Physics of Fluids,
Год журнала:
2024,
Номер
36(12)
Опубликована: Дек. 1, 2024
This
study
explores
a
hybrid
framework
integrating
machine
learning
techniques
and
symbolic
regression
via
genetic
programing
for
analyzing
the
nonlinear
propagation
of
waves
in
arterial
blood
flow.
We
employ
mathematical
to
simulate
viscoelastic
flow,
incorporating
assumptions
long
wavelength
large
Reynolds
numbers.
used
fifth-order
evolutionary
equation
using
reductive
perturbation
represent
behavior
tube,
considering
tube
wall's
bending.
obtain
solutions
through
physics-informed
neural
networks
(PINNs)
that
optimizes
Bayesian
hyperparameter
optimization
across
three
distinct
initial
conditions.
found
PINN-based
models
are
proficient
at
predicting
higher-order
partial
differential
equations
spatial-temporal
domain
[−1,1]×[0,2].
is
evidenced
by
graphical
results
residual
validation
showing
mean
absolute
residue
error
O(10−3).
thoroughly
examine
impacts
various
Furthermore,
combined
into
single
model
random
forest
algorithm,
achieving
an
impressive
accuracy
99%
on
testing
dataset
compared
with
another
artificial
network.
Finally,
analytical
form
estimated
provides
interpretable
square
These
insights
contribute
interpretation
cardiovascular
parameters,
potentially
advancing
applications
within
medical
domain.
Journal of Computational Physics,
Год журнала:
2024,
Номер
501, С. 112781 - 112781
Опубликована: Янв. 20, 2024
Physics-informed
neural
networks
(PINNs)
have
emerged
as
a
significant
endeavor
in
recent
years
to
utilize
artificial
intelligence
technology
for
solving
various
partial
differential
equations
(PDEs).
Nevertheless,
the
vanilla
PINN
model
structure
encounters
challenges
accurately
approximating
solutions
at
hard-to-fit
regions
with,
instance,
"stiffness"
points
characterized
by
fast-paced
alterations
timescale.
To
this
end,
we
introduce
novel
architecture
based
on
PINN,
named
loss-attentional
physics-informed
(LA-PINN),
which
equips
each
loss
component
with
an
independent
network
(LAN).
Feeding
squared
errors
(SE)
every
training
point
into
LAN
input,
attentional
function
is
then
built
and
provides
different
weights
diverse
SEs.
A
error-based
weighting
approach
that
utilizes
adversarial
between
multiple
LA-PINN
proposed
dynamically
update
of
SE
during
epoch.
Additionally,
mechanism
analysed
also
be
validated
performing
several
numerical
experiments.
The
experimental
results
indicate
method
displays
superior
predictive
performance
compared
holds
swift
convergence
characteristic.
Moreover,
it
can
advance
those
progressively
increasing
growth
rates
both
weight
gradient
error.
Abstract
Generative
Artificial
Intelligence
(GAI)
represents
an
emerging
field
that
promises
the
creation
of
synthetic
data
and
outputs
in
different
modalities.
GAI
has
recently
shown
impressive
results
across
a
large
spectrum
applications
ranging
from
biology,
medicine,
education,
legislation,
computer
science,
finance.
As
one
strives
for
enhanced
safety,
efficiency,
sustainability,
generative
AI
indeed
emerges
as
key
differentiator
paradigm
shift
field.
This
article
explores
potential
language
models
geoscience.
The
recent
developments
machine
learning
deep
have
enabled
model's
utility
tackling
diverse
prediction
problems,
simulation,
multi‐criteria
decision‐making
challenges
related
to
geoscience
Earth
system
dynamics.
survey
discusses
several
been
used
comprising
adversarial
networks
(GANs),
physics‐informed
neural
(PINNs),
pre‐trained
transformer
(GPT)‐based
structures.
These
tools
helped
community
applications,
including
(but
not
limited
to)
generation/augmentation,
super‐resolution,
panchromatic
sharpening,
haze
removal,
restoration,
land
surface
changing.
Some
still
remain,
such
ensuring
physical
interpretation,
nefarious
use
cases,
trustworthiness.
Beyond
that,
show
community,
especially
with
support
climate
change,
urban
atmospheric
marine
planetary
science
through
their
extraordinary
ability
data‐driven
modelling
uncertainty
quantification.
Computer-Aided Civil and Infrastructure Engineering,
Год журнала:
2025,
Номер
unknown
Опубликована: Фев. 18, 2025
Abstract
This
paper
introduces
a
novel
hybrid
multi‐model
thermo‐temporal
physics‐informed
neural
network
(TT‐PINN)
framework
for
thermal
loading
prediction
in
composite
bridge
decks.
Unlike
the
existing
PINN
applications
heat
transfer
that
focus
on
simple
geometries,
this
uniquely
addresses
multi‐material
domains
and
realistic
boundary
conditions
through
dual‐network
architecture
designed
structures.
The
further
incorporates
environmental
of
natural
convection
solar
radiation
into
loss
function
employs
learning
efficient
adaptation
to
varying
conditions.
Moreover,
mechanism
enables
rapid
new
states,
thus
markedly
reducing
computations
as
compared
conventional
finite
element
method
(FEM).
Through
noise‐augmented
training
parameter
identification,
TT‐PINN
effectively
handles
real‐world
monitoring
data
uncertainties
allows
material
property
calibration
with
limited
sensor
data.
framework's
ability
capture
complex
behavior
is
validated
by
studying
cable‐stayed
bridge.
It
significantly
reduces
computational
costs
traditional
FEM
approaches.
Mathematics,
Год журнала:
2025,
Номер
13(10), С. 1571 - 1571
Опубликована: Май 10, 2025
Physics-informed
neural
networks
(PINNs)
offer
a
mesh-free
approach
to
solving
partial
differential
equations
(PDEs)
with
embedded
physical
constraints.
Although
PINNs
have
gained
traction
in
various
engineering
fields,
their
adoption
for
railway
bridge
analysis
remains
under-explored.
To
address
this
gap,
systematic
review
was
conducted
across
Scopus
and
Web
of
Science
(2020–2025),
filtering
records
by
relevance,
journal
impact,
language.
From
an
initial
pool,
120
articles
were
selected
categorised
into
nine
thematic
clusters
that
encompass
computational
frameworks,
hybrid
integration
conventional
solvers,
domain
decomposition
strategies.
Through
natural
language
processing
(NLP)
trend
mapping,
evidences
growing
but
fragmented
research
landscape.
demonstrate
promising
capabilities
load
distribution
modelling,
structural
health
monitoring,
failure
prediction,
particularly
under
dynamic
train
loads
on
multi-span
bridges.
However,
methodological
gaps
persist
large-scale
simulations,
plasticity
experimental
validation.
Future
work
should
focus
scalable
PINN
architectures,
refined
modelling
inelastic
behaviours,
real-time
data
assimilation,
ensuring
robustness
generalisability
through
interdisciplinary
collaboration.
Communications Engineering,
Год журнала:
2025,
Номер
4(1)
Опубликована: Июнь 3, 2025
A
large
number
of
in-service
reinforced
concrete
structures
are
now
entering
the
mid-to-late
stages
their
service
life.
Efficient
detection
damage
characteristics
and
accurate
prediction
material
performance
degradation
have
become
essential
for
ensuring
safety
these
structures.
Traditional
methods,
which
primarily
rely
on
manual
inspections
sensor
monitoring,
inefficient
lack
accuracy.
Similarly,
models
materials,
often
based
limited
experimental
data
polynomial
fitting,
oversimplify
influencing
factors.
In
contrast,
partial
differential
equation
that
account
mechanisms
computationally
intensive
difficult
to
solve.
Recent
advancements
in
deep
learning
machine
learning,
as
part
artificial
intelligence,
introduced
innovative
approaches
both
This
paper
provides
a
comprehensive
overview
theories
models,
reviews
current
research
application
durability
structures,
focusing
two
main
areas:
intelligent
predictive
modeling
durability.
Finally,
article
discusses
future
trends
offers
insights
into
innovation
structure