Overview of the initial testing of Sonic Analysis of INfrastructure (SAIN) out infrastructure monitoring device.

1. Bridge vibrational response to a passing train
is a complicated physical process involving
energy transfer through the wheels, rail, ballast
and supporting substructure. Changes in
environmental conditions, such as amount of
rainfall and ambient temperatures can impact
the vibrational response of the structure. In this
paper there is no attempt to quantify or model
these individual processes, but rather take the
train, track, bridge as a ‘whole-body’ from which
the acoustic signatures are built. This approach
has the advantage of subsuming the smaller
environmental fluctuations.
These ‘whole-body’ signatures have been
developed for the single span bridge at Newton-
le-Willows on the UK West Coast mainline in
response to passing Virgin Pendolino trains. This
paper contains the details of how these signatures
have been developed using a simple acoustic
monitoring device and application of advanced
spectral analysis techniques. An evaluation is then
presented of how these signatures can be used to
provide long-term monitoring of both bridge and
train wheel-set health.
SAIN - SONIC
ANALYSIS OF
INFRASTRUCTURE
A R N M O R E
Limited
In the UK alone there are over 28,000 rail
bridges. Many of which have been in operation
since the Victorian era. Increasing traffic
and loads across these structures means
monitoring their condition is becoming more
important for safe operation of the railway
network.
07540 706994 mark.tovey@arnmore.com www.arnmore.com
There has been numerous studies undertaken on
remote condition monitoring of bridges. These
have, in the main, focussed on large structures
and have usually involved detailed instrumentation
and subsequent analysis of bridge dynamics.
Undertaking a network wide monitoring campaign
using this level of complexity would be extremely
costly. However, many bridges are simpler in design,
such as single and double span constructions. This
paper contains a study into an easy to implement
approach for monitoring these less complex
structures. The method uses the application of
acoustic spectral decomposition to produce ‘whole-
body’ signatures that can be used for measuring the
long-term health of these structures. Measurements
of the acoustic response of a single span bridge
on the UK West Coast mainline rail route have been
carried out to demonstrate how these signatures
can be developed.
Thousand railway
bridges in Europe
Year on year growth in
UK rail traffic alone
Bridge failures per year
estimated in USA
Bridges at risk of
failure in Italy 2019
>100 years
bridges in UK
Major bridge failures
worldwide in 2018Million+ per year spent
maintaining bridges in USA
300 10$200
3% 200
300 30%

2. 6. The equivalence of the substructure around both
the large and small bridges. A stretch of track
needed to be selected which contained two single
span bridges with differing size and residing
on, as close as possible, the same supporting
environment.
7. The construction of both bridges. Both bridges
were brick built single span bridges.
The location chosen was a stretch of track near
Newton-le-Willows on the UK West Coast mainline
route. This location has two single span bridges
within close proximity on a straight section of track.
The frequency of Pendolino train services on the
route provided a relatively high number of similar
trains for undertaking the study.
Acoustic recordings were taken using a VIDEOMIC
ME microphone and RODE muffler attachment to
Samsung mobile phones. Bridge recordings were
taken at a distance of 10m from each bridge as
close to the centreline of the bridge as possible.
To provide a measure of the rail acoustics without
the bridge readings were taken close to the rail line
around 100m from the bridges.
The recordings were then analysed using the
Audacity software package which contains a
number of functions including Fast Fourier
Transform (FFT) and Spectrograph applications.
FFT’s were performed on complete (end-to-end)
train recordings to develop the ‘whole-body’
signatures. Spectrographs were used to determine
train speeds and to count the number of axles for
train identification purposes.
CONCLUSIONS CONTRIBUTIONS
Although here have been many studies performed
on bridge vibrations, these have primarily focussed
on understanding the dynamics and condition of
larger bridges. The analysis presented in this paper
has been aimed at providing a simpler method for
monitoring smaller structures. The results of ‘whole-
body’ signatures for train-bridge interactions have
been presented to demonstrate how they can be
used for long-term evaluation of bridge and wheel-rail
condition. These signatures are based on acoustic
recording of trains passing over the bridge structure.
There are a number of advantages to undertaking
readings in this manner. One of the most significant
being that performing the recordings can be carried
out without the need for access to the tack. A study
of the Newton-le-Willows single span under bridge
on the UK West Coast mainline has been used to
demonstrate how this type of signature monitoring
can be applied to these types of structure.
The signature for the bridge has a broad set of
resonant frequencies up to 200 Hz which has
a defined shape around a peak at 100Hz. This
signature is dependent upon a number of factors
METHODS
A number of factors were taken into consideration
to when selecting a bridge to study. These included;
1. The number of trains passing over the bridge.
Which needed to be high enough to enable
enough recordings to be carried out within a
reasonable period of time.
2. The make up of the trains crossing over the
bridge. In this initial study it was essential to
limit any potential variability of acoustic signal
associated with different rolling stock.
3. The constancy of the train speeds across the
bridge. Again to limit the variability in the signal.
4. Ease of access to the bridge for making the
measurements. Given the number of recordings
that needed to be made it was essential to have
easy access to the road under the bridge from
which the recordings could be made.
5. No requirement for accessing the railway.
Measurements were taken at varying times of day
and year. It was necessary therefore not to have
to officially request access to the railway to make
the measurements.
the two most significant being the speed of the
passing train and the mass of the bridge. In this
study train speeds are calculated from the acoustic
recordings using the distance between the Pendolino
bogies and the spectrograph timing between bogies.
Speeds have been seen to be around 40 to 50 mph
over this stretch of track meaning the effect of train
speed variation on the signature has been minimal.
The mass dependency, which is key to the condition
monitoring approach presented here, has been
illustrated by comparing the small bridge signature
with that of a slightly bigger single span bridge
on the same section of track. The variation in the
signatures between the two bridges has been used
to illustrate how a change in a bridge’s structural
nature results in a change to the signature. This
paper has illustrated how long term monitoring of the
bridge signature using recordings after each passing
train can be used as a measure of changes to the
condition of the bridge should the signature vary
outside of a predetermined shape envelope.
The study also demonstrated how wheel-rail
signatures for the train bogies can also be monitored
in a similar way to check for wheel problems on
trains passing over the bridge.
INTRODUCTION
28,000Bridges on the UK rail network, many of which
were built in the Victorian era
The deterioration of railway bridges impacts
upon both rail safety and performance. In the UK
alone there are over 28,000 bridges on the rail
network, many of which were built in the Victorian
era. Increasing traffic volumes and weight are
accelerating the wear and tear on these structures.
Monitoring bridge health is therefore critical to
operation of the railway. Current techniques for
estimating bridge health tend to involve complex
instrumentation of the bridge for remote monitoring.
Instrumenting, recording and analysing all of the
28,000 bridges in this manner would be impractical.
However, smaller structures still need to be
monitored, ideally this would be by employing a
simpler analysis method. At present, small brick
built bridges are, generally, simply subjected to
regular visual inspections. Visual inspection of
the exterior of a bridge to estimate heath of the
structure is notoriously unreliable. The health of the
internal structural composition can therefore only be
made through more intrusive inspection techniques.
The pilot study described in this paper is to
provide a supplementary, real-time, assessment
of the bridge health through a simple continually
monitored ‘whole-body’ spectral signature.
This paper contains the results of a project to
demonstrate how these ‘whole-body’ signatures can
be built and how they can be used to monitor both
bridge and wheel-rail health.
The study was to develop a signature for a small
brick built bridge on the UK West Coast mainline
route from London to Scotland. Measuring the
acoustic spectral response of the bridge from
passing Virgin Pendolino services. These recordings
provided both bridge ‘whole-body’ acoustic
signatures as well as acoustic recordings for
each of the passing train wheel sets. Two sets of
readings have been made, one on the centre line of
the bridge and a second set to measure the acoustic
signature generated on the plain line around 100m
from the bridge. The first set therefore contains the
acoustic effects of the bridge while the second is
an estimate of the spectral background without the
bridge structure. The difference between these two
sets of recordings is used to develop the ‘whole-
body’ signature for the bridge structure.
A third set of recordings have been made at a
larger bridge located close by. These recordings
demonstrate how the signature changes as the
mass of the structure changes. Thus illustrating
how the signatures can be used to monitor for mass
and or structural changes in the bridge under study.

3. RESULTS
The results of the recordings for the bridges can
summarised as follows;
• Bridges of the nature analysed here were
estimated to provide vibrational signatures with
frequencies up to 200Hz.
• The location of the bridge under study was
important for this pilot study. The comparison
between the bridge signatures required the
geography and environment for both bridges to be
as similar as possible to remove as much of the
variability due to these factors as possible. The
site chosen contained two similar brick construct
bridges within 200m of each other.
• Both bridges also needed to be on a similar
stretch of track i.e. the train movement over the
bridges being as identical in speed and direction
as possible. The bridges analysed in this study
were located on the same stretch of straight track
reducing the variability from differences in train-
bridge interactions.
• Only recordings from Virgin Pendolino trains were
used in this study to eliminate variability between
types of train.
• Background train recordings were made on
the straight section of track and were used as
a measure of the ‘background’ signature for a
Pendolino train. Only trains travelling between
40mph and 50mph were used to develop both
bridge and background signatures.
These results show that bridge ‘whole-body’
signatures can be developed using simple acoustic
recording devices. This is a critical factor for
network-wide monitoring of structures of this
type. Signature accuracy can be increased with
increased number of recordings taken. The two
bridges considered in this pilot study were both
in good health. The results presented here show
how a healthy bridge responds to the impact of
trains moving over it. The variability in the signature
between two bridges of slightly different mass
can be measured using this simple assessment
technique. Monitoring a bridge signature over long
periods of time can therefore be used as an indicator
of changes in ‘mass’ structure. Thus providing
not only the potential for a real time monitoring
system but also aiding the maintenance inspections
associated with these smaller structures.
SMALL BRIDGE FINDINGS
• Recordings for the small bridge was made
using a simple acoustic monitor placed 10m
from the bridge.
• The ambient background signal was then
subtracted from these recordings. Leaving a
measure of the small ‘whole-body’ signature.
• For the small bridge the signature showed a
definite peak in the frequency profile around
100hz.
LARGE BRIDGE FINDINGS
• A similar set of readings were taken for the
large bridge.
• The signature for the large bridge was shown
to peak below 100hz
PENDOLINO SIGNATURE
• In addition to the bridge signatures the
spectrograph also picked up the wheel-rail
signatures for each of the Pendolino wheel
sets.
• Clear frequency peaks are shown in the
spectral analysis. These have not been
analysed in detail here but could be used to
check individual train wheel health.
• These signatures were used to ensure
measure train speeds to ensure the
recordings met the 0-50mph speed criteria.
07540 706994 mark.tovey@arnmore.com www.arnmore.com