EAI Endorsed Transactions on Energy Web,
Год журнала:
2025,
Номер
12
Опубликована: Май 2, 2025
INTRODUCTION:
High
energy
consuming
enterprises
continue
to
pay
increasing
attention
consumption.
Therefore,
designing
an
management
system
is
significant.OBJECTIVES:
To
improve
the
level
and
economic
benefits
of
enterprises,
a
high
enterprise
design
based
on
Internet
Things
technology
neural
network
algorithms
proposed.METHODS:
devices
are
used
for
data
collection
transmission.
The
combination
model
prediction
optimization
can
achieve
real-time
monitoring,
prediction,
control
consumption.RESULTS:
research
results
indicated
that
response
time
proposed
in
study
was
80.2
ms
when
number
people
600.
fluctuation
range
CPU
usage
within
24
hours
14%
45%.CONCLUSION:
A
integrates
networks
manage
more
efficiently
intelligently,
thereby
improving
production
efficiency
enterprise.
This
helps
companies
gain
greater
advantages
fierce
market
competition.
Sensors,
Год журнала:
2025,
Номер
25(3), С. 706 - 706
Опубликована: Янв. 24, 2025
This
study
focuses
on
the
diagnostic
analysis
of
cartilage
damage
in
knee
joint
based
acoustic
signals
generated
by
joint.
The
research
utilizes
a
combination
advanced
signal
processing
techniques,
specifically
empirical
mode
decomposition
(EEMD)
and
detrended
fluctuation
(DFA),
alongside
convolutional
neural
networks
(CNNs)
for
classification
detection
tasks.
Acoustic
signals,
often
reflecting
mechanical
behavior
during
movement,
serve
as
non-invasive
tool
assessing
condition.
EEMD
is
applied
to
decompose
into
intrinsic
functions
(IMFs),
which
are
then
analyzed
using
DFA
quantify
scaling
properties
detect
irregularities
indicative
damage.
separation
individual
frequency
components
allows
multi-scale
with
each
resulting
from
local
variations
amplitude
over
time
allowing
effective
removal
noise
present
signal.
CNN
model
trained
features
extracted
these
accurately
classify
different
stages
degeneration.
proposed
method
demonstrates
potential
early
pathology,
providing
valuable
preventive
healthcare
reducing
need
invasive
procedures.
results
suggest
that
EEMD-DFA
feature
extraction
offers
promising
approach
assessment
Heliyon,
Год журнала:
2024,
Номер
10(15), С. e35407 - e35407
Опубликована: Июль 30, 2024
In
the
context
of
burgeoning
industrial
advancement,
there
is
an
increasing
trend
towards
integration
intelligence
and
precision
in
mechanical
equipment.
Central
to
functionality
such
equipment
rolling
bearing,
whose
operational
integrity
significantly
impacts
overall
performance
machinery.
This
underscores
imperative
for
reliable
fault
diagnosis
mechanisms
continuous
monitoring
bearing
conditions
within
production
environments.
Vibration
signals
are
primarily
used
because
they
provide
comprehensive
information
about
equipment's
condition.
However,
data
often
contain
high
noise
levels,
high-frequency
variations,
irregularities,
along
with
a
significant
amount
redundant
information,
like
duplication,
overlap,
unnecessary
during
signal
transmission.
These
characteristics
present
considerable
challenges
effective
feature
extraction
diagnosis,
reducing
accuracy
reliability
traditional
detection
methods.
research
introduces
innovative
methodology
bearings
using
deep
convolutional
neural
networks
(CNNs)
enhanced
variational
autoencoders
(VAEs).
learning
approach
aims
precisely
identify
classify
faults
by
extracting
detailed
vibration
features.
The
VAE
enhances
robustness,
while
CNN
improves
expressiveness,
addressing
issues
gradient
vanishing
explosion.
model
employs
reparameterization
trick
unsupervised
latent
features
further
trains
CNN.
system
incorporates
adaptive
threshold
methods,
"3/5"
strategy,
Dropout
VAE-CNN
different
types
at
rotational
speeds
typically
reaches
more
than
90
%,
it
achieves
generally
acceptable
result.
Meanwhile,
augmented
model,
after
experimental
validation
various
dimensions,
can
achieve
satisfactory
results
compared
several
representative
network
models
without
augmentation,
improving
robustness
diagnosis.
Energies,
Год журнала:
2023,
Номер
16(22), С. 7680 - 7680
Опубликована: Ноя. 20, 2023
Effective
fault
detection,
classification,
and
localization
are
vital
for
smart
grid
self-healing
mitigation.
Deep
learning
has
the
capability
to
autonomously
extract
characteristics
discern
categories
from
three-phase
raw
of
voltage
current
signals.
With
rise
distributed
generators,
conventional
relaying
devices
face
challenges
in
managing
dynamic
currents.
Various
deep
neural
network
algorithms
have
been
proposed
location.
This
study
introduces
innovative
detection
methods
using
Artificial
Neural
Networks
(ANNs)
one-dimension
Convolution
(1D-CNNs).
Leveraging
sensor
data
such
as
measurements,
our
approach
outperforms
contemporary
terms
accuracy
efficiency.
Results
IEEE
6-bus
system
showcase
impressive
rates:
99.99%,
99.98%
identifying
faulty
lines,
99.75%,
99.99%
98.25%,
96.85%
location
ANN
1D-CNN,
respectively.
emerges
a
promising
tool
enhancing
classification
within
grids,
offering
significant
performance
improvements.
