This article has been accepted for inclusion in a future issue of this journal.

Content is final as presented, with the exception of pagination.

IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS 1

Self-Powered ZigBee Wireless Sensor Nodes

for Railway Condition Monitoring
Mingyuan Gao, Member, IEEE, Ping Wang, Yifeng Wang, and Lingkan Yao

Abstract— A track-borne energy transducer is a smart device

for harvesting energy of trains or rail transportation systems.
In this paper, the authors extend this application through
introducing a self-powered ZigBee wireless sensor node. The
proposed hardware prototype consists of a ZigBee coordinator
at road-side and a series of sensors (Accelerometer, temperature
sensor, humidity sensor, and infrared detector) connected to a
ZigBee end device at rail-side. The ZigBee end device is powered
by the magnetic levitation energy harvester and communicated
wirelessly with the ZigBee coordinator. The magnetic levitation
oscillator is selected due to its broad-band response character-
istics. The results indicate a peak–peak output voltage of 2.3 V
under the condition that the vehicle travels over the rail-borne
device at the speed of 105 km/h.
Index Terms— Energy harvesting, Internet of Things,
magnetic levitation, railway condition monitoring, wireless sensor
networks, ZigBee.
Fig. 1. Illustration of self-powered ZigBee wireless sensor networks for
railway condition monitoring.
I. I NTRODUCTION
A. Motivation wireless sensor network for railway condition monitoring. This
technology contributes to resolving the power supply problems
R AIL transport, as a means of conveyance of passenger
and cargo, plays an important role in our daily life. The
building of rail infrastructure has experienced a sustainable
of rail-side monitoring devices and benefits the development
of Internet of Things (IOT) in the area of intelligent transporta-
tion. Figure 1 illustrates the concept and approach of this paper
growth in the past decade, especially in some developing
for establishing a ZigBee stack-based wireless sensor network
countries. However, in remote area it is difficult to ensure
power supply of rail-side monitoring equipment because wired powered by track-borne energy harvesting system. The rail-
side sensors are connected to the ZigBee end device and
power supply is not available for railways deployed in these
powered by the track-borne energy harvester. The data can be
locations [1]. If we use wireless power supply such as bat-
teries, it is also a big challenge for the maintenance and transmitted to the ZigBee coordinator and be accessed through
webpage on the internet with cloud computing technology.
replacement of batteries in remote or inaccessible locations.
It is therefore necessary to develop a new energy strategy.
Actually, Germany has already engaged in covering more than B. Review of the State of the Art
one-third of its annual twelve billion kilowatt-hour energy ZigBee, as an IEEE 802.15.4-based specification, is a suit-
requirement for the railway network with renewable sources able communication protocol for creating wireless networks
by 2020 [2]–[4]. The saved energy is capable of powering with small, low-power digital radios [10], [11]. Compared
wireless sensor networks (WSNs), which can be used for with other wireless personal area networks (WPANs) such as
monitoring the railway infrastructure such as bridges, turnout, Bluethooth and Wi-Fi, the ZigBee specification is intended
rail tracks, cuttings and tunnels, and track beds [5]–[9]. This to be simpler, less expensive and consume less power [12].
motivates the authors in this paper to develop a self-powered The transmission distances are limited to 10–100 meters line-
of-sight and the ZigBee devices can also transmit data over
Manuscript received September 27, 2016; revised January 13, 2017 and long distances by establishing a mesh network of intermediate
May 15, 2017; accepted May 24, 2017. This work was supported by
the National Natural Science Foundation of China under Grant 51425804. devices to reach more distant ones [13]. Since ZigBee has a
The Associate Editor for this paper was K. Wu. (Corresponding author: defined transmission rate of 250 kbit/s, rendering it suitable for
Ping Wang.) intermittent data transmissions from sensors or input devices.
The authors are with the School of Civil Engineering, Southwest Jiaotong
University, Sichuan 610031, China, and also with the Key Laboratory of Browne et al. proposed a novel mobile antenna structure
High-speed Railway Engineering, Ministry of Education, Chengdu 610031, connected to a ZigBee wireless transceiver. The structure
China (e-mail: wping@home.swjtu.edu.cn). was used to relay stress and strain data on railway tracks.
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org. The power consumption was 1 mW for transceiver and the
Digital Object Identifier 10.1109/TITS.2017.2709346 communication distance reached to 91 m [14].
1524-9050 © 2017 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
2 IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
With regard to energy harvesters applied to transporta-
tion area, applications of piezoelectric pavement for highway
traffic monitoring had been reported [15]–[17]. Furthermore,
track-borne energy generation in terms of piezoelectric and
electromagnetic harvester was investigated [18]–[21]. Besides
track-borne energy harvester, Pasquale et al. designed a bogie-
borne energy harvester with magnetic suspensions [22]. The
device with package of 150 mm ×125 mm ×95 mm could
generate 100 mW power output on a scaled roller rig, which
simulated the freight train velocity of 80 km/h. In our previous
study, we also investigated the electrical characteristics, device
structure, and vibration response of track-borne electromag-
netic and piezoelectric energy transducers [23], [24].
In the field of self-powered railway condition monitoring,
the related work focused mainly on the energy harvesting
solutions; whereas there was some work for the monitoring of
truss bridge by using solar energy harvesting to power sensors.
Bischoff et al. [25] established a WSN for strain monitoring of
a truss railroad bridge. The base station was powered by a solar
energy rechargeable battery, while the end nodes were powered
by batteries alone. Sekula and Kolakowski [26] proposed
a Weigh-in-Motion (WIM) system using piezoelectric strain
sensors powered by photovoltaic modules. Solar power is a
proven way to generate energy. However, it has a number of
practical drawbacks for WSNs [1]. Solar panels need to be
cleaned frequently to maintain an optimal performance [27].
Solar power is highly dependent on the weather condition, and
sunlight varies on a daily basis. In addition, there is no sunlight
available for the railway transit in tunnel and the urban rail
transit, the latter is under a rapid growth in several mainland
Chinese cities [28].
C. Scope of Paper
Main challenges faced by railway engineers for railway
track monitoring are: 1) power supply problem of rail-side
equipment. Wired monitoring devices depend on the grid
infrastructure, which needs huge investments; whereas the
wireless ones require batteries which can only work for a
certain time period. The railway track is very long, cross
remote or natural region, rendering the replacement of batteries
impossible. 2) money-consuming to deploy the railway moni-
toring system. It costs much to deploy traditional monitoring
equipment.
Thanks to the rapid development of semiconductor man-
ufacturing process and large scale integrated circuit, Micro
Electromechanical Systems (MEMs)-based sensor technique
offers a low cost, low power, and integrated approach to struc-
tural healthy monitoring of railway track. ZigBee as low power
consumption, low cost, limited data rate, and high reliable
and secure communication protocol has well suited for the
application of wireless sensor networks for railway condition
monitoring. The objective of this paper is to solve the power
supply problem of rail-side ZigBee devices and build up a self-
powered railway monitoring prototype with ZigBee protocol
stack and magnetic levitation energy harvester.
Compared with the WSN system powered by solar energy,
the proposed solution is independent on the weather condition
and can be used for both the railway transit in tunnel and
Fig. 2. Illustration of types of track-borne electromagnetic energy harvester.
the urban rail transit. Compared with the other vibration-
based energy harvesting solution, the proposed one is capable
of energy harvesting at wide frequency range with small
displacement amplitude, which fits the requirement of the
vibration characteristics of the railway track. Section I of this
paper presents the review of the state-of-the-art of track-borne
energy harvester and application of ZigBee protocol in railway
industry, which inspired the authors to conduct this study.
Section II gives the principle of magnetic levitation oscillator
and introduces the track-borne electromagnetic energy har-
vester, which is used to power the ZigBee devices. Hardware
prototype of ZigBee wireless communication module, together
with selected monitoring sensor is introduced in Section III.
A DC-DC booster converter is developed to convert the
alternating input to a standard DC output. Section IV analyzes
the response of magnetic levitation harvester and reports the
test results of the proposed approach.
II. T HEORY
A. Magnetic Levitation Oscillator
The magnetic levitation oscillator is selected due to its
broad-band response characteristics. The dynamic system with
nonlinear restore force as in the levitation case can be written
down in form of Duffing equation as follows:
ẍ + 2ω0 ξ˙x + ω02 x + βx 3 = Fcos(ωt + φ) (1)
Where x is the displacement, t is the time, ξ is the damping
ratio, ω0 is the natural radian frequency, β is the nonlinear
stiffness parameter, F is the excitation amplitude, and φ is
the initial phase.
After implementing amplitude-frequency analysis, we find
that the amplitude of the input will change the frequency
response. It includes also unstable periodic solutions and
results will be discussed in the following parts.
B. Track-Borne Energy Harvesting
The energy harvesters can be connected to the rail in
two ways, namely, serial connection and parallel connection.
We classify the parallel-connected track-borne electromagnetic
energy harvester in terms of voice coil, spring resonant, and
magnetic levitation (as illustrated in Figure 2). In our case, as
illustrated in Figure 3, the harvester devices were mounted to
the rail foot and subjected to the vibration generated by the
wheelset/track system. For the serial configuration, the energy
harvester is normally located between the rail and sleeper,
or between the sleeper and ballast bed; therefore, the existence
of the serial-connected device will change the track stiffness.
For the parallel configuration, the energy harvester is rigidly
connected to the rail web or rail bottom; it is neither connected
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
GAO et al.: SELF-POWERED ZigBee WIRELESS SENSOR NODES FOR RAILWAY CONDITION MONITORING
Fig. 3. Encapsulated sensor node prototypes developed by SWJTU.
to the wheelset nor to the rail sleeper, so we treat the parallel
configuration as an added mass to the rail, and calculate the
resultant change of the rail mass and rail moment of inertia.
According to Lenz’ law, the direction of the induced current
always opposes the change that produced it. This explains
the resistance force in the electromagnetic energy harvester.
Assuming a harmonic excitation, the resultant damping force
can be calculated by
 2
P V2 N A B
Fdamping = = = ẋ (t) (2)
ẋ (t) (R + i ωL) ẋ (t) R + i ωL
Where P is the power, V is the induced voltage, R and L
 link quality to the adjacent ones.
are the resistance and inductance of the coil, and N, A, and B
indicate the number of coil turns, the coil wire cross-section
area, and the gradient of magnetic flux density, respectively.
The energy generated during a given time period (also
energy dissipated by damping) can be calculated by
 X0
E =2 c ẋ (t) d x = −2cωX 0 eiωt 2X 0 = −4cωX 20 eiωt (3)
−X 0
Because the harvesters were rigidly connected to the rail
and subjected to the vibration generated by the wheelset/track
system, we could utilize this excitation to enable the rail-borne
transducer for energy harvesting; however, the existence of
the rail-borne transducer could inversely exert an influence on
the track vibration response, especially when a large quantity
of transducers were installed onto the track. In such a case,
the accumulative contributions stemmed from the transducers
could be quite considerable to have an influence on the rail
deflection. Detailed analysis and device design of the track-
borne electromagnetic energy harvester could be found in
literature [23].
III. M ATERIALS AND M ETHODS
The proposed hardware prototype consists of a ZigBee coor-
dinator at road-side and a series of sensors (Accelerometer,
temperature sensor, humidity sensor, and infrared detector)
connected to a ZigBee end device at rail-side. The ZigBee end
device is powered by the magnetic levitation energy harvester
and communicated wirelessly with the ZigBee coordinator.
The hardware prototype is shown in Figure 3.
3
A. ZigBee Wireless Module
The ZigBee wireless module consists of a ZigBee coordina-
tor and several ZigBee end devices. The ZigBee coordinator is
a full-function device (FFD) which takes care for network for-
mation and maintenance. The ZigBee end device can be either
a FFD or reduced-function device (RFD) and participates into
the network.
The ZigBee hardware prototype includes a System-
on-Chip (SOP) chipset CC2530F256 (Texas Instruments),
a chipset PL2303, a SMA antenna connector, a 32 MHz
crystal, a USB port, an I2 C (Inter Integrated Circuit) and
a Serial Peripheral interface (SPI), several LEDs, power
on/off switch, program/debug connector, and other peripherals.
CC2530F256 is connected to the peripheral interfaces and to
a standard SMA (sub-miniature type A) connector with an
external RF antenna.
The development of the hardware prototype in this paper
referred to the schematics of ZigBee development kits of Texas
Instruments [29].
The Wireless Sensor Networks include three nodes: sensor
node, router node, and root node. The sensor node collects
the data, the root nodes as the data sink is linked to the base
station, and the router node is an intermediate node to relay the
data. The communication link between the WSN is established
by the ZigBee communication protocols, which is operated as
a multi-hop network. The Sensornet Protocol (SP) [30] defines
the routing tree. Its flexibility allows the choosing of the most
reliable nodes for transmission by periodically assessing the
B. Sensor Module
The proposed hardware prototype consists of a series of
sensors (Accelerometer, temperature sensor, humidity sensor,
and infrared detector) connected to a ZigBee end device at
rail-side.
The ADXL series MEMS accelerometers (ADXL345 and
ADXL1001) were used. Take ADXL345 for instance, it is a
small (3 mm × 5 mm × 1 mm LGA package) and ultralow
power (less than 350 μW) 3-axis MEMs accelerometer with
measurement up to ±16 g. Digital output data is accessible
through either a SPI (3 or 4 wires) or I2 C digital interface.
It can measure the static acceleration of gravity in tilt-
sensing applications and the dynamic acceleration resulting
from motion or shock. The ADXL345 is well suited for
mobile device application with high resolution (3.9 mg/LSB)
measurement of inclination changes less than 1.0° .
The DHT11 is a small-size (32 mm×14 mm) digital temper-
ature/humidity sensor. It includes a resistive humidity sensor
and a Negative Temperature Coefficient (NCT) thermometer.
The DHT11 has a measurement range of 20% – 90% relative
humidity and 0 – 50 degC. The measurement accuracy is
±5% RH and ±2 degC with resolution of 1.
The HC-SR501 is a high sensitivity and ultralow
power (325 μW) human infrared sensor module. It includes a
LHI778 probe and has two trigger modes. The HC-SR501 has
feature of automatic induction with sensing range of less than
120 degree cone angle at a detect distance less than 7 m.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

4 IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS

TABLE I
T YPICAL P OWER C ONSUMPTION OF S ENSOR D EVICES FOR R AILWAY M ONITORING

Besides aforementioned sensors which were used in the TABLE II

prototype test, we also make a summary of common-used T YPICAL P OWER C ONSUMPTION OF ZigBee N ODES
sensors for railway condition monitoring. Information such as
typical power consumption, package size, manufacturer, device
type is listed in Table I.

C. Power Supply Module

The ZigBee node is powered by the Lithium-batteries and
the energy harvester. In consideration of the volatility of real
traffic loads, the energy harvester does not power the ZigBee
nodes directly; instead, it powers the Lithium-batteries and
thus extends the operating time of overall system. Further-
more, the alternating induced voltage of the electromagnetic
transducer must be converted to the stable DC output. The
challenging of the energy conversion of electromagnetic har-
vesters is: 1) low induced voltage. Bridge rectifies always have
the forward voltage drop, and Schottky diodes could be used
to minimize the forward voltage to 0.3 V to 0.7 V; 2) power
tradeoff. Since the harvesters have limited power output, and
for series regulator, the voltage difference across transistors is
large and the resultant power dissipation always surpasses the
capability of the harvester; therefore the switching regulator
should be used, but the power distributed to the oscillator and
value of 2 V to 4 V induced voltage can be achieved by simply
comparator must be carefully treated to keep the capability of
the total power output. adjusting the parameters of the magnetic levitation harvester
We designed the energy conversion circuit with a full- (i.e., numbers of turn, coil geometry and arrangement, surface
magnetic flux density, mass of suspended magnets, etc.).
bridge rectifier and a DC-DC boost converter. The circuit was
connected to the electromagnetic transducer during the test.
Schottky barrier diodes BAT54LT1 were used to minimize D. Frequency-Sweeping Vibration Test
the voltage drop-off (Low forward voltage − 0.35 Volts Frequency sweeping vibration tests were conducted in Lab.
(Typical) @ IF = 10 mA). Table II summarizes the typical The purpose of this test was to characterize the perfor-
power consumption of the proposed ZigBee node. Table III mance of the magnetic levitation harvester over a wide range
records the work condition of DC-DC circuit. To enable the of frequency. Two frequency ranges (5 Hz–200 Hz and
DC-DC converter, the AC input voltage should be higher 5 Hz – 500 Hz) were chosen according to the standard
than 1.88 V. The magnetic levitation harvester needs to be IEC60068-2-6 [34]. The amplitude of acceleration was set
well designed to achieve a higher induced voltage for the better to 1 g, 2 g, 4 g, and 8 g. The standard specifies constant
performance of the system. According to the test, a peak-peak displacement amplitude below the cross-over frequency and
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

GAO et al.: SELF-POWERED ZigBee WIRELESS SENSOR NODES FOR RAILWAY CONDITION MONITORING 5

TABLE III
S UMMARY OF THE DC-DC C IRCUIT W ORK C ONDITION

constant acceleration amplitude above the cross-over fre-

quency. The number of sweep circles was 10 with the sweep
rate in 1 octaves/minute. The mass of suspended magnets was
specified as 60.4 g, 121.7 g, 182.9 g, and 288 g. Two integrated Fig. 4. Amplitude-frequency response of the magnetic levitation device (blue
electronic piezoelectric accelerometers (IEPE) were utilized. dot: stable periodic solution; red dot: unstable periodic solution; β = 0.5;
The one with sensitivity of 9.71 mV/(g) was fixed on the ξ = 0.05; F = 0.2).
vibration table as the control signal of vibration excitation,
and another one with sensitivity of 9.73 mV/(g) was attached
to the suspended magnets as the monitoring signal of dynamic
response.

E. On-Track Field Test

The ZigBee end device together with the magnetic levita-
tion harvester was mounted to the rail web on one section
of Chengdu railway track. The ZigBee nodes collected the
temperature, the humidity, and the acceleration of railway
track and communicated wirelessly with the ZigBee coor-
dinator. The ZigBee coordinator transferred the data to a
laptop computer, which can display the signals profile and
store the data to a log file. To record and observe the profile
of the generated voltage of magnetic levitation harvester,
an oscilloscope (Tektronix TDS2004C) with a voltage probe
of 10X attenuation was connected to the magnetic levitation
device. The sampling data from the oscilloscope were recorded
and transferred to a local computer for subsequent analysis.
The oscilloscope served as a data sampling device and was
connected to the laptop with Labview Signal Express software.
In this way, we can activate the auto trigger/record mode and Fig. 5. Frequency-sweeping vibration test setup (a) view of energy harvester
realize continuous data sampling. The data were stored on the configuration; (b) view of power and control unit; (c) overall view of exper-
laptop’s hard drive. imental setup. (1)monitoring IEPE; 2)suspended magnets; 3)coils; 4)control
IEPE; 5)power cabinet; 6)IPC with control software 7)vibration controller;
8)energy harvester; 9)vibration table; 10)oscilloscope; 11)laptop with Labview
IV. R ESULTS AND D ISCUSSION Signalexpress interface).
Three types of track-borne electromagnetic device were
compared. The rail vibration velocity had the direct corre- system locates away and no longer scales linearly. We could
lations with the generated voltage of electromagnetic devices. engage manually such nonlinearities through adjusting the
The signals are nearly same in amplitude profile and opposite non-linear parameter β and damping ratio ξ of the magnetic
in phase. levitation harvester. The nonlinear restore force of magnetic
Figure 4 shows the calculated results of amplitude- levitation setup depends on the space between magnets; and
frequency response of the magnetic levitation device. the damping force can be controlled by choice of coil parame-
Blue dot indicates the stable periodic solution, whereas red dot ters and magnetic flux density. Based on the aforementioned
represents the non-stable periodic solution. We take ξ = 0.05, analysis, magnetic levitation harvester is preferred due to its
β = 0.5 as calculation parameters. It is observed that a nonlinear characteristics of extending the devices’ response
jump from the larger response to the smaller response (for- range for track energy harvesting.
ward jump) or a jump from the smaller response to the Frequency sweeping vibration tests were conducted to char-
larger response (backward jump) might happen. The jump acterize the performance of the magnetic levitation harvester.
can be trigger by implementing a mechanical or electrical Test setup was shown in Figure 5. The harvester was mounted
perturbation. onto the vibration table. Two IEPE accelerometers were uti-
Another observation is that, compared with the linear reso- lized, one was fixed on the vibration table to control the
nance, the peak response of the nonlinear magnetic levitation vibration amplitude and frequency, and another was attached
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

6 IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS

Fig. 7. Main effects plot and interaction effects plot of the generated voltage
of the magnetic levitation harvester.

Fig. 6. Controlled input acceleration and monitored acceleration of suspended

magnets (suspended mass of 288 g, relative high position, 5000 numbers of
turn, and surface magnetic flux density of 0.5 T).

Fig. 8. Prototype of the magnetic levitation harvester and field test setup.
to the suspended magnets to monitor the response. Piezo-
electric signals were sampled and amplified in the vibration
controller (ECON UCON Series). A power cabinet provided
power supply to the vibration table and an industrial PC
displayed the software interface of the vibration controller.
An oscilloscope (Tektronix DPO2000) with a voltage probe
of 10× attenuation was connected to the magnetic levitation
device. The sampling data from the oscilloscope were recorded
and transferred to a local computer for subsequent analysis.
The oscilloscope served as a data sampling device and was
connected to the laptop with Labview Signal Express software.
Figure 6 shows the recorded data of frequency sweep-
ing test. Green solid line represents the specified control profile
of the acceleration excitation of vibration table. Red solid
line stands for the real control signals of acceleration during
up-sweeping, black solid line stands for those during down- Fig. 9. Power map of the energy harvester under 5-200 Hz frequency
sweeping test with excitation acceleration of 2g.
sweeping. The up and down-sweeping signals fit very well
and both signals indicate an unstable point at frequency
around 80 Hz. This is due to the resonance of aluminum
frame of the harvester. The blue solid line and black dash
line represent the response acceleration of suspended magnets
during up and down-sweeping. The response acceleration
signals of up and down-sweeping fit well at frequency range
from 20 Hz to 500 Hz.
Four factors and two levels’ design of experiment (DOE)
was conducted. As can be seen in Figure 7, mass of suspended
magnets, relative position between magnets and coils, vibra-
tion amplitude of displacement and acceleration could have an
obvious effect on the generated Vp-p of the energy harvester.
Figures 9-10 show the power generation map of the magnetic
levitation harvester with suspended mass of 121.7 g, relative
high position, 5000 numbers of turn, and surface magnetic Fig. 10. Power map of the energy harvester under 5-500 Hz frequency
flux density of 0.5 T. The DC-DC converter can be enabled sweeping test with excitation acceleration of 4g.
with a peak-peak voltage input larger than 1.88 V. The power
map records some of these points. The vibration velocity, usable voltage output over a wide frequency range from 7 Hz
displacement, and excitation frequency are also plotted. The to 500 Hz. It is known that the amplitude of rail acceleration is
operation range of the harvester extends with the increase of usually large than 10g and could reach 50 g or higher. So the
acceleration amplitude. At acceleration amplitude of 4 g and magnetic levitation harvester matches the requirement of rail
cross-over frequency of 31.5 Hz, the harvester could offers transportation.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
GAO et al.: SELF-POWERED ZigBee WIRELESS SENSOR NODES FOR RAILWAY CONDITION MONITORING
Fig. 11. Voltage profile of the magnetic levitation harvester under different
traffic loads.
Fig. 12. User interface of ZigBee communication module connected with
sensors of temperature/humidity and acceleration.
Figure 11 represents the voltage profile of the track-borne
magnetic levitation energy harvester under travelling loads.
The energy profile of the electromagnetic harvester indicates
a peak-peak output voltage of 2.3 V under the condition that
the vehicle of one carriage travels over the rail-borne device
at the speed of 105 km/h. Two traffic loads are compared. The
testing setup was shown in Figure 8.
Figure 12 shows the display interface of ZigBee communi-
cation module connected with sensors of temperature/humidity
and accelerometer. It should be noted that the proposed energy
harvester can generate enough power for ZigBee end device
when the vehicle passes; in case of no vehicle passing through,
there is no wheelset/track interaction energy for enabling
the system. Therefore, an energy storage circuit with batter-
ies and supplementary power source (e.g. wind turbine) are
necessary for round-the-clock and long-term monitoring. The
energy harvester can charge the batteries (either by under-track
wind turbine or by electromagnetic devices) and extend the
applications.
Serial port to WiFi adapter can be used to let ZigBee nodes
with serial port communication to any secure or open wireless
network or access point. Data can be transferred over TCP/IP,
UDP or HTTP using 802.11 b/g/n infrastructure or Adhoc
mode (Simple AP / Station Mode) networking. Once the
collected data from ZigBee nodes are available over WiFi,
Internet-based computing (cloud computing) can be enabled
access to a shared pool of configurable computing resources
(i.e., computer networks, servers, storage). The collected data
could be analyzed and visualized through cloud computing
and stakeholders at different sites can have access to these
data and make their own judgment for operation strategy of
railway track. The further investigation is current undergoing.
Figures 13-14 show the Received Strength Signal Indi-
cator (RSSI) of the ZigBee nodes. The Rx sensitivity of
7
Fig. 13. (a) RSSI in relation to TxPower (curve a) TxPower 4.5dbm; curve
b) TxPower 2dbm; curve c) TxPower −0.5dbm; payload 32 Byte; sunny;
height above ground 0m; distance 1-20m(x coordinate), 20 points); (b) RSSI in
relation to data payload (curve a) payload 48 Byte; curve b) payload 32 Byte;
curve c) payload 16 Byte; TxPower 4.5dbm; sunny; height above ground 0m;
distance 1-20m(x coordinate), 20 points).
Fig. 14. (a) RSSI in relation to weather condition (curve a) sunny; curve
b) hazy; curve c) windy; TxPower 4.5dbm, payload 32 Byte; height above
ground 0m; distance 1-20m(x coordinate); 20 points); (b) RSSI in relation
to height above ground (curve a) 0m; curve b) 1m; curve c) 1.5m; TxPower
4.5dbm, payload 32 Byte; sunny; distance 1-20m(x coordinate); 20 points).
the node was set to -90 dBm. Many factors could have
an influence on the RSSI, such as TxPower, data payload,
communication distance, and weather conditions. As shown
in Figure 13, the RSSI decayed in distance. At a distance above
20 m, the communication between the ZigBee end device
and coordinator became unstable. The value of TxPower had
a significant influence on the RSSI, in case of -0.5 dBm
TxPower, the communication became unstable at a distance
above 12 m. When the TxPower was set to 4.5 dBm, a reliable
communication can be guaranteed at a distance range from
0 m–20 m.
We also compared the RSSI of the nodes with different data
payload. It was observed that data payload had little contribu-
tion to the RSSI. Different weather conditions were also met
during the field test and three main weathers (i.e. sunny, windy,
and hazy) were analyzed. Heavy haze hit many places in China
in recent years during to unfavorable weather condition caused
by air pollution. The haze days were also very frequent at our
test field. It was observed that the hazy weather condition had
a negative influence on the RSSI. The deterioration of RSSI
in hazy weather might be caused by the high concentration of
PM2.5, which affected the free space signal transmission.
Unlike the wireless monitoring application in the Bridge
and elevated highway, whose mounting height of nodes is
very large; the track-borne ZigBee nodes are connected to the
rail foot and the railway track is quite close to the ground
foundation, so the effects of RSSI in relation to mounting
height should also be considered. As shown in Figure 14,
the larger the height above ground, the less attenuation of
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
8 IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
Fig. 15. (a) Packets loss rate in relation to data payload (curve a)
payload 48 Byte; curve b) payload 32 Byte; curve c) payload 16 Byte;
distance 5m, TxPower 4.5dbm; sunny; height above ground 0m; numbers of
node 1-3(x coordinate)); (b) Packets loss rate in relation to weather con-
dition (curve a) sunny; curve b) hazy; curve c) windy; distance 5m,
TxPower 4.5dbm, payload 32 Byte; height above ground 0m; numbers of
node 1-3(x coordinate)).
Fig. 16. (a) Data rate in relation to data payload (curve a) payload 48 Byte;
curve b) payload 32 Byte; curve c) payload 16 Byte; distance 5m, TxPower
4.5dbm; sunny; height above ground 0m; node number 1; time periods
1-20s(x coordinate)); (b) Data rate in relation to node numbers (curve a)
node number 1; curve b) node numbers 2; curve c) node numbers 3; distance
5m, TxPower 4.5dbm; payload 32 Byte; sunny; height above ground 0m; time
periods 1-20s(x coordinate)).
received signals. Based on the aforementioned results and
discussion, the installation distance between two track-borne
ZigBee nodes should be less than 20 m to ensure a reliable
communication.
Figure 15 indicates the packets loss rate in relation to
the number of nodes. The background of this test was to
observe the communication jamming when several end device
nodes would like to transmit the signals simultaneously to
a coordinator. This is important for sensor clustering of a
WSN. ZigBee nodes use the mechanism of Carrier Sense
Multiple Access with Collision Detection (CSMA/CD). This
mechanism allows transmission termination as soon as a
collision is detected. As shown in Figure 15, no packets were
lost when 1 node transmitted the signals; whereas when more
nodes joined the network and began to transmit the signals
simultaneously, the CSMA/CD mechanism worked, leading
to an increase of packets loss rate. As can be seen from
Figure 15(b), hazy weather could also have effects on packet
loss rate. In case of three nodes transmitting signals simulta-
neously under a haze day, the packet loss rate reached 20%.
Figure 16 shows the data rate of the ZigBee node. In radio
communications, the payload excludes any headers or meta-
data, is the part of transmitted data that includes the actual
intended message. Data rate with three different data pay-
load (i.e. 48 byte, 32 Byte, and 16 Byte) were tested and
the results scaled linearly. Above 5000 Byte (40 kbps) data
Fig. 17. Date rate in relation to weather condition (curve a) sunny; curve
b) hazy; curve c) windy; Distance 5m, TxPower 4.5dbm; payload 32 Byte;
node number 1; height above ground 0m; time periods 1-20s(x coordinate)).
TABLE IV
C OST E VALUATION OF THE S YSTEM
can be delivered for 1 ZigBee node. Figure 16(b) shows the
data rate in relation to numbers of node. It was observed that
the data rate did not scale linearly with the numbers of node.
At a data payload of 32 Byte for three nodes, the throughput
reached 7000 Byte (56 kbps). We also tested the data rate
under different weather conditions as shown in Figure 17.
It was noted that weather conditions had little influence on the
data rate once the communication link was reliably established.
Requirement of a feasible railway condition monitoring
system includes: 1) Reduced maintenance; 2) Save cost;
3) Constructability: the ease and efficiency with which the
monitoring system can be mounted; and 4) Maintenance strat-
egy: relied on the prediction of failure rather than maintenance
based on a regular schedule.
The cost of the proposed solution is summarized in Table IV.
The overall cost per km is around $55,000. Generally, it costs
¥0.1 – 0.2 billion per km to built a new line (e.g., the Beijing–
Tianjin Intercity Railway line is 117 km (72.7 statute miles)
delivered over 41 million rides. The line cost ¥20.42 billion
to build, and ¥1.8 billion per annum to operate, including
¥0.6 billion in interest payments on its ¥10 billion of loan
obligations [35]. Compared with the total cost to build a new
railway line, the cost per km of the proposed monitoring
system is acceptable.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
GAO et al.: SELF-POWERED ZigBee WIRELESS SENSOR NODES FOR RAILWAY CONDITION MONITORING
Fig. 18. Recorded rail acceleration and induced voltage under travelling
loads of a high speed train.
TABLE V
A CCURACY E VALUATION OF THE S YSTEM
The proposed system can be easily mounted to the track.
It has good constructability for both the upgrade pre-existing
rail lines and newly built rail lines. The supplementary power
source by energy harvester extents the battery exchange cycles,
and thus reduce the maintenance time and effort.
Figure 18 shows the rail acceleration and induced voltage
of the system under travelling loads of a high-speed train at a
speed of 240 km/h. The track consists of continuously welded
rail attached by Vossloh300-1 spring fasteners on concrete
double-block track. The vehicle consists of two power cars
and 6 passenger carriages. Each coach is supported by four
wheel-sets in two bogies with a wheel base of 2.5 m and a
rolling stock axle spacing of 18 m. The power car has a length
of 25.7 m and the passenger carriage has a length of 25 m.
The total length is 203 m with axle load of 15 ton.
To meet the requirement of prediction-oriented maintenance
strategy, the accuracy of the monitoring system must be
fulfilled. Table V evaluates the accuracy of the WSN. The
long-term stability of the system is still under evaluation.
V. L ESSONS L EARNED
1) Communication distance. In practical use, many factors
could affect the RSSI and limit the communication distance
of the sensor nodes. Given that the track-mounted device is
very close to the ground and subjected to different weather
conditions, the maximum installation distance between two
nodes should be less than 20 m.
9
2) Impacting of passing trains - geometric ones than
EM effect. By potting all electronic components into alu-
minum housing, the node is immune to the electro-magnetic
interference of passing trains; but the geometric blocking of
passing trains could have an influence on the sensor nodes. It is
thus recommended to mount the sensor nodes at the same side
of railway track.
3) Power supply strategy. Vibration-based harvester cannot
power the sensor nodes alone; battery together with power
management module is needed. The harvester could capture
the vibration energy induced by the travelling loads of vehi-
cle and charge the batteries. Alternative energy harvesting
approach such as wind turbine energy generation by train-
induced wind [36] could be considered as a supplementary
power source.
4) Application area. Track-borne system can be used to
vehicles by adding GPS module. Same sensor nodes can be
mounted to several carriages. If the abnormal signals are found
by all nodes at the same GPS coordinate, the railway track at
this point would be assumed as “unusual”; if the abnormal
signals are found by one node over a large range of region,
the relevant vehicle-bogie would be checked.
VI. C ONCLUSION
This paper investigated the possibility of establishing a self-
powered wireless sensor network by integrating the techniques
of ZigBee stack protocol and energy harvesting. The target
was aimed at the railway condition monitoring. Three types of
track-borne electromagnetic harvester were compared and we
found that energy harvesting by magnetic levitation is suitable
for powering the ZigBee end device due to its wide frequency
response. A hardware platform was developed and tested,
indicating the feasibility of the proposed approach. Further
investigation is current undertaken for evaluating the long-term
performance of the self-powered ZigBee sensor nodes for the
railway condition monitoring.
ACKNOWLEDGMENT
The authors would like to thank Linda Yao from GE
Healthcare for language modification.
R EFERENCES
[1] V. J. Hodge, S. O’Keefe, M. Weeks, and A. Moulds, “Wireless sensor
networks for condition monitoring in the railway industry: A survey,”
IEEE Trans. Intell. Transp. Syst., vol. 16, no. 3, pp. 1088–1106,
Jun. 2015.
[2] Siemens Online. (2016). Rail Electrification. [Online]. Available:
http://www.mobility.siemens.com/mobility/global/en/rail-solutions/
[3] K. W. Lee, A. J. Correia, K.-H. Lee, B. D. J. Neilan, and S. Gregor,
“Solar energy harvesting from roadways,” in Proc. 93rd Annu.
Meeting Transp. Res. Board, Washington, DC, USA, Jan. 2014,
pp. 4380-1–4380-18.
[4] G. Gatti, M. J. Brennan, M. G. Tehrani, and D. J. Thompson,
“Harvesting energy from the vibration of a passing train using a single-
degree-of-freedom oscillator,” Mech. Syst. Signal Process., vosl. 66–67,
pp. 785–792, Jan. 2016.
[5] L. Fagiano, M. Milanese, V. Razza, and M. Bonansone, “High-altitude
wind energy for sustainable marine transportation,” IEEE Trans. Intell.
Transp. Syst., vol. 13, no. 2, pp. 781–791, Jun. 2012.
[6] B. Ai et al., “Challenges toward wireless communications for high-speed
railway,” IEEE Trans. Intell. Transp. Syst., vol. 15, no. 5, pp. 2143–2158,
Oct. 2014.
[7] K. Kwong, R. Kavaler, R. Rajagopal, and P. Varaiya, “Real-time
measurement of link vehicle count and travel time in a road network,”
IEEE Trans. Intell. Transp. Syst., vol. 11, no. 4, pp. 814–825, Dec. 2010.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
10
[8] J. Zhou, C. L. P. Chen, L. Chen, and W. Zhao, “A user-customizable
urban traffic information collection method based on wireless sen-
sor networks,” IEEE Trans. Intell. Transp. Syst., vol. 14, no. 3,
pp. 1119–1128, Sep. 2013.
[9] T. T. Yang, Z. M. Zheng, H. Liang, R. L. Deng, N. Cheng, and
X. M. Shen, “Green energy and content-aware data transmissions in
maritime wireless communication networks,” IEEE Trans. Intell. Transp.
Syst., vol. 16, no. 2, pp. 751–762, Apr. 2015.
[10] A. Wheeler, “Commercial applications of wireless sensor networks using
ZigBee,” IEEE Commun. Mag., vol. 45, no. 4, pp. 70–77, Apr. 2007.
[11] J. Zheng and M. J. Lee, “Will IEEE 802.15.4 make ubiquitous net-
working a reality?: A discussion on a potential low power, low bit rate
standard,” IEEE Commun. Mag., vol. 42, no. 6, pp. 140–146, Jun. 2004.
[12] J.-S. Lee, Y.-W. Su, and C.-C. Shen, “A comparative study of wireless
protocols: Bluetooth, UWB, ZigBee, and Wi-Fi,” Proc. 33rd Annu.
Conf. IEEE Ind. Electron. Soc. (IECON), Taipei, Taiwan, Nov. 2007,
pp. 46–51.
[13] J.-S. Lee, “Performance evaluation of IEEE 802.15.4 for low-rate wire-
less personal area networks,” IEEE Trans. Consum. Electron., vol. 52,
no. 3, pp. 742–749, Aug. 2006.
[14] K. E. Browne, C.-C. Chen, and J. L. Volakis, “A novel radiator for a
2.4 GHz wireless unit to monitor rail stress and strain,” IEEE Trans.
Antennas Propag., vol. 56, no. 3, pp. 887–892, Mar. 2008.
[15] S. H. Kim, I. Stern, J. N. Shen, M. Ahad, and Z. Tucker, “Developing
an energy harvesting system using piezoelectric materials for high-
way sustainability,” in Proc. 95th Annu. Meeting Transp. Res. Board,
Washington, DC, USA, Jan. 2016, pp. 6506-1–6506-14.
[16] L. K. Guo and Q. Lu, “Potentials of piezoelectric and thermoelec-
tric methods on energy harvesting pavements,” in Proc. 95th Annu.
Meeting Transp. Res. Board, Washington, DC, USA, Jan. 2016,
pp. 5398-1–5398-18.
[17] A. Chiarelli, A. Dawson, and A. Garcia, “Analysis of the performance
of an air-powered energy-harvesting pavement,” Transp. Res. Rec.,
J. Transp. Res. Board, vol. 2523, no. 2523, pp. 156–163, 2015.
[18] T. Yuan, J. Yang, R. Song, and X. Liu, “Vibration energy harvesting
system for railroad safety based on running vehicles,” Smart Mater.
Struct., vol. 23, no. 12, pp. 1–14, 2014.
[19] M. Wischke, M. Masur, M. Kröner, and P. Woias, “Vibration harvesting
in traffic tunnels to power wireless sensor nodes,” Smart Mater. Struct.,
vol. 20, no. 8, p. 085014, Aug. 2011.
[20] C. A. Nelson, S. R. Platt, D. Albrecht, V. Kamarajugadda, and M. Fateh,
“Power harvesting for railroad track health monitoring using piezo-
electric and inductive devices,” Proc. SPIE, vol. 6928, p. 69280R,
Apr. 2008.
[21] J. J. Wang, G. P. Penamalli, and L. Zuo, “Electromagnetic energy
harvesting from train induced railway track vibrations,” in Proc.
IEEE/ASME Int. Conf. Mechatronics Embedded Syst. Appl. (MESA),
Suzhou, China, Jul. 2012, pp. 29–34.
[22] G. De Pasquale, A. Somà, and N. Zampieri, “Design, simulation, and
testing of energy harvesters with magnetic suspensions for the generation
of electricity from freight train vibrations,” ASME J. Comput. Nonlinear
Dyn., vol. 7, no. 4, pp. 614–620, 2012.
[23] M. Y. Gao, P. Wang, Y. Cao, R. Chen, and D. J. Cai, “Design and
verification of a rail-borne energy harvester for powering wireless sensor
networks in the railway industry,” IEEE Trans. Intell. Transp. Syst.,
vol. 18, no. 6, pp. 1596–1609, Jun. 2016.
[24] M. Y. Gao, P. Wang, Y. Cao, R. Chen, and C. Liu, “A rail-borne
piezoelectric transducer for energy harvesting of railway vibration,”
J. Vibroeng., vol. 18, no. 7, pp. 4647–4663, 2016.
[25] R. Bischoff, J. Meyer, O. Enochsson, G. Feltrin, and L. Elfgren,
“Event based strain monitoring on a railway bridge with a wireless
sensor network,” in Proc. 4th Int. Conf. Struct. Health Monitor. Intell.
Infrastruct., Zurich, Switzerland, Jul. 2009, pp. 1–8.
[26] K. Sekuła and P. Kołakowski, “Piezo-based weigh-in-motion system for
the railway transport,” Struct. Control Health Monitor., vol. 19, no. 2,
pp. 199–215, Mar. 2012.
[27] M. Gruden, A. Westman, J. Platbardis, P. Hallbjorner, and A. Rydberg,
“Reliability experiments for wireless sensor networks in train environ-
ment,” in Proc. Eur. Wireless Technol. Conf., Sep. 2009, pp. 37–40.
[28] Wikipedia. (2017). Urban Rail Transit in China. [Online]. Available:
https://en.wikipedia.org/wiki/Urban_rail_transit_in_China
[29] Texas Instruments. (2016). [Online]. Available: http://www.ti.com/lsds/
ti/wireless_connectivity/zigbee/tech_docs.page
[30] J. Polastre et al., “A unifying link abstraction for wireless sensor
networks,” in Proc. Int. Conf. Embedded Netw. Sensor Syst., Nov. 2005,
pp. 76–89.
IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS
[31] (2016). MEMS Accelerometer. [Online]. Available: http://www.colibrys.
com/
[32] (2016). Micro Displacement Sensors. [Online]. Available: https://www.
omron.com/ecb/products/photo/6/z4d-c1.html
[33] HBM. (2016). Rail Transport. [Online]. Available: https://www.
hbm.com/en/0469/rail-transport/
[34] Test FC: Vibration (Sinusoidal), document IEC 60068-2-6, 2007.
[35] Wikipedia. (2017). High Speed Railway in China. [Online]. Available:
https://en.wikipedia.org/wiki/China_Railway_High-speed
[36] P. Wang, Y. F. Wang, M. Y. Gao, Y. Cao, and Y. Yao, “Investigation on
the train aerodynamic performance and energy harvesting with track-
borne wind turbine,” in Proc. 96th Annu. Meeting Transp. Res. Board,
Washington, DC, USA, Jan. 2017, pp. 299-1–299-13.
Mingyuan Gao (S’09–M’10) received the M.S.
degree in mechanical manufacturing and automation
from Huazhong University of Science and Technol-
ogy, Wuhan, China, in 2010. From 2010 to 2011, he
was a Research Assistance with The University of
Tokyo, Japan. In 2012, he was a Research and Devel-
opment Engineer with the IPC Research Center,
Karlsruhe, Germany. His research interests include
railway condition monitoring, intelligent transporta-
tion system, train aerodynamics, and energy har-
vesting. He received the GE Foundation TECH
Award (First Place) in 2009 and 2010.
Ping Wang received the Ph.D. degree in high-
way and railway engineering from Southwest
Jiaotong University, Chengdu, China. He designed
and developed high-speed track turnout with a speed
of 350 km/h in China and resolved the techni-
cal challenge of laying track turnout over bridge.
He is currently a Professor with the School of Civil
Engineering, Southwest Jiaotong University, and the
Head of the Key Laboratory of High-Speed Rail-
way Engineering, Ministry of Education, China. His
research interests include the theory and dynamics
of high-speed track turnout and the track structure of high-speed railway.
Yifeng Wang was born in Heze, Shandong.
He received the B.S. degree in highway and railway
engineering from Southwest Jiaotong University,
Chengdu, China. He is currently working toward the
master’s degree. His research interests include intel-
ligent transportation system and structural health
monitoring.
Lingkan Yao received the Ph.D. degree in high-
way and railway engineering from Southwest
Jiaotong University, Chengdu, China. He is a Pro-
fessor with the School of Civil Engineering, South-
west Jiaotong University, and the Key Laboratory
of High-Speed Railway Engineering, Ministry of
Education, China. He has authored or co-authored
three books and over 150 articles. His research inter-
ests include the disaster prevention and maintenance
strategy of high-speed, freight, and urban railway
system. He was a recipient of the National Award
for Progress in Science and Technology in 2009.