Cutaneous
melanoma
is
the
most
aggressive
form
of
skin
cancer,
responsible
for
cancer-related
deaths.
Recent
advances
in
artificial
intelligence,
jointly
with
availability
public
dermoscopy
image
datasets,
have
allowed
to
assist
dermatologists
identification.
While
feature
extraction
holds
potential
detection,
it
often
leads
high-dimensional
data.
Furthermore,
datasets
present
class
imbalance
problem,
where
a
few
classes
numerous
samples,
whereas
others
are
under-represented.
Journal of Adhesion Science and Technology,
Год журнала:
2025,
Номер
unknown, С. 1 - 26
Опубликована: Янв. 16, 2025
Fibre-reinforced
polymer
(FRP)
composites
are
increasingly
favoured
for
strengthening
existing
structures
due
to
their
numerous
structural
benefits.
Nevertheless,
the
performance
of
such
technology
is
strongly
affected
by
behaviour
epoxy
resin
adhesive
layer,
which
largely
dependent
on
its
curing
conditions.
This
study
introduces
a
deep
learning
(DL)
framework
that
leverages
eXtreme
Gradient
Boosting
(XGBoost)
and
genetic
programming
(GP)
comprehensively
influence
scenarios
vitreous
transition
adhesive.
An
experimental
dataset
comprising
160
data
points
was
used
develop
predictive
models.
The
XGBoost
models
exhibited
high
accuracy
both
onset
temperature
peak
tan
δ
temperature,
achieving
R2
values
0.982
0.993
unseen
test
set,
respectively.
While
GP
lower
with
0.834
0.842,
they
provided
explicit
equations
enhance
interpretability
DL
model
facilitate
practical
application.
To
make
these
insights
accessible
engineers
without
expertise,
web-based
graphical
user
interface
software
developed,
incorporating
all
Additionally,
feature
assessment
conducted,
providing
visual
representations
impact
each
output
results,
thus
enhancing
engineering
applications.
Fibre
Reinforced
Polymers
(FRPs)
have
become
increasingly
popular
for
strengthening
concrete
structures
due
to
their
structural
benefits.
However,
a
major
concern
with
FRP-strengthened
members
is
poor
fire
resistance.
This
study
introduces
genetic
evolutionary
deep
learning
(DL)
approach
that
utilises
the
LightGBM
algorithm,
enhanced
Genetic
Algorithm
hyperparameter
optimisation,
alongside
Programming
(GP)
assess
resistance
performance
of
strengthened
reinforced
(RC)
beams.
A
substantial
dataset
comprising
20,000
data
points,
derived
from
numerically
modelled
results
validated
through
experimental
studies,
underpins
data-driven
DL
analyses.
The
model
demonstrates
high
predictive
accuracy
time
and
deflection
at
failure
RC
beams,
R2
values
0.923
0.789,
respectively.
Although
GP
shows
lower
(R2
0.642
0.643),
it
provides
explicit
equations
facilitate
deeper
understanding
ease
application.
graphical
user
interface
software,
incorporating
these
two
models,
has
been
developed
enable
engineers
apply
insights
in
practice
without
requiring
coding
skills.
Furthermore,
an
assessment
feature
influences
was
conducted,
visually
depicting
impact
on
output
results,
thus
enhancing
interpretability
engineering
applications.
