Measurement Science and Technology,
Год журнала:
2024,
Номер
35(4), С. 046119 - 046119
Опубликована: Янв. 17, 2024
Abstract
Transceiver
is
a
crucial
component
of
radar
system
that
allows
for
the
regulation
signal
phase
and
amplitude
as
well
amplification
both
transmitted
received
signals.
Its
operational
efficiency
has
significant
impact
on
whole
dependability
system.
To
ensure
safe
reliable
operation
system,
an
optimized
sparse
deep
belief
network
with
momentum
factor
developed
to
diagnose
potential
faults
transceivers.
Firstly,
term
added
into
parameter
update
enhance
anti-oscillation
ability
model
parameters
in
training,
while
regular
integrated
prevent
from
overfitting.
Secondly,
automatically
configure
hyper-parameters,
hybrid
sine
cosine
algorithm
(HSCA)
dynamic
inertia
weight
adaptive
strategies
proposed.
Thus,
effective
diagnostic
named
HSCA-MS-DBN
formed
by
combining
HSCA.
The
proposed
confirmed
using
actual-world
transceiver
dataset,
findings
experiments
reveal
this
surpasses
multiple
prominent
intelligent
models.
Electronics,
Год журнала:
2023,
Номер
12(10), С. 2263 - 2263
Опубликована: Май 16, 2023
Backscatter
communication
(BC)
is
a
promising
technology
for
low-power
and
low-data-rate
applications,
though
the
signal
detection
performance
limited
since
backscattered
usually
much
weaker
than
original
signal.
When
poor,
backscatter
device
(BD)
may
not
be
able
to
accurately
detect
interpret
incoming
signal,
leading
errors
degraded
quality.
This
can
result
in
data
loss,
slow
transfer
rates,
reduced
reliability
of
link.
paper
proposes
novel
approach
improve
systems
using
evolutionary
deep
learning.
In
particular,
we
focus
on
training
convolutional
neural
networks
(DCNNs)
BC.
We
first
develop
hybrid
algorithm
based
artificial
bee
colony
(ABC),
biogeography-based
optimization
(BBO),
particle
swarm
(PSO)
optimize
architecture
DCNN,
followed
by
large
set
benchmark
datasets.
To
ABC,
migration
operator
BBO
used
exploitation.
Moving
towards
global
best
PSO
also
proposed
exploration
ABC.
Then,
take
advantage
bit-error
rate
(BER)
studied
BC
system.
The
simulation
results
demonstrate
that
has
show
significantly
improves
signals
compared
existing
works.
IEEE Transactions on Instrumentation and Measurement,
Год журнала:
2023,
Номер
72, С. 1 - 12
Опубликована: Янв. 1, 2023
Effective
fault
diagnosis
of
critical
components
is
essential
to
ensure
the
safe
and
reliable
operation
entire
system.
This
paper
deals
with
transmitter/receiver
module,
which
a
component
in
phased
array
radar
system,
by
proposing
novel
deep
belief
network
learning
method.
A
sparse
based
on
Gaussian
function
first
constructed
automatically
learn
relationship
between
monitoring
data
health
conditions.
With
trained
network,
pseudo-labels
are
produced
for
unlabeled
samples,
while
information
entropy
employed
calculate
confidence
levels
reflecting
their
certainty
reduce
effect
pseudo-label
noise.
The
pseudo-labeled
samples
high
added
training
set
retrain
network.
Optimal
model
configuration
parameters
obtained
through
chaos
game
optimization
algorithm.
effectiveness
proposed
method
verified
real-world
dataset
from
certain
type
radar.
experiments
show
that
mean
identification
rate
this
can
reach
96.33%,
not
only
exceeds
some
network-based
modeling
methods,
but
also
other
intelligent
methods.
Processes,
Год журнала:
2023,
Номер
11(7), С. 1875 - 1875
Опубликована: Июнь 22, 2023
To
enhance
fault
characteristics
and
improve
detection
accuracy
in
bearing
vibration
signals,
this
paper
proposes
a
diagnosis
method
using
wavelet
packet
energy
spectrum
an
improved
deep
confidence
network.
Firstly,
transform
decomposes
the
original
signal
into
different
frequency
bands,
fully
preserving
signal’s
information,
constructs
feature
vectors
by
extracting
of
sub-frequency
bands
via
to
extract
information.
Secondly,
minimize
time-consuming
manual
parameter
adjustment
procedure
increase
diagnostic
accuracy,
sparrow
search
algorithm–deep
belief
network
is
proposed,
which
utilizes
algorithm
optimize
hyperparameters
networks
reduce
classification
error
rate.
Finally,
verify
effectiveness
method,
rolling
data
from
Casey
Reserve
University
were
selected
for
verification,
compared
other
commonly
used
algorithms,
proposed
achieved
100%
99.34%
two
sets
comparative
experiments.
The
experimental
results
demonstrate
that
has
high
rate
stability.
Measurement Science and Technology,
Год журнала:
2024,
Номер
35(12), С. 125012 - 125012
Опубликована: Сен. 4, 2024
Abstract
Traditional
diagnostic
methods
often
have
insufficient
accuracy
and
noise
reduction,
which
leads
to
errors.
To
address
these
issues,
this
paper
proposes
an
advanced
fault
diagnosis
model
that
combines
the
variational
mode
decomposition
(VMD)
improved
by
a
Variable-Objective
Search
Whale
Optimization
Algorithm
(VSWOA)
with
Pelican
(PO)-boosted
Kernel
Extreme
Learning
Machine
(KELM)
algorithm.
The
application
of
method
is
shown
here
in
rolling
bearings.
proposed
VSWOA
enhances
performance
VMD
incorporating
Sobol
sequence,
nonlinear
time-varying
factors,
multi-objective
initial
search
strategy,
elite
Cauchy
chaos
mutation
significantly
improving
reduction
vibration
signals.
Fault
information
precisely
extracted
using
waveform
sample
entropy,
composite
multiscale
fuzzy
enables
effective
feature
screening
dimensionality
reduction.
POA
fine-tunes
KELM
parameters,
increasing
classification
accuracy.
effectiveness
verified
through
experimental
evaluations
bearing
data
injected
Gaussian
(from
Case
Western
Reserve
University)
SpectraQuest
datasets,
where
significant
improvements
detection
are
achieved.
Fault
of
rolling
bearing
signal
is
a
common
problem
encountered
in
the
production
life.
Identifying
fault
helps
to
locate
location
and
type
quickly,
react
time,
reduce
losses
caused
by
failure
production.
In
order
accurately
identify
signal,
this
paper
presents
triple
feature
extraction
classification
method
based
on
multi-scale
dispersion
entropy
(MDE)
permutation
(MPE),
extracts
features
when
it
working,
uses
algorithm
determine
whether
there
fault.
Scale
2
MDE
combined
with
scale
1
MPE
as
three
required
for
experiment.
As
comparison
recognition
results,
(MSE)is
introduced.
Ten
scales
are
calculated,
all
combinations
obtained.
K
nearest
neighbor
used
recognition.
The
result
shows
that
combination
rate
proposed
reaches
96.2%,
which
best
among
combinations.
Computational Intelligence and Neuroscience,
Год журнала:
2022,
Номер
2022, С. 1 - 11
Опубликована: Окт. 15, 2022
Balancing
machine
is
a
general
equipment
for
dynamic
balance
verification
of
rotating
parts,
whether
it
breaks
down
or
does
not
determine
the
accuracy
verification.
In
order
to
solve
problem
insufficient
fault
diagnosis
balancing
machine,
method
based
on
Improved
Sparrow
Search
Algorithm
(ISSA)
optimized
Extreme
Learning
Machine
(ELM)
was
proposed.
Firstly,
iterative
chaos
mapping
and
Fuch
were
introduced
initialize
population
increase
diversity.
Secondly,
adaptive
factor
Levy
flight
strategy
also
update
individual
positions
improve
model
convergence
speed.
Finally,
feature
vector
input
ISSA-ELM
with
type
as
output.
The
experiment
showed
that
high
99.17%,
which
1.67%,
2.50%,
7.50%,
17.50%
higher
than
SSA-ELM,
HHO-ELM,
PSO-ELM,
ELM,
respectively,
further
improving
prediction
operation
state
machine.
2021 CAA Symposium on Fault Detection, Supervision, and Safety for Technical Processes (SAFEPROCESS),
Год журнала:
2023,
Номер
unknown, С. 1 - 6
Опубликована: Сен. 22, 2023
Transceiver
is
one
of
the
most
important
components
radar,
and
its
efficiency
greatly
affects
reliability
entire
radar
system.
To
improve
accuracy
fault
diagnosis
for
transceivers,
a
sparse
deep
belief
network
(DBN)-based
diagnostic
model
that
uses
adaptive
sine
cosine
algorithm
(SCA)
optimization
proposed.
Specifically,
regular
term
added
to
DBN
loss
function
prevent
overfitting
in
training.
At
same
time,
an
strategy
studied
realize
autonomous
switching
SCA,
so
as
precisely
optimize
hyper-parameters.
Therefore,
fusion
SCA
forms
effective
model,
named
ASCA-SDBN.
The
efficacy
proposed
ASCA-SDBN
validated
using
real-world
dataset
from
transceivers.
Experimental
results
show
this
outperforms
several
popular
intelligent
models.