AIP Advances,
Journal Year:
2024,
Volume and Issue:
14(7)
Published: July 1, 2024
This
study
aims
to
develop
predictive
models
for
accurately
forecasting
the
uniaxial
compressive
strength
of
concrete
enhanced
with
nanomaterials.
Various
machine
learning
algorithms
were
employed,
including
backpropagation
neural
network
(BPNN),
random
forest
(RF),
extreme
gradient
boosting
(XGB),
and
a
hybrid
ensemble
stacking
method
(HEStack).
A
comprehensive
dataset
containing
94
data
points
nano-modified
was
collected,
eight
input
parameters:
water-to-cement
ratio,
carbon
nanotubes,
nano-silica,
nano-clay,
nano-aluminum,
cement,
coarse
aggregates,
fine
aggregates.
To
evaluate
performance
these
models,
tenfold
cross-validation
case
prediction
conducted.
It
has
been
shown
that
HEStack
model
is
most
effective
approach
precisely
predicting
properties
concrete.
During
cross-validation,
found
have
superior
accuracy
resilience
against
overfitting
compared
stand-alone
models.
underscores
potential
algorithm
in
enhancing
performance.
In
study,
predicted
results
assessed
using
metrics
such
as
coefficient
determination
(R2),
mean
absolute
percentage
error
(MAPE),
root
square
(RMSE),
ratio
RMSE
standard
deviation
observations
(RSR),
normalized
bias
(NMBE).
The
achieved
lowest
MAPE
2.84%,
1.6495,
RSR
0.0874,
NMBE
0.0064.
addition,
it
attained
remarkable
R2
value
0.9924,
surpassing
scores
0.9356
0.9706
0.9884
indicating
its
exceptional
generalization
capability.
Mechanics of Advanced Materials and Structures,
Journal Year:
2024,
Volume and Issue:
unknown, P. 1 - 20
Published: Jan. 1, 2024
The
compressive
strength
(CS)
of
a
concrete
column
confined
with
spiral
stirrups
is
an
important
indicator
for
assessing
the
safety
and
stability
structures.
However,
achieving
accurate
CS
estimation
remains
challenging
due
to
complex
confinement
mechanism
provided
by
stirrups.
In
this
study,
three
robust
machine
learning
(ML)
algorithms—support
vector
regression
(SVR),
random
forest
(RF)
extreme
gradient
boosting
(XGBoost)—are
employed
predict
value
stirrup-confined
circular
columns.
hyperparameters
ML
models
undergo
fine-tuning
via
Bayesian
optimization
10-fold
cross-validation,
optimized
are
evaluated
their
predictive
capabilities.
Results
show
that
compared
SVR
RF,
XGBoost
exhibits
more
stable
generalization
performance,
average
coefficient
determination
(R2)
0.944
demonstrates
superior
accuracy
on
testing
dataset
R2
0.967.
To
provide
insights
into
relationship
between
input
features
output
value,
Individual
Conditional
Exception
(ICE)
plots,
one/two-dimensional
Partial
Dependence
Plots
(PDPs),
Shapley
Additive
Explanation
(SHAP)
techniques
utilized
interpret
model.
Additionally,
friendly
online
graphical
user
interface
(GUI)
has
been
specially
developed
based
model
facilitate
convenient
column.
Case Studies in Construction Materials,
Journal Year:
2024,
Volume and Issue:
21, P. e03510 - e03510
Published: July 11, 2024
Introducing
3D-concrete
printing
has
started
a
revolution
in
the
construction
industry,
presenting
unique
opportunities
alongside
undeniable
challenges.
Among
these,
major
challenge
is
iterative
process
associated
with
mix
design
formulation,
which
results
significant
material
and
time
consumption.
This
research
uses
machine
learning
(ML)
techniques
such
as
Extreme
Gradient
Boosting
(XGBoost),
Support
Vector
Machine
(SVM),
Decision
Tree
Regression
(DTR),
Gaussian
Process
(GPR),
Artificial
Neural
Network
(ANN)
to
overcome
these
A
dataset
containing
21
constituent
features
4
output
properties
(cast
printed
compressive
strength,
slump
flow)
was
extracted
from
literature
investigate
relationship
between
performance.
The
models
were
assessed
using
range
of
evaluation
metrics,
including
Mean
Absolute
Error
(MAE),
Root
Squared
(RMSE),
(MSE),
R-squared
value.
(GPR)
yielded
more
favorable
results.
In
case
cast
GPR
achieved
an
R2
value
0.9069,
along
RMSE,
MSE,
MAE
values
13.04,
170.12,
9.40,
respectively.
similar
trend
observed
for
strengths
directions
1,
2,
3.
exceeding
0.91
all
directions,
accompanied
by
significantly
lower
RMSE
(below
4.1).
also
validated
four
designs.
These
mixes
3D
tested
strength
flow.
GPR's
average
error
10.55
%,
while
SVM
slightly
9.38
%.
Overall,
this
work
presents
novel
approach
optimizing
3D-printed
concrete
enabling
prediction
flow
directly
design.
can
facilitate
fabrication
structures
that
fulfill
necessary
printability
requirements.
Scientific Reports,
Journal Year:
2024,
Volume and Issue:
14(1)
Published: May 25, 2024
One
of
the
major
challenges
in
civil
engineering
sector
is
durability
reinforced
concrete
structures
against
carbonation
during
physico-chemical
process
interaction
hydrated
cementitious
composites
with
carbon
dioxide.
This
aggressive
causes
penetration
into
reinforcement
part,
which
affects
behavior
structure
its
lifetime
due
to
corrosion
risk.
A
countermeasure
using
alternative
materials
improve
texture
and
resist
increased
depth
(CD).
Considering
that
CD
test
requires
a
long
time
skilled
technician,
this
study
strives
provide
an
approach
by
moving
from
traditional
laboratory-based
methods
towards
artificial
intelligence
(AI)
techniques
for
modeling
sustainable
containing
fly
ash
(CCFA).
Despite
development
single
AI
models
so
far,
it
undeniable
utilizing
metaheuristic
optimization
form
hybrid
can
their
performance.
To
end,
new
model
integration
biogeography-based
(BBO)
technique
neural
network
(ANN)
developed
first
estimate
CCFA.
The
error
distribution
results
revealed
59%
ANN
predictions
had
errors
within
range
(-
1
mm,
mm],
while
corresponding
percentage
ANN-BBO
was
70%,
indicating
11%
reduction
prediction
proposed
model.
Furthermore,
A10-index
highlighted
performance
improvement
78%
model,
met
closeness
predicted
values
observed
ones,
value
index
0.5019
0.8947,
respectively.
Analyzing
cross-validation
confirmed
reliability
generalizability
Also,
three
most
influential
variables
estimating
were
exposure
(27%),
dioxide
concentration
(22%),
water/binder
(18%),
Finally,
superiority
verified
comparing
previous
studies'
models.
