IEEE Transactions on Instrumentation and Measurement,
Journal Year:
2024,
Volume and Issue:
73, P. 1 - 9
Published: Jan. 1, 2024
In
actual
industrial
production,
differences
in
production
conditions
lead
to
variations
the
collected
data
distribution.
This
gives
rise
a
particular
problem:
while
one
set
of
has
complete
status
available,
another
only
possesses
from
healthy
state.
Differences
result
limitations
for
diagnosing
new
condition.
To
address
this
challenge,
method
based
on
envelope
order
spectra
generation
is
proposed.
Initially,
and
analysis
conducted
raw
vibration
align
across
different
domains
extract
domain-independent
signal
components—the
spectra.
Subsequently,
an
enhanced
Variational
Autoencoder
Generative
Adversarial
Network
(VAEGAN)
trained
using
The
model
then
employed
generate
synthetic
spectra,
serving
as
augmentation
working
conditions,
thereby
achieving
cross-domain
augmentation.
Next,
augmented
used
train
generic
fault
classification,
enabling
diagnosis.
Finally,
proposed
approach
validated
by
testing
it
with
real
Experimental
results
demonstrate
that
can
reliable
fake
under
diverse
accomplishing
diagnosis
preserving
privacy.
IEEE Transactions on Industrial Informatics,
Journal Year:
2024,
Volume and Issue:
20(8), P. 10008 - 10018
Published: May 1, 2024
Multisensor
information
fusion
techniques
based
on
deep
learning
are
crucial
for
machinery
fault
diagnosis.
However,
there
two
major
issues
in
previous
research.
First,
the
relationship
between
multisensor
samples
is
disregarded,
which
important
to
enhance
diagnostic
performance.
Second,
structure
of
algorithm
becomes
extremely
complex
with
prolonged
training
when
dealing
equipped
a
large
number
sensors.
To
address
aforementioned
issues,
our
study
proposes
new
mechanism
that
fuses
hypergraphs,
by
building
single-sensor
hypergraph
and
sensor
space
embed
as
nodes.
In
addition,
dual-branch
neural
network
designed
compute
hypergraphs
obtain
feature
representation
diagnose
faults.
The
validated
datasets
its
Sensors,
Journal Year:
2024,
Volume and Issue:
24(19), P. 6448 - 6448
Published: Oct. 5, 2024
Rolling
bearings
often
produce
non-stationary
signals
that
are
easily
obscured
by
noise,
particularly
in
high-noise
environments,
making
fault
detection
a
challenging
task.
To
address
this
challenge,
novel
diagnosis
approach
based
on
the
Kolmogorov-Arnold
Network-based
Hypergraph
Message
Passing
(KAN-HyperMP)
model
is
proposed.
The
KAN-HyperMP
composed
of
three
key
components:
neighbor
feature
aggregation
block,
fusion
and
KANLinear
block.
Firstly,
block
leverages
hypergraph
theory
to
integrate
information
from
more
distant
neighbors,
aiding
reduction
noise
impact,
even
when
nearby
neighbors
severely
affected.
Subsequently,
combines
features
these
higher-order
with
target
node's
own
features,
enabling
capture
complete
structure
hypergraph.
Finally,
smoothness
properties
B-spline
functions
within
Network
(KAN)
employed
extract
critical
diagnostic
noisy
signals.
proposed
trained
evaluated
Southeast
University
(SEU)
Jiangnan
(JNU)
Datasets,
achieving
accuracy
rates
99.70%
99.10%,
respectively,
demonstrating
its
effectiveness
under
both
noise-free
conditions.
Journal of Marine Science and Engineering,
Journal Year:
2023,
Volume and Issue:
11(3), P. 616 - 616
Published: March 14, 2023
Hydraulic
axial
piston
pumps
are
the
power
source
of
fluid
systems
and
have
important
applications
in
many
fields.
They
a
compact
structure,
high
efficiency,
large
transmission
power,
excellent
flow
variable
performance.
However,
crucial
components
easily
suffer
from
different
faults.
It
is
therefore
to
investigate
precise
fault
identification
method
maintain
reliability
system.
The
use
deep
models
feature
learning,
data
mining,
automatic
identification,
classification
has
led
development
novel
diagnosis
methods.
In
this
research,
typical
faults
wears
friction
pairs
were
analyzed.
Different
working
conditions
considered
by
monitoring
outlet
pressure
signals.
To
overcome
low
efficiency
time-consuming
nature
traditional
manual
parameter
tuning,
Bayesian
algorithm
was
introduced
for
adaptive
optimization
an
established
learning
model.
proposed
can
explore
potential
information
signals
adaptively
identify
main
types.
average
diagnostic
accuracy
found
reach
up
100%,
indicating
ability
detect
with
precision.
IEEE Transactions on Instrumentation and Measurement,
Journal Year:
2024,
Volume and Issue:
73, P. 1 - 9
Published: Jan. 1, 2024
In
actual
industrial
production,
differences
in
production
conditions
lead
to
variations
the
collected
data
distribution.
This
gives
rise
a
particular
problem:
while
one
set
of
has
complete
status
available,
another
only
possesses
from
healthy
state.
Differences
result
limitations
for
diagnosing
new
condition.
To
address
this
challenge,
method
based
on
envelope
order
spectra
generation
is
proposed.
Initially,
and
analysis
conducted
raw
vibration
align
across
different
domains
extract
domain-independent
signal
components—the
spectra.
Subsequently,
an
enhanced
Variational
Autoencoder
Generative
Adversarial
Network
(VAEGAN)
trained
using
The
model
then
employed
generate
synthetic
spectra,
serving
as
augmentation
working
conditions,
thereby
achieving
cross-domain
augmentation.
Next,
augmented
used
train
generic
fault
classification,
enabling
diagnosis.
Finally,
proposed
approach
validated
by
testing
it
with
real
Experimental
results
demonstrate
that
can
reliable
fake
under
diverse
accomplishing
diagnosis
preserving
privacy.
