BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:Identification of individual fibers from 3d digital images
DTSTART;VALUE=DATE-TIME:20180516T163500Z
DTEND;VALUE=DATE-TIME:20180516T165000Z
DTSTAMP;VALUE=DATE-TIME:20210928T200153Z
UID:indico-contribution-163-333@events.interpore.org
DESCRIPTION:Speakers: Andreas Grießer (Math2Market)\nFibrous structures a
re present in many materials\, including non-woven filter media used for f
iltration\, carbon-fiber reinforced plastics or glass-fiber reinforced pla
stics used in mechanical applications\, or gas-diffusion layers used in fu
el cells. Spatial distribution\, orientation\, length\, curvature and cent
er line of fibers in materials like these are essential characteristics ne
eded to know in modern material design. Being able to analyze these proper
ties from micro-CT scans is highly important to create precise models of e
xisting materials. \n\nMost algorithms currently used to extract the stati
stics of the fiber orientations [1\, 2\, 3] first split the image space in
to many small fiber segments and try to recombine the over-segmented fiber
segments afterwards. These methods lack accuracy\, because they often con
nect fiber center lines incorrectly. The center line determination is espe
cially challenging in places where two or more fibers touch.\n\nWe propose
a machine learning based algorithm to identify and extract the individual
fibers in segmented 3D images to be able to fully characterize the fibers
and the overall composition of a material. Machine learning based on deep
neural networks requires massive amounts of training data\, in our case k
nown fiber center lines. One approach could be to create these manually. H
owever\, this is not feasible for 3d data sets. Instead\, we use GeoDict
’s [4] fiber structure modelling capabilities and scripting capabilities
to generate training data sets\, which consist of voxelized 3d fiber mode
ls and analytically known fiber center lines.\n\nWhen applied to a micro-C
T scan\, the trained neural network first labels the contact voxels (or bo
nd points in the case of bonded fibers) between two fibers. Second\, the l
abeled contact voxels are removed from the segmented image. In the final s
tep\, the resulting connected components are analyzed\, and a skeleton-bas
ed approach is used to obtain analytic descriptions of every fiber. The ce
nter lines can then be used to analyze the material\, e.g. to find spatial
distributions\, fiber orientation\, fiber length\, fiber curvature\, etc.
It is also possible to create a beam element model of the original micro-
CT scan.\n\nhttps://events.interpore.org/event/2/contributions/333/
LOCATION:New Orleans
URL:https://events.interpore.org/event/2/contributions/333/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Image analysis\, microstructure generation and effective property
estimation of cement-based materials used as radioactive waste confinemen
t barriers
DTSTART;VALUE=DATE-TIME:20180516T171100Z
DTEND;VALUE=DATE-TIME:20180516T172600Z
DTSTAMP;VALUE=DATE-TIME:20210928T200153Z
UID:indico-contribution-163-346@events.interpore.org
DESCRIPTION:Speakers: Laurent Lemmens (SCK-CEN / KU LEUVEN)\nTransport thr
ough cement-based materials depends strongly on their3D microstructure. T
o evaluate transport and related processes at time and spatial scales beyo
nd experimental data\, and to gain in depth understanding of the critical
role of the microstructure\, numerical simulations are necessary. This\, h
owever\, requires an accurate description of the 3D microstructures involv
ed. However\, for cement paste\, 3D imaging techniques such as µCT do not
have sufficient resolution to appropriately describe the microstructure.
We therefore present a new 3D-multiphase stochastic reconstruction methodo
logy based on simulated annealing. Our methodology uses a single or set of
2D SEM images to calculate different 2D structural descriptors which are
then fitted by a multi-phase simulated annealing-based reconstruction algo
rithm. This approach is easily extendible for generating 3D microstructure
s\, by assuming isotropy. We demonstrate our approach for ordinary Portlan
d cement pastes with different water to cement ratios. We evaluate the rec
onstruction methodology by calculating diffusion coefficients with a pore-
scale transport model using a lattice Boltzmann approach and the generated
3D reconstructions. Results seem to agree very well with the measured dif
fusion coefficients for the same cement pastes. This builds confidence in
the adequacy of the proposed simulated annealing algorithm for generating
3D realizations of cement paste from a 2D image. The approach will furthe
r developed to simulate transport properties in evolving porous systems du
ring degradation to evaluate the confinement properties of cement-based ma
terials over larger time scales in e.g. the context of radioactive and haz
ardous waste management.\n\nhttps://events.interpore.org/event/2/contribut
ions/346/
LOCATION:New Orleans
URL:https://events.interpore.org/event/2/contributions/346/
END:VEVENT
BEGIN:VEVENT
SUMMARY:¬Numerical Design of Porous Materials Using Adjustable Level-Cut
Poison Field Method
DTSTART;VALUE=DATE-TIME:20180516T172900Z
DTEND;VALUE=DATE-TIME:20180516T174400Z
DTSTAMP;VALUE=DATE-TIME:20210928T200153Z
UID:indico-contribution-163-343@events.interpore.org
DESCRIPTION:Speakers: Ben Paisley (University of Cincinnati)\nDespite rece
nt advances in synthesis and manufacturing of porous materials and devices
\, producing porous structures with targeted properties is still an expens
ive\, trial-and-error procedure. Numerical porous media design is one of t
he possible ways to accelerate this process and to guide manufacturing.\nC
urrent numerical porous media design methodologies often include a random
microstructure generator nested within an optimization routine. At each it
eration\, the optimization algorithm compares properties of the microstruc
ture with desired target properties\, such as permeability\, porosity and
pore size distribution\, and adjusts the inputs to the generator according
ly. A considerable drawback of this approach is computational cost\, which
is mainly due to the time needed to generate a completely new microstruct
ure at each iteration. \nTo address this problem\, we propose the Adjustab
le Level-Cut Poison Field (ALCPF) method\, a new approach based on the lev
el-cut Poisson field theory [Grigoriu\, 2003]. First\, several initial dom
ains are generated based on set of filtered Poisson field parameters. Then
\, rather than generating a completely new filtered Poisson field\, the op
timization algorithm takes a weighted geometric average of the initial dom
ains to produce a new filtered Poisson field. The process is repeated unti
l an optimal domain is found\, with a dramatic reduction in computational
time. \nThe weighted geometric averaging also has an added advantage of sp
atial control over the material microstructure. By adjusting the weights o
f the geometric mean throughout the initial domains\, an inhomogeneous vir
tual material with desired microstructure can be generated.\n\nhttps://eve
nts.interpore.org/event/2/contributions/343/
LOCATION:New Orleans
URL:https://events.interpore.org/event/2/contributions/343/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Capillary Suction Response of Granular Materials from Computed Tom
ography and Direct Numerical Simulations
DTSTART;VALUE=DATE-TIME:20180516T165300Z
DTEND;VALUE=DATE-TIME:20180516T170800Z
DTSTAMP;VALUE=DATE-TIME:20210928T200153Z
UID:indico-contribution-163-338@events.interpore.org
DESCRIPTION:Speakers: Mohmad Mohsin Thakur (University of Tennessee)\nThe
mechanical and hydraulic response of granular media in partially saturate
d conditions can be highly intricate and requires proper understanding at
the pore scale. The response of partially saturated sands to complex loadi
ngs such as projectile penetration necessitates integrating capillary suct
ion in the constitutive model framework. In this study\, Soil Water Suctio
n Curve (SWCC) predictions of Ottawa sand were made using attenuation cont
rast based X-ray Computed Tomography (CT) data and Direct Numerical Simula
tions using the digital material library package GeoDict. This approach of
predicting SWCC is fast and less tedious compared to the existing experim
ental methods. The pore morphology method and Young Laplace equation were
used to simulate the drainage and imbibition processes. The resolution of
CT scans is critical in segmentation of pore phase from solid phase\, ther
efore\, CT data was acquired at three different resolutions and its effect
on SWCC predictions was observed. The drying and wetting behavior of soli
ds depends on the surface roughness and geometry which manifests variation
in contact angle. SWCC predictions indicate a significant variation in ai
r entry value of Ottawa sand as contact angle changes. Furthermore\, the d
istribution of wetting phase (water) and non-wetting phase (air) at differ
ent saturations is presented. The SWCC predictions obtained in this study
can be helpful to model coupling between multiphase flow and mechanical b
ehavior for granular materials such as sands for static and impact problem
s including projectile penetration studies.\n\nhttps://events.interpore.or
g/event/2/contributions/338/
LOCATION:New Orleans
URL:https://events.interpore.org/event/2/contributions/338/
END:VEVENT
BEGIN:VEVENT
SUMMARY:Capillary network simulations based on the centreline representati
on
DTSTART;VALUE=DATE-TIME:20180516T161700Z
DTEND;VALUE=DATE-TIME:20180516T163200Z
DTSTAMP;VALUE=DATE-TIME:20210928T200153Z
UID:indico-contribution-163-334@events.interpore.org
DESCRIPTION:Speakers: Rodrigo Neumann Barros Ferreira (IBM Research)\nTomo
graphic 3D imaging at the pore scale provides an accurate geometrical repr
esentation of the microstructure of porous networks in oil reservoir rocks
. Flow simulation models deployed on top of such geometrical representatio
n unveil a variety of phenomena and allow estimating oil recovery paramete
rs as part of reservoir assessment\, management and operation. Physical mo
dels based on sparsely-connected graphs have the advantages of high perfor
mance and memory efficiency. In this work\, we developed a centreline extr
action algorithm and used it to characterize the microstructure of the por
ous network and to build a capillary network simulation that enables the e
xtraction of relevant flow-related figures-of-merit from a 3D image. We do
that in a cloud solution environment that facilitates the consumption of
those computational tools and infrastructures by stakeholders in the indus
try.\n\nStarting from the segmented image\, the centreline extraction algo
rithm involves calculating the Image Foresting Transform [Falcão 2004] of
the pore space followed by a centrality-aware optimal path algorithm [Cor
men 2009] to find the most central path from a pore inlet to a pore outlet
. With the resulting centreline graph\, we build a capillary network in wh
ich every centreline voxel becomes a cylindrical capillary with diameter g
iven by the Euclidean Distance Transform\, therefore\, reducing drasticall
y (by a factor of 1000) the computational domain and number of variables i
nvolved. The Poiseuille’s law is used to establish the relationship betw
een imposed pressure gradient and flow rate in a capillary network simulat
ion [Man & Jing 1999]. Besides the flow-related simulation results\, one c
an extract morphological figures-of-merit from the centreline geometrical
representation\, such as connectivity\, fractal dimension\, diameter distr
ibution\, length distribution\, tortuosity distribution and the correlatio
n between pairs of such properties.\n\nhttps://events.interpore.org/event/
2/contributions/334/
LOCATION:New Orleans
URL:https://events.interpore.org/event/2/contributions/334/
END:VEVENT
END:VCALENDAR