use the larger value of DFs between the multiple
an example to check the linearity (Figure 6). The load increase is
design lanes loaded and single one.
observed to be 2.65 due ratio of the weight of T2 to T1 as shown in
In this study, due to the symmetry of this bridge, the DFs
Figure 6. It should be noted that the axle number and total length
of multiple lanes (two vehicles here) were also calculated and
of truck are not exactly the same. In addition, the placement of
shown in Figure 8A. The unit of displacement in the results is
the trucks in real life are somewhat different as one truck has
millimeter and the unit of strain is µε. From Figure 8A, it can
two and the other has three rear axles as shown in Figure 3.
be seen that the DFs calculated by experimental displacement
From Figure 6, it can be seen that the displacement and the
and strain at Girder 3 (G3) are very close to the ones calculated
strain increase ratios are quite consistent and gives 3.16 and 3.08,
by AASHTO (2017). Similarly, this agreement was also seen for
respectively. Considering the change of axle number and total
the DF results from FEM as shown in Figure 8A. It should be
length of truck, and also the similar trend between displacement
mentioned again that the multi girder DFs were calculated under
and strain (both in the range of 15 ± 1% to the difference of
two heavy loaded truck side by side and this would create a
weight increase), it is fair to say that the linearity is validated. This
load case more conservative than the AASHTO (2017) based DF
provides proof that the bridge behaves linearly under operational
results. In other words, considering a combined DF of 0.6 can be
loads in the range of 26–70 kip range.
regarded to be conservative. While the DFs of the other girders
calculated by experimental and FEM displacement and strain are
Distribution Factor Using Experimental much smaller than those calculated by AASHTO (2017). It means
Data and FEM that AASHTO (2017) gives more conservative DF, especially for
Using the responses across the bridge, we can obtain bridge load the girders away from the boundary of two adjacent lanes, e.g.,
distribution, which is critical for bridge response and load rating. G1, G2, G3, and G4. For the girder close to the boundary of two
With the displacement results of each girder, the DF can be adjacent lanes, e.g., G3, the DFs calculated from displacement
calculated based on Eq. (1) below: and strain are close to those calculated by AASHTO (2017),
αi but still smaller.
DFi = P10 (1)
j=6 αj
Load Rating
where i and j are the girder numbers with the range from 6 By utilizing the results of DFs from the experimental study and
to 10 and ai or aj are the strain or displacement of the girder FE model, a simple method yet widely used the load rating

Frontiers in Built Environment | www.frontiersin.org 6 May 2020 | Volume 6 | Article 46

Dong et al. Bridge Testing for Rating

FIGURE 4 | (A) Displacement results of static load case T2L1 (B) Distribution factor calculated from displacement results.

factor can be calculated using formulations given in AASHTO adjacent lanes, e.g., G1, G2, G3, and G4. For the girder close
(2017). The rating factors (RF) calculated by using different DFs to the boundary of two adjacent lanes, e.g., G3, the RFs from
and HL 93 truck for Strength Limit I are listed as shown in displacement and strain are close to those calculated by AASHTO
Figure 8B. General formulation for load rating from AASHTO (2017), but still larger. The experimental case also presents a more
(2017) is given with Eq. (2). The load factors such as ϕ, ϕs , conservative case due to low probability of having two such heavy
ϕc ,γDC ,γDW ,γp ,γL can be found in AASHTO (2017) standard. trucks side by side creating the most critical load case. As a result,
it can be concluded that the RFs in real life can even be considered
ϕc ϕs ϕR − γDC DC − γDW DW ∓ γp P to be even higher than reported in Figure 8B.
RF = (2)
γL (LL + IM)
As stated in the calculation of DF, the load rating using the DF Deflection of Limit Check
calculated from AASHTO (2017) also indicates that AASHTO The deflection limit check is to check whether the maximum
codes give the more conservative rating factors than the ones displacement of mid span is larger than the value calculated by
obtained by experimental and FEM method (RF-disp and RF- (L/800). This limit is commonly used to evaluate the serviceability
strain), especially for the girders away from the boundary of two of the bridge. Here L is the length of span, which is 15,849.61 mm

Frontiers in Built Environment | www.frontiersin.org 7 May 2020 | Volume 6 | Article 46

Dong et al. Bridge Testing for Rating

FIGURE 5 | FE Model of the bridge.

FIGURE 6 | Trucks loaded the left lane Truck-1 and Truck-2.

(52 ft). Here (L/800 = 52/800 = 0.065 ft) is 19.81 mm. As of the other objective of the modal testing is to use its results to
demonstrated in Figure 6, the maximum displacement of mid update the FE model of the bridge as developed in the previous
span of Girder 4 in loading case T2L1 under static vehicle load section above. These dynamic results can be tracked over time to
is 2.75 mm, which is considerably less than the serviceability determine any global changes. They were employed to validate
deflection limit. This finding can be somewhat expected due or calibrate FEM of the bridge. Here load case T2L1-55 (Truck 2
to the number of AASHTO girders and the span length of moved on Lane 1 with a speed of 55 mph = 80.7 ft/s) is considered
the bridge. In conclusion, the bridge fulfills AASHTO (2017) for the dynamic analysis. It is seen based on the speed of the truck
serviceability requirement. it takes about 0.64 s to cross the bridge. Figure 9A shows the
acceleration time histories of each measurement point for load
case T2L1-55. Figure 9B shows the FFT analysis of acceleration
DYNAMIC LOAD-TEST at P4 and possible modal frequencies were marked. Based
on all the collected time histories, operational modal analysis
Modal Testing methods, enhanced frequency-domain decomposition (EFDD)
To estimate the dynamic properties of the bridge, the and stochastic subspace identification-unweighted principal
accelerations versus time histories collected by fifteen component (SSI-UPC) as shown in Figure 9C, is employed to
accelerometers installed on the bridge were processed. One identify the modal frequencies, damping ratio and mode shapes

Frontiers in Built Environment | www.frontiersin.org 8 May 2020 | Volume 6 | Article 46

Dong et al. Bridge Testing for Rating

FIGURE 7 | (A) Strain results of static load case T2L1 (B) Distribution factor calculated from strain results.

TABLE 4 | Distribution factor calculation in AASHTO (2017).

DF for moment, interior girder DF for moment, exterior girder

One design lane loaded: One design lane loaded:

 0.4  0.3  0.1 5.5
mgSI S S Kg mgSE
moment = S
moment = 0.06 + 14 L 12Lts3

Two or more (multiple) design lanes loaded: Two or more (multiple) design lanes loaded:
moment = e(mgmoment )
mgME
 0.6  0.2  0.1 MI
S S Kg
mgMI
moment = 0.075 + 9.5 L 3 de
12Lts e = 0.77 + 9.1 ≥ 1.0
de is positive if girder is inside of barrier, otherwise negative
AASHT0 4.6.2.2.2b-l
AASHT0 4.6.2.2.2.1d-l

S = girder spacing (ft); L = span length (ft); tj , = slab thickness (in.); Kg = n(Ig + e2g A); n = modular ratio of girder and deck; Ig = moment of inertia of girder (in.∗ ), eg = girder
eccentricity which is the distance from girder centroid to middle centroid of slab, (in.); A = girder area (in.2 ).

Frontiers in Built Environment | www.frontiersin.org 9 May 2020 | Volume 6 | Article 46

Dong et al. Bridge Testing for Rating

FIGURE 8 | Comparison (A) DF (B) RF results.

(Artemis-Modal, 2015). Both methods are generally used to parts. The philosophy of this approach is to represent dynamics
identify modal parameters of linear systems using output-only of a structure is modeled as n × n state matrix (n: state space
measurements. In the SSI method, dynamic response of a dimension). Observation matrix can be estimated from a part of
structural system is assumed to consist of state and observation the state matrix. Thus, the system response vector that includes

Frontiers in Built Environment | www.frontiersin.org 10 May 2020 | Volume 6 | Article 46

Dong et al. Bridge Testing for Rating

FIGURE 9 | (A) Acceleration time-histories (B) FFT analysis of P4 acceleration (C) Stabilization diagram for SDDD (D) Estimated modal characteristics.

Frontiers in Built Environment | www.frontiersin.org 11 May 2020 | Volume 6 | Article 46

Dong et al. Bridge Testing for Rating

TABLE 5 | Comparison of modal frequencies. Impact Factor

No Type Experimental (Hz) FEM (Hz) Difference (%) Identification of the IM can be important for bridge operation
to possibly reduce live load effects on the bridge. This can be
SSI EFDD FEM-SSI FEM-EFBD achieved by improving the bridge surface, smooth expansion
joints, or limiting the traffic speed. As a result, it is critical to
1 Bending 8.349 8.307 7.97 4.54 4.06
determine the IM. In this study, the most conservative cases
2 Torsional 9.395 9.235 9.03 3.89 222
were selected to obtain the impact effect caused by moving
3 Torsional 12.453 12.879 15.56 24.95 20.82
loads. Table 6a shows the IM of three girders (G2, G3, and G4)
4 Torsional 24.529 26.821 28.96 18.06 7.98
calculated by the displacement data of T2L2, T2L2-35, and T2L2-
5 Bending 29.14 29.888 29.34 0.69 1.83
55. Here, the speed of the truck increased from static cases to 35
and 55 mph. From Table 6a, it can be seen that with the increase
TABLE 6 | (a) Impact factor Truck 2 for displacement (b) Truck 1 for strain. of the speed, the IM also increases. The maximum IM obtained
from all data sets is 15.94%, (Girder 3, 55 mph), which is much
(a) smaller than the value (33%) recommended by AASHTO (2017).
Girder no. Disp Disp Disp IM-35 (%) IM-55 (%) Table 6b shows the IM of three girders (G2, G3, and G4)
(static) (35mph) (55mph) calculated by the strain data of T1L1, T1L1-35, and T1L1-55.
One should note that the strain data has more signal noise than
G2 2.6489 2.7139 2.9762 2.45 12.36
displacements. In addition, the displacement data were cross-
G3 2.1229 2.1698 2.4613 2.21 15.94
validated using computer vision data. Nevertheless, the IM from
G4 1.0646 1.128 1.2075 5.96 13.42
strain measurements were also obtained. The maximum IM is
(b) 22.3%, (Girder 2, 35 mph), which is also much smaller than
the value (33%) recommended by AASHTO (2017). While the
Girder no. Strain Strain (35 Strain (55 IM-35 IM-55
(static) mph) mph)
impact factor of IM-35 of Girder 2, 22.3%, is larger than that
of IM-55, 10.5%, which is not reasonable as one would expect a
G2 6.98 8.53 7.71 22.3% 10.5% higher IM under higher speed vehicle, as also shown in Table 6a
G3 19.S5 20.45 20.47 3.0% 3.1% with deflections. For Girder 3, the IM are almost the same and
G4 25.46 23.83 22.55 −6.4% −11.4% for Girder 4, they are negative. The abnormal values here might
be caused by the signal noise received by the strain gauges.
Combining Table 6a and Table 6b, the author recommends
that the IM using deflections can be utilized. For the truck with
the modal characteristics of structure considered is obtained as 35 mph velocity, the IM is 5.96% and the truck with 55 mph
different version of observable part of the state matrix. More velocity, the IM is 15.94%.
details can be found in Van Overschee and De Moor (1996).
The main idea of EFDD approach is to decompose appropriately
the system response into a certain number of independent
SUMMARY AND CONCLUSION
single degree of freedom systems (SDOF). First, spectral density
matrices are predicted from raw experimental data. Singular In this study, static and dynamic load test were conducted for
value decomposition of the estimated spectral density matrices is an existing in-service concrete pre-stressed girder bridge through
then performed. The average singular values and corresponding the field test. Thus, the DF, load rating (RF), deflection limit
singular vectors present modal frequencies and mode shapes of check, modal characteristics and IM were obtained and compared
structural system, respectively. More details can be reached to with those from the conventional calculation methods and FE
the study of Brincker et al. (2000). model. Bridge behavior characterized with these indices (e.g., DF,
Figure 9D presents the estimated modal parameters and RF) are to be obtained as summarized in this paper to be able to
modes shapes. The frequency of the first bending mode (Mode make decisions about load posting, repair effectiveness, reducing
1) is 8.35 Hz and the frequency of the second bending mode traffic speed and ultimately major retrofit or replacement.
(Mode 5) is 29.14 Hz. The others are torsion modes. The A general framework for such a field study is presented along
outcomes from FE analysis of the bridge were also given in with an example on a typical highway bridge. Some of the specific
Table 5 with the comparison of those from the experimental. findings for this bridge are summarized in the following:
When compared with the updated FE model of the bridge, it
was observed to be difference only for the 3rd mode. However, • The DF of live load calculated by AASHTO standards
the modal participation mass ratio of this mode (3rd mode) was gives more conservative results when compared with the
obtained to be relatively lower, which means that the importance experimental and FEM approaches. The DF for a single
of this mode on a dynamic load can be neglected. Such data and load case is 0.4. The conservative experimental DF with
dynamic response can be collected efficiently under operating two heavy truck loads side by side gives 0.59, which is less
traffic and can be evaluated to track any stiffness or boundary than 0.62 of AASHTO code. This value was obtained as
condition change. 0.52 from the updated FE model.

Frontiers in Built Environment | www.frontiersin.org 12 May 2020 | Volume 6 | Article 46

Dong et al. Bridge Testing for Rating

• The load-rating factor (RF) of live load calculated by DATA AVAILABILITY STATEMENT
AASHTO standards gives more conservative results
comparing with the experimental and FEM approaches The datasets generated for this study are available on request to
using practical DFs. Rating factors of Strength Limit I the corresponding author.
are all larger than “2.0” for HL93 trucks for single lane
and RF for multilane is 1.10 which is slightly larger
than AASHTO code.
• Maximum deflections in static cases and dynamic
AUTHOR CONTRIBUTIONS
cases are within the limit calculated by (L/800) FC investigated the study. CD conducted the field studies.
and deflections are much less than the AASHTO SB conducted the field and analytical studies. NA and MD
code based L/800. supported field operations and document development. All
• Impact factors among all the cases are much authors contributed to the field studies and preparation of
smaller than the one recommended by AASHTO the manuscript.
standards (33%). We observe 16%, which
was observed under fully loaded truck at
55 mph on the bridge.
• Modal testing results were obtained through the FUNDING
experimental data. These results were used for
developing updated FE model of the bridge. The authors declare that this study received funding from Sanalil
The DF and RF outcomes from the updated FE Construction Inc. The funder was not involved in the study
model were obtained to be good agreement with design, collection, analysis, interpretation of data, the writing of
those form the field test. Hence, the updated this article or the decision to submit it for publication.
FE model can be adopted reliably for advanced
analysis of the bridge.
• The study showed that bridge condition assessment ACKNOWLEDGMENTS
could be conducted reliably fast with no need
for blocking traffic/interrupting bridge’s operation. The authors would like to acknowledge the input and field
Therefore, the framework given in the study can be support by Mr. Manny Cabrera, PE throughout the study. Mr.
practically implemented to a bridge in the same bridge Armon Rahmankhah, PE also provided field coordination for the
population. Computer vision methods and technology study, which facilitated the field applications. In addition, they
(camera, image, etc.) can be considerably effective for would like to acknowledge members of the Civil Infrastructure
this aim. Technologies for Resilience and Safety (CITRS) research group
• Based on the findings, it is shown that the bridge at University of Central Florida for their support in creation
has sufficient load carrying capacity and the retrofitted of this work. The authors acknowledge Mr. Wesley Shattenkirk
bridge can carry large truck loads. The full truck is and Mr. Ivan Del Barco of CITRS for their valuable support
∼70 kips, very comparable to 72 kips HL-93 AASHTO during the field execution of the study. The second and fourth
truck. For the load rating under multiple vehicles, the authors would like to kindly acknowledge The Scientific and
70 kip truck was employed by means of superposition Technological Research Council of Turkey (TUBITAK) through
due to linearity to obtain the rating factor for the grant number 2219. The authors would also like to thank Dr. Fuat
most critical condition. It is shown the AASHTO based Aras for his discussions on dynamic analysis particularly related
formulations satisfy the rating factor, and even the to the use of dynamic analysis software. The authors acknowledge
calibrated FE based load rating is even higher. As a the contributions of these individuals. The study presented here
result, the bridge can continue to serve and no load was supported by Sanalil Construction Inc. where Dr. Catbas
posting is necessary based on the results given in this served as the PI. The PI and his team thank Sanalil Construction
paper. for sponsoring this study.

REFERENCES Artemis-Modal (2015). User’s Guide, Structural Vibration Solutions. Aalborg:

Artemis-Modal.
AASHTO (2017). LRFD Bridge Design Specifications. Washington, Bas, S., and Catbas, F. N. (2019). “Bridge failures and mitigation using monitoring
DC: American Association of State Highway and Transportation technologies,” in Developments in International Bridge Engineering, eds A.
Officials. Caner, P. Gülkan, and K. Mahmoud (Cham: Springer).
AASHTO-MBE (2018). The Manual for Bridge Evaluation (MBE), 3rd Edn. Brincker, R., Zhang, L., and Andersen, P. (2000). “Modal identification
Washington, DC: American Association of State Highway and Transportation from ambient responses using frequency domain decomposition,” in
Officials. Proceedings of the 18th International Modal Analysis Conference (IMAC),
Agdas, D., Rice, J. A., Martinez, J. R., and Lasa, I. R. (2016). Comparison of visual San Antonio, TX.
inspection and structural-health monitoring as bridge condition assessment Catbas, F. N., Ciloglu, K., and Aktan, A. E. (2005). Strategies for condition
methods. J. Perform. Const. Facil. 30:04015049. doi: 10.1061/(ASCE)CF.1943- assessment of infrastructure populations: a case study on T-beam bridges.
5509.0000802 Struct. Infrastruct. Eng. J. SIE 1, 221–238. doi: 10.1080/15732470500031008

Frontiers in Built Environment | www.frontiersin.org 13 May 2020 | Volume 6 | Article 46

Dong et al. Bridge Testing for Rating

Catbas, F. N., Dong, C., Bas, S., and Alver, N. (2019). Indian River Bridge Test, Final Omar, T., and Nehdi, M. L. (2018). Condition assessment of reinforced concrete
Project Report by UCF CITRS. Fort Lauderdale, FL: Sanalil Construction Inc. bridges: current practice and research challenges. Infrastructures 3:36. doi:
Catbas, F. N., Gokce, H. B., and Gul, M. (2012). Practical approach for estimating 10.3390/infrastructures3030036
distribution factor for load rating: demonstration on reinforced concrete Sanayei, M., Reiff, A. J., Brenner, B. R., and Imbaro, G. R. (2016). Load rating
T-beam bridges. J. Bridge Eng. 17, 652–661. doi: 10.1061/(ASCE)BE.1943-5592. of a fully instrumented bridge: comparison of LRFR approaches. J. Perform.
0000284 Construct. Facil. 30:04015019. doi: 10.1061/(ASCE)CF.1943-5509.0000752
Catbas, F. N., Grimmelsman, K. A., Ciloglu, S. K., Burgos-Gil, I., and Coll-Borgo, Tawadrous, R., Morcous, G., and Maguire, M. (2019). Performance evaluation
M. (2006). Static and dynamic testing of a concrete T-beam bridge before and of a new precast concrete bridge deck system. J. Bridge Eng. 24:04019051.
after carbon fiber–reinforced polymer retrofit. Transport. Res. Rec. 1976, 76–87. doi: 10.1061/(ASCE)BE.1943-5592.0001422
doi: 10.1177/0361198106197600109 Torres, V., Zolghadri, N., Maguire, M., Barr, P., and Halling, M. (2019).
Catbas, F. N., Kijewski-Correa, T., and Aktan, A. E. (2013). Structural Identification Experimental and analytical investigation of live-load distribution factors for
of Constructed Systems: Approaches, Methods, and Technologies for Effective double tee bridges. J. Perform. Construct. Facil. 33:04018107. doi: 10.1061/
Practice of St-Id. Reston: American Society of Civil Engineers. (ASCE)CF.1943-5509.0001259
CSI (2019). SAP2000: Integrated Finite Element Analysis and Design of Structures. Van Overschee, P., and De Moor, B. (1996). Subspace Identification for Linear
Berkeley, CA: Computers and Structures Inc . Systems: Theory - Implementation - Applications. Dordrecht: Kluwer Academic
Davis, N. T., Hoomaan, E., Sanayei, M., Agrawal, A. K., and Jalinoos, F. (2018). Publishers.
Integrated superstructure-substructure load rating for bridges with foundation Zaurin, R., Khuc, T., and Catbas, F. N. (2016). Hybrid sensor-camera monitoring
movements. J. Bridge Eng. 23:04018022. doi: 10.1061/(ASCE)BE.1943-5592. for damage detection: case study of a real bridge. J. Bridge Eng. 21:05016002.
0001232 doi: 10.1061/(ASCE)BE.1943-5592.0000811
Dong, C.-Z., and Catbas, F. N. (2019). A non-target structural
displacement measurement method using advanced feature matching Conflict of Interest: The authors declare that the research was conducted in the
strategy. Adv. Struct. Eng. 22, 3461–3472. doi: 10.1177/136943321985 absence of any commercial or financial relationships that could be construed as a
6171 potential conflict of interest.
Gokce, H. B., Catbas, F. N., and Dan, M. F. (2011). System reliability and load rating
evaluation of a movable bridge. Transp. Res. Rec. J. Transp. Res. Board 2251, Copyright © 2020 Dong, Bas, Debees, Alver and Catbas. This is an open-access article
114–122. doi: 10.3141/2251-12 distributed under the terms of the Creative Commons Attribution License (CC BY).
Hiasa, S., Karaaslan, E., Shattenkirk, W., Mildner, C., and Catbas, F. N. (2018). The use, distribution or reproduction in other forums is permitted, provided the
“Bridge inspection and condition assessment using image-based technologies original author(s) and the copyright owner(s) are credited and that the original
with UAVs,” in Proceedings of the Conference on Structures Congress, (Reston: publication in this journal is cited, in accordance with accepted academic practice. No
ASCE), 217–228. use, distribution or reproduction is permitted which does not comply with these terms.

Frontiers in Built Environment | www.frontiersin.org 14 May 2020 | Volume 6 | Article 46

View publication stats