Machines,
Год журнала:
2024,
Номер
12(6), С. 368 - 368
Опубликована: Май 24, 2024
Because
of
the
difficulty
in
fault
detection
for
and
diagnosing
adjustment
hydraulic
servomotor,
this
paper
uses
feature
extraction
technology
to
extract
time
domain
frequency
features
pressure
signal
servomotor
splice
multiple
signals
through
Multi-source
Information
Fusion
(MSIF)
method.
The
comprehensive
expression
device
status
information
is
obtained.
After
that,
proposes
a
Algorithm
GA-SVDD-neg,
which
Genetic
(GA)
optimize
Support
Vector
Data
Description
with
negative
examples
(SVDD-neg).
Through
joint
optimization
Mutual
(MI)
selection
algorithm,
that
are
most
sensitive
state
deterioration
selected.
Experiments
show
MI
algorithm
has
better
performance
than
other
dimensionality
reduction
algorithms
field
abnormal
servomotors,
GA-SVDD-neg
stronger
robustness
generality
anomaly
algorithms.
In
addition,
make
full
use
advantages
deep
learning
automatic
classification,
realizes
diagnosis
based
on
1D
Convolutional
Neural
Network
(1DCNN).
experimental
results
same
superior
as
traditional
can
accurately
diagnose
known
faults
servomotor.
This
research
great
significance
intelligent
transformation
servomotors
also
provide
reference
warning
Electro-Hydraulic
(EH)
system
type
steam
turbine.
IEEE Transactions on Industrial Informatics,
Год журнала:
2024,
Номер
20(4), С. 6356 - 6368
Опубликована: Янв. 5, 2024
Machine
learning
models
have
been
widely
successful
in
the
field
of
intelligent
fault
diagnosis.
Most
existing
machine
are
deployed
static
environments
and
rely
on
precollected
datasets
for
offline
training,
which
makes
it
impossible
to
update
further
once
they
established.
However,
open
dynamic
environment
reality,
there
is
always
incoming
data
form
streams,
including
new
categories
that
constantly
generated
over
time.
In
addition,
operating
conditions
mechanical
equipment
time-varying,
results
continuous
stream
nonindependently
homogeneously
distributed.
industrial
applications,
diagnosis
problem
nonindependent
identically
distributed
streaming
referred
as
cross-domain
class
incremental
problem.
To
address
problem,
a
novel
broad
network
(CDCIBN)
proposed.
Specifically,
solve
domain-adaptation
loss
function
first
designed,
enables
conventional
handle
category
increment
task
well.
Then,
mechanism
learns
while
retaining
knowledge
old
well
enough
without
replaying
data.
The
effectiveness
proposed
method
evaluated
through
multiple
failure
cases.
Experimental
analysis
demonstrates
designed
CDCIBN
has
significant
advantages
variable
working
condition
application.
IEEE Access,
Год журнала:
2024,
Номер
12, С. 29345 - 29361
Опубликована: Янв. 1, 2024
Artificial
Intelligence
(AI)
is
a
key
component
in
Industry
4.0.
Rotating
machines
are
critical
components
manufacturing
industries.
In
the
vast
world
of
4.0,
where
an
IoT
network
acts
as
monitoring
and
decision-making
system,
predictive
maintenance
quickly
gaining
importance.
Predictive
method
that
uses
AI
to
handle
potential
problems
before
they
cause
breakdowns
operations,
processes
or
systems.
However,
there
significant
issue
with
models'
(also
known
"black
boxes")
inability
explain
their
decisions.
This
interpretability
vital
for
making
decisions
validating
model's
reliability,
leading
improved
trust
acceptance
AI-driven
strategies.
Explainable
solution
because
it
provides
human-understandable
insights
into
how
model
arrives
at
its
predictions.
this
regard,
paper
presents
AI-based
Industrial
rotating
machines.
The
proposed
approach
unfolds
four
comprehensive
stages:
a)
Multi-sensor
based
multi-fault
(5
different
fault
classes)
data
acquisition,
b)
Frequency-domain
statistical
feature
extraction,
c)
Comparison
results
multiple
algorithms,
d)
XAI
integration
using
"Local
Interpretable
Model
Agnostic
Explanation
(LIME)",
"SHapley
Additive
exPlanation
(SHAP)",
"Partial
Dependence
Plot
(PDP)"
"Individual
Conditional
Expectation
(ICE)"
interpret
results.
IEEE Open Journal of Industry Applications,
Год журнала:
2023,
Номер
4, С. 188 - 214
Опубликована: Янв. 1, 2023
This
review
article
systematically
summarizes
the
existing
literature
on
utilizing
machine
learning
(ML)
techniques
for
control
and
monitoring
of
electric
drives.
It
is
anticipated
that
with
rapid
progress
in
algorithms
specialized
embedded
hardware
platforms,
ML-based
data-driven
approaches
will
become
standard
tools
automated
high-performance
In
addition,
this
also
provides
some
outlook
toward
promoting
its
widespread
application
industry
a
focus
deploying
ML
onto
system-on-chip
field-programmable
gate
array
devices.
Decision Analytics Journal,
Год журнала:
2024,
Номер
10, С. 100425 - 100425
Опубликована: Фев. 15, 2024
Industry
4.0
denotes
smart
manufacturing,
where
rotating
machines
predominantly
serve
as
the
fundamental
components
in
production
sectors.
The
primary
duty
of
maintenance
engineers
is
to
upkeep
these
vital
machines,
aiming
reduce
unexpected
halts
and
extend
their
operational
lifespan.
most
recent
development
Predictive
Maintenance
(PdM).
Due
diversity
machinery
diverse
behaviour
each
machine
different
fault
conditions,
challenging
task
predictive
detect
fault,
diagnose
type
explain
why
a
particular
predicted.
This
study
proposes
an
effective
Explainable
strategy
considering
(1)
test
setup
building,
(2)
low-cost
Fast
Fourier
Transform
(FFT)
raw
data
using
multiple
sensors,
(3)
multi-sensor
fusion,
(4)
comparing
various
multi-class
classification
algorithms,
(5)
analysis
cases
concerning
versus
single
sensor
multi-location
location,
(6)
explainable
maintenance.
Quantitative
results
from
this
reveal
remarkable
multi-fault
detection
accuracy
classification,
with
highest
100%.
Furthermore,
fusion
significantly
outperforms
single-sensor
approaches,
demonstrating
enhancement
prediction
all
models.
Using
Artificial
Intelligence
methods
contributes
interpretability
diagnoses,
making
it
critical
advancement
Intelligent
Manufacturing
4.0.
study's
novelty
(Local
Interpretable
Model
Agnostic
Explanation
(LIME)
Random
Forest)
for
fusion.
