Acta of Bioengineering and Biomechanics Original paper

Vol. 24, No. 1, 2022 DOI: 10.37190/ABB-01989-2021-02

Discrete phase model of blood flow

in a roughness microchannel simulating the formation
of pseudointima

MAGDALENA KOPERNIK1*, KARINA DYRDA, PRZEMYSŁAW KURTYKA2, 3, ROMAN MAJOR3

1
AGH University Science and Technology, Kraków, Poland.
2
Foundation for Cardiac Surgery Development, Institute of Heart Prostheses, Zabrze, Poland.
3
Institute of Metallurgy and Materials Science, Polish Academy of Sciences, Kraków, Poland.

Purpose: The goal of the present study was the development of discrete phase model to simulate the phenomenon of backfilling
a morphologically complex surface by red blood cells (RBCs) in a flow microchannel and to anticipate the conditions of forming a pseu-
dointima. The objective of the experimental studies that inspired the development of the simulation was to create a surface that stimulates
the formation of the pseudointima layer. Methods: The finite volume method (FVM) and discrete particle method (DPM) were applied
to develop the target model. In addition, a mixture model and a roughness model of bottom layer were tested in the present study to show
their influence on simulation the phenomenon of backfilling a morphologically complex surface by RBCs in a flow microchannel.
Results: Numerical models were developed including: a) FVM models to compare the effect of applying boundary conditions
with/without roughness and cubes, as well as the analysis of their influence on blood velocity and shear stress; b) mixture models to
compare the effect of applying different boundary conditions and cubes on computed results; c) DPM models to compare the effect of
applying and not applying roughness as a boundary condition; d) DPM models with a morphologically complex surface and RBCs colli-
sions to present RBCs concentration, velocity and time distributions during flow in a channel. Conclusions: The analysis carried out for
the developed numerical models indicates that DPM model with cubes computes the best results. It also shows the backfilling of a mor-
phologically complex surface of the bottom microchannel with RBCs.

Key words: finite volume method (FVM), disperse particle method (DPM), mixture model, red blood cells (RBCs), blood models, micro-
channel, pseudointima

1. Introduction Experimental and numerical work on the modeling

of the flow in microchannels and ensuring good com-
pliance with the experiment is used, inter alia, mix-
The analysis of the phenomena related to the blood ture theory [27], discrete particle method [53] and
flow in the microchannels is focused on macroscopic SST k-ω turbulence model [53], [21]. The analysis of
flow phenomena such as Fahraeus effect and Fah- the flow in microchannels may also concern issues
raeus–Lindqvist effect [49]. For microtubes diameters related to blood-contacting devices, which cause he-
less than 300 µm, Fahraeus and Lindqvist observed molysis [53]. In this work, the turbulence model will
that the apparent blood viscosity decreases as the mi- be used due to the need to analyze the effects of in-
crotube diameter becomes smaller. The experimental duced surface roughness.
research tools used for these purposes are micro-PIV The selection of rheological models for blood flow
(particle image velocimetry) and confocal micro-PIV modeling in this work takes into account the relation-
techniques [52]. ship between the shear stress and the shear rate, for
______________________________
* Corresponding author: Magdalena Kopernik, AGH University Science and Technology, al. Mickiewicza 30, 30-059 Kraków,
Poland. E-mail: [email protected]
Received: November 22nd, 2021
Accepted for publication: February 10th, 2022
132 M. KOPERNIK et al.

which the differences are more visible for different havior of blood. RBCs are the main part responsible
roughnesses [44] when the rheological model with for blood viscosity.
well-separated yield stress is used [31]. The yield Influence on RBCs aggregation have [15]:
stress of blood is typically very small, on the order of – sedimentation rate,
1 mPa, and consequently difficult to measure, and can – fibrinogen and C-reactive protein;
be easily missed. – polysaccharides, dextran and hydroxyethyl starch
Fluids can be classified as compressible or incom- (HES);
pressible and depending on whether their viscosity – hydrodynamic radius that is inhibiting the aggre-
does not change applying a shear stress (Newtonian) gation below ca. 4 nm, while the above ca. 4 nm
or the viscosity changes applying a shear stress (non- the RBC aggregation is induced;
Newtonian). For example, the Newtonian blood models – colloidal osmotic pressure of the macromolecules
with well distinguished yield stress are Casson model, as the main reason for rouleaux formation (deple-
Herschel–Bulkley model and Bingham model [15]. tion theory);
However, the non-Newtonian character of blood rhe- – adsorption of macromolecules onto the RBC mem-
ology was recognized scientifically fairly early, more branes and hence a potential adhesion between two
than 100 years ago. The review of literature presented cells as the origin of rouleaux formations (bridging
in [14] and in the Authors’ previous works [28], [29] theory);
shows that it is recommended to use non-Newtonian – shape of contact zones and the buckling phenome-
blood models, like power law [29] and Herschel– non;
Bulkley [28] for small blood channels. Many papers – interparticle distance, which is low for depletion
state that blood cannot be considered as a single phase (close to zero) and high for bridging (in the range
homogeneous viscous fluid, especially in the narrowing of the radius of gyration of the macromolecules);
arteries [42]. Theory of interacting continua for mix- – temperature (could change the balance between
tures which treats blood as a two fluid mixture con- the fundamental mechanics of RBCs aggregation).
taining plasma, treated as a viscous fluid, and RBCs, Models that take aggregation of red blood cells
treated as a concentrated suspension with hematocrit into account are, for example, thixotropic models [19],
and shear rate dependent viscosity, with special mod- [22] or molecular dynamics models [3], [28], which
eling for the drag and lift forces [27]. This model has are not considered in this paper.
shown to have reasonable agreement with experi- Textured biomaterial surfaces in implantable left
ments conducted in a sudden expansion microchan- ventricular assist devices induce development of a non-
nel flow. Thus, the blood flow in small channels in thrombotic neointimal surface and allow elimination
the present paper will be also represented by the two- of anticoagulation therapy in device recipients [8],
phase models. [38], [47], [51], [55]. Characterization of the hemato-
Particle-based models are applied to simulate RBCs poietic cells formed within the neointimal surfaces of
membrane dynamics, margination effects and hetero- these devices may contribute to our understanding
geneities [3]. In the present paper, a discrete phase of this unique neointima. By implanting textured sur-
model (DPM) of backfilling a morphologically complex faces in contact with blood, a rapid clotting process oc-
surface by RBCs is developed to investigate the be- curs on the surface. While this may seem like a negative
havior of the RBCs from a Lagrangian view and a dis- for blood contact surfaces, the concept is that while
crete perspective. The blood behavior in Lagrangian clots form quickly on these surfaces, they are densely
view is examined on the basis of a particle tracking of adherent and do not appear to embolise into the
a RBC of blood plasma flow, whereas blood plasma bloodstream in a clinically relevant way. Over time,
behavior is considered in Eulerian view based on the additional blood cell interaction occurs, similar to an
assumption of a finite volume element in the fluid immune response. A heterogeneous layer containing
flow path. In modeling of RBC, primary blood plasma platelets, monocytes, macrophages, giant cells, lym-
interaction a two way approach is considered, in which phocytes and pluripotent hematopoietic cells is de-
RBC fluid flow and carrier blood plasma flow interact posited on the surface. It is postulated that pluripo-
with each other simultaneously. tent hematopoietic cells differentiate into fibroblasts,
The adhesion of RBCs has been the center of in- myofibroblasts and in some cases endothelial cells.
terest for years [15]. In vivo, these clustering proc- Fibroblast cells can then secrete extracellular matrix
esses could play a significant role in thrombi forma- components such as collagen, which is routinely
tions and in the blood coagulation, while the RBCs detected on textured surfaces after prolonged im-
aggregation is responsible for the shear thinning be- plantation.
 Discrete phase model of blood flow in a roughness microchannel simulating the formation of pseudointima 133

The goal of the present study is the development duced by cutting 150 μm cubes in the lower layer with
of DPM (discrete phase model) to simulate the phe- a distance of 200 μm between them.
nomenon of backfilling a morphologically complex
surface by RBCs in a flow microchannel and, in this 2.1.2. Physical model of surface morphology
way, to anticipate the hemodynamic conditions of
forming the pseudointima. The morphology of bottom channel was reached
by using commercially pure titanium powders sintered
in vacuum on Ti6Al7Nb substrate. Titanium powders
2. Materials and methods differed in size and morphology. Two types of grains
were used – in the form of regular spheres and irregular
crystals in three gradations, taking the diameters of
The chapter describes the FVM, mixture and DPM microspheres and the size of angular crystals into
models, taking the theoretical background and ex- account. The regular and irregular patterns of bottom
perimental research that constitute the basis for their layer of channel had target roughness from Ra = 20 to
development into account. 50 µm. The distance between deposited particles was
in the range of 10–200 µm.

2.1. Physical models

2.2. Numerical models
The section introduces motivation and experimental
basis for developing a microchannel blood flow model The method of describing numerical models included
with a modified surface. in this paper meets the criteria for numerical models
placed in [13]. Simulations shown in this paper were run
2.1.1. Physical model of microchannel using Ansys Fluent 2021 R1 an Ansys Fluent R2 soft-
ware due to the fact that the new version was launched
Microchannels allow for the simulation of various during developing simulations. These two versions do
physiological conditions in blood vessels, while the not have meaningful differences and they do not have
system applied in this work was also used in [9]. The influence on computed results presented in this paper.
developed system of microchannels [9] allows for
defined shear stresses in long-term culture, imaging 2.2.1. FVM and mixture models
of living cells and immunofluorescence for the with roughness and cubes
analysis of the response to shear stress, mimicking of
shear stress conditions in microcapillary, venous and The section presents detailed solutions for FVM
arterial flows, displacement and adhesion of sus- and mixture models, which describe the turbulence
pended cells in media, 3D cell culture (interstitial model and roughness model, as well as the assumed
flow). values of parameters.
The basic geometry of the system was designed in 2.2.1.1. Theoretical background of roughness
accordance with the data included in the instruction and mixture models
manual of the experimental setup, based on the µ-Slide I
Luer plate described in [9]. The developed model The k-ω SST turbulent model is focused on the
consists of a rectangular polymer plate with dimen- precision of the k-ω model in the areas close to the
sions of 25.5 × 75.5 mm and a thickness of 1.6 mm. wall. The k-ω SST model’s transport equations which
There are sleeves 3.5 mm high, 1 mm thick with an are used in the Ansys Fluent code are [26]:
outer diameter of 7 mm on the top wall. A tunnel was
∂ ( ρk ) ∂ ∂ ⎡ ∂k ⎤
cut inside the plate to allow the medium to flow. The + ( ρkui ) = ⎢ Lk ⎥ + Ak − Bk (1)
inlet and outlet ducts take the shape of a cylinder with ∂t ∂xi ∂x j ⎣⎢ ∂x j ⎥⎦
a diameter of 5 mm and a height of 4.2 mm. These
rollers are mounted on a rounded cuboid 50 mm long, ∂ ( ρω ) ∂ ∂ ⎡ ∂ω ⎤
+ ( ρωui ) = ⎢ Lω ⎥ + Aω − Bω + Cω ,
5 mm wide and 0.8 mm high. ∂t ∂xi ∂x j ⎣⎢ ∂x j ⎥⎦
On the basis of the first model, a modified layout
(2)
shape was also built. The modification concerned the
bottom wall of the channel, which was subjected to where: ui – mean velocity, Ak – the production terms
corrugation. The unevenness of the surface was intro- of k, Bk – production terms of k, Aω – the production
134 M. KOPERNIK et al.

terms of ω, Bk – dissipation terms of k, Bω – dissipation It is not physically meaningful to have a mesh size
terms of ω, Lk – the effective diffusivity of k, Lω – the such that the wall-adjacent cell is smaller than the
effective diffusivity of ω, and Cω – the diffusion term. roughness height.
For modeling surface roughness, the rough wall Eulerian multi-phase models can account for dis-
model is introduced in the Ansys Fluent which was persed-continuous phase interactions and continuous-
adapted from [50]. To implement surface roughness in continuous phase interactions. Dispersed phases can be
the Ansys Fluent, the roughness height (Hr) and the particles (solids), droplets (liquids) or bubbles (gas).
roughness constant (C) should be defined. The effects They are dissolved in the continuous fluid. Dispersed
of surface roughness can be properly accounted for in phases typically have micrometer to millimeter in dia-
CFD codes by modified wall law as follows [43]: meter. Continuous-continuous phase interactions form
a discrete interface. They are immiscible with each other.
U q u1 1 ⎡ u1 y q ⎤ Mixture models are a simplified version of the full Eule-
= ln ⎢9.793 ⎥, (3)
τω / ρ k1 ⎣⎢ μ ⎦⎥ rian model for dispersed-continuous phase interactions.
Volume-of-fluid (VOF) models are a simplified version
where: Uq – velocity at the wall near the cell at centre of full Eulerian model for continuous-continuous phase
point q, yq – height at the centre point q of the cells interactions. Thus, the mixture model was selected in
near the wall, μ – viscosity, ΔG – roughness function, simulation of a blood flow in the present paper.
τω – wall shear stress, k1 – Karman constant, and u*: Eulerian model equations can be characterized by:
the wall friction velocity which can be defined as – a continuity/mass conservation equation is solved
[39]: for each phase,
– the equation is multiplied by the volume fraction
u * = 0.55k q0.5 , (4) of the phase (rq)

where kq is the turbulent kinetic energy near the wall ∂ (rq ρ q )

+ ∇ ⋅ (rq ρ qU q ) = Σ Np=1 ( m•pq − mqp
•
) (8)
at point q of the cell. According to [10], ΔG for fully ∂t
rough regime can be defined as:
– we can get mass transfer between phases ( m •pq )
1
ΔG = ln(1 + CH S+ ) (5) during phase change processes (evaporation and
κ condensation),
where: H S+ – nondimentional roughness height, and – m •pq ≡ m•p → q (from p to q).
C – roughness constant. Dispersed-continuous phase interactions can take
The range of roughness constant is 0–1 and ac- any value of volume-fraction between 0 and 1. While,
cording to [26], who worked on impacts of roughness continuous-continuous phase interactions are re-
on the numerical centrifugal pumps using Ansys Fluent. stricted to a volume fraction of either 0 or 1 (except in
However, in this work, the value of C is taken as 0.5. the interface region).
Moreover, the nondimensional roughness height is A volume-fraction equation can be characterized by:
defined as: – A transport equation is solved for the volume-
fraction for each additional phase.
ρu * H r – N-1 volume fraction equations are solved:
H S+ = . (6)
μ
∂rq
+ ∇ ⋅ (rqU q ) = 0 (9)
In the Ansys Fluent, the whole roughness regime ∂t
is subdivided into the three regimes, and the formulas
proposed by Cebeci and Bradshaw based on Niku- – We can deduce the volume fraction of the primary
radse’s data [10] are adopted to compute ΔG for each phase, as the sum of volume-fractions must be
regime. equal to 1.
For the transitional regime (2.25 < H S+ ≤ 90) adopt- Σ qN=1rq = 1 . (10)
ed in this paper:
– This approach is common to Eulerian, mixture and
1⎡ H + − 2.25 ⎤ VOF models.
ΔG = ln ⎢ S + CH S+ ⎥ Interface capturing is adopted by the VOF and
κ ⎣ 87.75 ⎦ (7)
Eulerian models (if desired), and therefore will not be
sin{0.4258(ln H S+ − 0.811)}. discussed in this section.
 Discrete phase model of blood flow in a roughness microchannel simulating the formation of pseudointima 135

Eulerian model equations can be described as: AP Volume of spheres Surface area of spheres
= ∗ ,
– A separate momentum and energy equation are V Volume of cell Volume of spheres
solved for each phase.
(16)
– The phases share a common pressure filed (so p has
no sub-script). AP ⎛ π d p2 ⎞
= rp ⎜ ⎟, (17)
∂ (rq ρ qU q ) V ⎜ π d 3p / 6 ⎟
+ ∇ ⋅ (rq ρ qU qU q ) ⎝ ⎠
∂t (11)
• • AP 6rp
= − rq ∇p + ∇ ⋅ τ q + Σ p=1 ( D pq + m pqU pq − mqpU qp ).
N
= . (18)
V dp
– We get additional momentum transfer between
phase p and phase q. If the spheres have different sizes, an additional
transport equation for the interfacial area per unit vol-
– Momentum is transferred by mass ( m •pq ) and by
ume is solved.
other forces (Dpq). Summarizing, Eulerian multi-phase models solve
Interphase drag/momentum transfer can be briefly mass, momentum and energy equations for each phase
discussed as follows: individually. Inter-phase momentum and heat transfer
– As a separate momentum equation is solved for dictates how the phases interact with each other. Inter-
each phase, we get a separate velocity filed for phase drag is predominant momentum transfer
each phase at the cell centroid. mechanism.
– The velocities are different due to a drag and other
2.2.1.2. FVM and mixture model assumptions
inter-phase forces.
– The inter-phase momentum transfer from p → q to FVM (finite volume method) models and mixture
is denoted Dpq. models were developed in Ansys Fluent 2021 R1 soft-
Interphase drag can be described as follows: ware to compare the effect of applying different bound-
– The primary phase moves at velocity Uq. ary conditions with, without roughness and cubes, as
– The secondary phase moves at velocity Up. well as to investigate their influence on blood velocity
The drag force acting on the primary phase is (re- and shear stress.
1 The simulations were run on a standard PC (Proc-
member ρU 2 AC D ) : essor Intel® Core™ i7-10700 CPU, 2.9 GHz, 16 GB
2
RAM). The simulations lasted for 1.87 s (1500 itera-
1 tions) with time step of 0.00125 s. The FVM (finite
D pq = ρ q C D A(U p − U q ) | U p − U q | . (12)
2 volume method) mesh of a flow channel is composed
– The drag force per unit volume (V): of 488 078 cells and 954 329 nodes. The average ele-
ment size is 0.18 mm. The cubes generated in bottom
1 A layer of FVM model of flow channel had dimensions
D pq = ρ qC D P (U p − U q ) | U p − U q | , (13)
2 V of 150 μm and distance among them was 200 μm. The
AP mesh quality parameters (element quality – 0.56, as-
where: CD – drag coefficient, – interfacial area pect ratio – 9.64, skewness – 0.19, orthogonal quality
V
– 0.81) were very good and there were 6 layers of
per unit volume, ρ – density.
prism elements near walls of the FVM model. The
Interphase drag for dispersed-continuous phase in-
side view of FVM mesh of flow channel with indi-
teractions can be described using the most popular drag
cated bottom layer and cubes are shown in Fig. 1. The
model, which is the Schiller–Naumann drag model:
FVM model uses non-Newtonian blood power law [29]
⎧ 24 and the following coefficients of blood are assumed:
⎪ (1 + 0.15 Re 0.687 ) Re < 1000
C D = ⎨ Re , (14) density 1059 kg/m3 [12], specific heat 3617 J/kgK [25],
⎪⎩0.44 Re > 1000 thermal conductivity 0.52 W/mK [52]. The inlet ve-
locity of blood (FVM) and plasma (mixture model) is
ρq | U p − U q | d p set to 0.04 m/s [9].
Re = . (15)
μq The pressure-based solver and absolute velocity
formulation were selected for computational purpose.
The interfacial area per unit cell volume for dis- The pressure-velocity coupling scheme of solution
⎛A ⎞ was selected for computation with spatial discretiza-
persed spheres ⎜ P ⎟ is:
⎝V ⎠ tion using gradient (least squares cell based) and tran-
136 M. KOPERNIK et al.

sient formulation. The SST k-ω turbulence model solving the momentum, continuity, and energy equa-
[21], [53] was applied in all simulations. tions for the mixture, the volume fraction equations
for the secondary phases, and algebraic expressions
for the relative velocities. Parameters applied in com-
putations for two-phase blood model are presented in
Table 1 [1], [4], [18], [23], [24], [45], [48]. The mix-
ture model uses Schiller–Naumann drag coefficient
(compare Section 2.2.1.1).

2.2.2. DPM models with roughness

and cubes

The section presents detailed solutions for DPM

model, which describe the collisions model and La-
grangian Particle Tracking, as well as the assumed
values of parameters.
2.2.2.1. Theoretical background of DPM model
Development of numerical model of collisions was
preceded by experimental analysis of particle-wall colli-
sion process in a channel flow using particle tracking
velocimetry [49]. The effect of wall on the collision
process was analysed. Special attention was paid to the
influence of wall roughness. For example it was found
Fig. 1. Side view of FVM mesh of flow channel with indicated:
a) inlet and bottom layer, and b) cubes that wall roughness considerably alters the rebound
behaviour of the particles and causes a re-dispersion of
The bottom wall of the FVM and mixture models the particles. The experimental data were applied to
has a sand-grain roughness described by a standard develop an update of a wall collision model [30].
roughness model with parameters: roughness height Numerical models of particle-fluid flows can be
25 μm and 40 μm, and roughness constant 0.5. The categorized into three methods, according to the treat-
roughness model applied in the present paper has pa- ment of the particulate and the fluid phases [54]: Eule-
rameters of roughness based on experimental obser- rian-Eulerian methods, Lagrangian-Lagrangian methods
vation (see Section 2.1.2.) and they are correlated with and Eulerian-Lagrangian methods. In Eulerian-Eulerian
element size according to the roughness model as- methods, both the fluid and particulate phases are
sumptions shown in the present paper. defined as interpenetrating continuous phases. A dis-
crete phase model is used in the Lagrangian frame of
reference to track particles’ motion whereas the Eulerian
Table 1. Parameters used in the two-phase blood model formulation is simultaneously used for the continuous
with references
phase. Different forces (viscous drag, lift force, buoy-
parameters red blood cells blood plasma ancy etc.) acting on the Lagrangian particles along
density [kg/m3] 1 090 [48] 1 040 [48] their trajectories are taken into account as well as
specific heat [J/g°C] 0.87 [45] 3.93 [4] a stochastic behavior of the surrounding turbulent
thermal conductivity [W/mK] 2.36 [48] 9.93 [48]
flow.
In the Ansys Fluent software the particles are mod-
viscosity [kg/ms] 0.0075 [24] 0.0015 [23]
elled as spheres with density and a particle size distri-
molecular weight [kg/kmol] 0.0004 [18] 0.018 [1]
bution based on the experimental distributions. The
discrete phase is modelled using the Lagrangian ap-
In Ansys Fluent, three different Euler–Euler mul- proach with Fluent’s DPM. The Lagrangian tracking
tiphase models are available. In the present study the method (compare Section 2.2.2.2.) is used to solve the
mixture one has been selected to model the two individual trajectories of the theoretical particles by
phases of blood: plasma and RBCs. As in the Eulerian equating their inertia with external forces.
model, the phases are treated as interpenetrating con- Two-way coupling, the exchange of momentum
tinua. The mixture model [5] can simulate phases by between the particles and fluid, is accounted for. The
 Discrete phase model of blood flow in a roughness microchannel simulating the formation of pseudointima 137

physics includes: particle–wall collisions, particle– • The fictitious particle is discarded.

particle collisions and structure dependent drag. The • The trajectory calculation of the current real parti-
spherical drag model of Morsi and Alexander [39] is cle is continued until it is completed.
used to model the particles. The effects of particle • For all trajectories the averaging procedure is car-
clustering and channel diameter on the drag of an ried out to obtain local statistical moments in each
isolated particle is modelled using a pressure balance control volume.
model.
2.2.2.2. DPM models assumptions
Particle-particle collisions are modelled using the
Nanbu–Babovsky collision model [35]. This is a sto- DPM models were developed to compare the ef-
chastic collision model where the probability of a colli- fect of applying and not applying roughness as a bound-
sion between particles is based on the volume of ary condition, and finally the DPM models with cubes
particles within a cell and their relative velocities, as were implemented to present distributions of RBCs
follows: concentration, velocity and RBCs residence time in
the flow channel.
N ⎧π ⎫
Prcol = Σ j =j1 ⎨ (d p , i + d p , j ) | u p , j − u p , i | n p , j ⎬Δtcol , A DPM model of backfilling a morphologically
⎩ 4 ⎭ complex surface by RBCs is developed to investigate
(19) the behavior of the RBCs from a Lagrangian view and
a discrete perspective. The blood behavior in Lagran-
where: d – diameter, u – velocity, p – discrete phase, gian view is examined on the basis of a particle
i – index i, j – index j, Prcol – collision probability,
tracking of a RBC of blood plasma flow; Whereas
N – number of particles. blood plasma behavior is considered in Eulerian
The collision partner is determined at random from view based on the assumption of a finite volume
the other particles in the cell as:
element in the fluid flow path. In modeling of RBC
j = ⎣ΦN ⎦ + 1 , (20) – primary blood plasma interaction a two-way ap-
proach is considered, in which RBC fluid flow and
where ⎣...⎦ is the integer part of the argument and Φ is carrier blood plasma flow interact with each other
a random number between [0, 1]. The model requires simultaneously.
a limit on the collision time step, Lagrangian Particle Tracking (LPT) [36] in Ansys
Fluent 2021 R2 software is applied to show the
2 effect of backfilling the regularly morphologically
Δtcol , max = , (21)
max(vij N p ) complex surface of sample by RBCs. The numerical
procedure for carrying out Lagrangian Particle Track-
where: this limiting collision timestep is less than the ing for red blood cells in the Ansys Fluent is as
particle time step, the collision loop is repeated until following:
the sum of the collision time steps is greater. Thus, the – calculation of Reynolds number using RBC veloc-
model resolves the expected number of collisions. The ity from the previous time step / initial velocity,
post collision velocities are calculated from the im- – calculation of drag coefficient from Reynolds
pulse equations. number using the Schiller–Neumann drag model,
The general form of collision model algorithm [20] – solution of the force balance to calculate new RBC
compromises the following items: velocity,
• Call to the collision routine by the Lagrangian – update of the RBC position using the new velocity
solver supplying the necessary variables. and time step,
• Calculation of the instantaneous velocity of the – repetition of the procedure until the RBC times out
virtual collision partner. / terminates.
• Determination of the collision probability and The DPM simulations were run using computing
decision whether or not a collision occurs. unit DELL T7910 equipped with two processors (In-
– If a collision takes place tel® Xeon®CPU E5-2620 v3 @ 2.40GHz) in total of
• the position of the virtual collision partner is de- 12 cores, 128 GB DDR4 RAM and NVIDIA Quadro
termined and K5200 graphics card (8 GB GDDR5 (Hynix)).
• the binary collision is calculated deterministi- The FVM meshes, turbulence model, inlet velocity
cally; and boundary conditions are the same as for models
– if there is no collision the velocities remain un- introduced in Section 2.2.1.2.
changed.
138 M. KOPERNIK et al.

The diameter of RBCs is assumed from 5 μm to from the ground. The friction coefficient is set to 0.2
10 μm using statistical diameter distribution and the [17].
average diameter is 7.5 μm [11].
The value of relaxation time informs how the par-
ticles respond to changes in the flow [36]. Red blood 3. Results
cell relaxation time is equal:
ρ RBC d RBC
2
t RBC = , (22) The computed results shown in this section are
18μ
independent on size of time step, number of itera-
where: ρRBC – red blood cell density, dRBC – average tions and density or quality of computational mesh.
diameter of red blood cells, μ – viscosity of blood The results are considered as distributions of veloc-
plasma. The value of RBC maximal relaxation time is ity, shear stress for FVM, mixture and DPM mod-
4.037 μs. els, as well as distributions of particle residence
The rotation of RBCs is enabled. The density of time, concentration and diameter for DPM models.
plasma is 1040 kg/m3, specific heat is 3930 J/kgK, The results are analysed in selected cross-sections
thermal conductivity is 9.93 W/mK and viscosity is of flow channel and presented in Fig. 2 (bottom
0.0015 kg/ms [4], [48]. The density of RBCs is equal to plane of flow channel and cross-sections 1–3: be-
1090 kg/m3 and specific heat is equal to 870 J/kgK [45], ginning, middle and end of the flow channel).
[48]. Viscosity of red blood cells is set to 0.0075 kg/ms The aim of the analysis of the results is to com-
and their surface tension is set to 5·10-6 N/m [2], pare the influence of the boundary condition (rough-
[24]. ness, cubes) on the flow parameters (velocity, shear
The particles and plasma are injected in normal stress) in all models, as well as determining the in-
direction to the inlet. The total flow rate of RBCs is fluence of the boundary condition (roughness, cubes)
1·10–10 kg/s, RBCs time step is equal to 1.0·10–5 and on the parameters important for the backfilling proc-
the maximum number of iterations of DPM model ess (particle residence time, concentration) in the
is set to 100 000. The plasma velocity is set to 0.04 DPM models.
m/s [9]. The boundary condition of RBCs phase is The computed flow parameters (velocity and wall
assumed as escape, however RBCs can be reflected shear stress) for FVM, mixture and DPM models are

Fig. 2. Planes of model of flow channel considered in analysis

 Discrete phase model of blood flow in a roughness microchannel simulating the formation of pseudointima 139

shown in Figs. 3–5. The backfilling process parame- remain almost constant (several dozen pascals), the
ters (RBCs concentration, residence time) and flow highest shear stress values are computed in the flow
parameters (velocity and wall shear stress) for DPM channel for the model without roughness. The values
models are presented in Figs. 6–8. of shear stresses obtained in the flow channel with the
use of single-phase models are comparable to the val-
(a) ues presented in the literature, e.g., for flows through
similar microfluidic systems and through small ves-
sels with stenosis and (a dozen to over a hundred Pas-
cals) [32], [34].
From the point of view of blood contact devices, it
is therefore important to better understand the relation-
ship between shear stress, exposure time and blood
damage. An example of blood damage is not only he-
molysis or activation of platelets, but also thrombosis
and embolism as well as the destruction of von Wille-
brand factor (vWf). According to the literature, he-
molysis occurs at shear stress level (SSL) >150 Pa,
activation of thrombocytes >50 Pa and degradation of
von Willebrand factor >9 Pa [16].
(b) In the mixture model (Fig. 4a), the use of rough-
ness does not affect the flow velocity in the channel,
the analyzed models show a similar level of flow ve-
locity, the differences in values of flow parameters
among different parts of the flow channel are also
poorly visible. The use of cubes in the mixture model
increases the flow velocity along the bottom wall of
the flow channel compared to the other mixture mod-
els. In all mixture models, except for the mixture
model with cubes, the flow velocity along the bottom
wall of the channel is significantly lower than in the
other models (single-phase and DPM).
For the mixture model (Figure 4b), the shear
stresses remain at the level of a few Pa and are by an
Fig. 3. Flow results of FVM model: (a) velocity, order of magnitude lower than the shear stresses com-
(b) wall shear stress for different boundary conditions puted for the single-phase model. In the mixture model,
the shear stress level is the lowest on the bottom chan-
nel wall for the model without roughness, while the
4. Discussion highest values are for the model with cubes on the
bottom channel wall and in the middle of the flow
channel. In mixture models with roughness, the level
In the single-phase model, applying roughness de- of shear stresses is the same for a particular model in
creases the flow velocity along the bottom wall of the different sections of the flow channel. Two-phase
channel (Fig. 3a). The greatest reduction in flow ve- models show lower values of computed shear stresses
locity is visible for the model with cubes along the [29] than single-phase models, which also has oc-
bottom wall of the channel and the highest velocity curred in the present paper. Two-phase models are
values in the flow channel are computed for the model less prone to disturbance, which is visible in small
with cubes. differences between the calculated values of shear
In the single-phase model, the application of stresses and flow velocity for different parts of the
roughness slightly reduces the shear stresses along the channel for two-phase models with different boundary
bottom wall of the flow channel (Fig. 3b), however, conditions, except for the model with cubes. For ex-
the reduction of stresses is the greatest for the model ample, low shear stress values comparable to com-
with cubes. In all single-phase models and in all ana- puted for two-phase models in the present paper were
lyzed cross-sections of the channel, the shear stresses obtained in the experiment presented in [7] for micro-
140 M. KOPERNIK et al.

channels with blood and in multi-phase blood models stenoses they are about a dozen or so pascals [46].
applied for stenosed arteries in [6]. Thus, the values obtained in this paper for shear
stresses in the DPM model are comparable to the re-
(a) sults obtained by other authors for similar flow sys-
tems.

(a)

(b)

(b)

Fig. 4. Flow results of mixture model: (a) velocity,

(b) wall shear stress for different boundary conditions

Velocity analysis for DPM models (Fig. 5a) shows Fig. 5. Flow results of DPM model: (a) velocity,
that the differences in velocity values are significant (b) wall shear stress for different boundary conditions
for the DPM models compared to the single-phase and
mixture models. The highest flow velocities are ob- A comparative analysis of single-phase, mixture
served for the DPM model without roughness, while models and particle models shows that the influence
the lowest values, similar to the values for single- of particles on the flow parameters is significant and
phase and mixture models, occur for the DPM model the selection of the model with particles is appropriate
with cubes. in the context of micro-level phenomena, which one
For DPM models (Fig. 5b), the lowest values of or two-phase models cannot capture.
shear stresses occur at the bottom of the channel and The increase in roughness (Fig. 6) results in an in-
then increase but remain constant for the particular crease in red blood cell concentration, while the great-
model. The lowest shear stresses occur for the model est increase is observed for the DPM model with
with cubes and are comparable to the shear stress cubes.
level of the mixture model (a few Pa). Roughness In the DPM models, red blood cells are repre-
causes a decrease in shear stresses in the DPM mod- sented by spheres that lose their ability to deform and
els, the higher the roughness, the lower the shear they tend to increase the viscosity [33]. However, at
stress. In the DPM models in literature, the values of concentrations bigger than 50%, the displacement
shear stresses for microfluidic systems may reach decreases. At HCT >50%, although the crowded envi-
several hundred pascals [53], and for vessels with ronment leads to an increase of the cell deformation, it
 Discrete phase model of blood flow in a roughness microchannel simulating the formation of pseudointima 141

also limits the magnitude of the RBC radial dispersion. In the developed DPM model, the hematocrit is
However, in a healthy human body, HCT typically high, the concentration of blood cells is high, so the
ranges between 40.7–50.3% for men and 36.1–44.3% viscosity also increases. This also has to do with the
for women. In the suspension of spheres, at 50% con- reduction of shear stress (increased viscosity) and the
centration, the viscosity increases exponentially making formation of rolls [40].
them difficult to flow. Due to the fact that blood density Spheres have axial migration, however the spheres
above 50% RBC does not occur in natural conditions for not always migrate toward the centre. The spheres near
healthy patients, such a case was not considered. the wall move towards the centre whereas the ones in
the centre move towards the wall. At the end they reach
an equilibrium position. This effect is known as tubular
pinch effect [33].
The selected results of DPM model of backfilling
a morphologically complex surface by RBCs are shown
in Fig. 7 as distributions of RBCs diameter, residence

(a)

Fig. 6. RBCs concentration for different boundary conditions

(a)

(b)

(b)

(c)

(c)

Fig. 8. Results of DPM model of backfilling a morphologically

Fig. 7. Results of DPM model of backfilling complex surface by RBCs in selected cross-sections:
a morphologically complex surface by RBCs: (a) vector distribution of velocity in the bottom plane, (b) vector
(a) distribution of RBCs diameter, (b) distribution distribution of shear stress in the bottom plane, (c) distribution
of RBCs residence time, (c) distribution of RBCs velocity of RBCs concentration in the bottom plane
142 M. KOPERNIK et al.

time and velocity after 0.21 s of simulation when total the bottom plane of flow channel, especially for the
number of RBCs is equal to 4.2·106. The results pre- DPM model with cubes.
sented in Fig. 7 are generated according to the proce- The arrangement of cubes more than one hundred
dure called Lagrangian Particle Tracking. The selected micrometers high on the bottom plane of the flow chan-
results of DPM model of backfilling a morphologically nel organizes the flow of RBCs. The cubes strongly in-
complex surface by RBCs are also presented in Fig. 8 fluence the shear stress, velocity and concentration of
as distributions of velocity, shear stress and concen- particles. The values of shear stress and velocity are
tration of particles on the bottom of middle part of significantly decreased, whereas the values of particle
flow channel (exactly above cubes). concentration significantly increase in comparison with
The effect of “pouring” particles is clearly captured DPM models with roughness.
in the developed DPM model of backfilling a mor- Based on the simulation results, it is not possible
phologically complex surface by RBCs (Fig. 7). The to clearly determine which model most fully reflects
flow of particles in a channel with cubes runs orderly, the real conditions for the mechanism of pseudointima
because the particle movement trajectories are en- layer formation. However the most advanced model
forced by the system of cubes. The simulation was – DPM model with cubes is characterized by stress
carried out until the number of RBCs in the model values of several Pascals, suggesting the correctness
increased, which in the simulation means leaving the of the analysis. The next stage of the research is to
RBCs out of the inlet part of flow channel, going conduct laboratory validation tests with the use of full
through the middle part of flow channel and reaching human blood to verify the results obtained in this
the outlet part of flow channel. The distributions of study.
velocity and shear stress (Fig. 8) indicate that ar-
rangement of cubes on the lower surface of the chan-
nel organizes the flow of blood particles. Acknowledgements
The analysis of the parameters influencing the
backfilling process in the developed DPM model shows The research was partially financed by the Polish National Sci-
that the change in the shape of the bottom surface of ence Centre under Project PRELUDIUM 16 2018/31/N/ST8/01085.
Part of the work was executed under fundamental research no.
the flow channel has a very large impact on these 16.16.110.663 financed by AGH University of Science and Tech-
parameters, disproportionately greater than the change nology.
of the boundary condition.

References
5. Conclusions
[1] ADAMCZUK K., WOLAŃSKI W., KASPERA W., Analiza przepływu
krwi w tętnicach mózgowych, Aktualne Problemy Biome-
The introduction of roughness as a boundary condi- chaniki, 2016, 11, 9–14.
[2] ALIMOHAMADI H., SMITH A., NOWAK R., FOWLER V.,
tion in the FVM model of flow channel led to a small RANGAMANI P., Non-uniform distribution of myosin-mediated
reduction in relatively high values of shear stresses forces governs red blood cell membrane curvature through
and led to a slower flow, even in the case of small tension modulation, PLOS Comput. Biol., 2020, 16, 1–26.
surface roughness of the bottom plane of flow channel [3] BERIS A.N., HORNER J.S., JARIWALA S., ARMSTRONG M.J.,
(several dozen of micrometers). However, the influ- WAGNER N.J., Recent advances in blood rheology: A review,
Soft Condensed Matter, 2021, 1–22.
ence of cubes on the bottom layer in the FVM model
[4] BLAKE A.S.T., PETLEY G.W., DEAKIN C.D., Effects of changes
on flow parameters is not very significant. in packed cell volume on the specific heat capacity of blood:
The influence of roughness as a boundary condi- implications for studies measuring heat exchange in extracor-
tion while computing mixture model of blood is less poreal circuits, Brit. J. Anaesth., 2000, 84, 28–32.
visible, only introduction of cubes increased signifi- [5] BOWEN R.M., Incompressible porous media models by use of
cantly the computed values of blood velocities and the theory of mixtures, Int. J. Eng. Sci., 1980, 18, 1129–1148.
[6] BURADI A., MAHALINGAM A., Effect of stenosis severity on
shear stress. However, the mixture model leads to the wall shear stress based hemodynamic descriptors using
lowest values of shear stress of all the applied models. multiphase mixture theory, J. Appl. Fluid Mech., 2018, 11,
The introduction of the same roughness (several 1497–1509.
dozen of micrometers) as a boundary condition in the [7] CAI S., LI H., ZHENG F., KONG F., DAO M., KARNIADAKIS G.E.,
DPM model also led to a reduction in shear stresses SURESH S., Artificial intelligence velocimetry and micro-
aneurysm-on-a-chip for three-dimensional analysis of
and slowing down the flow, which additionally re-
blood flow in physiology and disease, PNAS, 2021, 118,
sulted in an increase of concentration of the RBCs on e2100697118.
 Discrete phase model of blood flow in a roughness microchannel simulating the formation of pseudointima 143

[8] CAI Q., LIAO W., XUE F., WANG X., ZHOU W., LI Y., ZENG W., fluidic chip technology (microViscTM) for clinical use and
Selection of different endothelialization modes and different determination of a new reference range, Ann. Clin. Biochem.,
seed cells for tissue-engineered vascular graft, Bioact. Mater, 2020, 57, 249–252.
2021, 6, 2557–2568. [24] JUNG J., LYCZKOWSKI R.W., PANCHAL CH.B., HASSANEIN A.,
[9] CAPACCIO A., CASERTA S., GUIDO S., RUSCIANO G., SASSO A., Multiphase hemodynamic simulation of pulsatile flow in a coro-
Dissolution of a surfactant-water lamellar phase investigated nary artery, J. Biomech., 2006, 39, 2064–2073.
by combining time-lapse polarized light microscopy and [25] KARAKI W., RAHUL, LOPEZ C.A., BORCA-TASCIUC D.A., DE S.,
confocal Raman spectroscopy, J. Colloid. Interf. Sci., 2020, A continuum thermomechanical model of in vivo electrosur-
561, 136–146. gical heating of hydrated soft biological tissues, Int. J. Heat
[10] CEBECI T., BRADSHAW P., Momentum transfer in boundary Mass Tran., 2018, 127, 961–974.
layers, Hemisphere Publishing Corp., New York, 1977. [26] KHANJANPOUR M.H., JAVADI A.A., Experimental and CFD
[11] DIEZ-SILVA M., DAO M., HAN J., LIM C.-T., SURESH S., analysis of impact of surface roughness on hydrodynamic
Shape and biomechanical characteristics of human red blood performance of a darrieus hydro (DH) turbine, Energies,
cells in health and disease, MRS Bull, 2010, 35, 382–388. 2020, 13, 928.
[12] EL-ARAGI G.M., Effect of electrohydraulic discharge on vis- [27] KIM J., ANTAKI J.F., MASSOUDI M., Computational study of
cosity of human blood, Phys. Res. Int., 2013, ID 203708. blood flow in microchannels, J. Comput. Appl. Math., 2016,
[13] ERDEMIR A., GUESS T.M., HALLORAN J., TADEPALLI S.C., 292, 174–187.
MORRISON T.M., Considerations for reporting finite element [28] KOPERNIK M., TOKARCZYK P., Development of multi-phase
analysis studies in biomechanics, J. Biomech., 2012, 45, 625–633. models of blood flow for medium-sized vessels with stenosis,
[14] FENG R., XENOS M., GIRDHAR M., KANG W., DAVENPORT J.W., Acta Bioeng. Biomech., 2019, 21, 63–70.
DENG Y., BLUESTEIN D., Viscous flow simulation in a steno- [29] KURTYKA P., WALKE W., KACZMAREK M., Fluid flow analy-
sis model using discrete particle dynamics: a comparison sis using finite element method, determining the effects of the
between DPD and CFD, Biomech. Model Mechanobiol., implantable mechanical heart valves on aortic blood flow,
2012, 11, 119–129. Adv. Intell. Syst., 2016, 472, 255–266.
[15] FLORMANN D.A.D., Physical characterization of red blood [30] LAIN S., ERNST M., SOMMERFELD M., Study of colliding
cell aggregation, Biological Physics, Universität des Saar- particle-pair velocity correlation in homogeneous isotropic
landes, 2017. turbulence, Appl. Sci., 2020, 10, 9095.
[16] FRASER K.H., ZHANG T., TASKIN M.E., GRIFFITH B.P., WU Z.J., [31] LEE B.-K., XUE S., NAM J., LIM H., SHIN S., Determination of
A quantitative comparison of mechanical blood damage pa- the blood viscosity and yield stress with a pressure-scanning
rameters in rotary ventricular assist devices: shear stress, capillary hemorheometer using constitutive models, Korea
exposure time and hemolysis index, J. Biomech. Eng., 2012, – Aust. Rheol. J., 2011, 23, 1–6.
134, 0810021. [32] LEE J., KIM I.G., OH Y.M., PARK C.-H., KIM C.S., Prelimi-
[17] HAYASE T., SHIRAI A., SUGIYAMA H., HAMAYA T., Meas- nary study for measurement of shear stress and hemocom-
urement of frictional characteristics of red blood cells mov- patibility using commercialized lab on a chip, J. Nanoscie.
ing on a plate in plasma due to inclined centrifugal force, Nanotechno., 2018, 18, 1123–1126.
Nippon Kikai Gakkai Ronbunshu, B Hen/Transactions of the [33] LIMA R., ISHIKAWA T., IMAI Y., YAMAGUCHI T., Blood flow
Japan Society of Mechanical Engineers, Part B, 2002, 68, behavior in microchannels: past, current and future trends,
3386–3391. Single and two-phase flows on chemical and biomedical en-
[18] HIMBERT S., ALSOP R.J., ROSE M., HERTZ L., DHALIWAL A., gineering, Dias R., Martins A.A., Lima R., Mata T.M. (Eds.),
MORAN-MIRABAL J.M., VERSCHOOR C.P., BOWDISH D.M.E., Bentham Science, 2012, 513–547.
KAESTNER L., WAGNER C., RHEINSTÄDTER M.C., The Mo- [34] LIU H., LAN L., ABRIGO J., IP H.L., SOO Y., ZHENG D., WONG K.S.,
lecular Structure of Human Red Blood Cell Membranes from WANG D., SHI L., LEUNG T.W., LENG X., Comparison of newto-
Highly Oriented, Solid Supported Multi-Lamellar Membranes, nian and non-newtonian fluid models in blood flow simulation in
Sci. Rep., 2017, 7, 39661. patients with intracranial arterial stenosis, Front. Physiol., 2021,
[19] HORNER J.S., ARMSTRONG M.J., WAGNER N.J., BERIS A.N., https://doi.org/10.3389/fphys.2021.718540
Investigation of blood rheology under steady and unidirec- [35] LOVE A.I.J., GIDDINGS D., POWER H., Gas-particle flow
tional large amplitude oscillatory shear, J. Rheol., 2018, 62, modeling: beyond the dilute limit, Procedia Engineer., 2015,
577–591. 102, 1426–1435.
[20] HUSSMANN B., PFITZNER M., ESCH T., FRANK T., A stochastic [36] MAHDAVIMANESH M., NOGHREHABADI A.R., BEHBAHANINEJAD M.,
particle-particle collision model for dense gas-particle AHMADI G., DEHGHANIAN M., Lagrangian particle tracking:
flows implemented in the Lagrangian solver of ANSYS CFX model development, Life Sci. J., 2013, 10, 34–41.
and its validation, 6th International Conference on Multi- [37] MITCHELL D., HONNERY D., SORIA J., Particle relaxation and
phase Flow, ICMF 2007, Leipzig, Germany, July 9–13, its influence on the particle image velocimetry cross-correlation
2007, 1–16. function, Exp. Fluids, 2011, 51, 933.
[21] JAMES M.E., PAPAVASSILIOU D.V., O’REAR E.A., Use of com- [38] MENCONI M.J., POCKWINSE S., OWEN T.A., DASSE K.A.,
putational fluid dynamics to analyze blood flow, hemolysis STEIN G.S., LIAN J.B., Properties of blood-contacting sur-
and sublethal damage to red blood cells in a bileaflet artifi- faces of clinically implanted cardiac assist devices: gene
cial heart valve, Fluids, 2019, 4, 1–19. expression, matrix composition, and ultrastructural char-
[22] JARIWALA S., HORNER J.S., WAGNER N.J., BERIS A.N., Appli- acterization of cellular linings, J. Cell. Biochem., 1995, 57,
cation of population balance-based thixotropic model to hu- 557–573.
man blood, J. Non-Newton Fluid, 2020, 281, 104294. [39] MENTER F., Two-equation eddy-viscosity turbulence mod-
[23] JØRGENSEN L.H., MØLLER V.S., REVSHOLM J., Plasma vis- els for engineering applications, AIAA J, 2002, 40, 254–
cosity: Evaluation of a new measuring method using micro- 266.
144 M. KOPERNIK et al.

[40] MORSI S.A., ALEXANDER A.J., An investigation of particle [48] ROSENTRATER K.A., FLORES R.A., Physical and rheological
trajectories in two-phase flow systems, J. Fluid Mech., 1972, properties of slaughterhouse swine blood and blood compo-
55, 193–208. nents, T ASAE, 1997, 40, 683–689.
[41] MURALI C., NITHIARASU P., Red blood cell (RBC) aggre- [49] SOUSA P.C., PINHO F.T., ALVES M.A., OLIVEIRA M.S.N.,
gation and its influence on non-Newtonian nature of A review of hemorheology: Measuring techniques and recent
blood in microvasculature, J. Model Mech. Mat., 2017, 1, advances, Korea – Aust. Rheol. J., 2016, 28, 1–22.
20160157. [50] SOMMERFELD M., HUBER N., Experimental analysis and
[42] NANDA S.P., BASU MALLIK B., A non-newtonian two-phase modelling of particle-wall collisions, Int. J. Multiphas. Flow,
fluid model for blood flow through arteries under stenotic con- 1999, 25, 1457–1489.
dition, Int. J. Pharm. Bio. Sci., 2012, 2, 237–247. [51] THYAGARAJAN B., KUMAR M.P., SIKACHI R.R., AGRAWAL A.,
[43] PATTANAPOL W., WAKES S.J., HILTON M.J., DICKINSON K.J., Endocarditis in left ventricular assist device, Intractable Rare
Modeling of surface roughness for flow over a complex vege- Dis. Res., 2016, 5, 177–184.
tated surface, Int. J. Math. Phys. Eng. Sci., 2008, 2, 18–26. [52] VAIDYA N., BARAGONA M., LAVEZZO V., MAESSEN R.,
[44] PICART C., PIAU J.-M., GALLIARD H., CARPENTIER P., Human VEROY K., Simulation study of the cooling effect of blood
blood shear yield stress and its hematocrit dependence, vessels and blood coagulation in hepatic radiofrequency
J. Rheol., 1998, 42, 1–12. ablation, Int. J. Hyperther., 2021, 38, 95–104.
[45] PONDER E., The specific heat and the heat of compression of [53] YU Z., TAN J., WANG S., Enhanced discrete phase model for
human red cells, sickled red cells, and paracrystalline rat red multiphase flow simulation of blood flow with high shear
cells, J. Gen. Physiol., 1955, 38, 575–580. stress, Sci. Prog., 2021, 104, 1–21.
[46] PRASHANTHA B., ANISH S., Discrete-phase modelling of an [54] ZAHARI N.M., ZAWAWI M.H., SIDEK L.M., MOHAMAD D.,
asymmetric stenosis artery under different Womersley num- ITAM Z., RAMLI M.Z., SYAMSIR A., ABAS A., RASHID M.,
bers, Ara. J. Sci. Eng., 2019, 44, 1001–1015. Introduction of discrete phase model (DPM) in fluid flow:
[47] RAFFI S., OZ M.C., SELDOMRIDGE J.A., FERRIS B., ASCH A.S., a review, AIP Conf. Proc., 2030, 2018, 020234-1-6.
NACHMEN R.L., SHAPIRO F., ROSE E.A., LEVIN H.R., Char- [55] ZILLA P., DEUTSCH M., BEZUIDENHOUT D., DAVIES N.H.,
acterisation of hematopoetic cells arising on the textured PENNEL T., Progressive Reinvention or Destination Lost? Half
surface on left ventriculat assist devices, Ann. Thorac. Surg., a Century of Cardiovascular Tissue Engineering, Front Cardio-
1995, 60, 1627–1632. vasc. Med., 2020, 7, 1–159.