IEEE SENSORS JOURNAL, VOL. 15, NO. 4, APRIL 2015
2255
Design and Application of a VOC-Monitoring
System Based on a ZigBee Wireless
Sensor Network
Changhai Peng, Kun Qian,
Member, IEEE,
and Chenyang Wang
Abstract—
Monitoring volatile organic compound (VOC)
pollution levels in indoor environments is of great importance
for the health and comfort of individuals, especially considering
that people currently spend
>80%
of their time indoors. The
primary aim of this paper is to design a low-power ZigBee
sensor network and internode data reception control framework
to use in the real-time acquisition and communication of data
concerning air pollutant levels from VOCs. The network consists
of end device sensors with photoionization detectors, routers that
propagate the network over long distances, and a coordinator
that communicates with a computer. The design is based on the
ATmega16 microcontroller and the Atmel RF230 ZigBee module,
which are used to effectively process communication data with
low power consumption. Priority is given to power consumption
and sensing efficiency, which are achieved by incorporating
various smart tasking and power management protocols. The
measured data are displayed on a computer monitor through
a graphical user interface. The preliminary experimental results
demonstrate that the wireless sensor network system can monitor
VOC concentrations with a high level of accuracy and is thus
suitable for automated environmental monitoring. Both good
indoor air quality and energy conservation can be achieved by
integrating the VOC monitoring system proposed in this paper
with the residential integrated ventilation controller.
Index Terms—
VOCs, monitoring, ZigBee, wireless sensor
networks, photoionization detector.
I. I
NTRODUCTION
V
OLATILE organic compounds (VOCs) are emitted as
gases from certain solids and liquids. VOCs include a
variety of chemicals, some of which can cause short- and
long-term adverse health effects. Many VOCs are found at
consistently higher (up to ten times higher) concentrations
Manuscript received October 10, 2014; revised November 18, 2014;
accepted November 19, 2014. Date of publication November 24, 2014; date of
current version February 5, 2015. The work was supported in part by the Open
Project Program of the Key Laboratory of Urban and Architectural Heritage
Conservation through the Southeast University, in part by the Ministry of
Education under Grant KLUAHC1212, in part by the National Natural Science
Foundation of China under Grant 51278107, in part by the Research and
Development Program, Ministry of Housing and Urban-Rural Development,
China, under Grant 2011-K1-2, and in part by the Key Program of the
Natural Science Foundation of Jiangsu Province under Grant BK2010061.
The associate editor coordinating the review of this paper and approving it
for publication was Prof. Subhas C. Mukhopadhyay.
C. Peng and C. Wang are with the School of Architecture, South-
east University, Nanjing 210018, China (e-mail: pengchanghai@gmail.com;
wangchenyang109@foxmail.com).
K. Qian was with the School of Automation, Southeast University,
Nanjing 210018, China. He is now with the School of Architecture, Southeast
University, Nanjing 210018, China (e-mail: kqian@seu.edu.cn).
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/JSEN.2014.2374156
indoors compared to outdoors [1]. In addition, VOCs are emit-
ted by thousands of products, including paints and lacquers,
paint strippers, cleaning supplies, pesticides, building materials
and furnishings, office equipment (such as copiers, printers,
correction fluids, and carbonless copy paper), and graphics
and craft materials (including glues and adhesives, permanent
markers, and photographic solutions) [2], [3].
Long-term exposure to VOCs can cause damage to the
liver, kidneys, and central nervous system. Short-term expo-
sure to VOCs can cause eye and respiratory tract irrita-
tion, headaches, dizziness, visual disorders, fatigue, loss of
coordination, allergic skin reactions, nausea, and memory
impairment [4]–[7].
Standards have yet to be set for VOCs in non-industrial
settings [8]. The Occupational Safety and Health Adminis-
tration (OSHA) of the United States regulates formaldehyde,
which is a VOC, as a carcinogen. OSHA has adopted a
permissible exposure level (PEL) of 0.75 ppm and an action
level of 0.5 ppm. The U.S. Department of Housing and Urban
Development has established a level of 0.4 ppm for mobile
homes. Based on current information, addressing the presence
of formaldehyde if it is present at levels higher than 0.1 ppm
is advisable [2].
Data on VOCs that are normally found in low concentrations
in indoor air are highly dependent on how the VOCs are
measured. All available measurement methods are selective in
what they can accurately measure and quantify, and none are
capable of measuring all of the VOCs present in an area. For
example, benzene and toluene are measured using different
methods than those used to measure formaldehyde and other
similar compounds. A wide range of measurement methods
and analytical instruments are available, and thus, the sensi-
tivity of the measurements varies greatly and exhibits a large
selectivity and bias. For this reason, any statement about the
VOCs present in a given environment must be accompanied
by a description of how the VOCs were measured so that
the results can be correctly interpreted by a professional.
In the absence of such a description, the statement would have
limited practical meaning [9].
Compared with traditional digital data loggers that
provide monitoring only at a single point, recent advances
in low-power wireless sensor networks (WSNs) allow for the
monitoring of spatially varying phenomena, such as the VOC
concentrations in buildings [10].
These WSNs allow environmental information to be mea-
sured and collected at distributed monitoring points, which
1530-437X © 2014 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.
2256
IEEE SENSORS JOURNAL, VOL. 15, NO. 4, APRIL 2015
provides awareness of the environmental conditions that
affect the overall uptime, safety, and compliance in built
environments and enables agile and flexible monitoring and
control systems [11], [12].
ZigBee-based WSNs are an emerging technology with a
wide range of potential applications, including environmental
monitoring and measurements. ZigBee-based WSNs consist of
several nodes that are equipped with processing, communica-
tion and sensing capabilities. These nodes are smaller and less
expensive than regular sensor devices. Such networks consist
of a large number of distributed sensor nodes that organize
themselves into a multi-hop wireless network [13].
These networks collect critical data from the system or the
environment and display these data on a monitoring server.
Thus, supervisors can interpret the data and, if necessary, take
immediate action in real time. Using the collected data, the
operational team can gain a better understanding of the envi-
ronment and can thus increase efficiency or prevent accidents
while reducing the total cost of data acquisition. These benefits
provide a distinct competitive advantage in the new industrial
revolution [14], [15].
This paper proposes a ZigBee WSN-based monitoring
system for measuring VOC concentrations. First, the system
hardware, which consists of a gateway, a base station, and
sensors, is designed. Specifically, a new hardware platform
for VOC sensor nodes is developed to collect both local
and remote VOC concentration measurements. Second, system
software is designed to enable data acquisition, processing
and transmission, and energy conservation strategies for the
entire system. Third, a preliminary experiment is performed
to verify the relative accuracy of the WSN sensor. Finally,
automatic action strategies are addressed to reduce VOC levels
and conserve energy.
II. R
ELATED
W
ORK
A. VOC Monitoring System With Sensor Network
Several studies have focused on improving the technolo-
gies of wireless sensor networks and monitoring systems
to enhance environmental monitoring [16], [17]. Monitored
data that are transmitted by the wireless sensor network are
analyzed by a backend server, and the analysis results are
passed to the monitoring area if the air quality characteristics
exceed acceptable levels. In [18], an air pollution monitor-
ing system was implemented in an outdoor environment.
Several sensor devices were set up, and the sensor-monitored
data were transmitted to the server. The server de-capsulated
and sorted the data, performed calculations according to the
domain knowledge and management rules in the database, and
transmitted the data to the alert system based on the results.
The alert system sent an alarm message to warn operators
that the concentration of a given gas was over the limit at
that location. In [19], a low-cost, low-complexity, convenient
and reliable system was designed for the implementation of
a real-time indoor VOC gas sensor network. After analy-
sis and comparison with other related studies, we selected
semiconductor sensors for sensing VOCs in our system. The
disadvantages of this type of sensor are that the target gas
is non-selective and easily influenced by temperature and
humidity, and thus, the resistance sensor (RS) value of clean
air must be known prior to the measurements (it must be
measured for a period of time in normal outdoor air). In [20],
the sensor’s hardware and software were integrated and used
in tests in a real environment. Instead of simultaneously
setting up different sensors, the authors integrated the different
sensors (e.g., humidity, temperature and VOC sensors) on
the same board. The different types of monitored air quality
data were transmitted to the monitoring system for further
analysis through a wireless sensor network. In the software
implementation, hop-by-hop communication was adopted with
medium access control (MAC), and efficient routing protocols
were used to forward the data. From source to sink, this
process adopted a tree-based method to establish the routing
table. The drivers of the devices used RETOS (Resilient,
Expandable and Threaded Operating System). Three compo-
nents were included in the applications: the multi-hop relay
was responsible for transmitting data, the sensor controller
controlled the sensors and accessed the monitored data, and
the data controller interpreted the monitored data to measure
harmful gases. A wireless sensor network for monitoring the
indoor air quality (IAQ) in buildings was presented in [21].
This network was composed of various IAQ sensors and
was capable of simultaneously measuring the IAQ levels at
several locations within a building. The network was integrated
with the building’s heating, ventilation, and air-conditioning
systems to ensure acceptable air quality in interior spaces.
Using the sensor network, various indoor air pollutants (CO,
CO
2
, VOCs, and airborne particles) can be measured in
spaces that contain potential sources of pollutants. When
the measured pollutant levels were above acceptable limits,
the network notified indoor occupants using an alarm and
activated the building’s climate control equipment to exhaust
the polluted air and bring in fresh air from outside.
B. Energy Efficiency in Wireless Networks
A WSN may consist of thousands of small and fully
autonomous nodes that make measurements, perform data
processing, and communicate with each other. The nodes route
data via multiple low-energy hops to sink nodes, which act
as gateways to other networks. Network management is fully
decentralized, which eliminates the need for a fixed controller
and infrastructure and prevents a single point of failure [22].
In most envisioned applications, nodes must operate with a low
energy budget and may have to scavenge supply energy solely
from their operational environment or operate for up to several
years on small batteries [23]. Hence, energy consumption is a
key performance metric for WSN realizations.
To satisfy the energy budget, WSN nodes operate with
constrained communication and computational resources.
Thus, a typical node processes data using a Micro-Controller
Unit (MCU) with a processing speed of several million
instructions per second (MIPS) and tens of kilobytes of
program and data memories [24]. Although advances in
radio frequency (RF) circuits have been significant, the radio
transceiver has the highest power consumption in a WSN
node. The power consumption of current radios is nearly the
same in the transmission and reception modes, and energy
PENG
et al.:
DESIGN AND APPLICATION OF A VOC MONITORING SYSTEM
2257
is conserved only in sleep mode, during which the radio
circuitry is completely switched off. The nodes should come
out of sleep mode only to transmit or receive packets that
are vital for node operation, which avoids idle listening to
unnecessary traffic [13]. Therefore, a pair of nodes should
be simultaneously active. Because global synchronization is
difficult to achieve in large networks using low-power and
low-cost components [25], nodes can reduce the amount of
idle listening and energy consumption by forming locally
synchronized communication links with their neighbors.
The WSNs are usually considered to be stationary in a
manner such that the established communication links are
unaltered for the entire network lifetime. In practice, even
an immobile network exhibits dynamic behavior [26] due
to random node failures and due to a dynamic operating
environment caused by the opening and closing doors, moving
objects, changing weather conditions, and interference with
other networks, all of which affect RF propagation. In addition,
numerous envisioned WSN applications (i.e., access control,
asset tracking, and interactive games) require mobile nodes,
which drive dynamic network topologies [13], [27]. Thus,
WSN nodes should be able to recognize topology changes
and update communication links rapidly, in an energy-efficient
manner, and with minimum interruptions in application data
routing. In current WSN proposals, new neighbors are typ-
ically discovered by listening on available RF channels for
their transmissions (network scan), which last up to tens of
seconds [12], [29]. As calculated in [29], each second of a
network scan consumes energy equal to 2800 frame trans-
missions using a typical low-power transceiver. In dynamic
networks, scans can increase the node energy consumption by
one order of magnitude. Thus, we need an energy-efficient
neighbor discovery process to substitute for network scans.
WSNs are often characterized by highly dense and large-
scale deployments in resource-constrained environments. The
constraints include limited capacity for processing, storage
and, especially, energy because the sensors are primarily
powered using batteries. Battery recharging in sensor networks
is sometimes impossible because of the number of nodes
but more often for the simple reason that it is practically or
economically unattainable. It is widely accepted that energy
limitations are an inevitable obstacle in the design of WSNs
because they impose strict constraints on network operations.
Moreover, the energy consumption of sensors plays an impor-
tant role in the network lifetime and has become the dominant
performance criterion in this area. If we want the system
to operate in a satisfactory mode for as long as possible,
these energy constraints require a compromise between several
activities at both the node and network levels [30].
Several studies have focused on optimizing node energy
consumption through the use of innovative conservation tech-
niques to improve network performance and maximize sensor
lifetimes. In general, energy conservation ultimately must
find the best trade-off between different energy-consuming
activities. The literature on WSNs recognizes that the radio
is a prominent consumer of energy [31], [32].
Minimizing energy consumption is a key goal in many
multi-hop wireless networking systems, especially if the nodes
of the network are battery powered. This requirement has
become increasingly important for WSNs. WSNs differ from
other types of multi-hop wireless networks because in most
cases, the sensor data must be delivered to a single sink or
base station (BS). Clearly, one of the primary concerns is the
lifetime of the network. Although there are several definitions
of the lifetime [33], a sensor network must be considered to
be dead whenever it is no longer able to forward any data to
the BS. We use a definition in which the network lifetime
is the time span from deployment to the instant at which
the network is considered to be nonfunctional. However, the
moment at which a network can be considered nonfunctional is
application-specific, e.g., at the instant that the first sensor dies,
a percentage of sensors die, or a loss of coverage occurs [34].
To design solutions that improve energy efficiency in wire-
less networks, it is essential to first identify the main sources
of energy consumption in wireless devices and understand
how wireless protocols and operations affect the energy
demand [35].
Several papers in the literature argue that network infrastruc-
ture is often unnecessarily powered on, especially when the
number of users and their traffic loads are low [36]. This
common practice guarantees high degrees of coverage and
service availability but is responsible for a large amount of
energy consumption and could be addressed by intelligent
sleep policies. Similar considerations apply at the client side,
where cellular phones could go into sleep mode by switching
their transceivers off during inactive periods [37]. Similarly,
although at different time scales, sleep mode operation could
be enabled in WLAN (wireless local area network) personal
devices that are engaged in lightweight communications [38]
or in wireless sensors that will only be requested to propagate
information after certain triggering events [39].
A second important cause of energy consumption is rooted
in the hardware design. Hardware inefficiencies have been
noted by many authors [40]. This body of work addresses
factors such as power-hungry processors, poor power ampli-
fier designs, and inefficient heat dissipation that demands
intense cooling schemes. Unfortunately, overcoming hardware
inefficiencies involves a complete redesign of the wireless
equipment, which is costly and time-consuming and is thus
subject to practical considerations.
Although hardware inefficiencies are difficult to address,
the various layers of a networking stack of wireless devices
have different degrees of flexibility. Several studies indicate
that most of these components also lack an energy-aware
design. For example, the physical (PHY) layer is frequently
implemented with sub-optimal rate selection and transmission
power control algorithms [41]. The MAC operation uses inap-
propriate contention parameters and yields non-optimal and
avoidable collision rates or unsatisfactory and unfair transmis-
sion schedules [42]. Similarly, routing protocols for wireless
networks neither use energy as a path metric nor lever-
age the node density to minimize energy consumption [43].
In addition, coordination between the PHY, MAC and routing
layers usually does not occur [44].
These papers demonstrate the intensive efforts that have
been made toward reducing the power consumption of WSNs
2258
IEEE SENSORS JOURNAL, VOL. 15, NO. 4, APRIL 2015
TABLE I
P
ERFORMANCE OF THE
PID-AH
Fig. 1.
Schematic of the WSNs.
and improving the accuracy of gas sensors. Using state-of-the-
art technology and with the expectations of using commercial
off-the-shelf photoionization detector (PID) gas sensors, we
develop our system with PID gas sensors. We also focus on
reducing the power consumption of the gas-monitoring WSN
by managing the sleep state of the nodes. This feature is
essential for obtaining sufficiently low average power con-
sumption for the network to operate for a substantial amount
of time. A favorable WSN network should be able to operate
continuously for several months, or even a year, if it operates at
a low capture rate. We evaluated the theoretical work lifetime
of a sensor node and tested the performance of its actual power
consumption using an experimental circuit for several days as
a preliminary attempt to quantify the energy consumption of
the designed WSN node.
III. D
ESIGN OF A
VOC M
ONITORING
S
YSTEM
B
ASED ON A
Z
IG
B
EE
WSN
A. System Hardware
The monitoring system consists of a gateway, a base station,
and sensors, as shown in Figure 1. The gateway, which
is also known as the client, connects the sensor nodes to
an existing Ethernet network. The base station provides the
connection between the sensor nodes and the gateway. The
sensors monitor the VOC concentrations in the building and
transmit data to the base station.
1) Gateway:
The gateway, which is a model SQ120 in this
implementation (depicted as “Client” in Figure 1), is based on
an Intel IXP420 XScale processor running at 266 MHz and
features one wired Ethernet port and two USB 2.0 ports. The
device is also equipped with 8 MB of programmable FLASH,
32 MB of random-access memory (RAM), and a 2 GB USB
2.0 system disk. The SQ120 runs the Debian Linux operating
system, which is a full-fledged standard Linux distribution for
the ARM architecture and comes preloaded with Crossbow’s
sensor network management and data visualization software
packages XServe and MoteExplorer. These programs are auto-
matically started during the boot time of the SQ120. To set up
a sensor network gateway configuration, a base station must
be plugged into the secondary USB port of the SQ120. The
sensor network management tool (XServe) can automatically
identify the types of sensor boards that are plugged into the
Fig. 2.
Linearity of the PID-AH sensor for VOC concentrations.
nodes of the WSN and instruct MoteExplorer to display the
data [45].
2) Base Station:
The base station, which is a full function
device (FFD), consists of the mote processor/radio platforms
(XM2110) and a gateway (MIB520CB) that are connected
via a 51-pin expansion connector. Thus, the base station is
configured as a ZigBee coordinator (ZC) of the WSNs. A base
station enables the aggregation of sensor network data onto
a PC or other computer platform. In particular, the MIB520
provides a serial/USB interface for both programming and data
communications. Therefore, the base station receives the data
sent by all of the nodes in the network and sends a message
across the USB connection to the computer.
3) Sensor Node:
The PID sensor is a typical detector for
measuring VOCs and other gases in concentrations from less
than one part per billion to 10,000 parts per million (ppm).
PID sensors are efficient and inexpensive detectors for many
gas and vapor analyses. They can operate continuously and
produce instantaneous readings. For this implementation, we
use the PID-AH sensor from Apollosense Ltd. The perfor-
mance of the PID-AH is shown in Table 1. Figure 2 shows
the linearity measured by the PID-AH for VOC concentrations,
and Figure 3 shows the PID-AH response to 1 ppm of
isobutylene [46].
The sensor node is composed of a mote processor/radio
platform (M) and a sensor (S) that are connected via a
51-pin and a 3-pin expansion connector, as shown in Figure 4.
Each node is a fully integrated and rugged sensor package that
uses energy-efficient radios and sensors to ensure an extended
battery life and performance. Each node is an FFD and inte-
grates Crossbow’s IRIS/IMOTE2 family processor/radio board
and antenna, which are powered by rechargeable batteries
and are configured as ZigBee end devices (ZEDs) or ZigBee
PENG
et al.:
DESIGN AND APPLICATION OF A VOC MONITORING SYSTEM
2259
Fig. 3.
Response of the PID-AH sensor to 1 ppm of isobutylene.
consumption by the radio. For example, energy depletion is
caused by the power consumption of the circuit and of the
transmitted signal. The consumption by the circuit is greater
than the transmission power at short distances, but the signal
power becomes dominant at longer ranges [48]. Transmission
Power Control (TPC) has been investigated for enhancing
energy efficiency at the physical layer by adjusting the radio
transmission power [49]. In Cooperative Topology Control
with Adaptation (CTCA) [50], regular adjustments of the
transmission power of every node are proposed to consider the
uneven energy consumption profile of the sensors. Therefore, a
node with a greater amount of remaining energy may increase
its transmission power, which will potentially enable other
nodes to decrease their transmission power and save energy.
However, the TPC strategy affects not only the energy but
also delays, link quality, interference and connectivity [47].
Directional antennas allow signals to be sent and received in
one direction at a time, which improves the transmission range
and throughput. Directional antennas may require localization
techniques for orientation, but multiple communications can
occur in close proximity, which results in the spatial reuse
of bandwidth. In contrast to omni-directional motes, which
transmit in unwanted directions, directional antennas limit
overhearing and require less power for a given range. Thus,
these devices can improve network capacity and lifetime while
positively influencing delay and connectivity [51].
B. System Software
Fig. 4.
Circuit structure of the sensor node.
routers (ZRs). A node has a radio range of 10 to 80 m (indoors)
and 50 to 300 m (outdoors) depending on its deployment.
The nodes form a wireless mesh network, and the range of
coverage can be expanded by adding additional wireless mesh
networks. A node comes pre-programmed and configured
with Crossbow’s XMesh low-power networking protocol. The
circuit of the sensor node is shown in Figure 4.
4) ZigBee Module:
The ZigBee module is composed
of mote processor/radio platforms (XM2110) and uses the
Atmel RF230, IEEE 802.15.4 compliant, ZigBee-ready radio
frequency transceiver integrated with an Atmega16 microcon-
troller. These enhancements provide up to three times the
radio range and twice the programmable memory of previous-
generation MICA motes [45]. In the sensor node, the XM2110
connects with the sensor board via a 51-pin expansion
connector, whose structure is shown in Figure 4.
ZigBee is a wireless network protocol that is owned by
the ZigBee Alliance and is adapted from the IEEE 802.15.4
standard, which defines the media layer and objective layer.
ZigBee provides a low transmission speed at low cost, low
power consumption, and high security and supports a large
number of web node operations.
The radio module is usually the main component that
causes the depletion of a sensor node’s battery. To reduce
energy dissipation due to wireless communications, we attempt
to optimize radio parameters such as coding and modula-
tion schemes, power transmission and antenna direction [47].
Modulation optimization attempts to find the optimal
modulation parameters that result in the minimum energy
1) ZigBee Sensor Node Firmware:
The firmware on the
sensor nodes sets the sensors in a wireless mesh network and
provides the algorithms that are necessary to operate the radio
and route the messages. The firmware is programmed and
compiled on a desktop computer using the nesC programming
language and is transferred to the on-board microprocessor of
a node [52].
The firmware on the sensor nodes primarily performs the
following three tasks:
a) Initialization of each hardware device:
To prepare an
operating system for operation, each hardware device must
initialize the hardware platform and software architecture after
powering on. The initialization primarily includes the initial-
ization of each hardware module, stack, system clock, flash
memory, several nonvolatile variables, MAC layer, application
framework layer and operating system and formation of the
terminal’s MAC address [53].
b) Data acquisition and processing:
The TinyOS imple-
mentation of the ZigBee protocol stack enables the efficient
processing of events, such as sensor data acquisition, data
processing, and radio transmission. The TinyOS platform can
be configured to ensure that the node uses little energy when
no action is required by the sensor or radio [52].
To save energy, the amount of data to be delivered to the
sink must be reduced. Generally, aggregation and adaptive
sampling can be adopted jointly during the data acquisition
and processing stages. In data aggregation schemes, nodes
along a path toward the sink perform data fusion to reduce
the amount of forwarded data. For example, a node could
京公网安备 11010802033920号
Copyright © 2005-2024 EEWORLD.com.cn, Inc. All rights reserved
评论