Scientific Reports,
Год журнала:
2023,
Номер
13(1)
Опубликована: Ноя. 14, 2023
Abstract
A
new
approach
that
combines
analytical
two-parameter
kinematic
theory
(2PKT)
with
machine
learning
(ML)
models
for
estimating
the
shear
capacity
of
embedded
through-section
(ETS)-strengthened
reinforced
concrete
(RC)
beams
is
proposed.
The
2PKT
was
first
developed
to
validate
its
representativeness
and
confidence
against
available
experimental
data
ETS-retrofitted
RC
beams.
Given
deficiency
test
data,
utilized
generate
a
large
pool
2643
samples.
aim
optimize
ML
algorithms,
namely,
random
forest,
extreme
gradient
boosting
(XGBoost),
light
machine,
artificial
neural
network
(ANN)
algorithm.
optimized
ANN
model
exhibited
highest
accuracy
in
predicting
total
strength
ETS-strengthened
ETS
contribution.
In
terms
beams,
achieved
R
2
values
0.99,
0.98,
0.96
training,
validation,
testing
respectively.
By
contrast,
could
predict
contribution
high
accuracy,
0.97
Then,
effects
all
design
variables
on
were
investigated
using
hybrid
2PKT–ML.
obtained
trends
well
appraise
reasonability
proposed
approach.
Journal of Composites Science,
Год журнала:
2025,
Номер
9(1), С. 23 - 23
Опубликована: Янв. 6, 2025
To
meet
the
increasing
demand
for
resilient
infrastructure
in
seismic
and
high-impact
areas,
accurate
prediction
reliability
analysis
of
performance
composite
structures
under
impact
loads
is
essential.
Conventional
techniques,
including
experimental
testing
high-quality
finite
element
simulation,
require
considerable
time
resources.
address
these
issues,
this
study
investigated
individual
hybrid
models
support
vector
regression
(SVR),
Gaussian
process
(GPR),
improved
eliminate
particle
swamp
optimization
hybridized
artificial
neural
network
(IEPANN)
predicting
failure
strength
concrete
developed
by
combining
normal
(NC)
with
ultra-high
(UHPC)
polyurethane-based
polymer
(PUC),
considering
different
surface
treatments
subjected
to
various
static
loads.
An
dataset
was
utilized
train
ML
perform
on
dataset.
Key
parameters
included
compressive
(Cfc),
flexural
load
U-shaped
specimens
(P),
density
(ρ),
first
crack
(N1),
splitting
tensile
(ft).
Results
revealed
that
all
had
high
accuracy,
achieving
NSE
values
above
acceptable
thresholds
greater
than
90%
across
datasets.
Statistical
errors
such
as
RMSE,
MAE,
PBIAS
were
calculated
fall
within
limits.
Hybrid
IEPANN
appeared
be
most
effective
model,
demonstrating
highest
value
0.999
lowest
PBIAS,
MAE
0.0013,
0.0018,
0.001,
respectively.
The
times
(N1
N2)
reduced
survival
probability
increased.
Scientific Reports,
Год журнала:
2024,
Номер
14(1)
Опубликована: Июль 3, 2024
Abstract
The
present
study
introduces
a
novel
approach
utilizing
machine
learning
techniques
to
predict
the
crucial
mechanical
properties
of
engineered
cementitious
composites
(ECCs),
spanning
from
typical
exceptionally
high
strength
levels.
These
properties,
including
compressive
strength,
flexural
tensile
and
strain
capacity,
can
not
only
be
predicted
but
also
precisely
estimated.
investigation
encompassed
meticulous
compilation
examination
1532
datasets
sourced
pertinent
research.
Four
algorithms,
linear
regression
(LR),
K
nearest
neighbors
(KNN),
random
forest
(RF),
extreme
gradient
boosting
(XGB),
were
used
establish
prediction
model
ECC
determine
optimal
model.
was
utilized
employ
SHapley
Additive
exPlanations
(SHAP)
for
scrutinizing
feature
importance
conducting
an
in-depth
parametric
analysis.
Subsequently,
comprehensive
control
strategy
devised
properties.
This
provide
actionable
guidance
design,
equipping
engineers
professionals
in
civil
engineering
material
science
make
informed
decisions
throughout
their
design
endeavors.
results
show
that
RF
demonstrated
highest
accuracy
with
R
2
values
0.92
0.91
on
test
set.
XGB
outperformed
predicting
0.87
0.80
set,
respectively.
capacity
least
accurate.
Meanwhile,
MAE
mere
0.84%,
smaller
than
variability
(1.77%)
previous
Compressive
sensitivity
variations
both
water-cement
ratio
(W)
water
reducer
(WR).
In
contrast,
exhibited
solely
changes
W.
Conversely,
input
features
moderate
consistent.
attributes
emerged
combined
effects
multiple
positive
negative
features.
Notably,
WR
exerted
most
significant
influence
among
all
features,
whereas
polyethylene
(PE)
fiber
as
primary
driver
affecting
capacity.
Structures,
Год журнала:
2024,
Номер
62, С. 106162 - 106162
Опубликована: Март 18, 2024
The
significance
of
Ultimate
Bond
Stress-Slip
(UBS-S)
in
reinforced
Ultra-High
Performance
Concrete
(UHPC)
structures
cannot
be
overstated,
as
it
directly
affects
their
load-carrying
capacity,
structural
integrity,
and
long-term
performance.
A
comprehensive
analysis
the
UHPC-Parallel
Micro
Element
System
(UHPC-PMES),
including
144
specimens,
evaluated
computational
efficiency
proposed
UBS-S
model.
To
this
end,
most
critical
settings
steel
bar
diameter,
concrete
cover,
bond
length,
UHPC
compressive
strength
(db,c,lb,fUHPC′)
were
directed
to
create
parametric
research.
Applying
hybrid
approach
three
different
optimization
techniques,
namely
Physics-Informed
Neural
Networks
(PINN),
Genetic
Algorithms
(GA),
Multiple
Linear
Regression
(MLR),
study
predicted
at
interface
bars.
It
formulated
hyper-parameters
effect
values
(a,m,β,G).
presented
research
used
these
algorithms
solve
an
inverse
problem
engineering.
Comparing
results
obtained
from
PINN,
GA,
MLR
demonstrated
that
machine
learning
techniques
PMES
model
could
effectively
accurately
investigate
ultimate
stress-slip
for
structures.
Journal of Theory and Practice of Engineering Science,
Год журнала:
2023,
Номер
3(12), С. 7 - 14
Опубликована: Дек. 29, 2023
Machine
Learning,
as
one
of
the
key
technologies
in
field
artificial
intelligence,
has
made
significant
advancements
recent
years.
This
study
provides
a
relatively
systematic
introduction
to
machine
learning.
Firstly,
it
gives
an
overview
historical
development
learning,
and
then
focuses
on
analysis
classical
algorithms
Subsequently,
elucidates
latest
research
aiming
comprehensively
explore
applications
learning
various
domains
discuss
potential
future
directions.
Buildings,
Год журнала:
2024,
Номер
14(1), С. 190 - 190
Опубликована: Янв. 11, 2024
Machine
learning
(ML)
algorithms
have
been
widely
used
in
big
data
prediction
and
analysis
terms
of
their
excellent
regression
ability.
However,
the
accuracy
different
ML
varies
between
problems
sets.
In
order
to
construct
a
model
with
optimal
for
fly
ash
concrete
(FAC),
such
as
genetic
programming
(GP),
support
vector
(SVR),
random
forest
(RF),
extremely
gradient
boost
(XGBoost),
backpropagation
artificial
neural
network
(BP-ANN)
adaptive
network-based
fuzzy
inference
system
(ANFIS)
were
selected
this
study;
particle
swarm
optimization
(PSO)
algorithm
was
also
optimize
structure
hyperparameters
each
algorithm.
The
statistical
results
show
that
performance
assembled
is
better
than
an
NN-based
addition,
PSO
can
effectively
improve
algorithms.
comprehensive
analyzed
using
Taylor
diagram,
PSO-XGBoost
has
best
performance,
R2
MSE
equal
0.9072
11.4546,
respectively.
