Non-Destructive Concrete Strength Prediction Using AI: A Comparative Study of Machine Learning and Deep Learning Models
Abstract
Accurate
prediction
of
concrete's
mechanical
properties
is
a
crucial
aspect
civil
engineering,
ensuring
the
structural
integrity
and
durability
constructions.
Traditional
destructive
testing
methods,
while
reliable,
are
time-consuming
resource-intensive.
This
study
presents
novel,
non-destructive
approach
for
predicting
compressive,
tensile,
flexural
strengths
concrete
using
only
two
input
parameters:
Ultrasonic
Pulse
Velocity
(UPV)
Electrical
Resistivity
(ER).
A
comparative
analysis
was
conducted
utilizing
five
machine
learning
deep
models:
Support
Vector
Regression
(SVR),
K-Nearest
Neighbors
(KNN),
Extreme
Gradient
Boosting
(XGBoost),
Multi-Layer
Perceptron
(MLP),
Convolutional
Neural
Networks
(CNN).
The
results
demonstrated
that
CNN
outperformed
all
other
models,
achieving
lowest
Root
Mean
Square
Error
(RMSE)
Relative
(MRE)
across
three
strength
predictions.
Specifically,
achieved
an
MRE
1.37%
compressive
strength,
1.25%
tensile
1.76%
highlighting
its
superior
predictive
accuracy
compared
to
traditional
models.
CNN's
strong
performance
stems
from
ability
learn
deep,
non-linear
feature
hierarchies
minimal
inputs.
By
capturing
complex
spatial
functional
dependencies
between
UPV
ER,
can
model
intricate
behavior
more
effectively
than
shallow
makes
it
particularly
suitable
tasks
involving
highly
physical
phenomena,
such
as
characteristics
indirect
measurements.
research
highlights
potential
AI-driven
efficient
alternative
offering
significant
advantages
in
terms
cost
reduction,
speed,
sustainability
construction
industry.
Research Square (Research Square), Год журнала: 2025, Номер unknown
Опубликована: Май 9, 2025
Язык: Английский