Gas
leakage
detection
in
different
industrial
sectors
is
enormously
important
for
safe
operation.
It
vital
to
quickly
and
automatically
detect
identify
the
type
of
gas
order
prevent
environmental
damage
protect
human
lives.
Existing
approaches
mainly
rely
on
electronic
noses
which
have
several
limitations
should
be
kept
within
region.
Lately,
novel
been
proposed
based
thermal
infrared
sensors
can
capture
heat
patterns
at
a
distance
far
from
leakage.
Motivated
by
success
artificial
intelligence
such
as
deep
learning
applications.
Combining
with
images
could
effectively
improve
accuracy.
In
this
study,
learning-based
pipeline
imaging
differentiate
between
categories.
Multiple
convolutional
neural
networks
(CNN)
models
are
used
feature
extraction
leading
spatial
features.
These
features
then
analyzed
via
fast
Walsh
Hadamard
transform
(FWHT).
Next,
these
integrated
using
principal
component
analysis
fed
machine
classifiers
detection.
The
accuracy
attained
98.0%
suggests
that
integration
method
has
improved
performance
Nano-Micro Letters,
Год журнала:
2024,
Номер
16(1)
Опубликована: Авг. 14, 2024
Abstract
As
information
acquisition
terminals
for
artificial
olfaction,
chemiresistive
gas
sensors
are
often
troubled
by
their
cross-sensitivity,
and
reducing
cross-response
to
ambient
gases
has
always
been
a
difficult
important
point
in
the
sensing
area.
Pattern
recognition
based
on
sensor
array
is
most
conspicuous
way
overcome
cross-sensitivity
of
sensors.
It
crucial
choose
an
appropriate
pattern
method
enhancing
data
analysis,
errors
improving
system
reliability,
obtaining
better
classification
or
concentration
prediction
results.
In
this
review,
we
analyze
mechanism
We
further
examine
types,
working
principles,
characteristics,
applicable
detection
range
algorithms
utilized
gas-sensing
arrays.
Additionally,
report,
summarize,
evaluate
outstanding
novel
advancements
methods
identification.
At
same
time,
work
showcases
recent
utilizing
these
identification,
particularly
within
three
domains:
ensuring
food
safety,
monitoring
environment,
aiding
medical
diagnosis.
conclusion,
study
anticipates
future
research
prospects
considering
existing
landscape
challenges.
hoped
that
will
make
positive
contribution
towards
mitigating
gas-sensitive
devices
offer
valuable
insights
algorithm
selection
applications.
The
role
of
artificial
intelligence
(AI),
machine
learning
(ML),
and
deep
(DL)
in
enhancing
automating
gas
sensing
methods
the
implications
these
technologies
for
emergent
sensor
systems
is
reviewed.
Applications
AI-based
intelligent
sensors
include
environmental
monitoring,
industrial
safety,
remote
sensing,
medical
diagnostics.
AI,
ML,
DL
can
process
interpret
complex
data,
allowing
improved
accuracy,
sensitivity,
selectivity,
enabling
rapid
detection
quantitative
concentration
measurements
based
on
sophisticated
multiband,
multispecies
systems.
These
discern
subtle
patterns
signals,
to
readily
distinguish
between
gases
with
similar
signatures,
adaptable,
cross-sensitive
multigas
under
various
conditions.
Integrating
AI
technology
represents
a
paradigm
shift,
achieve
unprecedented
performance,
adaptability.
This
review
describes
while
highlighting
approaches
AI–sensor
integration.
Applied Sciences,
Год журнала:
2025,
Номер
15(7), С. 3430 - 3430
Опубликована: Март 21, 2025
Indoor
transport
robots
are
currently
a
key
robotics
application
in
large
industrial
assembly
lines,
and
similar
future
deployment
as
indoor
mobile
delivery
horizontal
or
vertical
buildings
can
be
expected.
This
accelerated
if
the
robot
is
also
capable
of
performing
other
valuable
tasks
within
buildings.
In
this
direction,
paper
presents
first
results
obtained
by
embedding
compact,
low-power
electronic
nose
(also
known
an
eNose)
robot.
The
objective
implementation
evaluation
early
detector
gas
leaks.
general
advantage
using
sensing
capabilities
eNose
that
it
simultaneously
trained
to
detect
single
specific
complex
odor
composed
various
volatile
chemical
compounds.
Experimental
real
operation
conditions
have
confirmed
embedded
with
compact
ethanol
leaks
while
making
package
inside
building.
Sensors,
Год журнала:
2024,
Номер
24(18), С. 5904 - 5904
Опубликована: Сен. 11, 2024
The
identification
of
gas
leakages
is
a
significant
factor
to
be
taken
into
consideration
in
various
industries
such
as
coal
mines,
chemical
industries,
etc.,
well
residential
applications.
In
order
reduce
damage
the
environment
human
lives,
early
detection
and
type
are
necessary.
main
focus
this
paper
multimodal
data
that
were
obtained
simultaneously
by
using
multiple
sensors
for
thermal
imaging
camera.
As
reliability
sensitivity
low-cost
less,
they
not
suitable
over
long
distances.
overcome
drawbacks
relying
just
on
identify
gases,
camera
capable
detecting
temperature
changes
also
used
collection
current
dataset
comprises
6400
samples,
including
smoke,
perfume,
combination
both,
neutral
environments.
paper,
convolutional
neural
networks
(CNNs)
trained
image
data,
utilizing
variants
bidirectional
long-short-term
memory
(Bi-LSTM),
dense
LSTM,
fusion
both
datasets
effectively
classify
comma
separated
value
(CSV)
from
sensors.
can
valuable
source
research
scholars
system
developers
improvise
their
artificial
intelligence
(AI)
models
leakage
detection.
Furthermore,
ensure
privacy
client's
explores
implementation
federated
learning
privacy-protected
classification,
demonstrating
comparable
accuracy
traditional
deep
approaches.
IEEE Sensors Journal,
Год журнала:
2024,
Номер
24(9), С. 15598 - 15606
Опубликована: Март 18, 2024
With
the
extension
of
storage
period,
nutritional
components
soybeans
are
lost,
and
quality
loss
is
severe,
but
appearance
difference
not
obvious.
Low-quality
often
misrepresented
as
high-quality
soybeans.
In
this
work,
an
adaptive
deep
learning
approach
proposed,
integrating
with
electronic
nose
(e-nose)
system,
to
effectively
identify
different
periods.
First,
PEN3
e-nose
system
applied
obtain
gas
information
under
two
conditions.
Second,
a
multispace
self-attention
mechanism
(MSM)
proposed
selectively
import
features
influencing
classification
performance.
A
lightweight
network
based
on
attention
designed
(MSM-Net).
Finally,
by
conducting
ablation
experiments
comparing
state-of-the-art
methods,
MSM-Net
demonstrates
superior
results.
Under
temperature
25
°C
relative
humidity
75%
RH,
accuracy
98.50%,
precision
98.54%,
recall
98.48%
achieved.
45%
96.50%,
96.62%,
96.85%
The
findings
suggest
that
integration
offers
effective
detection
method
for
monitoring
Sensors,
Год журнала:
2024,
Номер
24(1), С. 302 - 302
Опубликована: Янв. 4, 2024
With
the
development
of
field
e-nose
research,
potential
for
application
is
increasing
in
various
fields,
such
as
leak
measurement,
environmental
monitoring,
and
virtual
reality.
In
this
study,
we
characterize
electronic
nose
data
structured
investigate
analyze
learning
efficiency
accuracy
deep
models
that
use
unstructured
data.
For
purpose,
MOX
sensor
dataset
collected
a
wind
tunnel,
which
one
most
popular
public
datasets
research.
Additionally,
gas
detection
platform
was
constructed
using
commercial
sensors
embedded
boards,
experimental
were
hood
environment
used
chemical
experiments.
We
investigated
networks,
convolutional
neural
long
short-term
memory,
well
boosting
models,
are
robust
on
data,
both
specially
datasets.
The
results
showed
had
faster
more
performance
than
series
models.
