An Occupational Health and Safety Monitoring

System
S.A. Ngubo, C.P. Kruger, G.P. Hancke and B.J. Silva
Department of Electrical, Electronic and Computer Engineering
University of Pretoria
Pretoria, South Africa
Corresponding author: [email protected]

Abstract—Hazardous environments at the workplace are a controls, monitoring of controls, review, and adjustment or
significant contributor to injuries due to accidents as well as redesign of controls as necessary [10].
chronic diseases. There are many occupational health and safety The problem addressed in this work is the design and
(OHS) systems, but they are costly or not flexible. This paper implementation of a low-cost wireless monitoring OHS system
presents a low-cost OHS. It consists of various sensors that can be that can be deployed in the work place, more specifically
used to monitor whether a safety helmet is being worn, the within harsh environments such as in mines or sawmills. To
worker is mobile and safety boots are being worn. The system this end, we design and implement an OHS system which
interfaces with a wireless sensor network and is suitable for consists of a Wireless Smart Transducer Interface Module
operation in environments such as underground mines and
(WSTIM) that is able to transmit and receive data periodically
sawmills.
and reliably, making the system capable of monitoring health
Keywords— activity monitory; occupational health and safety; and safety in a work environment [11]. This system monitors
smart helmet; wireless sensor network. whether a worker is wearing his/her safety helmet and safety
boots, and whether he/she is immobile or not.
I. INTRODUCTION The rest of the paper is structured as follows: the basics of
Recent advances in Internet-of-Things (IoT) and Wireless OHS and similar systems in literature are presented in Section
Sensor Networks (WSNs) have enabled monitoring for various II; an overview of the proposed system is given in Section III;
application scenarios [1], [2], where deployment and the design and implementation of the proposed system are
experimental evaluation of devices is desirable [3], [4]. In discussed in Section IV; the results are presented and discussed
particular, monitoring the well-being of subjects (e.g. people or in Section V; and in Section VI, the paper is finally concluded
animals) in general is of utmost important [5], [6], and this can and some future work is discussed.
be achieved using short range technologies [7] as well as long
range technologies which enable wider coverage [8]. A II. BACKGROUND
relevant application domain within this field is occupational
This section discusses some aspects of OHS and some
health and safety (OHS). The workplace is a significant
contributor to injuries due to accidents and chronic diseases related work.
due to environmental hazards. Although current health and A. Occupational health and safety
safety regulations may be regarded as significantly progressive
Occupational health and wellbeing of working people are
in improving worker protection, the systems in place for
crucial prerequisites for productivity and are of utmost
surveillance of health and safety suffer from a number of
importance for all socioeconomic and sustainable development.
limitations [9].
The workplace is a significant contributor to injuries due to
Several approaches to solving this problem are currently
accidents and chronic diseases due to environmental hazards.
available but they are not at a high level of technological
Preventing occupational illnesses, injuries and fatalities
development. A method for monitoring health and safety in the
requires immediate action. Over the past decade, there has been
workplace, referred to as ‘surveillance on self-report’, is one
a surge of accidents in mines across the globe [10].
type of OHS monitoring. This OHS monitoring method
The recent emergence of WSN technology is significantly
consists of a collection of data on certain aspects of health and
impacting the capabilities for automated distributed monitoring
safety obtained by dispensing self-administered questionnaires
of environments. The low manufacturing cost of sensors,
to employees or managers and supervisors in workplaces, and
increased coverage and accuracy in distributed sensing due to
subsequently administering regular feedback on the responses
their large deployments and their ability to operate in
[9]. Another method is risk management process, which is
inhospitable terrains has made them an amenable technology
fundamental to most OHS statutes. The underlying premise of
for various surveillance and monitoring applications [12]. The
risk management is that improvements in health and safety can
importance of WSNs in mining applications lies in its
be made by correctly identifying and addressing hazards or
simplicity coupled with great efficiency. In a disaster - such as
factors (which may be underlying or direct) that contribute to
a fire - conventional wired sensor networks may become
occupational risk. The main components of risk management
unreliable, necessitating a wireless radio system which is easier
are hazard identification, risk assessment, implementation of
and less expensive to deploy and maintain.

978-1-5090-2870-2/16/$31.00 ©2016 IEEE 966

B. Other systems in literature
Several OHS systems have been proposed recently.
Systems similar to the one proposed in this work include [13] -
[15]. The system in [13] consists of a smart safety helmet
(SSH) which is able to measure electroencephalography (EEG) Fig. 2. Description of the WNCAP.
and also has an inertial measurement unit (IMU) to detect
worker fatigue. The system can track head gestures and brain
activity to recognize anomalous behavior.
Similarly, the system in [14] is a wireless SSH that uses a
433 MHz module for wireless communication. It is equipped
with temperature, humidity and fire sensors. It connects
wirelessly to a gateway and is specifically aimed at mining
environments. The system in [15] is not aimed at OHS, but can
be easily adapted for this purpose. It measures force and speed
to detect bicycle accidents, and it has uses 315 MHz transceiver
to communicate with a gateway. The system proposed in this
paper relies on low-cost sensors and it easier to
expand/augment (i.e. add modules to) than similar systems in
literature. Fig. 3. Overview of the entire system.

III. OVERVIEW OF THE SYSTEM

IV. DETAILED DESIGN AND IMPLEMENTATION
This section contains general information on the proposed
system. This section contains information on the design and
implementation of the proposed system. Table I summarizes
A. Description of the system the characteristics of the wireless sensor node.
The system consists of the main sub-systems: a wireless TABLE I
smart transducer interface module (WSTIM) and a wireless WIRELESS SENSOR NODE SPECS
network capable application processor (WNCAP). The data
collected by the system’s sensors is relayed to the WNCAP via Specification Details
IEEE 802.15.4. The WNCAP is connected to a PC which hosts
a graphical user interface (GUI) so that the user can view the Microcontroller dsPIC30F4013
current status of the workers and make decisions based on this Transceiver (802.15.4) Dizic DZ-ZB-G
information. Accelerometer (3-axis) MMA7361L
The WSTIM consists of multiple sensors, namely, load Weight Sensor Strain gauge based
sensors, accelerometers, gyroscopes and light sensors, (Wheatstone bridge)
necessary to determine the current status of the worker, and Light sensors LDR (x 2)
these interface with a dsPIC microcontroller (MCU),
responsible for retrieving the data from the sensors and
reporting it wirelessly to the WNCAP, which then displays it The following sub-sections discuss the hardware and
and logs it. These are used to determine whether the miner is software specifications in detail.
wearing his/her safety helmet and is not immobile, for instance. A. WSTIM control unit
Two subsystems and the complete system are illustrated in Fig.
1, Fig. 2 and Fig. 3, respectively. The microcontroller (MCU) used for the sensor node is the
dsPIC30F4013. This 16-bit microcontroller executes all the
software responsible for connecting the sensor node with the
server, and well as sampling data from all sensors. Since it has
a 12-bit ADC, it is suitable for the purposes of this project. The
control unit is housed in a safety helmet, along with other
sensors. This serves as the core of the OHS.
The transceiver for the WSTIM is the DiZiC DZ-ZB-G RF
module, which has as on-board 32-bit ARM Cortex-M3
STM32W108CC MCU and is IEEE 802.15.4 compliant. This
module interfaces with the MCU via serial peripheral interface
(SPI). The PCB of this part of the wireless sensor node is
illustrated in Fig. 4. One of the reasons this module was
selected is the fact that it is an open platform which has been
Fig. 1. Description of the WSTIM. used in research [16] and lots of free resources exist.

967
Fig. 4. PCB of wireless sensor node (WSTIM).

The radio of the DiZiC module operates at 2.4 GHz and has
a data rate of up to 250 kb/s. The configuration selected for this
Fig. 5. Helmet showing 2 LDRs and a proximity sensor to determine
system is a mesh network. The Dizic module runs the Contiki
whether the worker is wearing the helmet or not.
operating system, enabling IPv6 over Low Power Wireless
Personal Area Networks (6LoWPAN). In this mesh network,
The proximity sensor also interfaces with an ADC channel
each wireless node serves as relay between the source node and
the other nodes that are part of the mesh. on the MCU and it keeps a low output by default, which only
goes high if an object is detected. The MCU samples the
B. WNCAP control unit values from the three sensors and based on the sensor
The wireless transceiver of the WNCAP is on the readings, the software on the MCU makes a decision on
client/receiving end of the network. It receives all the data sent whether the helmet is being worn by the worker or not. The
through the mesh network and is connected to a PC which circuit diagram used for the LDRs is illustrated in Fig. 6.
hosts the GUI. The module acts as a border router or gateway
router between the mesh network and the PC. The data
reaching the WNCAP is sent using RESTful resources in
CoAP message formats, therefore the WNCAP needs to be
able to decode and parse messages in this format. To this end,
the Californium CoAP framework in Java was used to interpret
the messages.
C. Helmet light and proximity sensors
The helmet uses two light dependent resistors (LDRs) and
a proximity sensor to sense whether the worker is wearing the
helmet or not. One of the LDRs is mounted on the left side of
the helmet and the other LDR is mounted on the right side,
both close to the location of the ears. The proximity sensor is Fig. 6. Schematic diagrams of LDR circuits.
placed on the top inside part of the helmet. D. Weight sensor
When the worker wears the helmet, both LDRs are
completely covered by the hat, making it a very dark This sensor is used to determine whether the worker
environment typically in the range of 0 – 500 lux. The assigned to a specific pair of boots is wearing them at all times.
proximity sensor is programmed to detect objects within 5 cm The sensor is based on strain gauges, which are variable
such that when the worker wears the helmet, the proximity resistors whose resistance changes according to the amount of
sensor is able to detect it. The placement of sensors in the strain/weight applied to it. These strain gauges are connected in
helmet is illustrated in Fig. 5. a full-bridge Wheatstone configuration, as shown in Fig. 7.
If the readings from the sensors are all positive it means the Resistive load cells work on the principle of piezo-resistivity.
worker is indeed wearing the helmet. Each LDR is connected When a stress/weight is applied to the sensor, it changes its
to a voltage divider. Since the resistance of the LDR changes resistance. This change in resistance leads to a change in output
with the amount of incident light, the output voltage Vout on voltage when an excitation voltage is applied. The resistive
the voltage divider can be used to detect the light and is load cell is not as accurate as a capacitive load cell but it is
connected to an ADC channel on the dsPIC MCU. much more reliable over a wider temperature range.

968
 1) Sensor data acquisition algorithms
A flowchart of the sensor data acquisition code is shown in
Fig. 9. The code was implemented on the dsPIC30F4013 using
C and is used to execute the operation of the WSTIM, such as
sampling the data from sensors, processing it, and sending it
wirelessly to the WNCAP.
If true
Start If (force
Get force forceStatus=7
reading < Else
reading
threshold) forceStatus=4
Set
configuration
bits

If (weight If true
Variable and Get weight shoeStatus=7
reading reading <
function Else
threshold)
declarations shoeStatus=4
Fig. 7. Weight sensor schematic diagram.
Initialize
variables and
I/O ports Get left LDR, If true
The strain gauges used here have a resistance of Rg = 1000 right LDR and
proximity
If (all 3 readings
< threshold)
hatStatus=7
Else
Initialize
Ω when no weight is applied to them. The full bridge peripheral readings hatStatus=4
modules
arrangement for a load cell is such that two strain gauges have NO YES

a positive change to tension, whereas the other two have a Receive instruction Acquire data
YES
positive change to compression [17]. Thus, when a weight is via SPI from
wireless module
SPI interrupt
flag bit set?
Instruction
received?
Send worker
status via SPI to
applied on the sensor, two of the sensors increase their NO wireless module

resistance while the other two decrease it. This change in Clear
resistance causes an unbalance in the bridge, which provides a interrupt flag
bit

differential output corresponding to the weight placed [18]. An

ADN620 amplifier was used to achieve a gain of 275 (by Fig. 9. Data acquisition program flow diagram.
setting Rx = 180 Ω) such that the output of the Wheatstone
bridge is conditioned by putting in a voltage range that can be Once all the initializations (e.g. I/O port, ADC and SPI)
detected by the ADC. The output of the ADN620 amplifier is have been done, the code then stays in an infinite loop that
connected to the dsPIC’s ADC. keeps waiting to receive an instruction via the SPI module from
the GUI through the wireless DiZiC modules attached to the
E. Force sensor dsPIC MCU. The SPI interface between the MCU and the
This sensor is used to determine whether a worker has Dizic module is based on SPI interrupts to mitigate effects
experienced a (harsh) force larger than 1.5 G and if this force from occasional noise on the SPI lines. Once all sensors have
has made the worker change orientation by making them fall. been sampled and the current status is known, this information
A Freescale Semiconductor MMA7361LC 3-axis is forwarded to the PC’s GUI wirelessly.
accelerometer was selected for this purpose. It is used to detect 2) Graphical user interface
any acceleration of 1.5 G or higher as well as any tilt This functional unit of the system displays the current state
disturbances imposed on the worker. The x-axis, y-axis and the of each worker that is being monitored, and to determine
z-axis outputs (XOUT, YOUT and ZOUT, respectively) of the whether they comply with the OHS regulations that this system
accelerometer interface with the ADC input channels on the is monitoring. The GUI was implemented in Java, since it has
dsPIC. The interface with the MCU is illustrated in Fig. 8.
to be integrated with the Californium framework, which is also
Java based. The main menu of the GUI is shown in Fig. 10.

Fig. 8. Accelerometer schematic diagram.

The LM3940 voltage regulator is used to regulate and step

down the 5V supply voltage to 3.3 V required to power the
accelerometer. Fig. 10. Main menu of the GUI.
F. Software
A brief description of the main software components which The GUI was designed with a few debugging features to
comprise the system follows. help the user while the application is running.

969
 V. RESULTS AND DISCUSSION limited the range that could be achieved because of costs and
This section discusses the results of all experiments. convenience. The distance of communication that was
measured was 152 meters between the wireless modules. This
A. Weight sensor test result was achieved mainly because the wireless DiZiC
The objective of this experiment was to determine whether modules used have power amplifiers which increase the
the weight sensor was able to detect weight within the required transmission range.
accuracy. This was tested by placing the sensor inside a scale The light sensors were conditioned to detect a light
and determining the weight of different gym weights. Readings intensity of between 0 lux and 2000 lux which is the light
were also compared to a bathroom scale using weights from 55 intensity of a bright fluorescent light similar to the lights used
to 80 Kgs. The results of this experiment are shown in Table II. in the mines to illuminate the work space.
TABLE II
RESULTS OF WEIGHT READINGS
Weight Off the shelf scale Weight sensor VI. CONCLUSION
Placed (kg) reading (kg) reading (kg) The goal of this work was to design and implement an
55 55 54.1 occupational health and safety monitoring system. A weight
60 60 60.9 sensor was designed and implemented to acquire the weight of
65 65 64.8 a worker. The weight sensor had to be designed to have an
70 70 69.8 accuracy of 95%. Light sensors also needed to be designed
75 75 74.5 and implemented. The light sensors were designed to sense a
80 80 80.9 light of between 0 and 2000 lux. The system’s main focus was
using sensors that were placed on the PPE of workers who
work in a mine to determine whether or not each specific
It is seen that the weight sensor can measure weight with an worker was complying with certain occupational health and
accuracy of 95%. safety regulations, and to test for any harsh forces that the
B. Light intensity and proximity sensors workers might have experienced that could result in them
The function of the light intensity sensors and proximity being knocked off balance. The health and safety regulations
sensors was tested by wearing the helmet and removing it that were monitored were whether the worker is wearing their
several times and observing the output on the MCU. It was assigned safety helmet as well as their assigned safety boots.
found that in 100% of the times the combination of these For future work, this system can be augmented with
sensors was able to detect whether the helmet was being worn tracking capabilities such as inertial sensors mounted on the
or not. worker’s boots or using wireless localization on the available
wireless infrastructure that the sensor nodes connect to [19],
C. Discussion of results [20]. This is particularly useful to locate miners in emergency
The performance of the system is mainly determined by situations. Another aspect that we will consider (which is often
how well the system is able to reflect the true conditions of the overlooked) is security of the sensors’ data, given its
workers in the mine. It was determined from the algorithm privacy implications [21]. Energy efficiency and explosion
testing that the system produced good results and met the GUI protection will also be investigated.
accuracy specification. The system was accurate although the
REFERENCES
latency between the wireless communication modules was
sometimes inconsistent. [1] L. M. Borges, F.J. Velez, and A.S. Lebres, “Survey on the
Characterization and Classification of Wireless Sensor Network
The performance of the weight sensors was found to be Applications,” IEEE Communications Surveys & Tutorials, vol.16, no.4,
within the required specification. The results were found to be pp.1860-1890, 2014.
inconsistent although the resistors that were used to design the [2] G.P. Hancke, B. de Carvalho e Silva, G.P. Hancke Jr., “The Role of
sensors were of the same type. This was due to tolerances that Advanced Sensing in Smart Cities”, MDPI Sensors, Feb. 2013.
are inherent in most resistors as well as the fact that the [3] C.P. Kruger and G.P Hancke, “Benchmarking Internet of things
devices,” IEEE International Conference on Industrial Informatics, pp.
voltages from the Wheatstone bridges were conditioned by 611-616, July 2014.
amplification. This meant that even the smallest differences [4] C.P. Kruger and G.P. Hancke, “Implementing the Internet of Things
before amplification became noticeable after. The tolerances vision in industrial wireless sensor networks,” IEEE International
were taken into account and their effect was reduced by using Conference on Industrial Informatics (INDIN 2014), pp. 627-632, July
1% tolerance resistors. This caused the present inconsistencies 2014.
to be minimal. The algorithm that was implemented on the [5] A. Kumar and G.P. Hancke, “A Zigbee-Based Animal Health
Monitoring System,” IEEE Sensors Journal, vol.15, no.1, pp. 610-617,
dsPIC30F4013 was also designed to further compensate for Jan. 2015.
the difference in the resistances, thus giving a result that was [6] A. Kumar and G.P. Hancke, “Energy Efficient Environment Monitoring
accurate within specifications. System Based on the IEEE 802.15.4 Standard for Low Cost
The wireless nature of the system implied that the Requirements,” IEEE Sensors Journal, vol.14, no.8, pp. 2557-2566,
integration of all the WSN nodes was easier than would have August 2014.
been with a wired implementation. It became clear from the [7] C.A. Opperman and G.P. Hancke, “Using NFC-enabled Phones for
Remote Data Acquisition and Digital Control”, in Proc. of IEEE Africon
wireless configuration that the wired configuration would have 2011, Sep. 2011.

970
 [8] W.D. Chen, G.P. Hancke, K.E. Mayes, Y. Lien and J-H. Chiu. “Using [14] S. Shabina, “Smart Helmet Using RF and WSN Technology for
3G Network Components to Enable NFC Mobile Transactions and Underground Mines Safety,” Int. Conf. Intelligent Computing
Authentication”, in Proc. of IEEE Conference on Progress in Applications (ICICA 2014), pp. 305-309, 6-7 March 2014.
Informatics and Computing, Dec. 2010. [15] M. Mohd Rasli, N.K. Madzhi and J. Johari, “Smart helmet with sensors
[9] M. Haque, “Surveillance on Self-report: A Trial of Health and Safety for accident prevention,” Int. Conf. on Electrical, Electronics and
Monitoring in Occupational Settings”, Occupational Medicine, vol. 50, System Engineering (ICEESE 2013), pp. 21-26, 4-5 Dec. 2013.
no. 3, pp. 182-184, 2000. [16] R. Fisher, L. Ledwaba, G.P. Hancke and C. Kruger, “Open Hardware: A
[10] M.A. Hermanus, “Occupational health and safety in mining-status, new Role to Play in Wireless Sensor Networks?,” Sensors, 15(3), pp. 6818-
developments, and concerns,” Journal of The South African Institute of 6844, 2015.
Mining and Metallurgy, vol. 107 no. 8, pp. 531-538. [17] National Instruments Corporation, “Strain Gauge Measurement – A
Tutorial”, National Instruments Corporation, 2014.
[11] P. Conway, D. Heffernan, B. O’Mara, P. Burton, and T. Miao, “Ieee
1451.2: An interpretation and example implementation,” IEEE [18] K. Sadasivam and S. Gupta, “Weighing scale design – measure signals
Instrumentation and Measurement Technology Conference (IMTC accurately, EE Times Industrial Control, pp. 1 - 7, 2011.
2000), pp. 535–540. [19] Adnan Abu-Mahfouz, G.P. Hancke and S.J. Isaac. “Positioning System
[12] A. Chehri, W. Farjow, H.T. Mouftah and X. Fernando, “Design of in Wireless Sensor Networks using NS-2”, Software Engineering, Vol.
wireless sensor network for mine safety monitoring,” Canadian Conf. on 2, No. 4, pp. 91-100, Oct. 2012.
Electrical and Computer Engineering (CCECE 2011), pp. 8-11, May [20] A. Abu-Mahfouz and G.P. Hancke, “An Efficient Distributed
2011. Localization Algorithm for Wireless Sensor Networks: Based on Smart
References Selection Criteria”, Inderscience International Journal of
[13] P. Li, R. Meziane, M. Otis, H. Ezzaidi and P. Cardou, “A Smart Safety Sensor Networks, Vol. 13, No. 2, pp. 94-111, May 2013.
Helmet using IMU and EEG sensors for worker fatigue detection,” IEEE
Int. Symposium on Robotic and Sensors Environments (ROSE 2014), , [21] G.P. Hancke, K. Markantonakis and K.E. Mayes, “Security Challenges
pp. 55-60, 16-18 Oct. 2014. for User-Oriented RFID Applications within the 'Internet of Things',”
Journal of Internet Technology, Vol. 11, No. 3, May 2010.

971

Powered by TCPDF (www.tcpdf.org)