Deleted Journal,
Journal Year:
2024,
Volume and Issue:
6(6)
Published: June 12, 2024
Abstract
Automatic
and
fast
rock
classification
identification
is
an
important
part
of
geotechnical
intelligent
survey
system.
Image
based
supervised
deep
learning
analysis,
especially
for
convolutional
neural
networks
(CNN),
has
potential
in
optimizing
lithologic
interpretation
using
borehole
core
images.
However,
the
accuracy
efficiency
lithology
models
are
low
at
present.
In
this
work,
a
systematic
enormous
data
framework
on
geological
system
firstly
established
to
provide
datasets.
The
dataset
composed
approximately
150,000
images
samples,
which
covers
igneous
rocks,
sedimentary
metamorphic
rocks.
Secondly,
CNN-deep
transfer
algorithm,
end-to-end,
image-to-label
established.
Finally,
generalization
proposed
model
field
drilling
verification
test
show
that
constructed
recognition
ability
identify
rocks
quickly
accurately,
12
kinds
common
engineering
more
than
95%.
provides
convenient
tool
geologists
scientific
researchers.
Linköping electronic conference proceedings,
Journal Year:
2024,
Volume and Issue:
208, P. 188 - 194
Published: June 14, 2024
We
explore
the
use
of
various
machine
learning
(ML)
models
for
classifying
lithologies
utilizing
data
from
X-ray
fluorescence
(XRF)
and
computed
tomography
(XCT).
Typically,
are
identified
over
several
meters,
which
restricts
ML
due
to
limited
training
data.
To
address
this
issue,
we
augment
original
interval
dataset,
where
marked
extensive
sections,
into
finer
segments
10cm,
produce
a
high
resolution
dataset
with
vastly
increased
sample
size.
Additionally,
examine
impact
adjacent
on
building
more
generalized
model.
also
demonstrate
that
combining
XRF
XCT
leads
an
improved
classification
accuracy
compared
using
only
data,
is
common
practice
in
current
studies,
or
solely
relying
Petroleum Science,
Journal Year:
2024,
Volume and Issue:
unknown
Published: July 1, 2024
It
is
of
great
significance
to
accurately
and
rapidly
identify
shale
lithofacies
in
relation
the
evaluation
prediction
sweet
spots
for
oil
gas
reservoirs.
To
address
problem
low
resolution
logging
curves,
this
study
establishes
a
grayscale-phase
model
based
on
high-resolution
grayscale
curves
using
clustering
analysis
algorithms
identification,
working
with
Shahejie
Formation,
Bohai
Bay
Basin,
China.
The
phase
defined
as
sum
absolute
relative
amplitude
well
their
features.
magnitude
gray
values
utilized
evaluating
material
composition
(mineral
+
total
organic
carbon)
shale,
while
difference
between
adjacent
used
structure
type.
research
results
show
that
can
well,
accuracy
applicability
were
verified
by
fitting
relationship
mineral
composition,
corresponding
relationships
amplitudes
laminae
development
shales.
Four
are
identified
target
layer
area:
massive
mixed
laminated
calcareous
shale.
This
method
not
only
effectively
characterize
but
also
numerically
degree
laminae,
solve
difficult
millimeter-scale
which
provide
technical
support
spot
complex
continental
lacustrine
basins.
