Provided by the author(s) and University College Dublin Library in accordance with publisher

policies. Please cite the published version when available.

Title Fleet Monitoring - Using Sensors in a Fleet of Passing Vehicles to Monitor the Health of
Bridges

Authors(s) O'Brien, Eugene J.; Ren, Yifei; Wang, Shuo; Keenahan, Jennifer; McCrum, Daniel

Publication date 2021-09-17

Conference details The 9th International Conference on Experimental Vibration Analysis for Civil Engineering
Structures, Online Event, 14-17 September 2021

Link to online version https://ec-intl.co.jp/evaces2021/

Item record/more information http://hdl.handle.net/10197/13161

Downloaded 2022-10-04T06:15:57Z

The UCD community has made this article openly available. Please share how this access
benefits you. Your story matters! (@ucd_oa)
 Fleet Monitoring – Using Sensors in a Fleet of Passing Vehicles to Monitor the Health of
Bridges

Eugene J. OBrien, Yifei Ren, Shuo Wang, Jennifer Keenahan, Daniel P. McCrum
University College Dublin, Ireland.

ABSTRACT
This paper proposes the use of a fleet of instrumented vehicles to monitor the condition of infrastructure and bridges. It is
anticipated that data from privately owned vehicles with low-cost accelerometer and GPS data, will be available for this purpose
in the future. An inverse version of the well known Newmark-Beta method is proposed to determine road/rail surface profile
from measured accelerations. Some results are reported from an instrumented train that made repeat runs on railway track over
a period of a month. For bridge health monitoring, the concept of a moving reference influence line is proposed as a damage
indicator. It is shown in simulation to give good indications of bearing damage in a simply supported bridge.
Keywords: Bridge, fleet monitoring, drive-by, SHM, structural health monitoring.

1. INTRODUCTION
There has been increased interest in recent years in methods of electronic health monitoring of road and rail bridges. However,
while individual sensors are inexpensive, a complete monitoring system installed on a bridge, with a power source and data
acquisition electronics, is relatively expensive. This has become commonplace on larger cable-supported bridges but would be
prohibitively expensive for the entire stock of small and medium-span bridges in a region or country. Some early results suggest
that the problem could be addressed with drive-by monitoring [1-3], i.e., where data from an instrumented vehicle crossing a
bridge would be used to infer information on the bridge’s condition. However, the challenge is considerable, given that a vehicle
at a typical highway speed, typically crosses a short-span bridge in one or two seconds. Fleet monitoring has the potential to
finally solve the bridge health monitoring problem by combining drive-by data from a large number of vehicle crossings.
Vehicles have ever greater numbers of sensors, including accelerometers, and it seems reasonable to assume that, in the medium
or long term, this data can be accessed for road and bridge management purposes. The concept is to harvest this data, from
regular vehicles using the network, and use it to infer information on the entire bridge stock.
Railway networks have particular potential for fleet monitoring as a relatively small number of trains make repeat runs over
the same bridges. An added advantage is that the trains are generally under one ownership, or a small number of ownerships.
Trains also generally come in a limited number of configurations, and generally return to the same stations where there is
potential to transfer data overnight. There are perhaps greater challenges for road bridges where the vehicles come with a great
variety of owners and operators, mostly from the private sector. To harvest such data would require regulation/legislation or an
opt-in process in return for some reward. For example, trucks could be allowed to carry greater load if they agreed to provide
access to on-board data.
This paper explores the concept of fleet monitoring for road and rail bridges. Inexpensive in-vehicle accelerometers are
assumed, combined with low-grade Global Positioning System data. Particular attention is given to the calibration challenge,
i.e., finding the dynamic properties of the vehicles, such as masses, centres of gravity and spring stiffnesses.

2. DRIVE-BY MONITORING
Drive-by or indirect monitoring of bridges was proposed by Yang et al in the 00’s [1,2] and validated by passing an instrumented
vehicle over a highway bridge in Taiwan [3]. The vehicle is considered as both exciter of the bridge and receiver of the resulting
vibration signal. The bridge vibrates as a result of the passing vehicle, in a way that reflects its dynamic properties and condition.
The bridge motions interact with those of the vehicle and its condition influences the vehicle vibrations. While small, the effect
has been shown to be measurable. An advantage is that the vehicle acts as a moving sensor, passing close to many potentially
damaged locations [4, 5].
2.1. Natural frequency, damping and mode shape
Malekjafarian et al [6], citing Yang et al [1], identify four key frequencies in any vehicle bridge crossing event: vehicle
frequency, moving vehicle pseudo-frequency and two shifted bridge frequencies – Fig. 1. Oshima et al [7] propose a specialist
excitation vehicle, incorporating an on-board oscillating mass. Kim et al [8] and Toshinami et al [9] validate the concept with
scaled laboratory tests though the vehicle mass was considerable, guaranteeing a high bridge excitation. Siringoringo and Fujino
[10] adopt a similar approach, this time validated with field experiments, and find the bridge fundamental frequency to an
accuracy of 11% for a vehicle travelling at 30 km/h. It should be noted that much greater levels of accuracy would be required
to detect bridge damage.

Figure 1. Acceleration spectrum of the vehicle (Finite Element model and Analytical); (1) vehicle frequency; (2) 2 v/L; (3)
b - v/L; (4) b + v/L; where v = speed, L = span length, b =bridge natural frequency. After [1].
Li et al [11] suggest that bridge frequency can be extracted with reasonable accuracy using a method of optimisation
(Generalised Pattern Search Algorithm) but do not consider a road profile. Malekjafarian and OBrien [12] apply frequency
domain decomposition to the power spectral density of the measured vehicle response in the presence of a road profile. In a
numerical study, they find bridge frequency, but only when the vehicle is moving slowly – 2 m/s. Malekjafarian et al [6]
identified a number of outstanding challenges in drive-by monitoring at that time: the influence of road profile and the variation
of the bridge frequency under a moving load [13, 14]. They recommend (i) low speed (below 40 km/h); (ii) multiple crossings
(at least 3) of the same vehicle; a heavy vehicle to excite the bridge; (iv) a small ratio of initial vehicle/bridge acceleration
amplitude and (v) prior calibration of the vehicle’s dynamic properties.
For concrete bridges in particular, the damping coefficient has been found to be damage sensitive [15, 16]. McGetrick et al [17]
investigate damping ratio using a drive-by approach. However, damping can be difficult to quantify in practice [18] and its
relationship with damage is uncertain. Mode shape and modal curvature have also been proposed for bridge health monitoring.
It is well established that there are discontinuities in modal curvature corresponding to local damage locations [19, 20]. Yang
et al [21] propose the Hilbert transform of the filtered response of the vehicle to estimate mode shape. Malekjafarian and OBrien
[22] propose a short time frequency domain decomposition to determine mode shapes. However, in both cases, very low speeds
were required to extract sufficient information for these approaches to be effective.
2.2. Damage Indicators
Apart from bridge dynamic properties, a number of authors have proposed other indices that they have correlated with bridge
damage. For example, McGetrick uses Moving Force Identification to determine the contact force history between vehicle and
bridge and suggests that any significant change in this signal is indicative of a change in bridge condition. Miyamoto and Yabe
[23, 24] install a sensor in a public bus and use data from multiple runs to monitor the condition of a bridge. They use a structural
anomaly parameter and a characteristic deflection as damage indicators and suggest that the concept is feasible for bridge health
monitoring provided the same bus is used for all measurements. Li and Au [25] propose modal strain energy and a Genetic
Algorithm optimisation to determine a damage indicator. Cerda et al [26, 27] use a support vector machine classifier to identify
damage in a laboratory scale model vehicle and bridge. Lederman et al [28] extend Cerda’s work, applying a regression to a
large dataset of damage locations and severities. Chen et al [29] also apply classification to the drive-by problem.
A number of authors use the Wavelet Transform to identify features in the vehicle vibration signal. Nguyen and Tran [30] apply
a Symlet wavelet to displacement responses which could presumably be inferred from the vibration signal. Khorram et al [31]
use a Gaussian 4 wavelet in a numerical study where the vehicle is represented by a simple moving force. McGetrick and Kim
[32, 33] apply a Continuous Wavelet Transform and show that damage manifests itself as peaks in the wavelet coefficients.
Malekjafarian et al [6] conclude that the main challenges for drive-by monitoring are (i) the excitation of the road profile which
tends to mask the influence of the bridge; (ii) the limited time of vehicle/bridge interaction which limits the quantity of relevant
data and (iii) environmental effects which can influence bridge properties. Fleet monitoring can address all three of these
challenges. The combination of data from many vehicle crossings overcomes the issue of a limited duration of data. While the
road profile remains present, the larger database improves the overall accuracy which allows the influence of the bridge to be
distinguished from the influence of the profile. Finally, the collection of data over an extended period of time reduces the
influence of time-dependent environmental effects as data will be combined from many different times of vehicle crossing.

3. INVERSE NEWMARK-BETA METHOD

An inverse Newmark-Beta method is proposed here to determine the road or rail profile using the vehicle response [34]. This
method is efficient and can calculate the profile with minimal computational effort. Here, the inverse Newmark-Beta method
is further developed using a simplified quarter-car model (Figure 2). The quarter car consists of a sprung mass, 𝑚𝑠 , and an
unsprung mass, 𝑚𝑢 , which represent the body and axle masses of the vehicle system respectively. A spring of stiffness 𝑘𝑠 and
a viscous damper of value 𝑐𝑠 connect the sprung mass and unsprung mass together. The axle mass connects to the road/rail
surface via a sprung of stiffness, 𝑘𝑡 .
The equations of motion of the vehicle can be defined as:
𝑀𝑢̈ + 𝐶𝑢̇ + 𝐾𝑢 = 𝑓𝑣 (1)
where 𝑀, 𝐶, and 𝐾 are the mass, damping and stiffness matrices of the vehicle respectively. The parameters, 𝑢̈ , 𝑢̇ and 𝑢 are
vehicle acceleration, velocity and displacement vectors respectively. The displacement vector of the vehicle is, 𝑢 = {𝑢𝑠 , 𝑢𝑢 }𝑇 .
The time-varying dynamic interaction force vector by 𝑓𝑣 where 𝑓𝑣 = {0, 𝐹𝑡 }𝑇 . The dynamic interaction force is, 𝐹𝑡 = 𝑘𝑡 × 𝑦 ;
where 𝑦 is the profile. The matrices are given by:
𝑚𝑠 0
𝑀=[ ] (2)
0 𝑚𝑢
𝑐𝑠 −𝑐𝑠
𝐶 = [−𝑐 𝑐𝑠 ] (3)
𝑠

𝑘𝑠 −𝑘𝑠
𝐾=[ ] (4)
−𝑘𝑠 𝑘𝑠 + 𝑘 𝑡

Figure 2. Quarter-car model and road model

The mass, damping and stiffness matrices of the vehicle 𝑀, 𝐶, and 𝐾 are assumed initially to be known. Also, the acceleration
of the sprung mass 𝑢̈ 𝑠 will be measured and therefore assumed to be known for each time step. Then, the displacement and
velocity of the sprung mass in each time step can be calculated using the Newmark Beta method:
𝑢𝑠,𝑡+𝛥𝑡 = (𝑢̈ 𝑠,𝑡+𝛥𝑡 + 𝑎2 × 𝑢̇ 𝑠,𝑡 + 𝑎3 × 𝑢̈ 𝑠,𝑡 )/𝑎0 +𝑢𝑠,𝑡 (5)
𝑢̇ 𝑠,𝑡+𝛥𝑡 = 𝑢̇ 𝑠,𝑡 + 𝑎7 × 𝑢̈ 𝑠,𝑡+𝛥𝑡 + 𝑎6 × 𝑢̈ 𝑠,𝑡 (6)
Taking Equations (3 - 6),
𝑚𝑠 × 𝑢̈ 𝑠,𝑡+𝛥𝑡 + 𝑐𝑠 × 𝑢̇ 𝑠,𝑡+𝛥𝑡 − 𝑐𝑠 × 𝑢̇ 𝑢,𝑡+𝛥𝑡 + 𝑘𝑠 × 𝑢𝑠,𝑡+𝛥𝑡 − 𝑘𝑠 × 𝑢𝑢,𝑡+𝛥𝑡 = 0 (7)
𝑚𝑢 × 𝑢̈ 𝑢,𝑡+𝛥𝑡 − 𝑐𝑠 × 𝑢̇ 𝑠,𝑡+𝛥𝑡 + 𝑐𝑠 × 𝑢̇ 𝑢,𝑡+𝛥𝑡 − 𝑘𝑠 × 𝑢𝑠,𝑡+𝛥𝑡 + (𝑘𝑠 + 𝑘𝑡 ) × 𝑢𝑢,𝑡+𝛥𝑡 = 𝑘𝑡 × 𝑦 (8)
In the Newmark Beta method, for unsprung mass:
𝑢̇ 𝑢,𝑡+𝛥𝑡 = 𝑎1 × (𝑢𝑢,𝑡+𝛥𝑡 − 𝑢𝑢,𝑡 ) − 𝑎4 × 𝑢̇ 𝑢,𝑡 − 𝑎5 × 𝑢̈ 𝑢,𝑡 (9)
Substituting Equation (9) into Equation (7) to replace 𝑢̇ 𝑢,𝑡+𝛥𝑡 :
𝑢𝑢,𝑡+𝛥𝑡 = (𝑚𝑠 × 𝑢̈ 𝑠,𝑡+𝛥𝑡 + 𝑐𝑠 × 𝑢̇ 𝑠,𝑡+𝛥𝑡 + 𝑘𝑠 × 𝑢𝑠,𝑡+𝛥𝑡 + 𝑐𝑠 × 𝑎4 × 𝑢̇ 𝑢,𝑡 + 𝑐𝑠 × 𝑎5 × 𝑢̈ 𝑢,𝑡 + 𝑐𝑠 × 𝑎1 × 𝑢𝑢,𝑡 )/(𝑐𝑠 × 𝑎1 + 𝑘𝑠 )
(10)
Hence, the displacement and velocity of the unsprung mass can be calculated as follows:
𝑢̇ 𝑢,𝑡+𝛥𝑡 = 𝑎1 × (𝑢𝑢,𝑡+𝛥𝑡 − 𝑢𝑢,𝑡 ) − 𝑎4 × 𝑢̇ 𝑢,𝑡 − 𝑎5 × 𝑢̈ 𝑢,𝑡 (11)
𝑢̈ 𝑢,𝑡+𝛥𝑡 = 𝑎0 × (𝑢𝑢,𝑡+𝛥𝑡 − 𝑢𝑢,𝑡 ) − 𝑎2 × 𝑢̇ 𝑢,𝑡 − 𝑎3 × 𝑢̈ 𝑢,𝑡 (12)
The effective stiffness matrix is:
̅] = [𝐾] + 𝑎0 × [𝑀] + 𝑎1 × [𝐶]
[𝐾 (13)
The effective force is:
𝐹̅𝑡+𝛥𝑡 = 𝐾
̅ × 𝑢𝑡+𝛥𝑡 (14)
𝐹𝑡+𝛥𝑡 = 𝐹̅𝑡+𝛥𝑡 − 𝑀 × (𝑎0 × 𝑢𝑡 + 𝑎2 × 𝑢̇ 𝑡 + 𝑎3 × 𝑢̈ 𝑡 ) − 𝐶 × (𝑎1 × 𝑢𝑡 + 𝑎4 × 𝑢̇ 𝑡 + 𝑎5 × 𝑢̈ 𝑡 ) (15)
Finally, the profile can be calculated as follows:
𝑦𝑡+𝛥𝑡 = 𝐹𝑡+𝛥𝑡 /𝑘𝑡 (16)
where, in the Newmark-Beta method, the integration constants are as listed here for time step, t:
1 𝛾 1 1 𝛾 𝛥𝑡 𝛾
𝛾 = 0.5, 𝛽 = 0.25 × (0.5 + 𝛾)2 , 𝑎0 = , 𝑎1 = , 𝑎2 = , 𝑎3 = − 1, 𝑎4 = − 1, 𝑎5 = × ( − 2) , 𝑎6 =
𝛽×𝛥𝑡 2 𝛽×𝛥𝑡 𝛽×𝛥𝑡 𝛽×2 𝛽 2 𝛽
(1 − 𝛾) × 𝛥𝑡, 𝑎7 = 𝛾 × 𝛥𝑡 (17)
To demonstrate the accuracy of the inverse Newmark-Beta approach using a quarter car, a ‘true’ road profile is used to generate
the accelerations. Then the generated accelerations are used to back-calculate the profile. A comparison of inferred results with
the ‘true’ profile show an almost perfect match – Figure 3. Clearly this is because the same vehicle/bridge interaction model is
used in the forward and inverse calculations and the vehicle properties are known exactly.
0.025
0.02
Profile (m)

0.015
0.01
0.005
0
-0.005 0 6 11 17 23 28 34 40 45 51 57 62 68 74 79 85 91 96
Position (m)
True profile Calculated profile

Figure 3. Calculated profile and true profile of quarter-car model

4. RAILWAY TRACK MONITORING

Railway bridges have particular potential for fleet monitoring due to the limited number of trains that traverse them and the
limited number of train owners/operators. A train carriage was instrumented for the purposes of drive-by monitoring in Ireland
in the 2010’s [35-40]. This was the carriage of a regular passenger train, with limited instrumentation, in contrast to heavily
instrumented and highly specialist Track Recording Vehicles (TRVs). The primary objective of this limited study was to
determine if the instrumented carriage on a regular passenger train could be used to determine track faults where ground
conditions have deteriorated. TRVs are commonly used by railway owners to monitor track but are expensive to operate and
there can often be delays between a speed restriction being imposed and the TRV visiting the site to determine track condition.
The sensors were installed in the trailer bogie of the first/last carriage (Figure 4) and data was collected over a period of a
month. The system can be calibrated to find carriage properties and used to determine track elevational profile. In simulations,
there is considerable variability in the inferred profile between runs but the main features of the track profile are clear – Figure
5.
 (a) Sensor installation locations on trailer bogie (dimensions in mm; drawing sourced (b) Bogie mounted gyrometer and
from Tokyu Car Corporation). accelerometer
Figure 4. Sensors installed in railway carriage

Figure 5. Reproducibility of inferred longitudinal profiles through case study area.

One part of the track passes over an area with peat soil where significant permanent deformation is known to exist – Figure
6(a). A survey shows significant settlement – Figure 6(b). An analysis of the acceleration data in this area indicated settlement
but the inferred magnitudes did not match the measured values – perhaps because the settlement increased under the weight of
the trains as they passed over.

Figure 6. a) Track settlement location (photograph: C. Bowe); b) Level survey of (unloaded) track longitudinal profile.
A preliminary investigation was also conducted into drive-by data as the railway carriage passed over a bridge [41]. In
simulations, the results looked promising – when some finite elements representing part of the bridge were removed to simulate
damage, the effect was evident in the measurements. However, this is much more difficult in practice with multi-axle vehicles,
especially for longer bridges where a combination of instrumented and non-instrumented carriages are present on the bridge at
any given time.

5. FLEET MONITORING OF BRIDGES

When on-bridge Quarter Car acceleration is used with the Inverse Newmark Beta method, the ‘apparent profile’ can be obtained.
Only Quarter Car vehicles are considered here – allowance for the interaction between axles is addressed elsewhere [42]. The
apparent profile (AP) is defined as the profile experienced by the vehicle when it passes over the bridge. This AP contains
profile elevations in the pavement/track and components of bridge deflection. As bridge deflection is affected by its health
condition (material stiffness, support conditions), it is possible to use AP to detect bridge damage. Here, a simple way is
presented of using the AP derived from a fleet of vehicles to detect bridge damage.
The relation between apparent profile and bridge profile elevations can be written as,
𝑝
𝑅𝑘𝑎 = 𝑅𝑘 + 𝛿𝑘 (18)

𝑝
where 𝑅𝑘𝑎 denotes the apparent profile elevation at location/scan k, 𝑅𝑘 denotes bridge surface profile elevation and 𝛿𝑘 denotes
the bridge deflection at the time and location corresponding to scan k. This latter is a moving reference deflection because its
location changes as the vehicle passes over the bridge. To describe the relation between moving reference deflection and vehicle
weights, a moving reference influence line (MRIL), Ik, is introduced. It is defined as the deflection response at a moving point,
k, due to a unit load at that point. Hence, the moving reference deflection at point k, due to vehicle n, is,

𝛿𝑛,𝑘 = 𝑊𝑛 𝐼𝑘 (19)

where Wn is the (Quarter Car) vehicle weight. To find the coordinates of the MRIL, an error minimisation process is used. This
process is similar to that used in Bridge Weigh-in-Motion theory to calculate a (fixed reference) influence line from direct
measurement [43]. The error function is defined as the sum of squared differences between measured deflections (AP calculated
from vehicle acceleration) and their theoretical equivalents (Eq.18),

𝑁 𝐾
𝑝 2
𝐸 = ∑ ∑{𝑅̃𝑛,𝑘
𝑎
− 𝑅𝑘 − 𝛿𝑛,𝑘 } (20)
𝑛=1 𝑘=1

where 𝑅̃𝑛,𝑘
𝑎
denotes the AP elevation ‘measured’ (indirectly) through accelerations, N is the number of vehicles in the fleet and
K is the number of scans. Substituting Eq.19 into Eq. 20 gives,

𝑁 𝐾
𝑝 2
𝐸 = ∑ ∑{𝑅̃𝑛,𝑘
𝑎
− 𝑅𝑘 − 𝑊𝑛 𝐼𝑘 )} (21)
𝑛=1 𝑘=1

To minimize the error function, partial derivatives are taken with respect to each MRIL coordinate:

𝜕𝐸
=0 (22)
𝜕𝐼𝑘

giving,
𝑁 𝑁
𝑝
(23)
𝐼𝑘 ∑ 𝑊𝑛 = ∑(𝑅̃𝑛,𝑘
𝑎
− 𝑅𝑘 )
𝑛=1 𝑛=1

Hence, the coordinates of the MRIL can be found by,

 ̃𝑎 𝑝
∑𝑁
𝑛=1(𝑅𝑛,𝑘 − 𝑅𝑘 )
𝐼𝑘 = (24)
∑𝑁𝑛=1 𝑊𝑛

𝑝
In Eq.24, the profile heights, 𝑅𝑘 are required. To simplify the calculation process, a ‘smooth profile’ will be assumed, which
𝑝
results in all profile heights (𝑅𝑘 ) being set equal to zero. Numerical simulation is used in this paper to validate the concept of
using MRIL to detect bridge damage. The beam model and vehicle fleet properties are similar to those used in the literature
[44]. Table 1 shows vehicle fleet properties. Bridge bearing damage is considered similar to Khan et al.’s work [45]. Vehicle
accelerations are sampled from a fleet of 200 vehicles (N=200) at a scan rate of 1000 Hz. The resulting MRILs are calculated
and plotted in Figure 7.

Table 1. Vehicle fleet properties

Property Symbol/Units Mean value Standard deviation
Sprung mass Ms: kg 9250 1850
Unsprung mass Mu: kg 750 150
Damping Cs: Ns/m 1×104 1×103
Unsprung stiffness Ks: N/m 7.5×105 7.5×104
Tyre stiffness Kt: N/m 3.5×106 3.5×105
Vehicle speed m/s 25 5

Figure 7. Inferred MRIL for healthy and two damage conditions.

Bearing damage is represented in the bridge model by inserting a rotational spring at the support. The stiffness of the additional
spring is 109 Nm, deemed to be a ‘low’ damage level in the literature [44]. In Figure 7, it is clear that bearing damage
significantly affects both the shape and area under MRIL functions. When the left bearing is damaged, MRIL trends to skew
to the right side and vice versa. MRIL can be seen to be a good damage indicator for bridge bearing damage.

6. CONCLUSIONS
This paper shows that an inverse version of the Newmark-Beta method of solving vehicle/bridge interaction equations can be
used to determine the road/rail profile associated with measured accelerations in a moving vehicle. For a road or railway, the
profile under a moving load is a valuable measure of condition and was found to be effective in detecting an area of poor
foundations on a railway track. For bridges, a Moving Reference deflection Influence Line can be found from measurements
taken in passing quarter cars. The moving reference influence line is shown to be damage sensitive, specifically for bearing
damage in a simply supported bridge. Ongoing work is addressing the issue of multiple axles in a fleet of instrumented vehicles.

ACKNOWLEDGMENTS
Mr. Yifei Ren and Mr. Shuo Wang acknowledge the Ph.D. scholarship awarded jointly by University College Dublin and the
China Scholarship Council.
REFERENCES
[1] Yang YB, Lin CW, Yau JD. Extracting bridge frequencies from the dynamic response of a passing vehicle. Journal of
Sound and Vibration. 2004;272:471-93.
[2] Yang YB, Lin CW. Vehicle-bridge interaction dynamics and potential applications. Journal of Sound and Vibration.
2005;284:205-26.
[3] Lin CW, Yang YB. Use of a passing vehicle to scan the fundamental bridge frequencies: An experimental verification.
Engineering Structures. 2005;27:1865-78.
[4] Gonzalez A, OBrien EJ, McGetrick PJ. Identification of damping in a bridge using a moving instrumented vehicle. Journal
of Sound and Vibration. 2012;331:4115-31.
[5] Keenahan J, OBrien EJ, McGetrick PJ, Gonzalez A. The use of a dynamic truck-trailer drive-by system to monitor bridge
damping. Struct Health Monit. 2014;13:143-57.
[6] Abdollah Malekjafarian, Patrick J. McGetrick, Eugene J. OBrien, A Review of Indirect Bridge Monitoring Using Passing
Vehicles, Shock and Vibration, 2015;2015:1-16.
[7] Oshima Y, Yamamoto K, Sugiura K, Yamaguchi T. Estimation of bridge eigenfrequencies based on vehicle responses
using ICA. Proceedings of the 10th International Conference on Structural Safety and Reliability, ICOSSAR2009. Osaka,
Japan13-17 September 2009.
[8] Kim CW, Isemoto R, Toshinami T, Kawatani M, McGetrick P, OBrien EJ. Experimental Investigation of Drive-by Bridge
Inspection. 5th International Conference on Structural Health Monitoring of Intelligent Infrastructure (SHMII-5). Cancun,
Mexico 2011.
[9] Toshinami T, Kawatani M, Kim CW. Feasibility investigation for identifying bridge’s fundamental frequencies from
vehicle vibrations. Fifth International Conference on Bridge Maintenance, Safety and Management, IABMAS2010.
Philadelphia, USA2010. p. 317-22.
[10] Siringoringo DM, Fujino Y. Estimating Bridge Fundamental Frequency from Vibration Response of Instrumented Passing
Vehicle: Analytical and Experimental Study. Adv Struct Eng. 2012;15:417-33.
[11] Li WM, Jiang ZH, Wang TL, Zhu HP. Optimization method based on Generalized Pattern Search Algorithm to identify
bridge parameters indirectly by a passing vehicle. Journal of Sound and Vibration. 2014;333:364-80.
[12] Malekjafarian A, OBrien EJ. Application of output-only modal method to the monitoring of bridges using an instrumented
vehicle. In: Nanukuttan S, Goggins J, editors. Civil Engineering Research in Ireland. Belfast, Northern Ireland 2014.
[13] Yang YB, Cheng MC, Chang KC. Frequency Variation in Vehicle-Bridge Interaction Systems. Int J Struct Stab Dy.
2013;13.
[14] Chang KC, Kim CW. Variability in bridge frequency induced by a parked vehicle. Proc of the twenty-fourth KKCNN
symposium on Civil Engineering. Hyogo, Japan2011. p. 75-9.
[15] Curadelli RO, Riera JD, Ambrosini D, Amani MG. Damage detection by means of structural damping identification.
Engineering Structures. 2008;30:3497-504.
[16] Modena C, Sonda D, Zonta D. Damage localization in reinforced concrete structures by using damping measurements.
Key Eng Mat. 1999;167-1:132-41.
[17] McGetrick PJ, Gonzalez A, OBrien EJ. Theoretical investigation of the use of a moving vehicle to identify bridge dynamic
parameters. Insight. 2009;51:433-8.
[18] Williams C, Salawu O. Damping as a damage indication parameter. Proceedings of the 15th international modal analysis
conference. Orlando, FL, USA: The Society for Experimental Mechanics; 1997. p. 1531–6.
[19] Zhu XQ, Law SS. Wavelet-based crack identification of bridge beam from operational deflection time history. Int J Solids
Struct. 2006;43:2299-317.
[20] Pandey AK, Biswas M, Samman MM. Damage Detection from Changes in Curvature Mode Shapes. Journal of Sound and
Vibration. 1991;145:321-32.
[21] Yang Y. B., Li Y. C., and Chang K. C., Constructing the mode shapes of a bridge from a passing vehicle: a theoretical
study, Smart Structures and Systems, 2014;13:797–819.
[22] Malekjafarian A, OBrien EJ. Identification of bridge mode shapes using Short Time Frequency Domain Decomposition of
the responses measured in a passing vehicle. Engineering Structures. 2014;81:386-97.
[23] Miyamoto A, Yabe A. Development of practical health monitoring system for short- and medium-span bridges based on
vibration responses of city bus. Journal of Civil Structural Health Monitoring. 2012;2:47-63.
[24] Yabe A, Miyamoto A, Isoda S, Tani N. Development of Bridge Monitoring System for Short- and Medium-Span Bridges
Based on Bus Vibration. Journal of Japan Society of Civil Engineers, Ser F4 (Construction and Management).
2013;69:102-20.
[25] Li ZH, Au FTK. Damage Detection of a Continuous Bridge from Response of a Moving Vehicle. Shock and Vibration.
2014;Vol. 2014.
[26] Cerda F, Garrett J, Bielak J, Barrera J, Zhuang Z, Chen S et al. Indirect structural health monitoring in bridges: scale
experiments. Bridge Maintenance, Safety, Management, Resilience and Sustainability. 2012:346-53.
[27] Cerda F, Chen S, Bielak J, Garrett JH, Rizzo P, Kovačević J. Indirect structural health monitoring of a simplified
laboratory-scale bridge model. Smart Structures and Systems. 2014;13.
[28] Lederman G, Wang Z, Bielak J, Noh H, Garrett JH, Chen S et al. Damage quantification and localization algorithms for
indirect SHM of bridges. Bridge Maintenance, Safety, Management and Life Extension. 2014:640-7.
[29] Chen SH, Cerda F, Rizzo P, Bielak J, Garrett JH, Kovacevic J. Semi-Supervised Multiresolution Classification Using
Adaptive Graph Filtering with Application to Indirect Bridge Structural Health Monitoring. Ieee T Signal Proces.
2014;62:2879-93.
[30] Nguyen KV, Tran HT. Multi-cracks detection of a beam-like structure based on the on-vehicle vibration signal and wavelet
analysis. Journal of Sound and Vibration. 2010;329:4455-65.
[31] Khorram A., Bakhtiari-Nejad F. and Rezaeian M. Comparison studies between two wavelet based crack detection methods
of a beam subjected to a moving load. International Journal of Engineering Science, 2012; Vol. 51: 204–215.
[32] McGetrick PJ, Kim CW. A Parametric Study of a Drive by Bridge Inspection System Based on the Morlet Wavelet. Key
Eng Mater. 2013;569-570:262-9.
[33] McGetrick PJ, Kim CW. An indirect bridge inspection method incorporating a wavelet-based damage indicator and pattern
recognition. 9th International Conference on Structural Dynamics, EURODYN 2014. Porto, Portugal 2014.
[34] Keenahan, J., Y.F. Ren, and E.J. OBrien, Determination of road profile using multiple passing vehicle measurements.
Structure and Infrastructure Engineering, 2020;16(9):1262-1275.
[35] OBrien, E.J., Bowe, C., Quirke, P., Cantero, D. Determination of longitudinal profile of railway track using vehicle-based
inertial readings, Journal of Rail and Rapid Transit. 2017;231(5):518-534.
[36] Quirke, P., Cantero, D., OBrien, E.J., Bowe, C. Drive-by detection of railway track stiffness variation using in-service
vehicles, Journal of Rail & Rapid Transit, 2017;231(4):498-514.
[37] OBrien, E.J., Quirke, P., Bowe, C., Cantero, D. Determination of railway track longitudinal profile using measured inertial
response of an in-service railway vehicle. Structural Health Monitoring. 2018;17(6):1425-1440.
[38] Malekjafarian, A., OBrien, E., Quirke, P., Bowe, C. Railway track monitoring using train measurements: an experimental
case study. Applied Sciences, 2019;9:4859.
[39] Malekjafarian, A., OBrien, E.J., Quirke, P., Bowe, C. Railway track monitoring using train measurements. Canadian
Journal of Civil Engineering. 2021.
[40] Malekjafarian, A., OBrien, E.J., Quirke, P., Cantero, D., Golpayegani, F. Railway Track Loss-of-Stiffness Detection Using
Bogie Filtered Displacement Data Measured on a Passing Train. Infrastructures, 2021;6:93.
[41] Quirke, P., Bowe, C., OBrien, E.J., Cantero, D., Antolin, P., Goicolea, J.M. Railway bridge damage detection using
vehicle-based inertial measurements and apparent profile. Engineering Structures, 2017;153: 421-442.
[42] OBrien, E.J., McCrum, D.P., Wang S. Fleet monitoring of bridges using a moving reference influence function, under
review.
[43] OBrien, E.J., Quilligan, M.J. and Karoumi, R. March. Calculating an influence line from direct measurements.
In Proceedings of the Institution of Civil Engineers-Bridge Engineering, 2006;159:31-34.
[44] OBrien, E., Khan, M.A., McCrum, D.P. and Žnidarič, A., 2020. Using statistical analysis of an acceleration-based bridge
weigh-in-motion system for damage detection. Applied Sciences, 2020;10:663.
[45] Khan, M.A., McCrum, D.P., OBrien, E.J., Bowe, C., Hester, D., McGetrick, P.J., O’Higgins, C., Casero, M. and Pakrashi,
V. Re-deployable sensors for modal estimates of bridges and detection of damage-induced changes in boundary
conditions. Structure and Infrastructure Engineering, 2021;2021:1-15.