Self-adaptive Sensing IoT Platform for Conserving Historic Buildings
and Collections in Museums
Rita Tse
1
, Marcus Im
2
, Su-Kit Tang
1
, Luís Filipe Menezes
3
,
Alfredo Manuel Pereira Geraldes Dias
4
and Giovanni Pau
5,6
1
School of Applied Sciences, Macao Polytechnic Institute, Macao, China
2
Macao Polytechnic Institute, Macao, China
3
Department of Mechanical Engineering, University of Coimbra, Coimbra, Portugal
4
University of Coimbra, Coimbra, Portugal
5
Department of Computer Science and Engineering, University of Bologna, Bologna, Italy
6
Department of Computer Science, University of California, Los Angeles, U.S.A.
Keywords: Self-adaptive Sensing, Internet of Things, World Heritage Conservation.
Abstract: As historic buildings and collections in museums are normally of deteriorated structure or materials, any
sudden change of weather or environment, such as oxygen level, temperature, humidity, air quality, etc., may
cause damages to them and it may not be recoverable. Internet of Things (IoT) is common in solving problems
by collecting environmental data using sensors. The data is live and immediate for visualizing the environment,
which is suitable for conserving the buildings and collections. However, there is no one-for-all IoT solution
for this conservation problem. In this paper, we propose the design of the sensor device in the IoT platform
for conserving historic buildings and collections in museums. The sensor device is self-adaptive, running
continuously without any interruption causing by the instability of power and network connection. The
platform is currently implemented for the conservation project in the Science museum, University of Coimbra,
Portugal. It has been running over a year and the conservation work is going well.
1 INTRODUCTION
Internet of things (IoT) is an architecture that allows
intelligent systems to collect environmental data from
sites at real time and to make immediate decision to
achieve the goals. In a typical IoT system, there are a
number of pervasive presence of things or objects,
called sensor devices in this paper, that are of
communicating capability to connect to edge servers
or backend servers in an IoT network. Data collected
and generated by the sensor devices will be sent to the
servers through wireless or wired connection,
creating a big data for further processing.
There is no one-for-all IoT solution for conserving
historic buildings and the collections in museums.
Depending on their types of materials and kinds of
environmental data to be collected, the sensor device
is designed. There are a number of existing solutions
(Tse et al, 2018) (Maksimović & Ćosović, 2019)
(Neri et al, 2009) (D'Amato et al, 2012) (Tse & Pau,
2016) (Pereira et al, 2017) (Aguiari et al, 2018) that
are engineered to deal with this materials and data
issues in their IoT systems.
To accurately collect the data (e.g. temperature,
humidity, location, CO2, pm2.5, etc.), the sensor
devices must be enabled with appropriate sensing
capabilities using sensors. As the devices generate
data continuously, the amount of data is unlimited but
the storage is limited that may raise a storage issue.
Moreover, deploying IoT systems in historic
buildings is very challenging as the buildings may be
short of power and network connection from time to
time. Data loss is likely to happen, resulting in the
failure of conservation.
To tackle the power and network connection
issues, this paper will disclose the design
consideration of the sensor device in the IoT platform.
The Self-adaptive sensing method will be proposed
which enables the sensor device to operate
continuously even though the power and network
connection are not stable. By estimating the size of
data generated by sensors and the amount of energy
consumed by the sensor device when backup power
392
Tse, R., Im, M., Tang, S., Menezes, L., Dias, A. and Pau, G.
Self-adaptive Sensing IoT Platform for Conserving Historic Buildings and Collections in Museums.
DOI: 10.5220/0009470203920398
In Proceedings of the 5th International Conference on Internet of Things, Big Data and Security (IoTBDS 2020), pages 392-398
ISBN: 978-989-758-426-8
Copyright
c
2020 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
is used, the design of the sensor device can be
optimized to work longer time until the main power
and network connection are back. This project is the
on-going research project for the Science Museum of
the University of Coimbra in Portugal. The IoT
platform has been deployed in the Museum and has
been running over a year.
The remaining of this paper is organized as
follows. In Section 2, we review the threats to historic
building and collections in museums and the types of
sensor to prevent the threats. After that, the self-
adaptive sensing IoT platform will be discussed,
which includes the architecture of the platform in
Section 3, and the self-adaptive sensing method in
Section 4. In Section 5, the ongoing research project
will be introduced as it implements the IoT platform.
Finally, we conclude this work in Section 6.
2 BACKGROUND
To conserve the historic buildings and the collections
in museums, it is crucial to study their structure,
quality and condition. There are many different kinds
of threats to the buildings and collections. Common
threats are fire, air pollution and visitor crowd that
can cause different levels of damage to them,
depending on their condition and types of material
(wooden, papers, textiles, photos and leathers, etc.).
Thus, the sensor device equipped with the oxygen
sensor, nitrogen oxide sensor and humidity sensor, or
some others sensors that can help against the threats,
is highly recommended.
2.1 Oxygen Sensor
Most of the historic buildings are of wooden structure
and they are flammable. After years of erosion, their
structure becomes deteriorated if their maintenance is
not properly done. When there is a fire, serious
damages would be caused to them and the result
would be unrecoverable. Thus, the fire safety
guideline is always in place. It describes the
principles and practices of fire safety for historic
buildings and for those who operate, use, or visit them
(Macau Cultural Affairs Bureau, 2019, October 5)
(National Fire Protection Association, 2019, October
5).
In China, burning incense is a Chinese custom.
Some Chinese temples allow visitors to burn
incenses. Staff are required to pay attention to any
unattended fire combustion caused by incense
burning. Moreover, clearance of the entrance of
temple is required as streets are narrow in some old
areas. Fire truck can hardly reach.
Therefore, in case of fire, controlling the oxygen
level is one of the feasible solutions to tackle the fire
problem (Jensen & Holmberg, 2006). The study
revealed that the air supply is one of the crucial
factors that can feed a fire, as oxygen in the air can
develop fire combustion. By reducing the oxygen
level in the air from 21% (normal) to 15%, the fire
development can be limited, while sufficient oxygen
is provided for evacuating people.
To control the level of oxygen, we may use the
oxygen sensor for monitoring the level at a site. The
sensor would be installed at some particular corners
near the places that are likely to have fire combustion.
The reason is that the oxygen level over there is
normally lower than other places. The sensor will
report the level constantly. If there is a fire, the level
is supposed to drop unexpectedly. In this case, some
procedures can be applied to control the air supply in
the site.
2.2 Nitrogen Oxide Sensor
Air pollution is an environmental problem in
developing cities that is mainly caused by highly
congested traffic (Pau & Tse, 2012). Tons of vehicle
exhausts are generated from the traffic every day,
which will produce tons of poisonous gas, called
nitrogen oxide. There are a number of studies
(Tétreault, 2003) (Szczepanowska, 2013) (Thickett &
Lee, 2004) (Florian, 2006) (Lavédrine, 2003) (Peng,
2017) conducted about how nitrogen oxide causes
damages to the materials commonly found in
museum, which include papers, textiles, photos and
leathers.
Nitrogen oxide will not cause damage to the
materials immediately. However, it is very likely that,
after exposing the materials for a certain period of
time, degree of damages can be developed. Nitrogen
oxide can weaken papers, textiles and photos, causing
the colour fading in intensity. For leathers, it can
create cracks on their surface. Therefore, protection is
needed (National Fire Protection Association, 2019,
October 5).
As nitrogen oxide is one of the key factors in
maintaining historic buildings and collections, it is
suggested to monitor the level of nitrogen oxide.
Similar to the monitoring of oxygen level, the
nitrogen oxide sensor can be used. By setting up an
acceptance level of nitrogen oxide, some procedures
can be applied if it is out of the level.
Self-adaptive Sensing IoT Platform for Conserving Historic Buildings and Collections in Museums
393
2.3 Humidity Sensor
Historic buildings and museums in World Heritage
are favourite tourist spots that would attract crowd of
visitors (UNESCO, 2019, October 5). If the visitors
are not well managed in a site, the current state of the
buildings or the collections inside the buildings may
be adversely affected by the humidity generated by
visitors, especially for books or textiles. High
humidity causes both dust mite populations and mold
colonies to grow (Public Broadcasting Service,
2019). Thus, limiting the number of visitors at a time
slot in a site becomes crucial. In order to limit the
number of visitors, the manual head count method at
the entrance or at the ticket booth is commonly used.
2.4 Thermal Sensor
Recently, existing solutions using thermal camera can
give the thermal map of a site (Abdelrahman &
Schmidt 2018). A thermal map can be used to detect,
track, and recognize humans without creating the
privacy issue (Gade & Moeslund, 2014). If a number
of thermal cameras are installed at somewhere giving
an overview of the place, we can compute the total
number of visitors at any time in a site (Tse et al,
2018). The humidity generated by visitors can be
estimated. In addition, the thermal camera can be
used for survey purpose, generating a statistic about
the favourite of visitors by counting their stopping
time for the collections. This minimizes the work
involved in staffing.
As the environment of developing cities is
complicated, conducting the conservation of historic
buildings and collections in museums is likely to be
harder. Using the right sensors in monitoring or
collecting data would be necessary. Once the sensor
device is designed and put in place, unlimited amount
of data will be generated by sensor devices.
3 THE ARCHITECTURE OF IoT
PLATFORM
The Self-Adaptive Sensing IoT Platform consists of a
number of sensor devices and servers locally and
remotely. The sensor devices are distributed in sites
and operate continuously. Figure 1 depicts the
architecture of the IoT platform.
In Figure 1, the architecture consists of a number
of sensor devices installed in a number of sites.
Depending on the conservation requirement for a site,
the sensor device is equipped with some sensors, such
as oxygen sensors, nitrogen oxide sensors and
thermal cameras. Each sensor device is implemented
by an Arduino or Raspberry Pie device. In order to
enable the sensor devices to perform their role
efficiently, each sensor device is equipped with a
battery for backup power. The battery capacity is
determined by the power consumption of the sensors
in the device. Currently, the battery can support the
operation of the sensor device for three days, which
is good enough for the weekends.
Figure 1: The Architecture of the IoT Platform.
To be scalable and stable, the sensor devices are
loosely connected to the backend server using the
wireless network connection, mainly relying on the
Wi-Fi network. Depending on the situation, if reliable
Wi-Fi network is not provided or not stable, 4G
cellular network is then preferred. For contingency
and efficiency purpose, an edge server is used if
needed. The edge server pre-processes the data and
check if the network connection is ready. If so,
processed data is then sent to the backend server for
further processing. The data is manipulated into a
general form that allows analysis before storing.
There is a local storage equipped in the sensor device
and its size is big enough for 7 days of data. When a
network connection is available, data will be
forwarded to backend servers. In the meantime, huge
amount of data is generated every day. It would raise
a performance issue on data retrieval in the IoT
platform. There are many solutions to resolve this
issue and data snapshots are used in the platform. To
minimize the data to be generated and extend the
operation time of the sensor devices on backup
power, we propose the self-adaptive sensing method.
IoTBDS 2020 - 5th International Conference on Internet of Things, Big Data and Security
394
4 SELF-ADAPTIVE SENSING
METHOD
The Self-adaptive sensing method in the IoT platform
enables the sensor device to operate continuously
even though the power and network connection are
not stable. Once the device is up and running,
unlimited amount of data will be generated. If a
network connection is available, the device may
forward the data to the edge server, if any, for
preliminary processing or to the backend server for
processing. If not, data will be stored in local storage.
As local storage size is limited, it may get full quickly
depending on the size of data generated by sensors
every time. As mentioned previously, sensors in the
IoT platform will be strategically selected against
threats. One of the selection criterions for sensors is
the data size in one generation.
4.1 Data Size
The data size in one generation for a sensor is fixed.
If a sensor generates b bits of data, for a number of
data generations, t, in one day, the total amount of
data, d, to be generated can be expressed as below.

(1)
If a sensor device is equipped with X = {x
1
, x
2
, …,
x
n
} sensors, the total amount of data, D, to be
generated in bits per day will be expressed as below.

(2)
It is noted that the local storage, S, in a sensor
device is suggested to be bigger than D in order to
maintain data integrity for a certain number of days,
c, such that

(3)
Since D is determined by the fixed outputs B =
{b
1
, b
2
, …, b
n
} from X = {x
1
, x
2
, …, x
n
} in one day,
for each sensor in X, t becomes a critical valuable for
D, affecting the determination of S, if c is constant.
To determine the value for t, it is necessary to
consider the wakeup time (another criterion) which is
about the power required by a sensor when it
operates.
4.2 Power Consumption
Another selection criterion for sensors is its power
consumption as it is directly relating to the capacity
required for backup power. In mathematical equation,
the amount of power consumption, P, depends on the
sensors selected to work in the sensor device, which
is calculated from the amount of energy, E = {e
1
, e
2
,
…, e
n
}, in kilowatt (kW), required by each sensor in
X = {x
1
, x
2
, …, x
n
} for t and the amount of energy, G,
required by the system of a sensor device, such that

(4)
To ensure the backup power is sufficient to
support the operation of the sensor device for a certain
number of days, c, the capacity required for backup
power, BP, in milliamp per hour (mAh) for the
raspberry pi at 5 volts is expressed below.

 1000
5
(5)
Nevertheless, the capacity of backup power is
limited. To minimize the energy consumption and
extend the operation time in offline, the running cycle
for a sensor in next section will be considered.
4.3 Running Cycle
In the Self-adaptive sensing IoT platform, there is a
running cycle for each sensor in the sensor device. In
the running cycle, there are five phases which are 1)
Data collection, 2) Data generation, 3) Data storing,
4) Sleep and 5) Wake up. Figure 2 shows the five
phases in the running cycle.
In phase 1, the sensor tries to collect
environmental data. Once, data is collected, the
sensor generates data in phase 2. In phase 3, the data
is then forwarded to servers if the network connection
is available. If not, the data will be stored at the local
storage (See Equation (3)) in the sensor device. If the
network connection becomes available in the next
cycle, all local data generated in previous cycles, will
be flushed and forwarded to the servers (edge server
or backend server). In phase 4, the sensor will be put
to sleep. When it is time out, a wake-up call will be
sent to the sensor in phase 5. One cycle is finished and
another cycle will begin.
Self-adaptive Sensing IoT Platform for Conserving Historic Buildings and Collections in Museums
395
Figure 2: The Running Cycle for a Sensor.
In each phase, it takes a certain amount of time to
complete its process when the power connection is
stable. If the power and network connection is not
stable in historic buildings, limited backup power is
then needed for the sensor device. In case of power
cut after hours, the backup power would operate its
sensor device for data transfer when the network
connection is available, and for data storing locally
when the connection is not available. To extend the
operation time of a sensor device, the device is better
put to sleep when the data variance is stable. It
minimizes the use of local storage, S, and saves the
backup power, BP.
To put sensors to sleep, an issue on the wakeup
time for a sensor will be raised. Another selection
criterion for sensors is the wakeup time it takes after
they sleep. Some sensors may take only a few milli-
seconds to wake up and some may take a minute. This
may create a time synchronization problem on the
data. To maintain the integrity and accuracy of data,
the amount of sleep time may be reduced to allow
more time for a wakeup call. Therefore, when
selecting a sensor, a balance between the sleep time
and the wakeup time in the overall operation is
needed. Figure 3 gives an overview of the operation
time in a running cycle.
In Figure 3, there are three sensors connected in a
sensor device, which are Sensor A, Sensor B and
Sensor C. Their running cycles are of the same length,
even though the portion of time for phases in a cycle
varies. In Sensor A, it requires 20% of the time to
wake up and 60% of the time to sleep. In Sensor C, it
takes 60% of time to finish a wakeup call, which is 3
times more than its sleep time. In this case, reducing
the sleep time to compensate the time taken for a
wakeup call can uniform the cycle length. It avoids
the data integrity and accuracy problems, and saves
energy consumption.
Figure 3: Operation Time in a Running Cycle for 3 Sensors.
5 ONGOING RESEARCH
PROJECT
This paper is based on the ongoing research project
for the Science Museum of the University of Coimbra
in Portugal, which is a collaboration between Macao
Polytechnic Institute, Macao SAR, China and
University of Coimbra, Portugal. The project is
designed for the conservation of the collections inside
the Museum using this self-adaptive sensing IoT
platform. Preliminarily the platform connects to a
number of sensor devices, monitoring the
temperature, humidity and indoor air quality (AIQ)
inside the Museum, as required by the Decree Law n.
78/2006 at April 4 in Portugal. As can be seen in
Figure 4, the sensor device (in green colour) was
being installed in the Museum in 2018.
The Science Museum is open to the public during
weekdays. When it is open, the power and network
connection are available for the Museum operation.
When it is closed, the power will be turned off and
network connection will be cut. The IoT platform
runs in the Museum and data is continuously
collected and generated by the sensor device. As the
platform is self-adaptive, the platform will switch to
run on the backup power when the power is down.
IoTBDS 2020 - 5th International Conference on Internet of Things, Big Data and Security
396
Figure 4: Installation of the Sensor Device in the Science
Museum of the University of Coimbra, Portugal.
The backup power for the sensor device is
sufficient to support its operation only for three days
over the weekend. As the running cycle suggested,
data is collected, generated and stored. When there is
no network connection, data is then stored in the
device locally. Afterwards, the device checks for the
network connectivity. If it is available, data in local
storage will be immediately flushed and forwarded to
servers.
Processed data in backend server is then
downloaded by an application in a mobile phone for
reference use. As can be seen in the mobile
application in Figure 5, the data, such as the current
PM figures, temperature and humidity, collected by
the sensor device inside the museum is shown. The
data is also illustrated in the line chart for a particular
period of time. It enables the management to have an
in-depth look of the environment in the Museum,
assisting them to design policies to meet the
conservation objectives.
6 CONCLUSIONS
The historic buildings and collections in museums
need immediate conservation as there are many
threats that can cause damages to them. The self-
adaptive sensing IoT platform provides a solution to
the conservation work. The platform implements the
self-adaptive sensing method which enables its sensor
devices to adapt itself to the environment with
unstable power and network connection. This is a
break-through in deploying IoT technologies in such
an environment without causing any interruption to
the conservation work. As the IoT platform has been
deployed in the Science Museum, University
of Coimbra, Portugal since 2018, it is shown that the
Figure 5: Monitoring Result of the Sensor Device.
conservation work is running well. This work is still
in-progress and the development of sensor device is
on-going as scheduled.
ACKNOWLEDGEMENTS
This work was supported in part by the Macao
Polytechnic Institute - Environmental Monitoring of
UNESCO Coimbra Science Museum (RP/ESAP-
01/2019). We would also like to thank Prof. Adélio
Manuel Rodrigues Gaspar and Prof. José Joaquim da
Costa for their support of the Coimbra Science
Museum.
REFERENCES
Tse, R., Aguiari, D., Chou, K.-S., Giusto, D., Tang, S.-K.,
& Pau, G. (2018). Monitoring Cultural Heritage
Buildings via Low-Cost Edge Computing/Sensing
Platforms: The Biblioteca Joanina de Coimbra Case
Self-adaptive Sensing IoT Platform for Conserving Historic Buildings and Collections in Museums
397
Study. In GOODTECHS, 4th EAI International
Conference on Smart Objects and Technologies for
Social Good. ACM Press. New York, NY, 148-152.
https://dx.doi.org/10.1145/3284869.3284876
Maksimović, M., & Ćosović, M. (2019). Preservation of
Cultural Heritage Sites using IoT. In INFOTEH, 18th
International Symposium INFOTEH-JAHORINA.
IEEE. https://dx.doi.org/10.1109/INFOTEH.2019.
8717658
Neri, A., Corbellini, S., Parvis, M., Arcudi, L., Grassini, S.,
Piantanida, M., & Angelini, E. (2009). Environmental
Monitoring of Heritage Buildings. In EESMS, IEEE
Workshop on Environmental, Energy, and Structural
Monitoring Systems. IEEE. https://dx.doi.org/10.1109/
EESMS.2009.5341308
D'Amato, F., Gamba, P., & Goldoni, E. (2012). Monitoring
Heritage Buildings and Artworks with Wireless Sensor
Networks. In EESMS, IEEE Workshop on
Environmental Energy and Structural Monitoring
Systems. IEEE. https://dx.doi.org/10.1109/EESMS.
2012.6348392
Tse, R., & Pau, G. (2016). Enabling Street-Level Pollution
and Exposure Measures. In MobiHealth, 6th ACM
International Workshop on Pervasive Wireless
Healthcare. ACM Press. New York, NY, 1–4.
https://dx.doi.org/10.1145/2944921.2944925
Pereira, L. D., Gaspar, A. R., & Costa, J. J. (2017).
Assessment of the Indoor Environmental Conditions of
a Baroque Library in Portugal. In Energy Procedia,
133, 257–267. ELSEVIER. https://dx.doi.org/10.1016/
j.egypro.2017.09.385
Aguiari, D., Delnevo, G., Monti, L., Ghini, V., Mirri, S.,
Salomoni, P., Pau, G., Im, M., Tse, R., Ekpanyapong,
M., & Battistini, R. (2018). Canarin II: Designing a
Smart e-Bike Eco-System. In CCNC, 15th IEEE Annual
Consumer Communications Networking Conference.
IEEE. https://dx.doi.org/10.1109/CCNC.2018.8319221
Macau Cultural Affairs Bureau. (2019, October 5). Macao
Temple Fire Safety Guideline. Retrieved from
http://www.culturalheritage.mo/cn/detail/2247/1/
National Fire Protection Association. (2019, October 5).
NFPA914 Code for the Protection of Historic
Structures. Retrieved from https://www.nfpa.org/
codes-and-standards/all-codes-and-standards/list-of-
codes-and-standards/detail?code=914
Jensen, G., & Holmberg, J. G. (2006). Hypoxic Air Venting
for Protection of Heritage. Riksantikvaren and Historic
Scotland. Technical Conservation, Research and
Education Group.
Pau, G., & Tse, R. (2012). Challenges and Opportunities in
Immersive Vehicular Sensing: Lessons from Urban
Deployments. International Journal of Signal
Processing: Image Communication, 27(8), 900-908.
ELSEVIER. https://dx.doi.org/10.1016/j.image.2012.
01.015
Tétreault, J. (2003). Airborne Pollutants in Museums,
Galleries, and Archives: Risk Assessment, Control
Strategies, and Preservation Management. Canadian
Conservation Institute, Ottawa.
Szczepanowska, H. M. (2013). Conservation of Cultural
Heritage, Key Principles and Approaches. Routledge.
London.
Thickett, D., & Lee, LR. (2004). Selection of Materials for
the Storage or Display of Museum Objects. British
Museum Press, London. In: Caple, C., 2011. Preventive
Conservation in Museums. Routledge. London.
Florian, M. E. (2006). The Mechanisms of Deterioration in
Leather. In: Kite, M., Thomson, R., 2006. Conservation
of Leather and Related Materials. Routledge. London.
https://dx.doi.org/10.4324/9780080454665
Lavédrine, B. (2003). A Guide to the Preventive
Conservation of Photograph Collections. The Getty
Conservation Institute. Los Angeles, USA.
Peng, J.-W. (2017). Review of Nitrogen Oxide Damage to
Cultural Heritage and Preventive Conservation
Strategies. Museology Quarterly, 31(4), 91-103.
http://dx.doi.org/10.6686%2fMuseQ.2017.31.4.4
National Fire Protection Association. (2019, October 5).
NFPA909 Code for the Protection of Cultural Resource
Properties - Museums, Libraries, and Places of
Worship. Retrieved from https://www.nfpa.org/codes-
and-standards/all-codes-and-standards/list-of-codes-
and-standards/detail?code=909
UNESCO. (2019, September 7). Managing Tourism at
World Heritage Sites: Practical Manual for World
Heritage Site Managers. Retrieved from
http://whc.unesco.org/uploads/activities/documents/act
ivity-113-2.pdf
Public Broadcasting Service. (2019, October 7). 8 Things
You Didn’t Know About Humidity. Retrieved from
https://www.pbs.org/newshour/science/8-things-didnt-
know-humidity
Abdelrahman, Y., & Schmidt, A. (2018). Beyond the
Visible: Sensing with Thermal Imaging. Interactions,
26(1), 76-78. Association for Computing Machinery.
https://dx.doi.org/10.1145/3297778
Gade, R., & Moeslund, T. B. (2014). Thermal Cameras and
Applications: A Survey. Journal of Machine Vision and
Applications, 25(1), 245-262. Springer.
https://dx.doi.org/10.1007/s00138-013-0570-5
IoTBDS 2020 - 5th International Conference on Internet of Things, Big Data and Security
398