steel research international,
Год журнала:
2023,
Номер
94(12)
Опубликована: Июнь 24, 2023
During
the
slab
sizing
press
(SP)
process,
pressing
force
corresponds
to
profile,
which
guides
production
schedule
design
and
final
profile
control.
To
accomplish
prediction
of
for
SP,
an
improved
ensemble
method
based
on
chaotic
Harris
hawks
optimizer
(CHHO)
stacking
is
proposed.
A
mechanistic
knowledge
introduced
feature
selection
that
enhances
rationality
input
features.
Subsequently,
11
machine
learning
models
are
compared
5
them
selected
as
candidate
learners
method.
Based
candidates
learners,
8
strategies
constructed,
stacked
model
with
extratree
regressor,
gradient
boosted
decision
trees,
kernel
ridge
regression
(KRR)
base‐learners
KRR
meta‐learner
performs
best.
The
R
2
,
MAE,
mean
square
error,
squared
log
absolute
percentage
error
test
dataset
0.9912,
0.0856,
0.0167,
0.0005,
2.00%,
respectively,
95%
errors
less
than
0.15
MN.
Then,
sensitivity
analysis
predictive
Shapley
Additive
Explanations
performed
demonstrate
good
alignment
proposed
physical
reality.
Furthermore,
cope
complexity
uncertainty
CHHO‐stacking
density
estimation
integrated
interval.
Materials Genome Engineering Advances,
Год журнала:
2023,
Номер
1(1)
Опубликована: Сен. 1, 2023
Abstract
With
the
development
of
new
information
technology,
big
data
technology
and
artificial
intelligence
(AI)
have
accelerated
material
research
industrial
manufacturing,
which
become
key
driving
a
wave
global
technological
revolution
transformation.
This
review
introduces
resources
databases
related
to
steel
materials.
It
then
examines
fundamental
strategies
applications
machine
learning
(ML)
in
design
discovery
materials,
including
ML
models
based
on
experimental
data,
manufacturing
simulation
respectively.
Given
advancements
AI/ML,
communication
technologies,
an
intelligent
mode
featuring
digital
twins
is
deemed
critical
guiding
next
revolution.
Consequently,
application
with
iron
industry
reviewed
discussed.
Furthermore,
service
performance
prediction
products
are
addressed.
Finally,
future
trends
for
data‐driven
AI
approaches
throughout
entire
life
cycle
materials
prospected.
Overall,
this
work
presents
in‐depth
examination
integration
technologies
industry,
highlighting
their
potential
directions.
