Symmetry,
Journal Year:
2025,
Volume and Issue:
17(3), P. 406 - 406
Published: March 8, 2025
To
enhance
the
maintenance
efficiency
and
operational
stability
of
rolling
bearings,
this
work
establishes
a
methodology
for
bearing
life
prediction,
employing
digital
twin
systems
to
evaluate
remaining
useful
bearings.
A
comprehensive
twin-integrated
model
entire
lifecycle
bearings
is
constructed
using
Modelica
language.
This
generates
sufficient
reliable
data
Due
symmetrical
physical
structure
generated
also
have
symmetry.
Based
on
characteristic
(RUL)
prediction
algorithm
developed
recurrent
neural
network
(RNN),
specifically
an
improved
gated
unit
(GRU)
model.
An
optimization
employed
adjust
hyperparameters
determine
initial
fault
point
bearing.
multi-feature
dataset
constructed,
effectively
enhancing
precision
reliability
lifespan
estimation.
existing
measured
bearing’s
lifecycle,
parameters
are
updated.
Through
parameter
degradation
component
twin,
generated.
By
combining
with
actual
measurement
data,
method
addresses
limitations
traditional
approaches
in
situations
where
complete
scarce,
providing
technical
support
intelligent
Measurement Science and Technology,
Journal Year:
2024,
Volume and Issue:
35(8), P. 086001 - 086001
Published: April 23, 2024
Abstract
Conventional
convolutional
neural
networks
(CNNs)
predominantly
emphasize
spatial
features
of
signals
and
often
fall
short
in
prioritizing
sequential
features.
As
the
number
layers
increases,
they
are
prone
to
issues
such
as
vanishing
or
exploding
gradients,
leading
training
instability
subsequent
erratic
fluctuations
loss
values
recognition
rates.
To
address
this
issue,
a
novel
hybrid
model,
termed
one-dimensional
(1D)
residual
network
with
attention
mechanism
bidirectional
gated
recurrent
unit
(BGRU)
is
developed
for
rotating
machinery
fault
classification.
First,
1D
optimized
structure
constructed
obtain
mitigate
gradient
exploding.
Second,
(AM)
designed
catch
important
impact
characteristics
samples.
Next,
temporal
mined
through
BGRU.
Finally,
feature
information
summarized
global
average
pooling,
fully
connected
layer
utilized
output
final
classification
result
diagnosis.
The
technique
which
tested
on
one
set
planetary
gear
data
three
different
sets
bearing
data,
has
achieved
accuracy
98.5%,
100%,
respectively.
Compared
other
methods,
including
CNN,
CNN-BGRU,
CNN-AM,
CNN
an
AM-BGRU,
proposed
highest
rate
stable
diagnostic
performance.
IEEE Journal of Radio Frequency Identification,
Journal Year:
2024,
Volume and Issue:
8, P. 282 - 321
Published: Jan. 1, 2024
In
the
rapidly
evolving
landscape
of
Industry
4.0,
digital
twins
have
emerged
as
a
transformative
technology
across
various
industrial
sectors.
This
paper
presents
comprehensive,
in-depth
review
twin
models
in
terms
concept
and
evolution,
fundamental
components
frameworks,
existing
based
on
their
functionalities.
The
also
discusses
how
are
used/adopted
different
industries
highlights
challenges
potential
solutions
to
address
current
issues.
aims
provide
researchers
industry
professionals
with
clear
insight
into
unique
benefits
applications
models.
will
help
comprehend
significance
for
specific
purposes
foster
advancement
state-of-the-art
techniques
this
field.
Machines,
Journal Year:
2025,
Volume and Issue:
13(1), P. 43 - 43
Published: Jan. 10, 2025
Bearings
are
critical
components
in
mechanical
systems,
and
their
degradation
process
typically
exhibits
distinct
stages,
making
stage-based
remaining
useful
life
(RUL)
prediction
highly
valuable.
This
paper
presents
a
model
that
combines
correlation
analysis
feature
extraction
with
Graph
Neural
Network
(GNN)-based
approach
for
bearing
stage
classification
RUL
prediction,
aiming
to
achieve
accurate
prediction.
First,
the
proposed
Pearson–Spearman
metric,
along
Kernel
Principal
Component
Analysis
(KPCA)
autoencoders,
is
used
group
fuse
health
indicators
(HIs),
thereby
obtaining
indicator
(HI)
effectively
reflects
process.
Then,
combining
Convolutional
(GCN)
Long
Short-Term
Memory
(LSTM)
networks
classification.
Based
on
results,
Adaptive
Attention
GraphSAGE–LSTM
(AAGL)
model,
also
introduced
this
study,
employed
precisely
predict
bearing’s
life.
Lubricants,
Journal Year:
2025,
Volume and Issue:
13(2), P. 81 - 81
Published: Feb. 12, 2025
Due
to
the
complex
changes
in
physical
and
chemical
properties
of
rolling
bearings
from
degradation
failure,
most
model-driven
data-driven
methods
generally
suffer
insufficient
accuracy
robustness
predicting
remaining
useful
life
bearings.
To
address
this
challenge,
paper
proposes
a
artificial
neural
network
method,
namely
CNN-LSTM
bearing
prediction
model
based
on
fruit
fly
optimization
algorithm
(FOA).
This
method
utilizes
deep
feature
mining
capabilities
convolutional
networks
(CNN)
long
short-term
memory
(LSTM)
effectively
extract
spatial
features
temporal
information
sequences
dataset.
In
addition,
introducing
FOA
enables
dynamically
adjust
hidden
layers
thresholds
while
optimizing
optimal
path,
thereby
finding
best
solution.
article
conducts
ablation
experiments
using
acceleration
dataset
IEEE
PHM
2012
The
experimental
results
show
that
FOA-CNN-LSTM
proposed
significantly
outperforms
other
comparative
RUL
stability,
verifying
its
effectiveness
innovation
dealing
with
processes.
helps
take
preventive
measures
before
faults
occur,
reducing
economic
losses
having
important
practical
significance
for
