International Journal of Pavement Engineering,
Journal Year:
2022,
Volume and Issue:
24(2)
Published: Oct. 29, 2022
Each
type
of
soil
has
different
optimal
stabilisation
additive
content.
To
design
the
component,
reliable
and
efficient
models
are
required.
The
study
proposes
Machine
Learning
(ML)
model
Support
Vector
Regression
(SVR)
to
predict
Unconfined
Compressive
Strength
(UCS)
stabilised
soil.
be
able
deliver
performance,
five
metaheuristic
algorithms:
Simulated
Annealing
(SA),
Random
Restart
Hill
Climbing
(RRHC),
Particle
swarm
optimisation
(PSO),
Hunger
Games
Search
(HGS)
Slime
Mould
Algorithm
(SMA)
integrated
with
SVR
model.
explore
effect
number
inputs
on
model's
data
was
divided
into
two
scenarios
input
variable
number.
ML
evaluated
by
K-Fold
numerical
indicators
R2,
RMSE
MAE.
results
show
that
in
Scenario
1,
SVR-HGS
a
higher
predictive
performance
than
other
models.
While
2,
SVR-PSO
gives
better
remaining
SHapley
Additive
exPlanation
(SHAP)
Partial
Dependence
Plots
2D
(PDP)
were
used
gain
insight
effects
variables
UCS,
cement
lime
variables.
Obtaining
have
an
important
influence
variation
which
is
considered
most
significant
variable.
detection
A-line
value
relatively
predictor
UCS.
At
suitable
value,
it
possible
reduce
content
chemical
stabilising
agents
(cement,
lime)
while
maintaining
UCS
at
relative
threshold.
Sustainability,
Journal Year:
2022,
Volume and Issue:
14(15), P. 9292 - 9292
Published: July 29, 2022
In
this
study,
we
investigated
the
flooding
accident
that
occurred
on
Metro
Line
5
in
capital
city
of
Zhengzhou,
Henan
Province,
China.
On
20
July
2021,
owing
to
an
extreme
rainstorm,
serious
inundation
Wulongkou
parking
lot
Zhengzhou
and
its
surrounding
area.
Flooding
forced
a
train
stop
during
operation,
resulting
14
deaths.
Based
our
preliminary
investigation
analysis
accident,
designed
three
main
control
measures
reduce
occurrence
similar
accidents
mitigate
impact
future,
given
increasing
number
storm
weather
events
recent
years:
(1)
conduct
subway
flood
risk
assessments
establish
early
warning
system,
involving
real-time
monitoring
meteorological
information
operation
construction;
(2)
improve
emergency
plans
response
mechanism
for
flooding;
(3)
strengthen
safety
awareness
training
ensure
orderly
evacuation
people
after
accidents.
Applied Sciences,
Journal Year:
2023,
Volume and Issue:
13(10), P. 6050 - 6050
Published: May 15, 2023
Measurement
while
drilling
(MWD)
data
reflect
the
rig–rock
mass
interaction;
they
are
crucial
for
accurately
classifying
rock
ahead
of
tunnel
face.
Although
machine-learning
methods
can
learn
relationship
between
MWD
and
mechanics
parameters
to
support
classification,
most
current
models
do
not
consider
impact
continuous
drilling-sequence
process,
thereby
leading
rock-classification
errors,
small
unbalanced
field
datasets
result
in
poor
model
performance.
We
propose
a
novel
deep
neural
network
based
on
Bi-directional
Long
Short-Term
Memory
(BILSTM)
extract
information-related
sequences
improve
accuracy
rock-mass
classification.
Two
optimization
modules
were
designed
model’s
generalization
Stratified
K-fold
cross-validation
was
used
datasets.
Model
validation
is
dataset
highway
Yunnan,
China.
Multiple
metrics
show
that
prediction
ability
significantly
better
than
those
multilayer
perceptron
(MLP)
support-vector
machine
(SVM),
exhibits
an
improved
The
reach
90%,
which
13%
15%
higher
MLP
SVM,
respectively.
