ABSTRACT
In
the
realm
of
earthquake
engineering,
response
spectra
play
a
crucial
role
in
characterizing
effects
site
dynamic
characteristics
under
seismic
activity.
Consequently,
accurately
predicting
is
paramount
importance.
We
have
developed
physics‐guided
bidirectional
long
short‐term
memory
neural
network
model
(Phy‐BiLSTM)
that
proficient
based
on
bedrock
records.
The
core
principle
Phy‐BiLSTM
to
improve
alignment
between
solution
space
and
ground
truth
by
integrating
physics
knowledge
obtained
from
physical
model.
introduced
this
study
utilized
5%‐damped
spectra,
which
were
derived
strong
motion
records
collected
at
KiK‐net
downhole
array.
results
substantiate
performance
enhancement
comparison
data‐driven
BiLSTM
Furthermore,
we
conduct
comparative
analysis
against
traditional
methods
(EQ,
SBSR)
as
well
other
architectures
(CNN
LSTM).
result
highlights
advantages
response.
Engineering Applications of Artificial Intelligence,
Journal Year:
2023,
Volume and Issue:
130, P. 107425 - 107425
Published: Dec. 22, 2023
In
recent
years,
the
European
Commission
and
International
Maritime
Organization
(IMO)
implemented
various
operational
measures
policies
to
reduce
ship
fuel
consumption
related
emissions.
The
effectiveness
of
these
relies
upon
developing
accurate
predictive
models
encompassing
influence
real
conditions.
This
paper
presents
a
deep
learning
method
for
prediction
consumption.
utilizes
big
data
analytics
from
sensors,
voyage
reporting
hydrometeorological
data,
comprising
266
variables
made
available
following
sea
trials
Kamsarmax
bulk
carrier
Laskaridis
Shipping
Co.
Ltd.
A
variable
importance
estimation
model
using
Decision
Tree
(DT)
is
used
understand
underlying
relationships
in
dataset.
Consequently,
developed
sailing
speed,
heading,
displacement/draft,
trim,
weather,
conditions,
etc.
on
(SFC).
achieved
by
incorporating
attention
mechanism
into
Bi-directional
Long
Short-Term
Memory
(Bi-LSTM)
network.
potential
new
demonstrated
training
streams
corresponding
rates
as
well
internal
external
comprehensive
comparison
with
existing
methods
indicates
that
Bi-LSTM
best
fit
when
high
frequency
data.
It
concluded
subject
further
testing
validation
could
be
development
decision
support
systems
monitoring
environmentally
sustainable
operations.
Applied Sciences,
Journal Year:
2024,
Volume and Issue:
14(15), P. 6658 - 6658
Published: July 30, 2024
Accurate
seismic
ground
response
analysis
is
crucial
for
the
design
and
safety
of
civil
infrastructure
establishing
effective
mitigation
measures
against
risks
hazards.
This
a
complex
process
due
to
nonlinear
soil
properties
complicated
underground
geometries.
As
simplified
approach,
one-dimensional
wave
propagation
model,
which
assumes
that
waves
travel
vertically
through
horizontally
layered
medium,
widely
adopted
its
reasonable
performance
in
many
practical
applications.
study
explores
potential
sequence
deep
learning
models,
specifically
1D
convolutional
neural
networks
(1D-CNNs),
long
short-term
memory
(LSTM)
networks,
transformers,
as
an
alternative
modeling.
Utilizing
motion
data
from
Kiban
Kyoshin
Network
(KiK-net),
we
train
these
models
predict
surface
acceleration
spectra
based
on
bedrock
motions.
The
data-driven
compared
with
conventional
equivalent-linear
SHAKE2000.
results
demonstrate
outperform
physics-based
model
across
various
sites,
transformer
exhibiting
smallest
average
prediction
error
ability
capture
long-range
dependencies.
1D-CNN
also
shows
promising
performance,
albeit
occasional
higher
errors
than
other
models.
All
exhibit
efficient
computation
times
less
0.4
s
estimation.
These
findings
highlight
approaches
