Anda di halaman 1dari 6

A Low Cost Wireless Sensor Node for Building Monitoring

S. Valenti˚ , M. Conti˚ , P. Pierleoni˚ , L. Zappelli˚ , A. Belli˚ , F. Gara: , S. Carbonari: , and M. Regni:


˚ Department of Information Engineering, Università Politecnica delle Marche, Ancona, Italy
: Department of Civil and Building Engineering, and Architecture, Università Politecnica delle Marche, Ancona, Italy
s.valenti@univpm.it

Abstract—In this paper a medium-high performance wireless node are based on the mechanical resonance of a vibrating
sensor node based on low-cost Micro Electro-Mechanical Systems quartz element to create a periodic electric signal. As time
(MEMS) accelerometers for Structural Health Monitoring (SHM) goes by, clocks on different nodes will undergo frequency
applications is proposed. The objective of the work is the
estimation of the performances of the low cost MEMS system variations due to temperature differentials, statistical variations
for SHM with a possible application on a pervasive permanent and aging. These variations cause the drift in time of each
monitoring of buildings in seismic prone regions often affected by node with respect to the other nodes. To compensate for this
earthquakes. The work aims at improving the sensing technology drift, the system therefore requires a global time reference.
involved in the dynamic identification and SHM, with the intent to The goal of a time synchronization algorithm is to ensure, by
reduce the overall cost of acquisition’s system and to make faster
the tests procedure. Preliminary experimental tests performed adjusting local clocks to the global clock at regular intervals,
to compare the performance of the proposed system with those that logical clocks of any two nodes in the WSN are as closely
of a typical high-performance and expensive wired system, are synchronized as possible at all times.
presented and discussed. Many are the applications of WSN on SHM or earthquake
Index Terms—MEMS Accelerometer, Structural Health Mon- monitoring. Torfs et al. [4] presented a wireless system for
itoring, Smart Sensor, Wireless Sensor Network.
the hearthquake monitoring. Low power custom sensors and
readout for the monitoring of strain and accelerometers are
I. I NTRODUCTION
presented. A star topology is proposed over an 802.15.4 MAC
The estimation of a structure’s state is a challenging task. in the 900-MHz band. The paper does not face the aspect of
Structural Health Monitoring (SHM) is an engineering pro- time synchronization of the sensors required to perform modal
cess that consists of the use of sensors to extract response analysis.
parameters due to actions (both ambient, anthropic and nat- Chintalapudi et al. [5] presented a WSN for SHM using
ural actions) such as stress, strain, displacement, velocity Mica-2 or Mica-Z mote sensor platforms from Crossbow
and acceleration [1]. Its purpose is to detect, through the Technologies. It focuses on data storage in the node to face
analysis of the variation in time of above parameters, unusual the high data rate and high packet loss of the networks. No
behaviors of the structure due to damage by determining information on sensor accuracy is given. Since the application
approximately the position and extent of the problem on example uses a forced excitation, a not high accurate sensor is
the structure and establishing the corrective actions to be required. The base station calculates the time of the samples’
undertaken. Evaluations obtained through SHM are generally generation by subtracting the residence time from its local
used to predispose maintenance activities, increase safety and time. But the details are not presented and no information on
reduce uncertainty. The monitoring can be short-term or long- synchronization accuracy is given.
term and the acquisition system can makes use of a few sensors Araujo et al. [6] presented a WSN architecture applied in the
up to hundreds of them according to the demands of the monitoring of a real bridge. An accurate (less than 120 µs) time
structure to monitor. synchronization is obtained using a direct call from physical
The technology improvements of sensors, data acquisition, layer to application layer, avoiding all the intermediate stack
signal processing and big data analysis produced great ad- layers. The measured data are stored in an internal SD memory
vantages to the SHM sector. In particular, Wireless Sensors and the communication packets are sent via WiFi using its
Network (WSN) avoids the high cost of cabling required to large bandwidth. So the system is complex using at the same
set up a wired SHM network [2]. On the other hand, many are time WiFi for data transfer and Zigbee for synchronization.
the challenges to be faced in the design of a wireless SHM A survey of SHM Using WSNs is presented in [7]. Some
system: high data transmission rate, time synchronization aspects are faced: sensor type (not detailed), damage detection
among nodes, network scalability, power consumption, local and localization algorithm, network scalability, data trans-
data storage availability, protection against packet loss. mission rate, data storage availability, power consumption,
Most of recent development in the WSNs focuses on energy time-synchronization error, and processing algorithms, sensor
consumption, ad hoc wireless protocols and time synchro- placement optimization and a review of experimental work
nization algorithms. One of the main problems to be faced with real structures.
in developing wireless solutions for SHM is to synchronize The sensors used in SHM are strain, acceleration and veloc-
the whole network so that all nodes can acquire data at ity. Micro Electro-Mechanical Systems (MEMS) are typically
the same time [3]. The crystal oscillators present on each used mainly due to the low power consumption [8]. A deep

978-1-5386-6405-6/18/$31.00 ©2018 IEEE


review of accelerometer sensors used for SHM is reported ‚ Local data storage: in order to avoid a bottleneck on
in [9]. The frequencies of dominant modes are of bridges real time data transmission, the data are stored in a
and buildings typically below 10 Hz; therefore sampling local memory, and transmitted after the acquisition with
frequencies can be chosen at values upwards 100 Hz. Sample a polling scheme.
frequency, output resolution and signal to noise ratio are ‚ Low-Power: using modules with low-power consumption
characteristics that strongly affect the cost of the sensor and and a platform that supports the implementation of an
they must be carefully selected since their values affect modal external battery is fundamental in order to extend the
analysis results. SHM techniques based on ambient excitation autonomy of the sensor.
are easier to deploy but require high accuracy sensors.
A. Hardware
A permanent wired SHM of a building requires that a
relevant number of high accuracy sensors must be permanently The implemented system consists of 2 wireless sensor nodes
placed in the building with an high cost of wiring in addition with the sensing unit on board and a wireless concentrator
to the cost of the sensors. This cost makes the permanent node (border router) that acts as a network master. The
wired SHM possible only for buildings of particular regard. individual wireless sensor node prototype is mainly composed
In many other cases the monitoring is performed for a fixed of a development board which, through its interfaces, commu-
limited period of time and than the system dismounted and nicates with a radio module, a storage unit and an acceleration
placed in another building. The wireless connection reduces sensor, as shown in Fig. 1. The wireless sensor node prototype
the cost of the installation, but a permanent SHM requires consists of the following components:
a low cost sensor node [10]. The monitoring of a building
requires dozens of 2D-accelerometer sensors, and the cost of
an accurate high resolution and low noise with filtering 2D-
accelerometer sensor and 24-bit ADC is extremely high (from
2000 to 5000 USD).
In this paper a medium-high performance wireless sensor
node based on low-cost MEMS accelerometers for Structural
Health Monitoring applications, is proposed. Experimental
tests carried out on the Tower of the Engineering Faculty of
the Università Politecnica delle Marche, allowed to compare
the performance of the proposed system with those of a typical Figure 1: Block scheme of the proposed low-cost wireless
high-performance and expensive wired system. The objective accelerometric unit.
of the work is the estimation of the performances of the low
cost WSN for SHM with a possible application on a pervasive 1) Development Board: The STM32 Nucleo board (STMi-
permanent monitoring of buildings in seismic prone areas, like croelectronics) is a development platform mainly used to build
the center of Italy. prototypes with any STM32 microcontroller line, choosing
from the various combinations of performance, power con-
II. I MPLEMENTATION OF THE P ROPOSED S YSTEM
sumption and features. The board provides various hardware
A wireless sensor network for building SHM is implemented interfaces (USART, SPI, I2C, analog pins, etc.) that can be
adopting very low cost nodes with medium-high performance used for the specific application. In this study, the STM32L se-
components. The following requirements have been taken into rie development board was adopted. It is equipped with MCU
account for the node implementation: STM32L152RE, that fully reflects our requirement of low cost
‚ Low-Cost: each component of the sensor node (hardware and good performance with very low power consumption. The
and firmware) must be inexpensive and reliable at the maximum consumption of the complete board is 100 mA, in
same time. stand by the power consumption is 1.5 µA.
‚ Wireless connectivity: WSNs need smart sensor nodes 2) Wireless Shield: The X-NUCLEO-IDS01A4 (STMicro-
to monitor, acquire and transmit data through the net- electronics) and STEVAL-IKR001V8D (STMicroelectronics)
work. To date the WSN idea converges with the Internet are expansion boards of STM32 Nucleo based on the Spirit1
of Things (IoT) thanks to the high capacity of nodes RF module. Both modules interface with the STM32 micro-
allowing the implementation of a lot of protocols and controller via SPI connections and GPIO pins. Spirit1 is a low
applications inside them. power and low cost transceiver and is particularly suitable
‚ High Performance: thanks to the growing potentialities for WSN applications. In this work the chosen frequency
of actual 32 bit microcontrollers, the trend of WSNs for radio transmission is 868MHz, using GFSK modulation
evolution is moving to the integration of TCP/IP protocols and 100 kbps data rate. The open field distance covered
into embedded devices. is about 200 m for the X-NUCLEO-IDS01A4 (low power
‚ Real-time data acquisition: for a SHM system, the device RF module) and 2 km by using the STEVAL-IKR001V8D
must be able to receive orders from the data acquisition (high power RF module). This distance allows to build up
module according to predetermined commands. a wireless network with a star topology for the monitoring
of the building considered for the tests. The star topology
simplifies the architecture of the network and it avoids hop-
by-hop connection with unpredictable connection delays. The
maximum consumption in Tx mode is 21 mA (@ +11 dBm)
for the low power RF module and 350 mA (@ +27 dBm)
for the high power RF module. In this work the low power
RF module (X-NUCLEO-IDS01A4) is used as the distances
between the WSN nodes were short.
3) Sensing Unit: The EVAL-ADXL355Z (Analog Device)
is a small size board that allows evaluation of the performance
of the ADXL355 low noise, low power, 3-axis MEMS ac-
celerometer. The accelerometer communicates via I2C with
the STM32L NUCLEO board. To make the accelerometric
Figure 2: Prototype of the proposed low-cost wireless accelero-
measurements more reliable, the sensing unit was fixed to
metric unit.
a brass cylinder weighing approximately 0.5 kg. The main
features of ADXL355 are:
˝
‚ 0 g offset vs. temperature: 0.15 mg/ C maximum; and X-NUCLEO-IDS01A4). It is also composed of:
‚ Low power: 660 µW in measurement mode and 69.3 µW 6) PC application: The border router is connected via a
in standby mode; USB cable to a PC on which a simple graphical user interface
‚ Output of ˘2 g to ˘8 g full scale range (FSR); (GUI) has been developed for the management of the WSN.
?
‚ Ultra low noise density: 25 µg/ Hz; Through the management software you can view the list of
‚ 20-bit analog-to-digital converter (ADC); nodes connected to the border, send commands and messages
‚ maximum sensitivity of 3.9mg/LSB @ ˘2g; including those useful for maintaining the synchronization of
‚ Up to 4 KHz Output Data Rate (ODR); the WSN.
‚ Programmable high- and low-pass digital filters; 7) GPS Shield: The GPS Logger Shield is based around
‚ Integrated temperature sensor; a GP3906-TLP GPS which is is a POT (Patch on Top) GPS
˝ ˝
‚ Operating temperature range: -40 C to 125 C. module, specially designed for ultra low power consumption
4) Storage Module: A Micro SD Card is used for trans- purpose solutions. It provides high position and speed accuracy
ferring and storing data to and from a standard SD card. The as well as high sensitivity and tracking capabilities in urban
SPI interface allows to directly communicate with STM32L conditions. In this work, the border router uses GPS times-
Nucleo board. The consumption of the module is 40 mA in tamps to maintain global time and all other nodes synchronize
writing. In order to reduce the consumption of this module, with it.
the acquired accelerometric data have been saved in a buffer
B. Firmware
which will be written in the memory only when it is filled.
One of the main problem of a SHM WSN is the high data An operating system (OS) is essential in WSN applications.
transmission rate when the sensor is transferring data to the The proposed system is based on Contiki OS. It is open source
central node. On the other hand, the latency of a SHM system and one of the major operating systems used for connecting the
is usually not a relevant aspect. The proposed solution is the trillions of devices of the Internet of Things. Contiki allows
local storage of the samples (20 min of measurements), the to realize custom applications and is suitable for embedded
data with the local time stamp are send to the central node at devices with low computing and memory capacities thanks to
the end of the measurement phase in using a polling procedure. its highly portability, modularity and adaptability [11].
The time synchronization procedure allows a time alignment The Contiki netstack is composed by:
of all the data of all the sensors in the central node. ‚ Physical & Data Link layer: based on 802.15.4 standard.
5) Power Supply: The sensor node is powered by an CSMA protocol is used at MAC Layer and Radio Duty
external 5 Volt power supply. Being the nodes placed inside Cycle Layer is available for energy saving;
the building the power supply is not a problem. The power ‚ Adaptation layer: a 6lowPAN implementation is used to
dissipation of the complete node is in any case low, the adapt IPv6 to 802.15.4 and for encapsulation and header
maximum power dissipation is 580,7 mW: 360 mW (main compression mechanisms;
board) + 76 mW (wireless tx), 0.7 mW (sensor), 144 mW ‚ Routing: the RPL protocol allows to automatically forms
(memory). wireless IPv6 network and is used for an efficient man-
The complete prototype node is shown in Fig. 2. The total agement of routes;
cost of the prototype is approximately US$ 65,00. If compared ‚ Network layer: compatibility with the IP protocol is
with the US$ 5000 of a high accuracy sensors, the cost guaranteed thanks to the implementation of uIP. The
reduction is relevant. addressing is with IPv6 standard;
The border router is based on the same development plat- ‚ Transport layer: UDP or TCP are available as transport
form and radio module used for sensor nodes (STM32 Nucleo protocol;
‚ Application layer: COAP or HTTP are used to exchange
data through the wireless network.
Custom applications are implemented at the highest level
of firmware and refer to the underlying layers APIs. This
allows a fast prototyping of the system. The network im-
plemented consists of many sensors nodes (nucleo+wireless
transceiver+SD+sensor), a border router (nucleo+wireless
transceiver+GPS) connected to a PC that allows to send
the data in a cloud database. The network topology is a
star topology and a polling scheme is used. Once the WSN
synchronization procedure has been started, the border node
sends to each sensor node the message containing the start
and end of data acquisition time. At the end of the test each
sensor node sends the accelerometric data saved in its SD
card to the border node which saves them on its database. (a)
The time synchronization is performed at application layer,
this in general do not allow the same accuracy achievable
if performed at physical layer. Preliminary results suggest a
maximum error of 6 ms, that is acceptable considering that
the oscillation frequencies of the building under consideration
are much lower than 10 Hz.
III. E XPERIMENTAL S ETUP
In this work, the Tower of the Engineering Faculty of the
Università Politecnica delle Marche, shown in Fig. 3(a), has
been used as demo structure. The performance of the devel-
oped wireless sensor node is evaluated through comparisons
with a reference highly accurate and costly system, currently
used by researchers of the Department of Civil and Building (b)
Engineering, and Architecture of the Università Politecnica
delle Marche to perform structural monitoring applications. In Figure 3: Photo of the Tower of the Engineering Faculty of
the next few months a complete monitoring system consisting the Università Politecnica delle Marche (a) and orientation of
of 12 mono-axial sensors will be installed in different positions the three mono-axial sensors of reference system on the top
of the building, in order to monitor the structural dynamic re- floor of the tower (b).
sponse due to ambient excitations. Preliminary results reported
in this work consist in the comparison between the two SHM
systems with 3 1D accelerometers placed at the last floor of ‚ Broadband Resolution: 0.000001 g rms
the building. ‚ Measurement Range: 0.5 g pk
‚ Frequency Range: (˘5%) 0.1 to 200 Hz
A. Reference System ?
‚ Spectral Noise(1 Hz) 0.06 µg/ Hz
The reference system, installed on the Tower in August ?
‚ Spectral Noise(10 Hz) 0.01 µg/ Hz
2017, is a continuous monitoring system that uses a limited
‚ Cost about 2000 USD.
number of sensors, designed for the identification of the fun-
damental frequencies and of the relative modal damping ratios The main features of DAQ NI 9234 are:
of the tower. The system is completely wired and consists of 3 ‚ 4 channels
mono-axial piezoelectric high sensitivity accelerometers (PCB ‚ 24 bit resolution
393B31 of PCB piezotronics), 3 low-noise coaxial conductors, ‚ Sample rate: 51.2 kS/s/ch
a data acquisition unit (NI 9234 of National Instruments) and ‚ Noise at maximum sample rate: 50 µV rms
a PC. The 3 accelerometers are installed on the top floor of ‚ Dynamic range: 102 dB
the tower (195 m of altitude) as shown in Fig. 3(b). The ‚ Cost about 2500 USD.
computer, with the task of acquiring and recording the signals,
is connected to a modem and can be consulted remotely, to B. Proposed System
download the recorded data and check for any malfunctions. The proposed system includes two wireless accelerometer
The main features of the 1D accelerometer PCB 393B31 nodes and a wireless node that acts as a WSN concentrator
are: and which has the task of sending commands, maintaining the
‚ Sensitivity: (˘5%) 10.0 V/g synchronization of the WSN and identifying any malfunctions.
The two sensor nodes were positioned next to the wired
accelerometers of the reference system, so as to acquire
along the same directions: the first measured the accelerations
along the x and y axes (Fig. 4(a)), the second measured the
acceleration along the x axis (Fig. 4(b)). In the test performed,
in both nodes, the ADXL355 sensor was configured to have a
full-scale of ˘2g and an ODR of 50 Hz. The transmission
power of each node was set in such a way that the three
nodes making up the WSN were directly visible and that the
concentrator node could send broadcast messages. In order
to maintain synchronism between the nodes, the Reference
Broadcast Synchronization (RBS) was implemented. It is
receiver-receiver-based synchronization technique and aims to
provide synchronization among a set of nodes, which are
located within the single-hop broadcast range of a beacon (a)
sensor node. RBS was selected for this study because of its
low complexity, performance cited in the literature, and ex-
tensive use [3]. Once the WSN synchronization procedure has
been started, the border node periodically sends the reference
packets called beacon to the sensor nodes in broadcast mode.
The beacon packet receive event is used as a time reference
for the offset and skew estimate of neighboring nodes. When
a sensor node receives a beacon packet, it records the arrival
time by time-stamping according to its local clock. Then it
exchanges his timestamp with the rest of the nodes. Each node
uses packets containing timestamps and linear regression to
estimate offsets and drifts related to neighboring nodes.
IV. E XPERIMENTAL T EST AND R ESULTS
The reference system allows the identification of the first
(b)
three vibration frequencies and the relevant damping ratios
(as monitoring is limited to the last floor of the tower). These
are obtained through the well-known Subspace Stochastic Figure 4: Photos of the wireless sensor nodes positioned next
Identification (SSI-Cov) technique from accelerometer signals. to the reference sensors during the tests carried out on the top
Fig. 5 (a) and (b) report the stabilization diagram obtained floor of the tower. The nodes measure the accelerations along
starting from the accelerometric measurements of proposed the x and y axes in (a) and the acceleration along the x axis
WSN and of the wired high cost reference SHM system, in (b).
respectively. The peaks of the Frequency Response Function
(FRF), blue line in Fig. 5 (a) and (b), correspond to the natural Table I: Natural frequencies, damping ratios and last floor
oscillation frequencies of the buildings. Table I and Table II modal displacement of the building obtained by WSN.
report the values of the first three natural frequencies obtained f[Hz] ξ[-] 9AX 9AY 9BX
by the time measurements of the proposed WSN and of the
wired high cost reference SHM system, respectively. The error 1,00 2,33 -0,62 -0,76 1,00
between the two estimations of the frequencies are less than 1,24 2,09 1,00 0,36 0,35
7%.
1,52 7,36 -0,76 1,00 0,51
Fig. 6 shows modal displacements of the last floor of the
building relevant to the first three frequencies. Fig. 6 (a) and
Table II: Natural frequencies, damping ratios and last foor
(b) reports the results obtained by the proposed WSN and
modal displacement of the building obtained by the reference
Fig. 6(b) reports the results obtained by the wired high cost
wired system.
reference SHM system. Even this visualization evidences the
small difference between the two SHM systems. The prelimi- f[Hz] ξ[-] 9AX 9AY 9BX
nary results show that the proposed low cost wireless system
gives accurate enough results, at least in some applications. 1,07 2,55 0,21 -0,18 1,10
Further investigations will be carried out to correlate the effect 1,21 1,62 1,09 0,85 0,95
of the accuracy of the acceleration sensors and of the time
1,51 1,86 -0,95 1,14 0,65
synchronization on the modal analysis.
V. C ONCLUSION
In this work a low cost wireless sensor node for SHM
applications has been presented. We evaluated the performance
of the sensor node through comparison of results, in terms
of modal frequencies and displacements, with those of a
typical wired system. The preliminary test was carried out
on the Tower of the Engineering Faculty of the Università
Politecnica delle Marche. Preliminary results are positive.
Future investigations will be carried out with a network of
at least 12 nodes to monitor the complete building.
Given the performance of the proposed wireless sensor
node, future studies will also focus on its possible use in the
quantification of earthquakes on a local scale, in monitoring
(a) the shaking at the base of buildings and in Earthquake Early
Warning (EEW) systems.
ACKNOWLEDGMENT
The authors gratefully acknowledge support from the Sci-
entific Research Project 2016 “Structural Health Monitoring
of Constructions with Wireless Sensor Network”, funded by
the Università Politecnica delle Marche.
R EFERENCES
1 Balageas, D., Fritzen, C.-P., and Güemes, A., Structural health monitoring.
John Wiley & Sons, 2010, vol. 90.
2 Lynch, J. P. and Loh, K. J., “A summary review of wireless sensors and
sensor networks for structural health monitoring,” Shock and Vibration
Digest, vol. 38, no. 2, pp. 91–130, 2006.
(b) 3 Sivrikaya, F. and Yener, B., “Time synchronization in sensor networks: a
survey,” IEEE network, vol. 18, no. 4, pp. 45–50, 2004.
4 Torfs, T., Sterken, T., Brebels, S., Santana, J., van den Hoven, R.,
Figure 5: Stabilisation diagram: (a) wireless sensors; (b) Spiering, V., Bertsch, N., Trapani, D., and Zonta, D., “Low power wireless
reference sensors. sensor network for building monitoring,” IEEE Sensors Journal, vol. 13,
no. 3, pp. 909–915, 2013.
5 Chintalapudi, K., Fu, T., Paek, J., Kothari, N., Rangwala, S., Caffrey, J.,
Govindan, R., Johnson, E., and Masri, S., “Monitoring civil structures with
a wireless sensor network,” IEEE Internet Computing, vol. 10, no. 2, pp.
26–34, 2006.
6 Araujo, A., García-Palacios, J., Blesa, J., Tirado, F., Romero, E.,
Samartín, A., and Nieto-Taladriz, O., “Wireless measurement system for
structural health monitoring with high time-synchronization accuracy,”
IEEE Transactions on instrumentation and measurement, vol. 61, no. 3,
pp. 801–810, 2012.
7 Noel, A. B., Abdaoui, A., Elfouly, T., Ahmed, M. H., Badawy, A.,
and Shehata, M. S., “Structural health monitoring using wireless sensor
networks: A comprehensive survey,” IEEE Communications Surveys &
Tutorials, vol. 19, no. 3, pp. 1403–1423, 2017.
8 Pierleoni, P., Belli, A., Maurizi, L., Palma, L., Pernini, L., Paniccia, M.,
(a) and Valenti, S., “A wearable fall detector for elderly people based on
ahrs and barometric sensor,” IEEE Sensors Journal, vol. 16, no. 17, pp.
6733–6744, 2016.
9 Sabato, A., Niezrecki, C., and Fortino, G., “Wireless mems-based ac-
celerometer sensor boards for structural vibration monitoring: a review,”
IEEE Sensors Journal, vol. 17, no. 2, pp. 226–235, 2017.
10 Sazonov, E., Janoyan, K., and Jha, R., “Wireless intelligent sensor network
for autonomous structural health monitoring,” in Smart Structures and
Materials 2004: Smart Sensor Technology and Measurement Systems, vol.
5384. International Society for Optics and Photonics, 2004, pp. 305–315.
11 Palma, L., Pernini, L., Belli, A., Valenti, S., Maurizi, L., and Pierleoni, P.,
“Ipv6 wsn solution for integration and interoperation between smart home
(b) and aal systems,” in Sensors Applications Symposium (SAS), 2016 IEEE.
IEEE, 2016, pp. 1–5.

Figure 6: Plan view of the last floor modal displacements: (a)


wireless sensors; (b) reference sensors.

Anda mungkin juga menyukai