Helicopter Turboshaft Engines’ Neural Network System for Monitoring Sensor Failures
Sensors,
Journal Year:
2025,
Volume and Issue:
25(4), P. 990 - 990
Published: Feb. 7, 2025
An
effective
neural
network
system
for
monitoring
sensors
in
helicopter
turboshaft
engines
has
been
developed
based
on
a
hybrid
architecture
combining
LSTM
and
GRU.
This
enables
sequential
data
processing
while
ensuring
high
accuracy
anomaly
detection.
Using
recurrent
layers
(LSTM/GRU)
is
critical
dependencies
among
time
series
analysis
identification,
facilitating
key
information
retention
from
previous
states.
Modules
such
as
SensorFailClean
SensorFailNorm
implement
adaptive
discretization
quantisation
techniques,
enhancing
the
input
quality
contributing
to
more
accurate
predictions.
The
demonstrated
detection
at
99.327%
after
200
training
epochs,
with
reduction
loss
2.5
0.5%,
indicating
stability
processing.
A
algorithm
incorporating
temporal
regularization
combined
optimization
method
(SGD
RMSProp)
accelerated
convergence,
reducing
4
min
13
s
achieving
an
of
0.993.
Comparisons
alternative
methods
indicate
superior
performance
proposed
approach
across
metrics,
including
0.993
compared
0.981
0.982.
Computational
experiments
confirmed
presence
highly
correlated
sensor
method's
effectiveness
fault
detection,
highlighting
system's
capability
minimize
omissions.
Language: Английский
New Method for Improving Tracking Accuracy of Aero-Engine On-Board Model Based on Separability Index and Reverse Searching
Aerospace,
Journal Year:
2025,
Volume and Issue:
12(3), P. 175 - 175
Published: Feb. 22, 2025
Throughout
its
service
life,
an
aero-engine
will
experience
a
series
of
health
conditions
due
to
the
inevitable
performance
degradation
major
components,
and
characteristics
deviate
from
their
initial
states.
For
improving
tracking
accuracy
self-tunning
on-board
engine
model
on
output
variables
throughout
new
method
based
separability
index
reverse
search
algorithm
was
proposed
in
this
paper.
By
using
method,
qualified
training
set
neural
networks
created
basis
eSTORM
(enhanced
Self
Tuning
On-board
Real-time
Model)
database,
problem
that
is
reduced
or
even
process
not
convergent
can
be
solved.
Compared
with
introducing
sample
memory
factors,
paper
makes
maintain
higher
whole
simple
enough
for
implementation.
Finally,
center
generated
calculation
could
used
real-time
monitoring
gas
path
parameters
without
additional
calculations.
commonly
sliding
window
avoids
low
efficiency
caused
by
fewer
abnormal
data
samples.
Language: Английский
An Intelligent Self-Validated Sensor System Using Neural Network Technologies and Fuzzy Logic Under Operating Implementation Conditions
Big Data and Cognitive Computing,
Journal Year:
2024,
Volume and Issue:
8(12), P. 189 - 189
Published: Dec. 13, 2024
This
article
presents
an
intelligent
self-validated
sensor
system
developed
for
dynamic
objects
and
based
on
the
concept,
which
ensures
autonomous
data
collection
real-time
analysis
while
adapting
to
changing
conditions
compensating
errors.
The
research’s
scientific
merit
is
that
has
been
integrates
adaptive
correction
algorithms,
fuzzy
logic,
neural
networks
improve
sensors’
accuracy
reliability
under
operating
conditions.
proposed
provides
error
compensation,
long-term
stability,
effective
fault
diagnostics.
Analytical
equations
are
described,
considering
corrections
related
influencing
factors,
temporal
drift,
calibration
characteristics,
significantly
enhancing
measurement
reliability.
logic
application
allows
refining
scaling
coefficient
adjusts
relationship
between
measured
parameter
utilizing
inference
algorithms.
Additionally,
monitoring
diagnostics
implementation
states
through
LSTM
enable
detection.
Computational
experiments
TV3-117
engine
demonstrated
high
data-restoring
during
forced
interruptions,
reaching
99.5%.
A
comparative
with
alternative
approaches
confirmed
advantages
of
using
(Long
Short-Term
Memory)
in
improving
quality.
Language: Английский