International Journal of Control and Automation

Vol.7, No.1 (2014), pp.401-408

http://dx.doi.org/10.14257/ijca.2014.7.1.36

Research on the Railroad Bridge Monitoring Platform Based on the

Internet of Things

Ying Sun1
1
Department of Electronic and Information Engineering,
LiaoningRailway Vocational and Technical College, 121000
[email protected]

L.
Abstract

A
As railroad bridge is one of the most important infrastructures of railway, the safety of

EG
bridge has a direct effect on railway operation efficiency and safety. In this paper, we

s I ly.
propose a distributed railway bridge monitoring platform based on the Internet of Things
(IOT) to monitoring the state of bridge safety. We made detailed research on the sensor nodes

ei n
LL
deployment, sensing data updating and early warning mechanisms of the wireless sensor

fil O
network. The proposed system achieved the desired purpose of real-time monitoring of
is on
railroad bridge safety during test in China Railway Corp.
th si
Keywords: railroad bridge safety; Internet of Things; Wireless sensor network; monitoring
y er

platform
eb eV

1. Introduction
As an important infrastructure of the country, as the artery of national economy and as a
ad in

popular traffic tool, railway plays an important role in modern logistics system. As one of the
important infrastructure of railway, railroad bridge directly affect the safety of railway traffic
m nl

status. So the research on the safety of railroad bridges has important practical significance
O

[1]. Railway bridge has the characteristics of large load, large impact force, traffic density,
high requirements of the standard for resistance to natural disaster. Therefore, periodic rapid
detection of infrastructure is a necessary measure to master the railway bridge safety
condition and ensure the safety.
At present, there are many long-span bridges have deployed large health monitoring
systems, mainly for real-time monitoring of vibration response, strain and deformation of
ok

structural mechanics, thus to evaluate the operation of the safety status of the bridge. There
are a lot of traditional methods for bridge monitoring such as the low strain reflected wave
Bo

method, the sound wave transmission method, the core method and so on. The existing
methods of bridge safety monitoring has many limitations and inconvenience such as the
incomplete judgment of multiple defects, the inaccurate judgment of the nature of the defect,
the long time and high cost of monitoring, which has a risk of false positives or false
negatives .
Compared with the traditional measurement methods, satellite positioning of the bridge
monitoring method can not only achieve the 3D high precision measurement of a day of
uninterrupted, but also has the characteristics of large range, short detention, strong real-time
and synchronous measurement. However, the above researches and tests are all for the road
bridges and none for railway bridges [2]. With the development of China's high-speed

ISSN: 2005-4297 IJCA

Copyright ⓒ 2014 SERSC
International Journal of Control and Automation
Vol.7, No.1 (2014)

railway, higher requirements are put forward on security of railway bridges, so the study of
railway bridge monitoring is imperative.
The emerging technology of wireless sensor network (WSN), which consists of a large
number of sensing nodes each capable of sensing, processing and transmitting environmental
information, has the advantages of sensor nodes highly integrated, low cost, low power
consumption, network flexibility and so on. As a new technology of the IOT, WSN has broad
application prospects in environmental monitoring. Using wireless sensor network for
monitoring of railway bridges can help to reduce costs, increase flexibility and intelligent.
This paper proposed a railroad bridge monitoring platform based on the Internet of Things to
monitoring the operation state of the bridge, and we performed detailed study of sensor
deployment, upload and and warning mechanism thus to ensure design rationality.
The structure of the paper is as follows: in this section, we give the background of the

L.
Security research status of railway bridges and review of our proposed monitoring platform.

A
Section 2 describes the system framework. The main discussing of this paper is described in
section 3-5, which includes the deployment of sensor nodes, data collection and uploading,

EG
s I ly.
and early warning data processing. Section 6 provides a summary and some future work of
the paper.

ei n
LL
fil O
2. System Framework
The railroad monitoring platform system designed in this paper can be divided into
is on
two parts, namely are outfield and monitoring center. The outfield refers to railroad
bridges and railway line where we deployed sensors to collect data, targeted and rational
th si

sensor deployment strategy are used to collect sensing data of different types of sensors ;
y er

we use different ways of data uploading for emergency data and daily data. The
monitoring center consists of the monitoring sever and database, which is used to
eb eV

perform centralized data processing thus to achieve real-time monitoring of the railway
bridge safety sate.
ad in

The proposed monitoring system uses a hybrid deployment of multimedia sensor nodes
and ordinary sensor nodes. The specific deploy method will be introduced in the Section 3. It
m nl

can communicate with each other between nodes of one type or the two nodes of different
types. It use Ad Hoc network to build a network, multimedia nodes are used to collect image
O

information such as bridge cracks, and ordinary nodes are used to sense data such as pressure
of bridge deck, soil temperature, and piers displacement information. After data compression,
encryption and fusion， the data are sent to the sink node through multi-hop communication,
and finally sent to the monitoring host via train or internet depending on the urgency of
different data. The specific transmission and upload process will be described in Section 4.
ok

The monitoring center is to receiving, storage, analysis, process and display the collected
data, users can perform monitoring, querying and processing the entire bridge information by
Bo

handling the monitor host. The monitoring platform also has early warning mechanism and
security level assessment mechanism to ensure the safe use of the bridge. The specific early
warning mechanism will be introduced in the Section 5. System architecture of the entire
railroad monitoring platform is shown in Figure 1.

402 Copyright ⓒ 2014 SERSC

 International Journal of Control and Automation
Vol.7, No.1 (2014)

:Ordinary node

:Multimedia node

Monitoring center :Sink node

Internet
Dail
y da
te

Sink node
te
y da
train genc
Emer Railroad Bridge

L.
A
EG
Figure 1. Framework of monitoring system

s I ly.
3. The Deployment of the Wireless Sensor Network

ei n
LL
fil O
As the basis of the Railway bridge monitoring platform, wireless sensor network provide
bridge monitoring information collection, pre-processing and transmission for the Railway
is on
bridge monitoring platform. Whether the topology of the wireless sensor network is
reasonable or not will make a great impact on the performance of the whole network. In this
th si
paper, the sensor deployment of the wireless sensor network is divided into two parts: the
sensor node deployed on the bridge and the sensor node deployed around railway line.
y er
eb eV

3.1. The sensor node deployed on the bridge

The main task of the sensor nodes deployed on the bridge is to monitor the security state of
the bridge. The sensor nodes include pressure sensors, humidity sensors, temperature sensors,
ad in

deformation sensors and even wireless multimedia sensors. They can give a multi-level and
m nl

comprehensive monitor of the bridge.

Pressure sensors are mainly deployed on the pier and the underside of the bridge deck
O

bracket, especially the major weight-bearing position. They are used to real-time monitor
whether the pressure has enormous changes. The sudden change of the pressure implies that
the bridge structure is undergoing a certain deformation. Usually, the position where the
pressure becomes larger is more prone to damage.
Humidity sensors are mainly deployed on the bridge near the water or the culvert. When
the humidity is too high, the bridge surface and the internal steel stent are prone to corrosion,
ok

results in a great damage to the bridge. Railway safety department should monitor the
humidity changes of the key position of the bridge timely. And then give a maintenance and
Bo

drainage timely.
The temperature of the soil surrounding the bridge has a great impact of the bridge
security, especially when the temperature is too low, this easily leads to the change of the soil
and results in the change of the whole position of the bridge. In extreme cases, the entire
bridge will has a deviation. Hence, temperature sensors are mainly deployed on the lower
bridge and the inside of the surrounding soil to monitor the change of the soil in extreme
weather.
Deformation sensors are deployed on the road surface of the bridge where is more likely to
have a greater deformation. The deformation of the bridge road surface usually implies the
change of the bridge support structure. Railway safety department should combine with

Copyright ⓒ 2014 SERSC 403

International Journal of Control and Automation
Vol.7, No.1 (2014)

pressure sensors and other sensors to give an overall judgment of the security state of the
bridge.
Bridge bearing is an important part of the bridge structure, to which the shift can easily
occur by the influence of concrete shrinkage, temperature changes, and the traffic load. The
displacement sensor can implement the real-time detection of bridge bearing shift, which
should be combined with pressure sensors and other sensors' data for the integrity judgment
of bridge safety.
Wireless multimedia sensor is a new type of sensor which has the function of collecting
and processing multimedia information such as audio and image information. In the bridge
monitor network, we mainly use the wireless image sensor. Image sensor is a positive
complement for other types of sensors. It plays an supervisory role on malicious behavior
against the bridge safety.

L.
A
3.2. The sensor node deployed around railway line

EG
According to the features of the railway, [3] designed node deployment scenarios and it

s I ly.
achieved the goals of energy saving and extending the network lifetime. In this paper, the
main task of the sensor node deployed around railway lines is to deliver the data monitored by

ei n
LL
the bridge sensors safely and quickly to the nearest monitor center so that the monitor center

fil O
can have a real-time control of the bridge security. Because of the simplicity of its
application, this paper intends to use the homogeneous communication node. There is no
is on
difference between sensors on the communication range, energy and other aspects. Besides,
considering the communication tasks of the sensors deployed on every position are the same,
th si

this paper deploys the sensor symmetrical on the both side of the railway line, for the purpose
y er

of load balancing and network life extending.

eb eV

4. Information Collection based Wireless Sensor Nodes

Our monitoring platform system of railway bridges based Internet of Things is divided into
ad in

two parts: outfield node information acquisition and monitoring center. Information collecting
is realized by deploying corresponding sensors on the bridge or around the railway line
m nl

depending on the monitoring requirements. Sensor bridge monitoring platform mainly detects
O

the two diseases about railway bridges: bridge support displacement detection and bridge
crack detection. Bridge support displacement detection is achieved by displacement sensors
[4], and bridge crack detection is achieved by multimedia sensor nodes which can collect,
process, transfer crack picture and other multimedia information.
The original monitoring data real-time collected by sensor node can be divided by the local
threshold determination to two types: the emergency data and daily data. The emergency data
ok

is the bridge real-time monitoring data that indicates a serious threat to the bridge’s safety, for
example that the displacement measured by the crack displacement sensors or the crack size
Bo

measured by multimedia sensor is over a pre-set threshold, which is the emergency data; daily
data is the data which do not indicate a threat to the security of the bridge. The transmission is
different for urgent data and daily data.

4.1. Emergency data transmission and uploading

Because the emergency data indicates a serious threat to the bridge’s safety, so it is
necessary to send back the data to the monitoring center for threat verdiction. External sensor
monitoring network formed by the sensor nodes sends the emergency data directly to the
monitoring center via the Internet. Monitoring center parses and processes the emergency
data, and transfers the collected data to the displacement of bridge support or crack decision

404 Copyright ⓒ 2014 SERSC

 International Journal of Control and Automation
Vol.7, No.1 (2014)

level. When the result of verdiction is beyond the predefined normal range, the alarm will
automatically alarm, so as to truly achieve the remote real-time monitoring and improve the
efficiency of support inspection maintenance and replacement maintenance.

4.2. Daily data transmission

In order to reduce the redundant information of sensor in space and time, so that to reduce
unnecessary data transmission thus to save sensor network node energy consumption, the
daily data that is not urgent to process, will be process for data fusion before transmitted to
the monitoring center. Data fusion process is shown in Figure 2. The sensor nodes are
clustered based on different sensor types. The sensor nodes in the same cluster monitor the
same type of data information. There is a certain degree of information redundancy because

L.
of the close distance between the nodes [5]. Each node uploads its own perception data to the
cluster head nodes, cluster head nodes upload the fusion data by local computing to the sink

A
nodes, then the sink node transmits the final fusion results to the monitoring center database.

EG
The database with the accumulation of the essential data in a long time, can accurately reflect

s I ly.
the working state of the monitored railway bridge, and timely find development tendency for
diseases of bridge body and bridge support, which enable to provide decision basis for

ei n
LL
inspection maintenance especially jack-up replacement and other plans.

fil O
is on
th si
y er

Internet
eb eV

train
Em data

a
erg

dat
ad in
enc

ily
y

Da
m nl

Sink node
O

cluster cluster
cluster

temperature
ok

sensor
multimedia displacement
sensor
Bo

sensor

Figure 2. Data fusion process

5. Early Warning Data Processing

After the monitoring center collected the sensing data uploaded by the sink node, a certain
pretreatment process will be preceded to gain the current safety status grade of the bridge.
Security level of the bridge should be set according to the actual demand. For example, it can
be classified as normal, yellow warning, orange early-warning and red warning, and each
security level has a corresponding threshold. If the fusion results exceeded the preset

Copyright ⓒ 2014 SERSC 405

International Journal of Control and Automation
Vol.7, No.1 (2014)

threshold, the alarm monitoring system shall raise the alarm and inform the appropriate
warning levels promptly. When maintenance personnel received warning, hazard would be
identified and appropriate measures would be taken to eliminate the causes of the threat.

5.1. Repeatedly warning mechanism

In the practical application of bridge safety monitoring wireless sensor networks, sensor
malfunction is an inevitable phenomenon. The accuracy of sensing data measured by sensors
can directly affect decision of the system. In extreme circumstances, some sensors are
completely broken will cause mistakes in the operating condition assessment of bridge, which
may even lead to security incidents [6]. Sensor malfunction modes are mainly biased failure,
shock failure, short circuit fault, open circuit failure and drift failure [7]. Monitoring wireless

L.
sensor network consists of various types of sensors, such as displacement sensors,
temperature sensors and multimedia sensors. In the actual bridge safety monitoring network,

A
the number of these sensors is often great. When the transmitting data fluctuation of a sensor

EG
network node is large, the monitoring system may cause an error level of security

s I ly.
assessments, and thus gives an incorrect warning message.
To avoid early warning errors information misleading regulators, repeatedly warning

ei n
LL
approach should be taken to determine the correctness of the warning message. After early

fil O
warning monitoring system informed, measures need not be taken immediate to deal with the
problem. But can wait for a certain time, to observe whether the data received warning again.
is on
If the warning data consecutive time exceeds a predetermined value, a security problem could
be determined. And the problem should be solved by human intervention. The way that
th si

repeatedly alert would largely avoid the occurrence of wrong positives. In addition, there’re
y er

other means to provide system stability and accuracy for example employing sensors have
superiority in stable performance or adopting accurate fault detecting techniques [8].
eb eV

With growing of using time, bridges and other infrastructure will be gradually maturing,
and the security situation will be gradually reduced. Obviously, we should not use the security
level for new bridge to assess that of an old bridge. Therefore, the appropriate threshold
ad in

should be set for bridges of different using time to evaluate the security level of different
m nl

bridges. For example, the crack on the bridge is an important indicator to determine the safety
level of a bridge [9]. Monitoring indicators of bridge cracks include length, width, and the
O

crack growth rate and so on. For a newly built bridge, assessments standard tend to be very
strict, even a quite short crack is not allowed. While assessing a bridge built 20 years ago, the
standard willbe lowered a lot for the old bridge cracks are inevitable.

5.2. Automatically threshold adjusting based on the feedback learning

ok

Referencing certain standards, security threshold level can be manually set according to the
actual situation. However, many practical situations complicated than imagined, artificial
Bo

setting is often not accurate. If a dynamic threshold adjustment algorithm can be built, the
current security level of bridge will be automatically assessed according to the actual situation
of the bridge, and this algorithm should have robustness and adaptability. The neural network
algorithm is a good way to solve this problem.
Neural network algorithm has self-learning ability, and is widely used in pattern
recognition, optimization, data compression, and other areas [10].After inputting certain priori
data, artificial neural networks using some learning algorithms can automatically determine
the weight and give an optimal classification results. Figure 3 illustrated the automatically
threshold adjusting processing used in this paper.

406 Copyright ⓒ 2014 SERSC

 International Journal of Control and Automation
Vol.7, No.1 (2014)

Figure 3 illustrates the threshold level automatic adjustment process based on feedback
learning mechanism (in this case, the "feedback" is not the error back propagation (BPF)
neural network). When setting an initial threshold, enough sample data should be input to
ensure the neural network adjust the network weights according to the setting of the initial
threshold automatically and achieve enough accuracy. In the late process of monitoring, as
time goes on, the bridge monitoring system will compute and input the collected data to the
neural network repeatedly, and then the final evaluation results will be achieved. If the result
is normal, the results will be stored as the training data. After normal monitoring data is
sufficient, the recent input data (for example, 100) will be the training data to train the neural
network. With the circulation the neural network weights will change gradually, which is
suitable for bridge safety status. As the training process is relatively slow, even there is an
error during the process, the neural network can still give an accurate judgment and send

L.
warning information.

A
EG
Set initial

s I ly.
threshold

ei n
LL
fil O
Input data to train
Normal result
neural network
is on
th si

Adjust neural Input monitoring Achieve

y er

network data assessment result

eb eV
ad in

Abnormal result
m nl
O

Send warning
information

Figure 3. A threshold automatic adjustment process based on feedback

learning mechanism
ok

6. Conclusion
Bo

The railway bridge monitoring platform based on the Internet of Things proposed in this
paper mainly includes three key modules: wireless sensor’s deployment, wireless sensor
nodes’ information collection and warning information’s processing. During the test period in
China Railway Corp, the platform’s performance is excellent, for it can detect various kinds
of common bridge damage accurately, which is significant for the safety of the railway
bridge. As the next step of our work, we will set up a complete set of railway safety
monitoring system, which includes the real-time safety monitoring of infrastructures along the
railway such as the railway station, bridge, tunnel, line and so on. It will provide the railway
system a full range of security early warning, not only confined to the railway bridge any
more.

Copyright ⓒ 2014 SERSC 407

International Journal of Control and Automation
Vol.7, No.1 (2014)

References
[1] J. Zheng, “Railway Passenger Transportation Development in China”, Urban Transport of China, vol. 8, no. 1,
(2010) January.
[2] T. W. Hou, S. Guo and T. L. Han, “Simulations of safety monitoring for single railway bridge”, Sensor World,
vol. 18, no. 3, (2012).
[3] J. S. Fu, Z. J. Zhang and Y. Liu, “Research and Simulation on An New Algorithm RTST for Energy-Efficient
Coverage in URail Traffic Wireless Sensor Networks”, Journal of the China Railway Society, (2013).
[4] Z. H. Xia, S. F. Zhu and Z. H. Zong, “Assessment of support displacement of a continuous rigid-frame bridge
with a large span based on long-term health monitoring”, Proceedings in 18th National Conference on
Structural Engineering, (2009), GuangZhou, China, pp. 542-548.
[5] R. Wu, “Research on Data Fusion Technology of Wireless Multimedia Sensor Networks”, Nanjing University
of Posts and Telecommunication, (2013).
[6] T. H. Guo and J. Nurre, “Sensor failure detection and recovery by neural networks”, Neural Networks,

L.
IJCNN-91-Seattle International Joint Conference on IEEE, vol. 1, (1991), pp. 221-226.
[7] S. K. Yung and D. W. Clarke, “Local sensor validation”, Measurement and Control, (1989).

A
[8] M. Henry, “Sensor validation and fieldbus”, Computing & Control Engineering Journal, vol. 6, no. 6, (1995),

EG
pp. 263-269.

s I ly.
[9] I. Abdel-Qader, O. Abudayyeh and M. E. Kelly, “Analysis of edge-detection techniques for crack
identification in bridges”, Journal of Computing in Civil Engineering, vol. 17, no. 4, (2003), pp. 255-263.
[10] C. Baladrón, J. M. Aguiar and L. Calavia, “Performance study of the application of artificial neural networks

ei n
LL
to the completion and prediction of data retrieved by underwater sensors”, Sensors, vol. 12, no. 2, (2012), pp.

fil O
1468-1481. is on
th si
y er
eb eV
ad in
m nl
O
ok
Bo

408 Copyright ⓒ 2014 SERSC