A Concept for an Ultra-low Power Sensor Network
Detecting and Monitoring Disaster Events in Underground Metro Systems
Jonah Vincke, Scott Kempf, Niklas Schnelle, Clemens Horch and Frank Schäfer
Fraunhofer Institute for High-Speed Dynamics, Ernst-Mach-Institut, EMI, Eckerstrasse 4, Freiburg, Germany
Keywords: Ultra-low Power, Wireless Sensor Network, Energy Harvesting, Tunnel System.
Abstract: In this paper, the concept for an ultra-low power wireless sensor network (WSN) for underground tunnel
systems is presented highlighting the chosen sensors. Its objectives are the detection of emergency events
either from natural disasters, such as flooding, or from terrorist attacks. Earlier works have demonstrated that
the power consumption for the communication can be reduced such that the data acquisition (i.e. sensor sub-
system) becomes the most significant energy consumer. By using ultra-low power components for the smoke
detector, a hydrostatic pressure sensor for water ingress detection and a passive acoustic emission sensor for
explosion detection, all considered threats are covered while the energy consumption can be kept very low in
relation to the data acquisition. The total average consumption for operating the sensor sub-system is
calculated to be less than 35.9 µW.
More than half of the planet’s population now lives in
urban areas. This creates the need for various mass
rapid transport systems including metro systems.
New vulnerabilities for society that arise due to
disaster events, such as terrorist attacks, flooding or
fire, are increased by a higher population density and
current political processes. To address these
challenges – in particular for underground metro
systems – as part of the bi-national research project
SenSE4Metro (Sensor-based Security and Emergen-
cy management system for underground Metro
systems during disaster events) (SenSE4Metro,
2016), a concept for a wireless sensor system is
introduced that can detect the most significant threats
and which provides rescue forces with the relevant
necessary information in case of an emergency. The
particular operation site leads to the requirement for
the WSN that each node must be energy autarkic,
which in turn necessitates the application of ultra-low
power (ULP) components. The focus in this work lies
on the needed sensors to achieve these goals and fulfil
the project requirements.
First, the state-of-the-art of wireless sensor
networks for different kind of tunnels is presented.
After that, an overview of the proposed wireless
sensor network is given, taking the special linear
topology for a tunnel system into account. Finally, a
concept for the needed sensors that can cover all
addressed threats is presented that focuses on the
power consumption and takes the applicability for a
metro tunnel system into account.
The use of wireless sensor networks for tunnel
systems has been investigated in several works. Of
the systems described, most are designed primarily
for road tunnels (e.g. M. Ceriotti, 2011; L. Mottola,
2010; Z. Sun, 2008) or mine tunnels, (e.g. D. Wu,
2010; H. Jiang, 2009). Of the latter, some focus on the
radio transmission in tunnels (e.g. D. Wu and H.
Jiang) and others on special protocols designed to
increase robustness against underground collapses
(e.g. M. Li, 2007). Of the former, Ceriotti et al.
presents a WSN consisting of 40 nodes to monitor the
light conditions of a 260 m long tunnel. Mottola et al.
compared the data of a traffic tunnel (here, railroad
tunnels are proposed as analogous with the assessed
road tunnel) with a WSN for a vineyard to make
suggestions for the communication in road tunnels.
For underground rail tunnels, only a few works
exist. Wischke et al. (2011) discuss the generation
side of the energy autarkic nodes by proposing a
vibration energy harvesting solution for maintaining
the wireless sensor nodes in rail tunnels. Bennett et al.
Vincke J., Kempf S., Schnelle N., Horch C. and SchÃd’fer F.
A Concept for an Ultra-low Power Sensor Network - Detecting and Monitoring Disaster Events in Underground Metro Systems.
DOI: 10.5220/0006186901500155
In Proceedings of the 6th International Conference on Sensor Networks (SENSORNETS 2017), pages 150-155
ISBN: 421065/17
2017 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
(2010) used MICAz boards as wireless sensor nodes
to monitor cracks in a 170 m long part of the Prague
Metro and in a 115 m long section of the London
Metro. Sivaram Cheekiralla (2005) used a WSN to
monitor the deformation of a train tunnel during
construction using 18 nodes using only a star
Raza et al. introduced a combination of an ULP
wake-up receiver with model-based sensing to reduce
the power consumption of the rail tunnel WSN by
Ceriotti et al. (2011). They simulated an increased
lifetime of the nodes by a factor of over 2000 due to
their method, which demonstrated the influence of a
data model to reduce necessary data transmission and
the use of a wake-up receiver.
The optimization of wireless communication in
tunnel systems has been widely discussed and
especially Raza et al. (2016) showed that only the
power consumption of the data acquisition is of
interest when using a wake up receiver. Therefore, the
focus of this paper lies on the conception of the sensor
system based on an initial layout and requirement
specification for the WSN.
For the purposes of defining the requirements of the
WSN, an assessment of past terrorist attacks on
underground, tunnel and rail infrastructure was
performed (S. Kempf, 2016). The assessment results
highlighted several distinct differences between
attacks on underground systems as compared to
above-ground networks, with respect to tactics and
effectiveness (in terms of casualties). Ultimately,
event scenarios were defined based on explosive and
arson attacks targeting the trains and tunnels
themselves. While the assessment also highlighted
the threat of biological/chemical attacks, due to
previous studies (A. Pflitsch, 2010), these were
explicitly omitted from the scope of the project.
Adding recent historical flooding in underground
networks (Prague 2002, New York 2012) and
accidents (Valencia 2006, Moscow 2014): the
following events should be detected and the
corresponding data acquired by the WSN:
Train passage (positioning and movement),
Fire (temperature and smoke presence),
Explosion (impact peak pressure and
specific impulse)
Flooding (water presence and depth).
The combined data will be applied to determine
danger levels and traversability of tunnel segments
and to coordinate paths of access (rescue forces) and
escape (passengers).
In order to acquire the necessary data above,
sensors must be positioned both at ground and ceiling
level and all along the tunnel segment. It was decided
that the implementation would be performed as a
parallel linear topology (see Figure 1) with cable
wired gateway nodes at each metro station. The
topology provides added robustness via path
redundancy, as both nodes can be used to forward
messages. Its relatively long path lengths however
require special tuning and adaptation of routing
algorithms. While classic tree based routing
protocols, such as Contiki Collect or CTP (O.
Gnawali, 2009), can in principle be applied directly
nodes at the end of long paths would have to forward
all the messages generated by preceding nodes in the
path thus creating significant load and using
disproportionally more energy. To counter this data
from different nodes may be combined or filtered to
only update for significant changes instead of simply
forwarding all generated data towards the nearest
gateway. This however remains an area of active
research and development.
The upper nodes can be used to measure smoke
and impact pressure and are powered using a wind
energy harvester. The lower nodes measure water
ingress and temperature and are powered using a
piezoelectric vibration energy harvester attached to a
rail. In standard (non-emergency) operation,
situational status messages are transmitted during
Figure 1: Applied WSN topology in underground tunnel.
A Concept for an Ultra-low Power Sensor Network - Detecting and Monitoring Disaster Events in Underground Metro Systems
train passage events, which means the energy
required for data acquisition and transmission is
provided directly via energy harvesting processes.
For emergency event detection, an energy storage
system is provided to ensure a constant power supply.
For all nodes, the CC2650 from Texas
Instruments is chosen as the MCU due to its very low
power consumption and integrated RF module.
To achieve low power consumption, a holistic
concept has been developed. This includes the
application of modern ultra-low power sensors,
enabled only when necessary, as well as the re-
application of the same sensors for various disaster
events if possible. The precision of the sensors is of
less importance in contrast to the power consumption.
A robust detection of a dangerous event is sufficient.
4.1 Water Ingress
There are several methods for the detection of water
ingress and determination of the resulting water level.
These vary from mechanical solutions using floats to
change a resistance, a capacitance or to close a contact
to pure capacitance or resistance measurements as
well as hydrostatic, ultrasonic and radar methods.
Many can be realized with an ultra-low power
consumption but vary with respect to their robustness,
dependence on the medium and the tunnel’s shape.
Mechanical solutions have the disadvantage that
their dimensions need to be in the same range as the
measurable water level and their shapes are limited.
On the other hand they are independent on the media
and can be realized as ultra-low power systems.
Optical or ultrasound distance sensors are not
dependent on such limitations nor on the media or the
shape of the tunnel. But they lack on the measurable
distance and power consumption. As an example the
infrared distance sensor GP2Y0A710 from Sharp
needs above 1 mW for one measurement every 5
seconds while only covering a distance of up to 5 m.
Ultrasonic sensors that can measure distances of up to
8 m or more have commonly a power consumption of
over 1 Watt during operation and need several
hundreds of milliseconds until the first measurement
is possible. As an example the UC30-2 from SICK
needs up to 1.2 W for approximately 450 ms until a
measurement can take place. Sensors for smaller
distances such as the LV-MaxSonar-EZ have a power
consumption of about 10mW for half a second for a
measurable distance of 6.45 m.
Of the other solutions, measuring the hydrostatic
pressure seems most promising for achieving very
low power consumption. Here the pressure caused by
the water ingress at the bottom of the tunnel has to be
measured as well as that above the water level. The
disadvantage of this principle is that calculating the
water depth according to the induced pressure
difference is dependent on the mediums density by
design. Also the system’s robustness in a harsh
environment such as an underground tunnel has to be
investigated. Since passing trains induce pressure
disturbance, the measurement has to be adjusted
during train passage events. The simulations
(ThermoTun Online, 2016) have shown that a train
with a cross-sectional area of 8 m² in a 5 m high
tunnel with a speed of 200 km/h creates a pressure
difference of up to 1.36 kPa in the tunnel. This would
be equivalent to a water level of 138.7 mm. How the
pressure is disturbed over the cross section and over
the time in a real tunnel has to be investigated.
As an example for the advantages of pressure
sensors, two MS5806 can be used to measure the
pressure at the bottom and above the water level.
Using these values, water levels of up to 9 m with a
precision of 0.13 cm can be measured in theory while
consuming less than 3 µW for each sensor when
measuring once per second. A temperature sensor is
also included that can be used for the other sensors in
addition, reducing the overall power consumption.
Because of the independence on the tunnel’s shape,
the water ingress detection will be based on
measuring the pressure induced by the water. Since in
most cases the media will be ground water the density
of the media will be similar in most cases and
therefore the dependence of the system on the media
can be neglected. The sensor will be located at the
wall. The pressure at the bottom of the tunnel is
measured using a tube mounted to the wall that is
connected to the sensor and goes to the ground as
shown in Figure 2.
Figure 2: Schematic of the water level measurement system
using a tube to measure the pressure at the ground of the
SENSORNETS 2017 - 6th International Conference on Sensor Networks
The water ingress depth
can be determined
using the pressure at the base of the tube, assuming
that the media is water:
In order to determine the pressure p
, it is
necessary to consider the air compression induced by
the water ingress within the tube. Using the ideal gas
law (pV=nRT), the reference pressure (external
sensor) and assuming constant mass, temperature and
cross-sectional area within the tube, the height of the
air column can be determined:
The pressure at the tube base is a summation of
the air pressure due to compression and the water
ingress within the tube:
Reapplying the water depth equation, the total
pressure p
can be determined:
With (2):
Finally, inserting back into (1):
4.2 Fire
To detect fires, heat and smoke are measured. Heat is
measured using the integrated sensor of the pressure
sensor. Three classical smoke detection systems
exists. While measuring the concentration of carbon
monoxide either consumes too much power or is
limited in life time, ionic sensors can reach a power
consumption as low as 25 µW (Z. Moktari, 2013) but
consist of a radioisotope. Common photoelectric
smoke detectors consume approximately 90 µW. This
power can be reduced down to less than 6 µW by
using ultra-low power microcontroller units (MCUs)
and operational amplifiers and by reducing the
sampling ratio down to one sample each 8 s (M.
Mitchell, 2012). Based on this and because of the
German regularities regarding ionic materials, the
smoke detectors are realized based on the
photoelectric effect. An infrared LED emits light in a
smoke chamber that has to be reflected by particles
such that the reflection can be measured by a
photodiode. To increase the robustness of the sensor
against disturbances such as ambient light, the sensor
measures the output of the photodiode when the
IRLED is turned off additionally. As in (M. Mitchell,
2012), a measurement is done every 8 seconds. If
smoke is detected, the interval is reduced down to 4
seconds. After three detections an alarm is triggered.
4.3 Explosion
Due to the physical sensor node layout and the
limitations of ultra-low power WSNs with regard to
time resolution and synchronization, an exact
determination of explosion position is not feasible.
Using empirically determined thresholds based on the
results of in-house explosion experiments at the EMI
(A. Stolz, 2010; O. Millon, 2013; F. Schäfer, 2014),
the sensor system can however determine the
remaining structural capacity of tunnel walls based on
the peak pressure and specific impulse observed at the
node. The expected variable impact distance of 0-50
meters (based on 100 m node spacing) can be taken
into account when establishing these thresholds.
To measure the blast wave, an acoustic emission
sensor, the VS150-M, has been chosen which has
been tested successfully in previous projects
performed by researchers at the EMI (O. Millon,
2013; M. Erd, 2016). It generates an electrical signal
from the deformation of the sensor using a
piezoelectric element and therefore needs no supply
power. Nevertheless, to measure its signal, an electric
subsystem is needed. As in previous work, the most
power saving system is to measure the exceedance of
several thresholds. For this purpose three MOSFETs
are used that will consume up to 9 µW. As opposed
to the water ingress and fire detection sensors, the
signal of the VS150-M can be used to wake up the
MCU in case of an explosion. This allows measuring
the start time of the event, the exceedance of different
thresholds and the duration of the blast. Using several
nodes this is enough to estimate the severity of the
blast as well as the energy released.
The validation of the system is ongoing. A single
pressure sensor increases the power consumption by
14.79 µW when no measurement is done and
A Concept for an Ultra-low Power Sensor Network - Detecting and Monitoring Disaster Events in Underground Metro Systems
40.08 µW if the pressure and temperature is measured
once per second. In Figure 3 a depth measurement in
a water pipe filled with tap water is shown. The
hydrostatic pressure at the bottom is measured using
a hose connected to a MS5806 pressure sensor
located above the pipe.
Figure 3: Calculated depth from pressure readings.
The results of the system are shown in Figure 3.
The water depth is calculated according to equation
(6) using the measured pressure. The compression of
the air is compensated for the 645 mm long hose.
Here the mean error between the nominal and
measured depth increases from 1.2 mm for a real
depth of 0 mm up to -14.6 mm for a real depth of 500
mm. This corresponds to a relative error of up to
4.32 % as shown in Figure 4.
Figure 4: relative difference between the nominal and
measured depth.
The power consumption of the smoke sensor is
4.06 µW in the case of one measurement every 8
seconds. The response of the sensor to the application
of a test spray is shown in Figure 5. Here a
measurement is done every 0.25 seconds in contrast
to normal operation. The response of the smoke
detector when the IRLED is turned on is shown in red.
This can be compared with the response when the
IRLED is turned off as shown in orange. The induced
difference between both measurements is increased
from an average of 203.77 mV up to 758 mV.
Figure 5: Response of the smoke sensor when using a test
spray as a function of voltage (mV) over time (s). Red: IR
LED turned on, orange: IR LED turned off.
A concept for a wireless sensor network for
monitoring underground metro systems, specifically
focusing on the requirements and design of the
sensors themselves, has been presented. As energy
autonomy is desired, low energy consumption of the
components, while maintaining the minimum sensing
integrity and resolution, is of highest priority.
Using a highly integrated MEMS pressure sensor,
ULP components for a smoke detector with a reduced
sampling rate and an acoustic sensor for the explosion
detection, the expected average power consumption
for the sensor system in normal operation can be
reduced to 13.06 µW for upper nodes and 35.9 µW
for lower nodes. In this case, smoke and water depth
are measured every eight seconds. The accuracy of
the depth measurement is capable for giving the
rescue forces a situational awareness. The smoke
detector is very sensitive as its output is increased by
a factor of more than three. Therefore, the
combination of both nodes provides the capability of
detecting explosion, fire and water ingress, while only
consuming very low power.
In further work, the sensor system will be
integrated and validated. For the pressure sensor, a
hose ending has to be developed that ensures the
robustness of the system in such a harsh environment
like a metro. In addition, the pressure disturbances in
the tunnel systems induced by passing trains also has
to be analyzed.
Using the presented sensor system, a larger
security management and emergency response
system will be developed, whereby all interested
parties, such as metro network operators and rescue
forces, will be informed in real-time of critical
developments, for instance degradation of tunnel
structural integrity and impairment of traversability
of tunnel segments due to emergency events in order
SENSORNETS 2017 - 6th International Conference on Sensor Networks
to minimize the secondary damage (e.g. in terms of
human life) of these events.
This work is part of the research project “Sensor-
based Security and Emergency management system
for underground Metro system during disaster
events” (SenSE4Metro) and is funded by BMBF
(German Federal Ministry of Education and
Research) through the joint program “International
cooperation in civil security research: Cooperation
between Germany and India” under Grant No.
P. J. Bennett, K. Soga, I. Wassell, P. Fidler, K. Abe, Y.
Kobayashi, and M. Vanicek, 2010. Wireless sensor
networks for underground railway applications: case
studies in Prague and London. In Smart Structures and
Systems, vol. 6, no. 5–6, pp. 619–639.
M. Ceriotti, M. Corrà, L. D. Orazio, R. Doriguzzi, D.
Facchin, G. P. Jesi, R. L. Cigno, L. Mottola, A. L.
Murphy, M. Pescalli, et al, 2011. Is there light at the
ends of the tunnel? Wireless sensor networks for
adaptive lighting in road tunnels. In Information
Processing in Sensor Networks (IPSN), 2011 10th
International Conference, pp. 187-198
M. Erd, F. Schäfer, M. Kostic, and L. M. Reindl, 2016.
Event monitoring in emergency scenarios using energy
efficient wireless sensor nodes for the disaster
information management. In International Journal of
Disaster Risk Reduction, vol. 16, pp. 33–42
O. Gnawali, R. Fonseca, K. Jamieson, D. Moss, and P.
Levis, 2009. Collection tree protocol. In Proceedings of
the 7th ACM Conference on Embedded Networked
Sensor Systems (SenSys '09). ACM, New York, USA,
pp. 1-14.
H. Jiang, L. Chen, J. Wu, S. Chen, and H. Leung, 2010. A
Reliable and High-Bandwidth Multihop Wireless
Sensor Network for Mine Tunnel Monitoring. In IEEE
Sensors Journal, vol. 9, no. 11, pp. 1511–15173
J. Johansson, M. Gustafsson, and J. Delsing, 2006. Ultra-
low power transmit/receive ASIC for battery operated
ultrasound measurement systems. In Sensors and
Actuators A: Physical, vol. 125, no. 2, pp. 317–328
S. Kempf, F. Schäfer, G. Sitharam, P. Kleist, W. Gräfling,
N. Ferguson, T. Stuchtey, T. Chakraborty, V. Matsagar,
N. Gebbeken, 2016. SenSE4Metro: A bi-national multi-
disciplinary project for monitoring underground metro
environments in disaster events. At 11th Future
Security, Berlin, Germany
M. Li and Y. Liu, 2007. Underground Structure Monitoring
with Wireless Sensor Networks. In Proceedings of the
6th International Conference on Information
Processing in Sensor Networks, New York, USA, 2007,
pp. 69–78.
Measurement Specialties, Inc., 2010. MS5806-02BA
Miniature Altimeter Module. In Datasheet
000058061951 ECN2205
O. Millon, F. Schäfer, A. Stolz, and W. Riedel, 2013.
Automatisierte Informationsgewinnung und Schutz
kritischer Infrastruktur im Katastrophenfall (AISIS). In
Fraunhofer IRB Verlag Stuttgart.
M. Mitchell, 2006. Study of an ionic smoke sensor. In Texas
Instruments Incorporated, Texas, Application Report
Z. Mokhtari, S. Holé, and J. Lewiner, 2013. Study of an
ionic smoke sensor. In Meas. Sci. Technol., vol. 24, no.
5, p. 55006.
L. Mottola, G. P. Picco, M. Ceriotti, Ş. Gunǎ, and A. L.
Murphy, 2010. Not All Wireless Sensor Networks Are
Created Equal: A Comparative Study on Tunnels. In
ACM Trans Sen Netw, vol. 7, no. 2, p. 15:1–15:33
A. Pflitsch, M. Brüne, J. Ringeis, and M. Killing-Heinze,
2010. ‘OrGaMIR‘– Development of a safety system for
reaction to an event with emission of hazardous
airborne substances – like a terrorist attack or fire –
based on subway climatology. In 4th Intl. Symposium
on Tunnel Safety and Security, Frankfurt a. Main,
U. Raza, A. Bogliolo, V. Freschi, E. Lattanzi, and A. L.
Murphy, 2016. A Two-Prong Approach to Energy-
Efficient WSNs: Wake-Up Receivers plus Dedicated,
Model-Based Sensing. In arXiv:1601.06664
F. Schäfer, A. Stolz, L. Reindl and K. Thoma, 2014.
Energieautarkes Funksensornetzwerk zur Erhöhung der
Sicherheit in Bahntunnels bei terroristischen
Anschlägen mit Sprengstoffen. In SCIENCE^2 – Safety
and Security, Homeland Security – Nationale
Sicherheit und Bevölkerungsschutz, vol. 01/2014, ISSN
SenSE4Metro, 2016, Sensor-based Security and
Emergency management system for underground
Metro system during disaster events. At
A. Stolz, W. Riedel, C. Mayrhofer, M. Nöldgen, and K.
Dörendahl, 2010. Tunnel Structures subjected to
Explosions. At 5th Future Security, Berlin, Germany.
Z. Sun and I. F. Akyildiz, 2008. Not All Wireless Sensor
Networks Are Created Equal: A Comparative Study on
Tunnels. Channel Modeling of Wireless Networks in
Tunnels. In IEEE GLOBECOM 2008 - Global
Telecommunications Conference, 2008, pp. 1–5.
ThermoTun Online, 2016. Rail Tunnel Ventilation and
Aerodynamics Software. At http://www.thermotun-
online.com/index.php. (Accessed: 09-Jun-2016).
M. Wischke, M. Masur and P. Woias, 2011. Vibration
harvesting in traffic tunnels to power wireless sensor
nodes. In Smart Materials and Structures, 20(8).
D. Wu, L. Bao, and R. Li, 2009. A holistic approach to
wireless sensor network routing in underground tunnel
environments. In Comput. Commun., vol. 33, no. 13, pp.
A Concept for an Ultra-low Power Sensor Network - Detecting and Monitoring Disaster Events in Underground Metro Systems