9th European Workshop on Structural Health Monitoring

July 10-13, 2018, Manchester, United Kingdom

Low-cost and High-performance Wireless Sensor Nodes for

Structural Health Monitoring Applications

Niusha Navabian1 and Sherif Beskhyroun 2

More info about this article: http://www.ndt.net/?id=23262

1 PhD Candidate, New Zealand, niusha.navabian@aut.ac.nz

2 Senior Lecturer, New Zealand, sherif.beskhyroun@aut.ac.nz

Abstract

Although damages throughout structures are considered as a main cause of structural

collapse, many civil engineering structures keep to be used despite associated possible
damage accumulation. Regular condition Monitoring of important civil structures such
as bridges and buildings is crucially significant to provide the community with an
adequate level of safety for the structures they use every day. Moreover, the need for
quick damage assessment of structures after severe events such as earthquakes is
essential to avoid unnecessary disorder and reduce economical and downtime impacts.
Regarding these important needs, the use of sensor technology integrated within
structures could extremely increase the inspection procedure through rapid in-situ data
collection and processing. However, utilisation of wired sensory system could increase
the installation and maintenance costs of structural health monitoring systems especially
for large civil infrastructures.
In the past decades, improvement in sensor technology provides researchers with great
opportunities to develop wireless sensor nodes for a reliable Structural Health
Monitoring (SHM). However, due to some drawbacks of available wireless sensory
systems for SHM applications, two series of wireless sensor nodes are designed in this
study to collect structural vibrations from different types of large-scale civil structures.
The first series of the sensors is a low-cost wireless sensor node and has designed to
collect large-amplitude earthquake-induced vibrations from affected structures
immediately after earthquakes. The second series of wireless sensor nodes is high-
performance and highly sensitive sensor which is designed to record very low-
amplitude ambient vibrations from different large and complex civil structures, such as
long-span bridges and high-rise buildings. To show the performance of wireless sensors
to measure structural vibrations, some field testing are carried out. To this end, several
experimental setups are performed on a footbridge and two high-rise office buildings
using both versions of wireless sensor nodes and an expensive wired sensory system as
the reference system. Based on the results obtained, it can be concluded that both
versions of wireless sensor nodes are competitive in cost and accuracy with the
commercially available wired sensory systems. The accuracy of the low-cost sensor
node matches very well with the expensive reference wired system and the resolution
and sensitivity of the second version of wireless sensor node is highly promising and
more accurate than the reference system to measure very low-amplitude structural
vibrations from very large and stiff civil infrastructures.
1. Introduction

In today’s world, the ability to regularly monitor the structural integrity is becoming
considerably significant from both economic and life-safety viewpoints. Moreover, the
need for quick damage assessment of structures after severe events such as earthquakes
and severe winds is essential to avoid unnecessary disorder and reduce economical and
downtime impacts. There are many techniques in the literature, which have been used
recently for structural assessment and damage identification. Among these methods,
vibration-based methods are among most widely-used techniques by researchers to
assess the integrity of civil structures [1-3]. To use such valuable methods for reliable
and accurate structural assessment, the dynamic characteristics of structures such as
natural frequencies, mode shapes, and modal damping need to be precisely measured.
To accomplish this task, ambient vibration testing, forced vibration testing and
earthquake response measurement can be carried out.
Utilisation of sensor technology integrated within structures could extremely increase
the efficiency of inspection procedure through rapid in-situ data collection and
processing. However, traditional sensory systems are wire-based connecting the sensors
to a central server by running wires throughout the structure. Such system usually incurs
high installations costs and long setup times especially for large civil infrastructures. An
example of such monitoring system is a wired monitoring system installed on the Tsing
Ma suspension bridge in Hong Kong, which has a cost of $8 million ($27,000 per
sensing channel) [4]. Due to high cost and complexity of traditional wired sensory
systems, the improvement in Micro-Electro-Mechanical System (MEMS) provides
engineers and researchers with great opportunities to develop sensor nodes with sensing
capabilities, wireless communication and data processing options for reliable SHM
applications. This improvement could provide a successful implementation of large
sensing networks throughout the large civil structures by limiting the high cost of
installing and maintaining the extensive lengths of wiring [5, 6]. In the literature, there
are several research for wireless sensor development [7-9]. However, several issues
were highlighted for vibration monitoring of large-scale civil infrastructures. These
issues include sensing instability due to unregulated battery voltage, the lack of
accuracy and sensitivity for micro-vibration measurements, difficulty in managing large
volume of monitoring data, and restrictions in high-speed data sampling [10, 11].
In order to address the limitations associated with the current wireless sensors
developed for SHM applications, the contribution of this paper is to develop two series
of wireless sensor nodes, including low-cost and high-performance sensors, to collect
earthquake-induced vibrations and low-amplitude ambient vibrations from different
types of civil structures. The first series of sensors is a low-cost wireless sensor node
and has been designed to collect large-amplitude earthquake-induced vibrations from
affected structures after earthquake. The second series of nodes is a high-performance
and highly sensitive sensor which is designed to record very low-amplitude ambient
vibrations from large civil structures such as long-span bridges and high-rise buildings
for continuous monitoring. To show the performance and accuracy of the wireless
sensors to measure structural vibrations, several experimental setups are performed on a
footbridge and two high-rise buildings using both versions of wireless sensor nodes and
an expensive wired sensory system. The results of the tests show that both versions of
wireless sensor nodes are competitive with the commercially available wired sensory
system in accuracy and cost. Moreover, it can be seen that the precision of both series of

2
sensors is highly promising to measure large-amplitude and very low-amplitude
structural vibrations from large and stiff civil infrastructures.

2. Development of Wireless Sensor Nodes

In this section, the component and layout of the wireless sensor boards designed in this
study are presented. As mentioned before the first version is a low-cost wireless sensor
to measure high-amplitude earthquake-induced vibrations immediately after
earthquakes. The second version is a high-performance and highly sensitive wireless
sensor to collect very low-amplitude ambient vibrations from large civil structures. The
initial circuit design of the first version of wireless sensor node (WSN-V1) is shown in
Figure 1.

Figure 1. Block diagram of WSN-V1 for earthquake-induced vibration measurement.

As is obvious in figure above, the components of the first version of wireless sensor
node are: 1) a pico-power microcontroller, 2) a wireless transceiver
(transceiver/receiver) module, 3) a humidity and temperature sensor, 4) an external
memory for data storage, 5) a real-time clock (RTC) for time synchronization between
nodes, 6) a 3-axis digital-output accelerometer with an integrated 10-bit Analogue-to-
Digital Converter (ADC) and digital filter, and 7) a battery power supply circuit.
An Atmel ATmega Series Microcontroller was selected as the key component of the
node because of its low cost and low power consumption in active and sleep modes.
Also, the selected chip has user-selectable clock frequency ranging from 4 MHz to 20
MHz which is adequate for SHM applications. This MCU could provide a great trade-
off between accuracy, cost and power consumption, which is the main goal of the board
design. For short-range wireless communication, a standard low-power wireless
communication format of IEEE802.15.4 called ZigBee was used. This protocol is a low-
cost and easy-to-employ module, which could provide quick and seamless
communication between nodes. Although ZigBee has smaller capacity for data
transmission than that of Bluetooth, it could offer longer distance for wireless
transmission, multi-hoping functions for transmitting data, mesh network topology, and
more flexibility in network configuration [7]. For environmental parameters, a fully
calibrated humidity and temperature sensor with low power consumption and high
accuracy was selected. In addition, because the memory capacity of the selected
microcontroller is not enough for high-volume vibration measurements, a 2-Mbit I2C-

3
compatibale EEPROM (Electrically Erasable PROgrammable Memory) was chosen as
an external memory for the temporal data storage. The most important component of the
sensor board is a 3-axis digital-output accelerometer with an integrated ADC chip with
10-bit resolution measurement up to ±4 g. This chip was selected due to its low cost and
relatively high resolution, enough to collect high-amplitude vibrations induced by small
to large earthquakes. The noise density of the accelerometer in x and y axes is 150
µg/√Hz and in z axis is 250 µg/√Hz, which could provide enough accuracy for high-
amplitude vibration measurement. A is obvious in Figure 1, this accelerometer has a
built-in digital filter to improve the measurement resolution. Low cost and low power
consumption features of this board allow us to install dense array of sensors throughout
large structures to collect more data from different parts of structures to precisely
monitor the condition of the whole structure after earthquakes. The WSN-V1 will be
configured in event-triggered wakeup mode to save power for long-term monitoring
applications. In this mode, the sensor nodes will be put into sleep mode and once an
application-specified event like earthquake happens, the nodes will be awakened to
record data associated with the event. The data collected during the events are really
informative to detect any possible changes in structures as a result of earthquakes [12].

Figure 2. Block diagram of WSN-V2 for low-amplitude ambient vibration measurement.

The initial design of the second version of wireless sensor node (WSN-V2) is shown in
Figure 2. The components of this sensor board are similar to the first series of wireless
sensor node except the sensing component and the memory chip. This version of
wireless sensor node will be set into time-triggered or periodic sampling mode
according to which they will wake up every day for a user-specified interval to record
ambient vibrations for continuous and long-term monitoring. Therefore, they will create
large sets of measurement which need to be stored in a temporary storage space before
transmission. A 64-Mbit external flash memory was integrated into the board design to
address this issue. In addition, due to the fact that the vibrations caused by traffic, wind
and other environmental (ambient) factors are very low-amplitude, a low-noise density
and low-drift MEMS accelerometer is required. To this end, a 3-axis digital-output
MEMS accelerometer was selected to satisfy the requirements for low-amplitude
vibration measurements. This accelerometer supports ±2.048 g, ±4.096 g, and ±8.192 g
ranges with minimal offset drift over temperature and long term stability enabling
precision applications with minimal calibration. This low-power and low-cost chip has
an ultralow noise density of 25 µg/√Hz in 3 axes, which is especially ideal for all

4
wireless product designs. As can be seen in Figure 2, the accelerometer uses an
analogue, low-pass, antialiasing filter to reduce out of the band noise to limit bandwidth
for a high-resolution signal. Further digital filtering option is integrated to the
accelerometer design to maintain excellent noise performance at various bandwidths.
Moreover, it provides an internal 20-bit ADC to digitize the filtered analogue signals.
Thanks to high resolution and performance of the accelerometer, it was selected for the
design of the high-performance series of wireless sensor nodes. More details of the
components selected for the two versions of wireless sensor nodes are presented in
Table 1.

Table 1. Specifications of the sensor board components.

Components Manufacturer Active Current sleep Current
(mA) (µA)
Low-power microcontroller Atmel Inc. 0.2 0.1

Receiver-Transmitter module Digi International Inc. TX: 33 <1

RX: 28
Real-Time Clock Microchip Inc. 1.2×10-3 < 0.2
EEPROM memory STMicro. Inc. 2 5
Temperature sensor Amphenol Inc. 0.75 0.6
Digital accelerometer Analog Devices Inc. 0.17 0.1
Total power requirements for the WSN-V1: 36 mA 7 µA

Low-noise digital accelerometer Analog Devices Inc. 0.2 21

Flash memory Cypress Inc. 10 20
Total power requirements for the WSN-V2: 45 mA 43 µA

It is obvious from the table that the power requirements for the WSN-V1 in active and
sleep modes are less than those for the WSN-V2 allowing dense array instrumentation
of the first series of sensors throughout large infrastructures. However, the WSN-V2 has
higher resolution and accuracy for low-amplitude ambient vibration measurement at the
expense of higher power consumption in active and sleep modes. The software of sensor
nodes is based on Atmel Studio 7, an Integrated Development Environment (IDE) for
developing and debugging all AVR and Atmel applications. It has an easy-to-use
environment to write, built, and debug the applications in C/C++ or assembly
languages. It should be noted that the software part of the sensor nodes is not completed
yet and needs more development and debugging for long-term monitoring applications.
The top and bottom views of initial prototypes of WSN-V1 and WSN-V2 are depicted
in Figures 3 and 4, respectively.

5
 (a) (b)
Figure 3. a) Top side, and b) Bottom side of WSN-V1 (Low-cost sensor).

(a) (b)
Figure 4. a) Top side, and b) Bottom side of WSN-V2 (high-performance sensor).

3. Field testing on Large-scale Structures

To show the performance and accuracy of the wireless sensors to measure various levels
of structural vibrations, some preliminary full-scale testing have been performed. To
this end, several experimental setups were carried out on a pedestrian footbridge and
two high-rise office buildings. The sensory systems used for the tests include WSN-V1,
WSN-V2 and a high cost wired sensory system as the reference system. The wireless
sensors associated with their battery packs and the wired sensory system are shown in
Figures 5 and 6, respectively.

(a) (b) (c)

Figure 5. Wireless sensory system: a) WSN-V1 with battery pack, b) WSN-V2 with battery pack,
c) gateway node.

6
 (a) (b) (c)
Figure 6. a) Wired sensory system, b) NI DAQ system, c) GP Series wired accelerometer.

As is obvious in Figures 5 (a) and (b), current prototypes of WSN-V1 and WSN-V2 use
three alkaline 1.5V AA batteries and three alkaline 1.5V AAA batteries respectively as
the power source. The XBee STM Grove Development Board compatible with XBee
PRO S2C 802.15.4 RF module is used as the gateway node for both series of wireless
sensor nodes (Figure 5 (c)). This board is a simple unit, which could connect any
XBee/XBee PRO device to a PC, to evaluate the performance of the XBee device. The
wired sensory system is shown in Figure 6. The system includes NI cDAQ-9184, 4-Slot
Ethernet National Instrument CompactDAQ Chassis, and a high-performance 1-axis
analogue-output accelerometer from Crossbow GP Series. The total cost of the wire-
based sensory system including sensing devices, acquisition system and cables for data
transmission is around 4,000 NZD for one accelerometer, while the cost of the wireless
sensory systems including sensing components, acquisition and transmission systems is
around 70 NZD and 150 NZD per node respectively for WSN-V1 and WSN-V2.

4. Analysis and Test Results

4.1 Footbridge Tests
The first setup, including ambient and free vibration testing, was performed on the
Wellesley footbridge with an approximate length of 15 m. The footbridge is located in
Wellesley Street East, Auckland, New Zealand, which connects AUT University to the
Albert Park. Two wireless sensor nodes with corresponding gateway nodes and one
wired accelerometer connected to the NI DAQ system were utilised for the tests. Both
the wireless and wired sensory systems were connected to a Laptop computer as the
analysis platform. For the first test, WSN-V1, WSN-V2 and the wired accelerometer
were aligned and installed on the mid-span of the footbridge to measure the vertical
ambient vibrations induced by traffic and people on the structure during peak hours.
Free vibration testing was also carried out on the bridge to compare the free vibration
structural responses and the acceleration amplitudes recorded by sensory systems. For
this test, jumping on the bridge at several intervals was performed and considered as the
excitation source. Figure 7 shows the bridge and the test setup on the structure.

(a) (b)
Figure 7. a) Wellesley footbridge, b) test setup for the bridge test.

7
Power Spectral Density (PSD) values of acceleration data recorded on the bridge by
wired sensory system, WSN-V1 and WSN-V2 are shown in Figure 8 for ambient
vibration testing. The PSD values of acceleration were obtained using the vibration data
recorded in vertical direction, i.e. channel 1 of the wired sensor and channel 3 of the
wireless sensors. As is clear from the figures, both wireless sensor nodes could measure
the first two vertical natural modes of the bridge with magnitudes of around 3.1 Hz and
7.2 Hz quite similar to those of measured by the reference wired sensory system. It
should be mentioned that for all vibration tests performed in this study, the data was
sampled at frequency of 25, 62.5 and 100 Hz respectively by WSN-V1, WSN-V2, and
the wired sensory system during sampling period of 300 seconds.

fs1 = 3.174 Hz fs1 = 3.223 Hz fs1 = 3.174 Hz

fs2 = 7.227 Hz fs2 = 7.373 Hz fs2 = 7.202 Hz
Figure 8. PSD values of vertical acceleration due to ambient vibration testing on the footbridge.

fs1 = 3.174 Hz fs1 = 3.223 Hz fs1 = 3.204 Hz

fs2 = 7.275 Hz fs2 = 7.275 Hz fs2 = 7.324 Hz
Figure 9. PSD values of vertical acceleration due to free vibration testing on the footbridge.

Figure 9 shows the PSD values obtained from free vibration testing. Similar structural
peaks for the first two natural modes were measured using the vibrations induced by
jumping on the bridge. As is obvious, the frequency responses of the bridge measured
by wireless sensors in both ambient vibration and free vibration testing are quite
constant. This consistency indicates the performance and accuracy of the sensors to
record structural vibrations for a constant and precise estimation of modal
characteristics under various excitation sources.

8
 Figure 10. Time-series vertical acceleration due to free vibration testing on the footbridge.

Figure 10 shows time-series responses of the bridge in free vibration testing to compare
the acceleration amplitudes recorded by two wireless sensors and the reference wired
sensor due to the jumps on the footbridge. As is obvious, the acceleration amplitude
recorded by the wired and wireless sensors ranges from -0.036 g to 0.038 g that match
very well. It shows the performance and accuracy of the wireless sensors for structural
vibration measurement. The free vibration structural responses recorded on the bridge
due to the first jump are also depicted in Figure 11. As is clear, the free structural
response recorded by WSN-V2 is very clean and without measurement noise in
comparison to WSN-V1 and the reference wired sensor showing high resolution and
accuracy of WSN-V2 for structural vibration measurement. It should be noted that such
noise-free and clean free vibration structural response is very hard to measure during
large-scale field testing.

Figure 11. Free vibration structural response due to the first jump on the footbridge.

4.2 Building Tests

In this part, ambient vibrations induced by wind and operational uses in Buildings WZ
and WG of AUT University are measured to show the performance of WSN-V2 for
measuring very low-amplitude vibrations from large and stiff civil infrastructures. To
do, the 12-storey WG building as high-rise building and 6-storey WS building as more
stiff building were instrumented with the three sensory systems. The sensors were
installed on floor 11 of WG Building and floor 6 of WS Building and the data was
recorded between 12-13 pm, the time with peak operational uses of the building and
traffics on the street as the excitation sources. The time-series acceleration recorded in
longitudinal direction of WG Building and the corresponding frequency-domain
responses are shown in Figures 12 and 13, respectively.

9
 Figure 12. Longitudinal acceleration recorded due to operational loading from WG Building.

fs = N.A fs = N.A fs = 1.709 Hz

Figure 13. PSD values of longitudinal acceleration recorded due to operational loading from WG
Building.

As is obvious from time-domain ambient vibrations in Figure 12, the noise level of
WSN-V2 is very small in comparison to that of WSN-V1 and the reference wired
sensor. Also, PSD values of acceleration recorded in WG Building by WSN-V2 shows a
peak of 1.709 Hz, the first longitudinal vibration mode of the building due to
operational uses. The time-series acceleration recorded in transverse direction of WS
Building and the corresponding frequency-domain responses are also shown in Figures
14 and 15. Similar result can be observed for the stiff building. As is clear in Figure 15,
there is a structural peak of 2.625 Hz in PSD values of acceleration recorded by WSN-
V2 as the first transverse vibration mode of the building under operational loading. As
is obvious from the results, WSN-V1 and the wired sensor could not measure any
structural peaks for the buildings, however those ones measured by WSN-V2 indicate
its sensitivity and performance to precisely measure very low-amplitude ambient
vibrations from large and stiff civil infrastructures.

Figure 14. Transverse acceleration recorded due to operational loading from WS Building.

10
 fs = N.A fs = N.A fs = 2.625
Figure 15. PSD of transverse acceleration recorded due to operational loading from WS Building.

To show the consistency of the results obtained for WSN-V2, another sets of data were
recorded from the buildings at a different time interval to measure the vibrations
induced by wind loading. To this end, the buildings were instrumented by WSN-V2
between 22-23 pm on Saturday night with minimum operational uses (no people and no
traffic) and maximum level of wind loading. The PSD values of acceleration recorded
by WSN-V2 in x and y directions from WG and WS Buildings are presented in Figure
16. As is obvious, the sensor could measure the first longitudinal and transverse natural
frequencies of WG building due to wind loading with magnitude of 1.709 Hz. However,
the first longitudinal natural frequency of 2.22 Hz was only measured for WS Building.
Due to the fact that the building is stiff in transverse direction, no structural peak was
measured under low-level excitation of wind loading. It can be observed that the natural
frequencies measured by WSN-V2 under operational and wind loading are similar and
consistent that shows the performance of the sensor to precisely record very low-
amplitude ambient vibrations under different excitation sources.

WG Building-Longitudinal direction WG Building-Transverse direction

fs = 1.709 Hz fs = 1.709 Hz

WS Building-Longitudinal direction WS Building-Transverse direction

fs = 2.228 Hz fs = N.A
Figure 16. PSD of acceleration recorded by WSN-V2 due to wind loading form buildings.

11
5. Conclusions
In this study, two series of wireless sensor nodes have been developed to measure
different levels of structural vibrations from large-scale civil infrastructures. The first
series of sensors is a low-cost wireless sensor to measure large-amplitude earthquake-
induced vibrations from affected structures after earthquakes. The second series of
wireless sensors is a high-performance and highly sensitive sensor to record very low-
amplitude ambient vibrations from large civil structures. To show the performance of
the wireless sensors, some field testing are carried out. To do, different experimental
setups were performed on a footbridge and two large office buildings using both
versions of wireless sensors and an expensive wired sensory system as the reference
system. The results obtained from various setups show that the accuracy of first version
of wireless sensors is quite promising for large-amplitude vibration measurements and
comparable in cost with available costly wired sensory systems. In addition, it was
observed that the second version of wireless sensors is highly sensitive and precise to
measure very low-amplitude structural vibrations for accurate estimation of modal
parameters for large and stiff civil structures. For more precise field testing, large
numbers of wireless sensors will be installed on a full-scale concrete bridge in Auckland
in an earthquake prone zone to evaluate the capability of the sensory network for
vibration-based condition monitoring of large-scale civil infrastructures.
Acknowledgements
The financial support of the New Zealand Earthquake Commission (EQC) is gratefully
acknowledged. The authors would like to especially thank Justin Matulich for his great
assistance in designing and manufacturing the wireless sensor boards.
Reference
1. MS Huang, M Gül, and HP Zhu, "Vibration-Based Structural Damage Identification under
Varying Temperature Effects", Journal of Aerospace Engineering. 31(3), pp 04018014, 2018.
2. N Navabian, M Bozorgnasab, R Taghipour, and O Yazdanpanah, "Damage identification in
plate-like structure using mode shape derivatives", Archive of Applied Mechanics. 86(5), pp
819-830, 2016.
3. S Beskhyroun, T Oshima, S Mikami, and Y Miyamori, "Assessment of vibration-based damage
identification techniques using localized excitation source", Journal of civil structural health
monitoring. 3(3), pp 207-223, 2013.
4. C Farrar, "Historical overview of structural health monitoring", Lecture notes on structural
health monitoring using statistical pattern recognition, 2001.
5. M Çelebi, "Seismic instrumentation of buildings", US Department of the Interior, US Geological
Survey, 2000.
6. SN Pakzad, GL Fenves, S Kim, and DE Culler, "Design and implementation of scalable wireless
sensor network for structural monitoring", Journal of infra. systems. 14(1), pp 89-101, 2008.
7. MJ Chae, HS Yoo, JY Kim, and MY Cho, "Development of a wireless sensor network system
for suspension bridge health monitoring", Automation in Construction. 21, pp 237-252, 2012.
8. A Sabato, MQ Feng, Y Fukuda, DL Carni, and G Fortino, "A novel wireless accelerometer
board for measuring low-frequency and low-amplitude structural vibration", IEEE Sensors
Journal. 16(9), pp 2942-2949, 2016.
9. MD Kohler, S Hao, N Mishra, R Govinda, and R Nigbor, "ShakeNet: A portable wireless sensor
network for instrumenting large civil structures", 2015.
10. JP Lynch, AS Kiremidjian, KH LLaw, T Kenny, and E Carryer,"Issues in wireless structural
damage monitoring technologies", Proc. of the 3th World Conference on Structural Control, 2,
pp. 667-672, 2002.
11. M Leopold, "Sensor network motes: Portability & performance", 2008.
12. J Cao, and X. Liu, "Wireless sensor networks for structural health monitoring", Springer, 2016.

12