IEEE Consumer Electronics Magazine - March 2018 - 26

SIGNAL PROCESSING

This module directly connects to
the broker with connection failover
and message-buffering mechanisms, preventing data loss when
connectivity issues arise on the
Internet Protocol network.

Due to the low resolution of the used MEMS, we need to
perform a signal-processing procedure to enhance the signalto-noise ratio of the acquired signal. In particular, we used
finite impulsive response filters to filter out the zero-g bias
introduced by the MEMS to limit the bandwidth of the
acquired data in the 0-5-Hz range and decrease the electrostatic noise. Hence, we obtained only the signal variations
that carry the useful information about the natural frequencies of the structure.

DATA GATHERING
SYSTEM DEFINITION
WS nodes developed for our system have been endowed with
low-cost microelectromechanical systems (MEMS). Resolution R represents the strictness constraint of the system (see
Table 1). This requirement is strictly correlated with the
acceleration noise density (AND) and the bandwidth B
through R = AND B . AND defines the minimum detectable acceleration above the background noise within the
bandwidth of the transduced signal. Thus, based on the equation and the values in Tables 1 and 2, AND must be strictly
below 10 -3 6(m/s 2) / Hz @ .
Among all available commercial accelerometers, the
LIS344 [15] and the M4030 [16] have been considered. The
features of both LIS344 and M4030 are shown in Table 2,
and we chose to endow the WS nodes with LIS344 because
of its best quality-to-price ratio. The sensor nodes are based
on the single-board computer Galileo by Intel [17], and it is
designed to be compliant with Arduino hardware and compatible with Arduino IDE software, making it suitable for IoT
applications [3].
Accelerometer outputs are then connected to an external
three-channel 16-b analog-to-digital converter through an
interintegrated circuit interface. We adopted sampling-rate
adjustment techniques to reduce the data-sampling rate as
much as possible to save energy without affecting the resolution reported in Table 1, and temperature and humidity transducers have been connected through the analog interface to
the board.

Table 1. The requirements of the monitoring system.
Range Acceleration
(m/s2)

Resolution
(m/s2)

Bandwidth (Hz)

±20

±1−5 . 10−3

25 − 50

Data gathering involves many kinds of technologies to perform operations such as data sampling, data transmission,
and data storage. During data sampling, it is essential to synchronize over time the acceleration data of all nodes so that
the algorithms used by the SHM tools can control the health
of the building while analyzing the temporal correlation of
the data acquired in different sections of the structure.
In our system, the synchronization is jointly guaranteed by
using the internal clock of the Galileo board with the NTP.
Each node is endowed with an NTP client, and, just before
the acquisition, each node sends a timing request to the local
NTP server. To receive and store data gathered from the WS
nodes, we configured a back-end messaging service based on
single-core virtual machines provisioned with an MQTT broker (Mosquitto) [23]. The chosen communication module
then leverages a message-oriented system based on the IoT/
M2M MQTT protocol [18], which is an extremely lightweight
publish/subscribe messaging transport.
Moreover, this module directly connects to the broker with
connection failover and message-buffering mechanisms,
preventing data loss when connectivity issues arise on the
Internet Protocol network. The sensing information received
through the messaging service is stored in two different databases. In particular, a MySQL database stores part of the WS
information, while a Mongodb database stores the other part.
The MySQL database is used to store sensor identification, the
type of data acquired, and the medium access control address
of the sensor nodes. These data need to be related to also supply information for applications different from the structural
monitoring based on the context of the IoT paradigm provided.
Instead, the MongoDB nonrelational is used to store all of the
sensing information gathered from the WS nodes.
The proposed IoT-based infrastructure needs to be supported through top-end services such as monitoring, analytics,
backup, and rest. The monitoring service implements two
operations. Primarily, it offers virtual machine statistics such as

Table 2. The features of the accelerometers.
Range Acceleration (m/s2)

Sensor Nodes [V/ (m/s 2)]

AND [(m/s 2) / Hz ]

Bandwidth (Hz)
1,500
200

LIS344

±20 or ±40

0.066

5 . 10−4

M4030

±20 or ±60

0.1

5 . 10−4

26 IEEE Consumer Electronics Magazine

march 2018

Table of Contents for the Digital Edition of IEEE Consumer Electronics Magazine - March 2018