International Journal for Numerical and Analytical Methods in Geomechanics,
Год журнала:
2024,
Номер
unknown
Опубликована: Дек. 4, 2024
ABSTRACT
With
the
acceleration
of
urbanization,
stability
foundation
is
being
more
crucial
to
performance
and
service
superstructure.
As
our
understanding
factors
influencing
soil's
physical
mechanical
behavior
deepens,
it
becomes
increasingly
challenging
for
traditional
limit
equilibrium
analysis
methods
accurately
consider
complex
affecting
stability,
such
as
initial
fabric
anisotropy
caused
by
particle
morphology
geological
deposition
in
sand.
Although
some
scholars
had
used
advanced
constitutive
models
finite
element
method
(FEM)
investigate
influence
on
responses
foundations,
this
approach
failed
reveal
microscopic
information
underlying
shear
failure
sandy
soil
foundations.
In
study,
ultimate
bearing
capacity
mode
shallow
studied
using
hierarchical
FEM
discrete
(DEM)
coupling
method.
Four
representative
volume
elements
(RVEs)
with
varying
bedding
plane
angles
are
constructed
DEM
characterizing
different
anisotropies,
specific
stress–strain
RVEs
directly
passed
into
corresponding
Gauss
points
replace
conventional
model.
Numerical
results
show
that
affects
foundations
significantly,
micromechanical
behaviors
at
local
have
been
explored,
which
advances
micromechanisms
progressive
significantly.
Journal of Rock Mechanics and Geotechnical Engineering,
Год журнала:
2024,
Номер
unknown
Опубликована: Апрель 1, 2024
To
efficiently
predict
the
mechanical
parameters
of
granular
soil
based
on
its
random
micro-structure,
this
study
proposed
a
novel
approach
combining
numerical
simulation
and
machine
learning
algorithms.
Initially,
3500
simulations
one-dimensional
compression
tests
coarse-grained
sand
using
three-dimensional
(3D)
discrete
element
method
(DEM)
were
conducted
to
construct
database.
In
process,
positions
particles
randomly
altered,
particle
assemblages
changed.
Interestingly,
besides
confirming
influence
size
distribution
parameters,
stress-strain
curves
differed
despite
an
identical
gradation
statistic
when
position
varied.
Subsequently,
obtained
data
partitioned
into
training,
validation,
testing
datasets
at
7:2:1
ratio.
convert
DEM
model
multi-dimensional
matrix
that
computers
can
recognize,
3D
models
first
sliced
extract
multi-layer
two-dimensional
(2D)
cross-sectional
data.
Redundant
information
was
then
eliminated
via
gray
processing,
stacked
form
new
representing
soil's
fabric.
utilizing
Python
language
Pytorch
framework,
convolutional
neural
networks
(CNNs)
developed
establish
relationship
between
constrained
modulus
from
The
mean
squared
error
(MSE)
function
utilized
assess
loss
value
during
training
process.
When
rate
(LR)
fell
within
range
10-5-10-1,
batch
sizes
(BSs)
4,
8,
16,
32,
64,
stabilized
after
100
epochs
in
validation
dataset.
For
BS
=
32
LR
10-3,
reached
minimum.
set,
comparative
evaluation
predicted
CNNs
versus
simulated
reveals
minimum
absolute
percentage
(MAPE)
4.43%
under
optimized
condition,
demonstrating
accuracy
approach.
Thus,
by
CNNs,
variation
characteristics
related
fabric
would
be
evaluated
directly
tracking
assemblages.
Applied Sciences,
Год журнала:
2025,
Номер
15(10), С. 5567 - 5567
Опубликована: Май 16, 2025
To
address
the
limitations
in
conventional
granular
morphology
characterization
where
excessive
emphasis
has
been
placed
on
elongation
index
(EI)
while
neglecting
flatness
(FI)
and
their
coupled
interactions,
this
study
establishes
an
EI/FI
co-regulated
dual-parameter
morphological
framework.
Through
integrated
triaxial
compression
experiments
discrete
element
simulations,
we
systematically
investigate
multi-scale
mechanical
responses
spanning
macroscopic
stress–strain
behavior
to
microscopic
force-chain
evolution.
The
results
show
that
(1)
regulation
of
pore
structure
by
parameters
presents
non-linear
characteristics,
(2)
evolution
peak
shear
strength
is
predominantly
governed
anisotropy.
(3)
parabolic
relationship
between
maximum
dilatancy
angle
shown.
(4)
micro
analysis
reveals
have
limited
influence
statistical
distribution
characteristics
contact
force
chain,
but
a
significant
regulatory
effect
anisotropic
network.
International Journal for Numerical and Analytical Methods in Geomechanics,
Год журнала:
2024,
Номер
unknown
Опубликована: Ноя. 26, 2024
ABSTRACT
A
two‐way
coupling
numerical
framework
based
on
smoothed
particle
hydrodynamics
(SPH)
is
developed
in
this
study
to
model
binary
granular
mixtures
consisting
of
coarse
and
fine
grains.
The
employs
updated
Lagrangian
SPH
simulate
grains,
with
configurations
at
each
time
step,
total
efficiently
grains
without
configurations.
Riemann
solver
utilized
introduce
dissipation
facilitate
their
To
enhance
computational
efficiency,
a
multiple
time‐stepping
scheme
initially
applied
manage
the
integration
between
Several
experiments,
including
column
collapse,
low‐speed
impact
craters,
flow
impacting
blocks,
are
conducted
validate
stability
accuracy
proposed
algorithm.
Subsequently,
two
more
complex
scenarios
involving
soil–rock
mixture
slope
considering
irregular
shapes,
bouldery
debris
flows
natural
terrain,
simulated
showcase
potential
engineering
applications.
Finally,
detailed
analysis
performed
evaluate
efficiency
advantages
present
approach.
findings
consistent
previous
experimental
results,
implementation
can
improve
by
up
600%,
thereby
providing
significant
for
large‐scale
simulations.
International Journal for Numerical and Analytical Methods in Geomechanics,
Год журнала:
2024,
Номер
48(16), С. 3909 - 3932
Опубликована: Авг. 9, 2024
Abstract
The
current
study
presents
superposition‐based
concurrent
multiscale
approaches
for
porodynamics,
capable
of
capturing
related
physical
phenomena,
such
as
soil
liquefaction
and
dynamic
hydraulic
fracture
branching,
across
different
spatial
length
scales.
Two
scenarios
are
considered:
superposition
finite
element
discretizations
with
varying
mesh
densities,
peridynamics
(PD)
method
(FEM)
to
handle
discontinuities
like
strain
localization
cracks.
approach
decomposes
the
acceleration
rate
change
in
pore
water
pressure
into
subdomain
solutions
approximated
by
models,
allowing
high‐fidelity
models
be
used
locally
regions
interest,
crack
tips
or
shear
bands,
without
neglecting
far‐field
influence
represented
low‐fidelity
models.
coupled
stiffness,
mass,
compressibility,
permeability,
damping
matrices
were
derived
based
on
framework.
proposed
FEM‐FEM
porodynamic
coupling
was
validated
against
analytical
numerical
one‐
two‐dimensional
consolidation
problems.
PD‐FEM
model
applied
liquefaction‐induced
accumulation
near
a
low‐permeability
interlayer
layered
deposit
fracturing
branching.
It
has
been
shown
that
offers
modeling
flexibility
efficiency
taking
advantage
FEM
complex
domains
PD
ability
resolve
discontinuities.
