Buildings,
Год журнала:
2024,
Номер
14(8), С. 2492 - 2492
Опубликована: Авг. 12, 2024
This
study
is
focused
on
the
punching
strength
of
fiber-reinforced
polymer
(FRP)
concrete
slabs.
The
mechanical
properties
reinforced
slabs
are
often
constrained
by
their
shear
at
column
connection
regions.
Researchers
have
explored
use
reinforcement
as
an
alternative
to
traditional
steel
address
this
limitation.
However,
current
codes
poorly
calculate
FRP-reinforced
aim
was
create
a
robust
model
that
can
accurately
predict
its
strength,
thus
improving
analysis
and
design
composite
structures
with
In
study,
189
sets
experimental
data
were
collected,
six
machine
learning
models,
including
linear
regression,
support
vector
machine,
BP
neural
network,
decision
tree,
random
forest,
eXtreme
Gradient
Boosting,
constructed
evaluated
based
goodness
fit,
standard
deviation,
root-mean-square
error
in
order
select
most
suitable
for
study.
optimal
obtained
compared
models
proposed
researchers.
Finally,
explainability
conducted
using
SHapley
Additive
exPlanations
(SHAP).
results
showed
forests
performed
best
among
all
outperformed
existing
suggested
effective
depth
important
proportional
strength.
not
only
provides
guidance
but
also
informs
future
engineering
practice.
Scientific Reports,
Год журнала:
2024,
Номер
14(1)
Опубликована: Июнь 20, 2024
Abstract
Graphene
nanoplatelets
(GrNs)
emerge
as
promising
conductive
fillers
to
significantly
enhance
the
electrical
conductivity
and
strength
of
cementitious
composites,
contributing
development
highly
efficient
composites
advancement
non-destructive
structural
health
monitoring
techniques.
However,
complexities
involved
in
these
nanoscale
are
markedly
intricate.
Conventional
regression
models
encounter
limitations
fully
understanding
intricate
compositions.
Thus,
current
study
employed
four
machine
learning
(ML)
methods
such
decision
tree
(DT),
categorical
boosting
(CatBoost),
adaptive
neuro-fuzzy
inference
system
(ANFIS),
light
gradient
(LightGBM)
establish
strong
prediction
for
compressive
(CS)
graphene
nanoplatelets-based
materials.
An
extensive
dataset
containing
172
data
points
was
gathered
from
published
literature
model
development.
The
majority
portion
(70%)
database
utilized
training
while
30%
used
validating
efficacy
on
unseen
data.
Different
metrics
were
assess
performance
established
ML
models.
In
addition,
SHapley
Additve
explanation
(SHAP)
interpretability.
DT,
CatBoost,
LightGBM,
ANFIS
exhibited
excellent
with
R-values
0.8708,
0.9999,
0.9043,
0.8662,
respectively.
While
all
suggested
demonstrated
acceptable
accuracy
predicting
strength,
CatBoost
exceptional
efficiency.
Furthermore,
SHAP
analysis
provided
that
thickness
GrN
plays
a
pivotal
role
GrNCC,
influencing
CS
consequently
exhibiting
highest
value
+
9.39.
diameter
GrN,
curing
age,
w/c
ratio
also
prominent
features
estimating
This
research
underscores
accurately
forecasting
characteristics
concrete
reinforced
nanoplatelets,
providing
swift
economical
substitute
laborious
experimental
procedures.
It
is
improve
generalization
study,
more
inputs
increased
datasets
should
be
considered
future
studies.
Buildings,
Год журнала:
2025,
Номер
15(3), С. 408 - 408
Опубликована: Янв. 28, 2025
This
paper
presents
a
comprehensive
study
of
the
mechanical
properties
lime-based
mortar
in
an
acidic
environment,
employing
both
experimental
analysis
and
machine
learning
to
model
techniques.
Despite
extensive
use
construction,
particularly
for
strengthening
structures
as
externally
bonded
materials,
its
behavior
under
conditions
remains
poorly
understood
literature.
aims
address
this
gap
by
investigating
performance
prolonged
exposure
environments,
laying
groundwork
further
research
critical
area.
In
phase,
commercial
hydraulic
was
subjected
varying
environmental
conditions,
including
solution
immersion
with
pH
3.0,
distilled
water
immersion,
dry
storage.
Subsequently,
specimens
were
tested
flexure
following
durations
1000,
3000,
5000
h.
modeling
extreme
gradient
boosting
(XGBoost)
algorithm
deployed
predict
h
exposure.
Using
data,
models
trained
capture
complex
relationships
between
stress-displacement
curve
(as
output)
various
properties,
density,
corrosion,
moisture,
duration
input
features).
The
predictive
demonstrated
remarkable
accuracy
generalization
(using
4-fold
cross-validation
approach)
capabilities
(R2
=
0.984
RMSE
0.116,
testing
dataset),
offering
reliable
tool
estimating
mortar’s
over
extended
periods
environment.
comparative
that
samples
exposed
environment
reached
peak
values
at
3000
exposure,
followed
decrease
contrast,
exhibited
earlier
onset
strength
increase,
indicating
different
material
responses
conditions.
