Physics of Fluids,
Journal Year:
2024,
Volume and Issue:
36(12)
Published: Dec. 1, 2024
In
the
field
of
geological
exploration,
accurately
distinguishing
between
different
types
fluids
is
crucial
for
development
oil,
gas,
and
mineral
resources.
Due
to
scarcity
labeled
samples,
traditional
supervised
learning
methods
face
significant
limitations
when
processing
well
log
data.
To
address
this
issue,
paper
presents
a
novel
fluid
classification
method
known
as
Resilient
Semi-Supervised
Meta-Learning
Network
(RSSMLN)
based
on
wavelet
transform
K-means
optimization,
which
combines
advantages
few-shot
semi-supervised
learning,
aiming
optimize
recognition
in
Initially,
study
employs
small
set
samples
train
initial
model
utilizes
pseudo-label
generation
clustering
prototypes,
thereby
enhancing
model's
accuracy
generalization
ability.
Subsequently,
during
feature
extraction
process,
preprocessing
techniques
are
introduced
enhance
time-frequency
representation
data
through
multi-scale
decomposition.
This
process
effectively
captures
high-frequency
low-frequency
features,
providing
structured
information
subsequent
convolution
operations.
By
employing
dual-channel
heterogeneous
convolutional
kernel
extractor,
RSSMLN
can
capture
subtle
features
significantly
improve
accuracy.
Experimental
results
indicate
that
compared
various
standard
deep
models,
achieves
superior
performance
identification
tasks.
research
provides
reliable
solution
oilfield
applications
offers
scientific
support
resource
exploration
evaluation.
Sensors,
Journal Year:
2024,
Volume and Issue:
24(14), P. 4590 - 4590
Published: July 15, 2024
Aiming
at
the
complex
characteristics
of
negative
pressure
waves
in
low-pressure
pipelines
inside
buildings,
we
proposed
an
estimation
method
fluctuation
trends
based
on
robust
Kalman
filter
and
improved
VMD,
which
can
be
used
for
leakage
detection.
The
reconstructed
baseline
signal
accurately
describe
trend
wave
after
drop,
quantitatively
express
characteristic
difference
between
condition
gas
usage
condition.
was
to
estimate
fluctuations.
parameters
VMD
were
adaptively
calculated
WAA
discrete
scale
space.
components
contained
IMFs
separated
by
a
reconstruction
Fourier
series.
Based
simulation
signal,
restore
component
signal.
actual
signals,
accuracy
small
detection
is
96.7%
large
73.3%.
Journal of Marine Science and Engineering,
Journal Year:
2024,
Volume and Issue:
12(8), P. 1384 - 1384
Published: Aug. 13, 2024
The
speed
control
of
a
pipeline
inspection
gauge
(PIG)
directly
affects
the
quality
comprehensive
submarine
pipelines.
However,
mechanism
gas
flow
behavior
in
under
influence
pig
valve
is
not
well
understood.
In
this
study,
driving
differential
pressure
was
modeled
based
on
building
block
method
and
numerical
simulations.
For
first
time,
rate
torque
opening
process
studied.
results
show
that
when
angle
increased
from
4.5°
to
22°,
reduced
1325
73
kPa,
realizing
94.5%
reduction.
addition,
bypass
7.7
2470
Nm
during
closing
process.
increases
were
correlated
with
torque.
established
experimental
system
for
measurement
confirmed
analysis
results.
By
clarifying
law
variation,
study
provides
theoretical
guidance
structural
design
scheme
unit.
Journal of Marine Science and Engineering,
Journal Year:
2024,
Volume and Issue:
12(8), P. 1430 - 1430
Published: Aug. 19, 2024
Hydraulic
pumps
are
the
core
components
that
provide
power
for
hydraulic
transmission
systems,
which
widely
used
in
aerospace,
marine
engineering,
and
mechanical
their
failure
affects
normal
operation
of
entire
system.
This
paper
takes
a
single
axial
piston
pump
as
research
object
proposes
small-sample
fault
diagnosis
method
based
on
model
migration
strategy
situation
only
small
number
training
samples
available
diagnosis.
To
achieve
end-to-end
diagnosis,
1D
Squeeze-and-Excitation
Networks
(1D-SENets)
was
constructed
one-dimensional
convolutional
neural
network
combined
with
channel
domain
attention
mechanism.
The
first
pre-trained
sufficient
labeled
data
from
source
conditions,
then,
strategy,
some
underlying
parameters
were
fixed,
amount
target
conditions
to
fine-tune
rest
model.
In
this
paper,
proposed
validated
using
an
dataset,
experimental
results
show
can
effectively
improve
overfitting
problem
sample
recognition
accuracy.
Physics of Fluids,
Journal Year:
2024,
Volume and Issue:
36(12)
Published: Dec. 1, 2024
In
the
field
of
geological
exploration,
accurately
distinguishing
between
different
types
fluids
is
crucial
for
development
oil,
gas,
and
mineral
resources.
Due
to
scarcity
labeled
samples,
traditional
supervised
learning
methods
face
significant
limitations
when
processing
well
log
data.
To
address
this
issue,
paper
presents
a
novel
fluid
classification
method
known
as
Resilient
Semi-Supervised
Meta-Learning
Network
(RSSMLN)
based
on
wavelet
transform
K-means
optimization,
which
combines
advantages
few-shot
semi-supervised
learning,
aiming
optimize
recognition
in
Initially,
study
employs
small
set
samples
train
initial
model
utilizes
pseudo-label
generation
clustering
prototypes,
thereby
enhancing
model's
accuracy
generalization
ability.
Subsequently,
during
feature
extraction
process,
preprocessing
techniques
are
introduced
enhance
time-frequency
representation
data
through
multi-scale
decomposition.
This
process
effectively
captures
high-frequency
low-frequency
features,
providing
structured
information
subsequent
convolution
operations.
By
employing
dual-channel
heterogeneous
convolutional
kernel
extractor,
RSSMLN
can
capture
subtle
features
significantly
improve
accuracy.
Experimental
results
indicate
that
compared
various
standard
deep
models,
achieves
superior
performance
identification
tasks.
research
provides
reliable
solution
oilfield
applications
offers
scientific
support
resource
exploration
evaluation.
