Highlights in Science Engineering and Technology,
Год журнала:
2024,
Номер
98, С. 323 - 336
Опубликована: Май 16, 2024
The
aim
of
this
paper
is
to
conduct
an
assessment
study
illegal
wildlife
trade
projects
using
LSTM
models
and
Monte
Carlo
simulation
techniques.
Firstly,
based
on
the
data
from
1994-2023,
we
predicted
number
animal
plant
in
next
five
years
model,
results
showed
that
although
was
a
decreasing
trend,
it
still
high,
indicating
problem
needs
attract
global
attention.
Subsequently,
used
Kendall
correlation
coefficient
analyse
relationship
between
counts
economic,
environmental
climate
indicators,
found
positive
with
economic
losses
natural
disasters
extreme
weather
events.
Finally,
identified
seven
key
parameters
affecting
project
success
simulated
posterior
distributions
these
Markov
Chain
method,
then
conducted
simulations
estimate
probability
as
93.12%.
Sensitivity
analyses
indicate
most
sensitive
level
financial
support
monitoring
technology.
Overall,
data-driven
approach,
provides
important
reference
for
assessing
projects.
Computer-Aided Civil and Infrastructure Engineering,
Год журнала:
2025,
Номер
unknown
Опубликована: Май 19, 2025
Abstract
Real‐time
prediction
of
subgrade
settlement
caused
by
shield
tunneling
is
crucial
in
engineering
applications.
However,
data‐driven
methods
are
prone
to
overfitting,
while
physical
rely
on
certain
assumptions,
making
it
difficult
select
satisfactory
parameters.
Although
there
currently
physics‐data‐driven
methods,
they
typically
require
extensive
iterative
calculations
with
models,
which
makes
them
unavailable
for
real‐time
prediction.
This
paper
introduces
a
lightweight
method
predicting
tunneling.
The
core
concept
involves
using
single
calculation
the
model
provide
weak
constraint.
A
deep
learning
network
then
designed
capture
spatiotemporal
correlations
based
ConvLSTM.
By
iteratively
incorporating
data,
constraints
further
enhanced.
combines
predictive
power
reasonable
laws,
validated
good
performance
practical
project.
results
demonstrate
that
this
meets
requirements
engineering,
achieving
an
coefficient
determination
0.980,
root
mean
square
error
0.22
mm,
and
absolute
0.15
mm.
Furthermore,
outperforms
both
models
demonstrates
generalization
performance.
study
provides
effective
guidance
practices.
