When David Celeita Rodríguez, former director
of the Energy Systems Engineering
Program, promoted by the School
of Engineering, Science and Technology
at Universidad del Rosario, began to
conceive his research, he recalled Marie
Curie. The Polish scientist, a pioneer in
the study and description of radioactivity,
was the first woman to receive a Nobel Prize and
the only person to date to win it in two different categories
of science: physics (1903) and chemistry (1911).
Thanks to their findings, the development of x-rays was
achieved. Celeita Rodríguez evoked that revolutionary
scientist because, if with electromagnetic waves it was
possible to see the deepest part of human anatomy, her
study allowed her to listen inside energy systems, as if
it were an x-ray to electrical equipment to examine its
state and diagnose or forecast possible damage.
At the beginning of 2023, while working
as a professor of Universidad del Rosario, Celeita
Rodríguez gathered an interdisciplinary
group of undergraduate students in Applied
Mathematics and Computer Science, as well
as in Energy Systems Engineering, with the
purpose of exploring the field of classification
and intelligent detection of faults in electrical
equipment by taking thermographic images,
that is, photographs taken at a distance that
allow the temperature of energy systems to
identify defective functions. In his role as
leader during that year, he guided students
to be an essential part of the process of observation,
analysis and disclosure, in an area of
study that has gained relevance in the world.
Thermographic
images of
engines
organized in
five groups,
analyzed with
grayscale and
corresponding
Jaccard index,
taken from
the scientific
paper Smart
equipment
failure detection
with machine
learning applied
to thermography
inspection
data in modern
power systems.
The idea to forge this project arose from a
call from the Institute of Electrical and Electronic
Engineers (IEEE), an organization
based in New York (United States), that promotes
scientific advances and innovations
within these fields of knowledge. The primary
requirement demanded by the American
entity was that students should play a stellar
role within the process.
In addition, the research team had the scientific backing
of the assistant professor of Electronic Engineering and researcher
of the French university Centrale Supélec, Trung
Dung Le. This is how students Ana María Garzón, Natalia
Laiton and Víctor Sicachá dared to contribute from their
knowledge with the purpose of the study.
Soon the work deployed would begin to provide rewards
for his still new academic life. The findings of his
research activity were initially published in the IEEE indexed
journal, through an article called “ Smart equipment failure detection with machine learning applied to thermography inspection data in modern power systems”.
One more will be released, two others have already
been accepted and one remains under review. With this,
the socialization stage of the project is in full consolidation
and seeks a high impact within the industry in the short,
medium and long term.
However, what is the research about, and why is it so
relevant? As a first step, the team built a state of the art
from which they were able to identify those similar studies
that have been carried out around the world. From these
works a database and images were set up.
Then came the experimental stage, which consisted
of using infrared cameras to measure the temperature of
electrical and electronic equipment, such as transformers and motors, in enclosed spaces. Once the thermographic
images were achieved, artificial intelligence and machine
learning techniques were used, with the intention of processing
these photographs and introducing an algorithm
that allowed their classification into clusters.
Ana María Garzón, a student of Applied Mathematics
and Computer Science at URosario and member of
the team that presented the research advances at the
conference held in France, argues that “we are doing
cutting-edge things, not only for Colombia and Latin
America, but for the world...”
Within the study sample, 624 thermographic images
were obtained (369 of induction motors and 255 of transformers).
From an unsupervised machine learning method
called vector support machine it was possible to classify the
photographs of electrical equipment into two categories: hot
and cold, which allowed to determine abnormal temperatures
in the systems.
By using a grayscale, in addition to observing the pixels, it
was possible to establish what type of damage could be identified.
After that, the images were segmented to find the regions
of interest. There, when removing the background of
the photos obtained and focusing on a specific area of the devices,
those that presented defects became evident; likewise,
the characteristics found according to the grayscale configured
elements to anticipate and predict eventual failures.
As a result of the procedure, defects were found in the systems
with an accuracy of 90 percent, according to the Jaccard
index, which measures the rate of similarity between
two data sets. According to experts, this achievement contributes
to the prediction of errors in electrical systems long
before they occur and allows decisions to be made towards
the future.
Celeita Rodríguez, who holds a master's degree and doctorate
in Electrical Engineering, explains that the research
is born from the idea of contributing to the development of
the "internet of things" (a concept used to define a system
of electronic devices interconnected through a wireless
network and that allows sharing data or information) within
power plants, especially in sectors such as industry and
commerce, whose operation depends in large part on devices
vulnerable to different types of disruptions.
“Many decades ago we worked only with analogous electrical
signals, such as the use of thermostats to regulate temperature.
Today, with the conjuncture of industry 4.0 (also
called the fourth industrial revolution), which involves the
use of artificial intelligence, digitalization, interconnectivity
and data capture in real time, we can take infrared photographs
and see how the equipment is inside, which allows
us to obtain valuable information,” explains the researcher.
Therefore, it is no longer just a question of capturing images
of the present, but it is also possible to predict defects, in order
to make proactive decisions, one of the main objectives
of the project that he leads.
The value of detection and prediction
Humans have built a contemporary civilization that is based
on electrical systems. In order for the dynamics of society to
continue to develop smoothly, its proper functioning must
be monitored and regular maintenance carried out. In this
scenario, an inconsistency in any device, derived, for example, from an abnormal condition that alters
the voltage or current, is capable of generating
serious disturbances at large economic, social
and human scales. A fire caused by a defect in
a factory’s equipment not only leads to material
losses, but also to a risk to workers’ lives.
Víctor Sicachá, a student of Applied Mathematics
and Computer Science, adds that his contribution to
the project has constituted a breaking point in his
professional perspectives. “If I was not a researcher
before, I may now have an open door to be one later.
This team has given me that possibility.”
According to Professor Celeita, the conventional
maintenance of electrical systems
means the interruption in about 45 percent
of operations, which also implies an economic
investment in the evaluation of equipment.
This has become a constant concern for the
industrial sector. Therefore, the classification
and detection of faults by thermal imaging
are an effective and non-invasive option as a
predictive method and prevention of defects
that can scale and cause major damage.
When it comes to protection and inspection
in the safety of operators, “human life is
what prevails in the face of a possible alteration
that can generate explosions of transformers
and large volumes of energy,” Celeita
emphasizes. Second is the team: “If we
can protect it before catastrophe strikes, so
much the better.” Identifying defects from
thermal imaging thus helps to avoid disruption
to essential electrical systems, such as
housing, retail and hospital infrastructure.
“By analyzing the photographs and exploring
the inside of the devices, we become a
kind of physician of the energy systems, to
make preventive and proactive decisions,” he
complements.
For Camilo Salazar Palacio, an electronic
engineer from the Julio Garavito Colombian
School of Engineering and a professor in the
School of Exact Sciences and Engineering at
Universidad Sergio Arboleda, “a system for
predicting and detecting breakdowns early
through the use of thermographic images,
and with an adequate level of training, can
substantially reduce accidents in any scenario.”
It highlights that, for example, high temperatures
can cause incidents such as short
circuits and explosions, so being able to identify
these anomalies in the operating ranges
of the devices allows to reduce the occurrence
of possible thermal accidents.
The impact of research
Although the exploration of thermographic
images as a method of classifying and detecting
faults in electrical equipment has gained
space in various research centers around the
world, in Colombia this approach has only
just begun to occupy a preponderant place in
energy systems engineering laboratories. It is
for this reason that the study led by Universidad del Rosario
has aroused the interest of academic institutions in Europe
and the United States.
The project was driven by theParis Electronic Engineering
Group (GeePs)of the Centrale Supélec University, and
co-funded by theZucker Faculty Grant del IEEE, as well
as by the School of Engineering, Science and Technology of
Universidad del Rosario, the Society of Industry Applications
(IAS) of the IEEE, and theIEEE Foundation, and the IEEE
Foundation.
The awards for this team have not been long in coming.
In June 2023, they received the Outstanding Young Scientist
Award, as part of the 11th International Conference on Smart
Grid(icSmart Grid) held in Paris.
Applied
Mathematics
and Computer
Science
students
Ana María
Garzón and
Natalia Laiton
received the
Outstanding
Young Scientist
Award for their
research at
the Smart Grid
2023 in Paris,
France.
“It is a milestone that undergraduate students achieve a
certain prominence or visualization in international recognition,”
Celeita Rodríguez highlights while sharing that academics
and doctoral candidates from French engineering institutions
were surprised to learn about the level of scientific
work that is being developed at Universidad del Rosario. “It is
very significant to realize that these industry interests can be
linked from very early on during undergraduate training, in
courses, in nurseries and in research groups,” he emphasizes.
On her part, Ana María Garzón, a student of Applied Mathematics
and Computer Science at the institution and member
of the team that presented the advances of research at the congress
held in France, argues that “we are doing cutting-edge
things, not only for Colombia and Latin America, but for the
world. I believe that high-level research is being done at Universidad
del Rosario. Being able to participate in the conference
I learned a lot from the industry.” In addition, she notes
that thermographic imaging is relatively new, so finding datasets
that show faults and the information required is not so
straightforward. All this, in her opinion, is a great reward for
the work ahead.
Meanwhile, Víctor Sicachá, also a student of Applied Mathematics
and Computer Science, adds that his contribution to
the project has constituted a breaking point in his professional
perspectives. “If I was not a researcher before, I may now have
an open door to be one later. This team has given me that possibility.”
Natalia Laiton, student of Applied Mathematics and
Computer Science, was also part of the research team
and decided to contribute from her knowledge for the
purpose of the study.
Prospects for the future
One of the objectives of the project is to extend its focus of
study towards energy systems in open spaces, such as photovoltaic
solar panels, which are currently in vogue today due to
the imminent energy transition. This implies
an effort in terms of exploration, but also an
economic challenge.
Replicating the project's thermal imaging
model in open space systems means greater
investment in research and technology.
Drones and thermal cameras are needed,
among other tools that are relatively expensive.
Today, says Professor Celeita Rodríguez,
those who develop thermography have
enough infrastructure to monitor equipment
in closed spaces, but the field of action is migrating
to open areas. “We are going there, we
are in the process of acquiring the thermal
camera and thus being able to work with our
own equipment,” he concludes.
In October of last year, the team was invited
to a conference in Nashville (United States)
to present the progress of their project and socialize
a new publication that describes their
scientific findings, with which the scope of
the study continues to expand. Meanwhile, in
the URosario Renewable Energy Laboratory, which operates with solar panels and a wind
turbine located on the building’s roof, students
continue to debate and build the course
of their research together.
