Advances in Structural Engineering,
Journal Year:
2024,
Volume and Issue:
unknown
Published: Oct. 1, 2024
Steel
circular
hollow
section
(CHS)
members
are
widely
utilized
as
axial
force-resisting
structural
in
civil
engineering
structures.
The
buckling
strength
under
loads
is
one
of
the
critical
parameters
to
determine
performance
steel
CHS
members,
which
significantly
affected
by
discreteness
introduced
geometries,
material,
and
initial
imperfections.
However,
reduction
factor
employed
modern
design
codes
(i.e.
Chinese
EC3)
only
accounts
for
caused
all
kinds
does
not
reflect
impacts
every
single
imperfection.
To
fill
gap,
this
paper
proposed
an
interpretable
machine-learning
method
provide
probabilistic
prediction
result
a
distribution
form
with
consideration
detailed
discreteness.
model
predict
nominal
was
first
developed
utilizing
ten
machine
learning
algorithms
after
sufficient
numerical
simulations,
where
verified
using
test
results.
artificial
neural
network
(ANN)
selected
developing
due
its
highly
reliable
testing.
ANN
models
were
further
interpreted
Shapley
Additive
exPlanations
(SHAP)
interrelationship
different
parameters.
Then,
bucking
established
based
on
models,
Latin
hypercube
sampling
applied
address
generated
model’s
effectiveness
evidence
that
results
can
match
results'
probability
density
function
from
while
reducing
computation
time.
Finally,
parameters’
impact
strength’s
evaluated
global
sensitivity
analysis
(GSA)
method.
shows
substantially
influences
accurate
prediction.
