AI in Civil Engineering,
Год журнала:
2024,
Номер
3(1)
Опубликована: Ноя. 14, 2024
Abstract
This
study
is
primarily
aimed
at
creating
three
machine
learning
models:
artificial
neural
network
(ANN),
random
forest
(RF),
and
k-nearest
neighbour
(KNN),
so
as
to
predict
the
crippling
load
(CL)
of
I-shaped
steel
columns.
Five
input
parameters,
namely
length
column
(
L
),
width
flange
b
f
thickness
t
web
w
)
height
H
are
used
compute
(CL).
A
range
performance
indicators,
including
coefficient
determination
R
2
variance
account
factor
(VAF),
a-10
index,
root
mean
square
error
(RMSE),
absolute
(MAE)
deviation
(MAD),
assess
effectiveness
established
models.
The
results
show
that
all
ML
(machine
learning)
models
can
accurately
load,
but
ANN
superior:
it
delivers
highest
value
=
0.998
lowest
RMSE
0.008
in
training
phase,
well
0.996
smaller
0.012
testing
phase.
Additional
methods,
rank
analysis,
reliability
regression
plot,
Taylor
diagram
matrix
employed
models’
performance.
index
β
calculated
by
using
first-order
second
moment
(FOSM)
technique,
result
compared
with
actual
value.
Additionally,
sensitivity
analysis
performed
check
impact
variables
on
output
(CL),
finding
has
greatest
followed
,
order.
demonstrates
techniques
useful
for
developing
a
reliable
numerical
tool
measuring
It
found
proposed
also
be
other
kinds
failures
different
perforated
Scientific Reports,
Год журнала:
2024,
Номер
14(1)
Опубликована: Ноя. 6, 2024
This
paper
presents
a
non-linear
finite
element
model
(FEM)
to
predict
the
load-carrying
capacity
of
three
different
configurations
elliptical
concrete-filled
steel
tubular
(CFST)
short
columns:
double
tubes
with
sandwich
concrete
(CFDST),
and
inside
inner
tube,
single
outer
tube
concrete.
Then,
parametric
analytical
study
was
performed
explore
influence
geometric
material
parameters
on
CFST
columns.
Furthermore,
current
investigates
effectiveness
machine
learning
(ML)
techniques
in
predicting
These
include
Support
Vector
Regressor
(SVR),
Random
Forest
(RFR),
Gradient
Boosting
(GBR),
XGBoost
(XGBR),
MLP
(MLPR),
K-nearest
Neighbours
(KNNR),
Naive
Bayes
(NBR).
ML
models
accuracy
is
assessed
by
comparing
their
predictions
FE
results.
Among
models,
GBR
XGBR
exhibited
outstanding
results
high
test
R2
scores
0.9888
0.9885,
respectively.
The
provided
insights
into
contributions
individual
features
using
SHapley
Additive
exPlanations
(SHAP)
approach.
from
SHAP
indicate
that
eccentric
loading
ratio
(e/2a)
has
most
significant
effect
columns,
followed
yield
strength
(
$$\:{f}_{yo}$$
)
width
$$\:2{a}_{ii}$$
).
Additionally,
user
interface
platform
been
developed
streamline
practical
application
proposed
ML.
Scientific Reports,
Год журнала:
2025,
Номер
15(1)
Опубликована: Апрель 15, 2025
The
current
study
investigates
the
application
of
artificial
intelligence
(AI)
techniques,
including
machine
learning
(ML)
and
deep
(DL),
in
predicting
ultimate
load-carrying
capacity
strain
ofboth
hollow
solid
hybrid
elliptical
fiber-reinforced
polymer
(FRP)-concrete-steel
double-skin
tubular
columns
(DSTCs)
under
axial
loading.
Implemented
AI
techniques
include
five
ML
models
-
Gene
Expression
Programming
(GEP),
Artificial
Neural
Network
(ANN),
Random
Forest
(RF),
Adaptive
Boosting
(ADB),
eXtreme
Gradient
(XGBoost)
one
DL
model
Deep
(DNN).Due
to
scarcity
experimental
data
on
DSTCs,
an
accurate
finite
element
(FE)
was
developed
provide
additional
numerical
insights.
reliability
proposed
nonlinear
FE
validated
against
existing
results.
then
employed
a
parametric
generate
112
points.The
examined
impact
concrete
strength,
cross-sectional
size
inner
steel
tube,
FRP
thickness
both
DSTCs.The
effectiveness
assessed
by
comparing
models'
predictions
with
results.Among
models,
XGBoost
RF
achieved
best
performance
training
testing
respect
determination
coefficient
(R2),
Root
Mean
Square
Error
(RMSE),
Absolute
(MAE)
values.
provided
insights
into
contributions
individual
features
using
SHapley
Additive
exPlanations
(SHAP)
approach.
results
from
SHAP,
based
prediction
model,
indicate
that
area
core
has
most
significant
effect
followed
unconfined
strength
total
multiplied
its
elastic
modulus.
Additionally,
user
interface
platform
streamline
practical
DSTCs.
