This
thesis
proposes
a
novel
and
effective
physics-informed
machine
learning
framework
to
explore
microscale
variations
in
plant-based
foods
during
drying.
By
initiating
fundamental
numerical
investigations
at
the
cellular
level—which
significantly
influence
bulk-level
changes—this
research
highlights
framework's
robustness
flexibility
over
traditional
methods
like
FEA
meshfree
particle-based
methods.
The
work
demonstrates
considerable
potential
for
employing
this
learning-based
computational
modelling
technique
complex,
nonlinear
showcasing
its
superiority
analysing
predicting
drying
process
of
efficiently
accurately.
International Journal of Mechanical Sciences,
Год журнала:
2024,
Номер
275, С. 109267 - 109267
Опубликована: Апрель 7, 2024
This
paper
introduces
a
novel
Physics-Informed
Neural
Network-based
(PINN-based)
multi-domain
computational
framework
to
analyse
nonlinear
and
heterogeneous
morphological
variations
of
plant
cells
during
drying.
Here,
two
distinct
models
are
involved:
PINN-MT
simulate
mass
transfer;
PINN-NS
shrinkage.
The
coupled
examine
cellular
changes
resulting
from
moisture
loss
Firstly,
the
framework,
in
tandem
with
homogeneous
conditions,
operates
parallel,
allowing
mutual
parameters
update
between
models.
approach
demonstrates
ability
approximate
shrinkage
within
tissue,
factoring
influence
surrounding
cells.
Secondly,
non-uniform
cell
wall
properties
boundary
conditions
incorporated
into
this
through
domain
decomposition.
Inherent
capabilities
neural
networks
allow
for
seamless
integration
multiple
domains,
additional
terms
introduced
at
interfaces.
shows
capacity
account
drastic
even
under
extreme
drying
which
is
key
novelty
has
been
challenging
task
existing
traditional
methods.
Hence,
proposed
offers
an
innovative
avenue
understanding
not
only
cells,
but
also
soft
matter
general.
