Powder Technology 249 (2013) 463–474

Contents lists available at ScienceDirect

Powder Technology
journal homepage: www.elsevier.com/locate/powtec

New correlations for heat and fluid flow past ellipsoidal and cubic
particles at different angles of attack
Andreas Richter a,⁎, Petr A. Nikrityuk a,b
a
CIC Virtuhcon, Department for Energy Process Engineering and Chemical Engineering, Technische Universität Bergakademie Freiberg, Fuchsmühlenweg 9, 09596 Freiberg, Germany
b
Department of Chemical and Materials Engineering, University of Alberta, 9107-116 Str., Edmonton, AB T6G 2V4, Canada

a r t i c l e i n f o a b s t r a c t

Article history: This work considers heat and fluid flow past ellipsoidal and cubic particles at different angles of attack. Although
Received 11 March 2013 numerous works have investigated drag forces for spherical and non-spherical particles, there are very few works
Received in revised form 12 June 2013 about Nusselt number relations for such particles, especially in cases where the particles are not oriented in a
Accepted 30 August 2013
streamwise direction. Motivated by this fact, three-dimensional numerical simulations of heat and fluid flow
Available online 12 September 2013
past non-spherical particles at different angles of attack were performed. In particular, a laminar-steady flow
Keywords:
past ellipsoidal and cubic particles was considered and the results compared against those for a spherical particle.
Heat transfer In order to take into account the angle of attack in semi-empirical models, new correlations for the drag, lift and
Drag coefficient torque coefficients and for the Nusselt number were developed and the accuracy of the closures is discussed in
Nusselt number comparison with published models.
Cube © 2013 Elsevier B.V. All rights reserved.
Ellipsoid
Angle of attack

1. Introduction In the past decades, numerous works have been devoted to under-
standing the physics of flows past particles of different shapes, and to es-
Particle-laden flows play an important role in many technical and timating correlations for the drag force as a function of the Reynolds
natural processes. Due to the high computational effort it is not feasible number (Re) and the particle shape. Hottovy and Sylvester [2] measured
to directly resolve the heat and fluid flow past each individual particle. drag coefficients of irregularly shaped particles for Reynolds numbers be-
Therefore, Euler–Lagrange (e.g. Discrete Particle Model, DPM) or tween 7 and 3000. Haider and Levenspiel [3] used archival measurement
Euler–Euler based approaches have become well-established tools for data to develop correlations for spheres, isometric and non-isometric
understanding and optimizing particle-laden flows. However, these solids and disks. The Reynolds number ranges from 0.1 to 106. Ganser
models rely on closure relations for the heat transfer and the forces [4] introduced new relations based on two factors: Stokes' and Newton's
and torques acting on the particle. One of the most popular assumptions shape factors. Tran-Cong et al. [5] measured the steady-state free-fall con-
in such models is that all particles are perfectly spherical. It is obviously ditions of irregular-shaped particles, which are modeled as agglomerates
that in practice the particles feature a more complex geometry, which of smaller spheres, and developed correlation formulas to describe the
results in different drag coefficients and Nusselt numbers. As an exam- drag coefficient for groups of particles. A new, simple regression formula
ple, Fig. 1 illustrates the shape of char particles in pulverized-coal reac- for regular particles that comprises only the sphericity of the particle is
tors at one stage of conversion [1]. If the particles are not oriented in a given by Yow [6]. Based on literature data, Gabitto and Tsouris [7] devel-
streamwise direction, drag force and heat transfer change and addition- oped relations for cylindrical particles. Furthermore, Loth [8] studied nu-
al forces and torques are present. All of these effects influence the trajec- merical and experimental data from the literature, stating in summary
tory of each individual particle, which modifies the particle distribution that ellipsoids can mainly be described by their aspect ratio, whereas the
and hence the whole technological process in which the particles are in- drag coefficient for regular particles depends on the surface ratio and for
volved. The advantage of Euler–Lagrange approaches in contrast to irregular particles on the minimum, maximum and averaged area ratio.
Euler–Euler approaches is the possibility to take the angle of attack Most of the publications have focused on particles oriented in the di-
into account more easily if single particles are tracked. From this point rection of the stream. As an extension, Hölzer and Sommerfeld [9]
of view, the results of this work are closer to Euler–Lagrange models. performed a comprehensive analysis of experimental data published
in the literature and of their own numerical results. The authors devel-
oped a regression formula for cd based on the standard, and on the
⁎ Corresponding author. Tel.: +49 3731 394801; fax: +49 3731 394555.
crosswise and lengthwise sphericity, which allows the particle orienta-
E-mail addresses: a.richter@vtc.tu-freiberg.de (A. Richter), nikrityu@ualberta.ca tion to be taken into account. The resultant correlation has a mean rela-
(P.A. Nikrityuk). tive deviation between measurements and the regression model of only

0032-5910/$ – see front matter © 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.powtec.2013.08.044
464 A. Richter, P.A. Nikrityuk / Powder Technology 249 (2013) 463–474

Nomenclature

a,b constants
2
A reference area, π4d
cd drag coefficient, 0:5ρjFu j A
x
→ 2
∞
cl lift coefficient, 0:5ρjuF j A
y
→ 2
∞
cm torque coefficient (torsional moment), 0:5ρjuM j A d
z
→ 2
∞
d diameter of the volume-equivalent sphere
f function
!
F force vector, (Fx, Fy, Fz)T
h specific enthalpy
!
M torque (torsional moment) vector, (Mx, My, Mz)T
!
n normal vector, (nx, ny, nz)T
Nu Nusselt number, αd λ
p static pressure
Fig. 1. Example of different coal–particle shapes. Spherical particles are marked as blue,
Pr Prandtl number
! cubic particles as green and ellipsoidal particles as red.
r direction vector Picture was taken from Brix et al. [1].
!
Re Reynolds number, ρj uμ jd
Ncv number of control volumes
S boundary surface
T temperature and plates, while Sparrow et al. [13] performed a review of closure rela-
!
u velocity vector, (ux, uy, uz)T tions for Nu applied to circular and non-circular cylinders. The latter two
pffiffiffiffi
u* friction velocity, τρw
works are based on data sets taken from the literature. Saha [14] carried
x, y, z coordinates out numerical investigations into fluid flow and heat transfer around a
yw wall distance cube. Richter and Nikrityuk [15] focused on ellipsoidal and cuboidal
y+
∗
non-dimensional wall distance, u ρy
μ
w
particles in laminar flow regimes that are oriented in a streamwise
direction, and provided relations for the drag coefficient and the Nusselt
number. As a first step towards understanding the relation between
Greek symbols particle orientation and heat transfer, Wittig et al. [16] studied the influ-
α surface-averaged heat transfer coefficient ence of particle orientation on heat and fluid flow past a cube placed in a
ε relative error homogeneous flow.
λ thermal conductivity Based on this analysis it can be said, in summary, that closure rela-
τw wall shear stress tions for lift and drag forces together with heat transfer and torques
ρ density are not well described in the literature applied to non-spherical parti-
μ dynamic viscosity cles with arbitrary orientation, but such correlations are essentially for
ψ angle of attack the adequate modeling of particulate flows.
e
ψ transformed angle of attack The intention of this work is to overcome this gap. In this work, first,
φ rotation around one body axis a comprehensive data set was provided for heat and fluid flow past
ellipsoidal and cubic particles at different angles of attack by utilizing
the commercial Finite Volume solver ANSYS FLUENT™ V 14.0 [17] to
Subcripts solve the three-dimensional flow problem together with heat transfer.
max maximum As illustrated in Fig. 1, the mentioned particle shapes can be found e.g.
n non-dimensionalized in pulverized-coal reactors. Second, based on the numerical results,
part particle new correlations were developed for drag, lift and torque coefficients
∞ free stream and for the Nusselt number. The development of such correlations
using regression analyses is consistent with previously published ap-
proaches (see e.g. [3,4,9]).
Contrary to various approaches in the literature, the main goal in this
work is not to cover a large range of Reynolds numbers, which causes
14.1% for all the data considered. This is a significant improvement com-
pared to the widely-used correlation by Haider and Levenspiel [3]
(383% deviation) or by Ganser [4] (384% deviation). At very low particle
Reynolds numbers, the particle motion is dominated by drag forces. At
higher Reynolds numbers, side forces and torques can influence the par-
ticle trajectories [10]. To take this into account, Hölzer and Sommerfeld
[11] also determined flow characteristics as a function of the angle of at-
tack at Reynolds numbers between 0.3 and 240, and examined rotating
spheres and spheroids as well as spheres and spheroids placed in a
shear flow. Unfortunately the authors provided no closure relations
for lift forces and torques.
The situation is slightly different for arbitrary particles involving
heat transfer phenomena. The particle shape and orientation in the
gas flow significantly affect the heat transfer. This fact is often not
taken into account, thus the number of references related to the heat
transfer of non-spherical particles is relatively low. For example, Fig. 2. Computational domain and boundary conditions (not true to scale). The origin of
Whitacker [12] developed correlations for circular cylinders, spheres coordinates is located in the center of the particle.
 A. Richter, P.A. Nikrityuk / Powder Technology 249 (2013) 463–474 465

Table 2
Grid convergence. Flow past a cubic particle at Reynolds number 100. ψ = 30 deg.

grid Ncv cd εcd/% Nu εNu/%

1 1.60 · 105 1.3843e + 00 2.37 6.1786 2.16

2 6.50 · 105 1.3947e + 00 1.64 6.1164 1.13
3 1.30 · 106 1.4101e + 00 0.55 6.0777 0.49
4 1.97 · 106 1.4199e + 00 0.14 6.0537 0.10
5 3.56 · 106 1.4179e + 00 – 6.0478 –

not influence the velocity field, so the temperature (or energy) behaves
like a passive scalar.

3. Numerical scheme, discretization and software validation

Fig. 3. Computational mesh for flow past an ellipsoidal particle. Angle of attack ψ =
45 deg. The sliding interface is marked in red. 3.1. Numerical scheme

For this study, the commercial Finite Volume solver ANSYS FLUENT™
V 14.0 [17] was utilized to solve the three-dimensional Navier–Stokes
relatively large errors in different flow regimes, but to provide more equations with heat transfer (Eq. (1)). The Pressure-Based Coupled Al-
precise results for small particle Reynolds numbers (≤200) as can be gorithm was used, so the momentum and continuity equation are
found in pulverized-coal reactors and many other technical processes. solved together, whereas the energy equation was solved separately.
In this Reynolds number range the flow remains steady-state, but may Since the equations are non-linear, they are linearized first using an im-
exhibit an unsymmetrical flow field [18,15]. The transport properties plicit method. A point-implicit linear equation solver (Gauss–Seidel)
correspond to air at 300 K. is then applied in conjunction with an algebraic multigrid method
(AMG) to solve the resultant system of linear equations. For the spatial
2. Model setup and governing equations discretization of convective terms the QUICK scheme [19] was applied.
This scheme is based on a weighted average of a second-order-upwind
The present work studies heat and fluid flow past non-spherical par- scheme and a central interpolation scheme. The implementation in
ticles that are placed in a cross-flow of air at 300 K. The particle surface ANSYS FLUENT uses a variable, solution-dependent value to blend
temperature is assumed to be constant at 400 K, so the temperature dif- between second-order-upwind scheme and central interpolations. De-
ference between the particle and the fluid is 100 K. At the inflow bound- tails are given in [17,19].
ary the flow and the temperature are defined as uniform and constant In the numerical setup, inflow boundaries were defined as a ve-
over time, so the inflow conditions directly determine the overall flow locity inlet, where the velocity vector and the temperature are kept
field. In a long distance normal to the flow direction the influence of constant. Outflow boundaries were defined as a zero-diffusion-flux con-
the flow field on the particle is assumed to be negligible, so all surround- dition for all flow variables, together with an overall mass balance cor-
ing boundaries excluding the in- and outflow are defined as symmetry rection, and symmetry boundaries by applying a zero flux condition
boundaries. for each conserved variable.
Several assumptions can be made: The fluid behaves as a nearly in-
compressible medium since the velocity magnitudes are much lower 3.2. Discretization
than the local speed of sound, so the density is assumed to be constant.
All remaining material properties of the fluid are defined as constant It is a well-known fact that for heat and mass flow problems the
and correspond to air at 300 K, with Pr = 0.74. Additionally, viscous quality of the numerical results depends strongly on the quality of the
heating and buoyancy effects are assumed to be negligible. Since only numerical grid. For example, the use of unstructured grids may lead to
steady-state regimes are considered, the time derivatives for the conser- increased numerical diffusion [20] and very often does not guarantee
vation properties can be neglected. the proper accuracy and convergence of a flow solution. Thus, although
Based on the assumptions made, the governing equations read as unstructured meshes are easy to use, they should be applied very

!
 ∇  u ¼ 0 1 μ !
! ! 5
u  ∇ u ¼ − ∇p þ Δ u
ρ ρ ð1Þ Haider and Levenspiel (1989)
  λ
! Saha (2004)
u  ∇ h ¼ ΔT:
ρ 4 present study

!
Here u denotes the velocity vector, ρ the density, p the pressure, μ 3
cd

the viscosity, h the enthalpy, λ the thermal conductivity and T the tem-
perature. As a result of the constant density, the temperature field does 2

Table 1 1
Numerical grid for different particle shapes.

Particle shape Ncv ymax+ 0 50 100 150 200

Spherical 1.84 · 106
0.07
Re
Ellipsoidal 1.88 · 106 0.10
Fig. 4. Comparison of the drag coefficient of a cube against experimental [3] and numerical
Cubic 1.30 · 106 0.14
[24] results.
466 A. Richter, P.A. Nikrityuk / Powder Technology 249 (2013) 463–474

Table 3 10
Regression formulae for the drag coefficient depending on the particle shape.
9
Particle shape Correlation
8
cd;sphere ¼ 17:9 pffiffiffiffi
Re þ Re þ 0:162
Spherical 7:47

cd;ellipsoid ¼ 18:9 pffiffiffiffi 7

Re þ Re þ 0:0316
Ellipsoidal 6:25

Nu
cd;cube ¼ 20:4 pffiffiffiffi
Re þ Re þ 0:216
Cubic 8:19
6
5
4
carefully. In order to ensure that the mesh was of high quality with suf- 3
ficiently fine resolution near the wall, body-fitted, hexahedral meshes 0 25 50 75 100 125 150 175 200
were applied. Since the preparation of such numerical grids is difficult, Re
if a series of different angles of attack are considered, sliding surfaces
were used. To achieve this, the whole domain was divided into two Fig. 6. Comparison of numerical data (symbols) and new regression curves (solid lines,
parts, which are separated by a spherical surface with a radius of 6 d. see Table 4) for Nusselt numbers for different particle shapes. The particles are oriented
The inner domain and the outer domain feature a structured grid, but parallel to the direction of flow. Numerical data are taken from [15].

the meshes on the inner and outer sides of the interface surface are
not identical. This means that on one surface of a finite volume that is
located at the inner side of the interface, one or more surfaces from fi- the number of control volumes Ncv are constant for each dedicated par-
nite volumes outside the interface are attached. At the interface the nu- ticle shape.
merical flux is calculated from the weighted average of all fluxes coming A study of the grid independence for spherical particles can be found
from the neighboring volumes. This technique allows us to rotate the in the previous work [15]. To guarantee grid-independent results even
inner domain and hence to change the angle of attack without needing for non-spherical particles with arbitrary ψ, a grid study was performed
to fully create a new mesh. In [21] the authors discretized an irregularly- for a cubic particle at Reynolds number 100. The particle is oriented in
shaped particle with two structured, hexahedral meshes, one mesh the flow field with ψ = 30 deg. Four different meshes were applied.
with and the other one without a sliding interface. A comparison of Each mesh features double the number of control volumes on a particle
the calculated drag coefficients and Nusselt numbers demonstrated circumference, and double the number of control volumes in directions
the validity of this approach. normal to the particle surface, compared to the next coarser mesh. The
Following the study in [15], the computational domain extends 16 d resultant drag coefficients and Nusselt numbers are listed in Table 2. The
in the crosswise direction (y, z direction) in order to avoid blockage data for the finest grid were chosen as reference values. From Table 2 it
effects caused by an inadequate domain size. In the flow direction (x di- can be seen that the maximum error rate is approximately 2% for the
rection), the domain size is 32 d, with d as the diameter of the volume- coarsest mesh. In this work Grid 3 was selected for the following studies,
equivalent sphere. The center of the sphere is located at the origin of the since it provides an error rate that is significantly below one percent,
coordinate system. Fig. 2 illustrates the computational domain. and has only a moderate number of finite volumes.
To ensure that velocity and temperature gradients are of a sufficient-
ly fine resolution, the non-dimensional wall distance y+ features an av-
pffiffiffiffiffiffiffiffiffiffiffi 3.3. Software validation
eraged value of only 0.06 (y+ = u⁎ρyw / μ, with u∗ ¼ τw =ρ the
friction velocity and τw the wall shear stress). An example of such a The CFD software package used in this work is closed-source, so the
grid is depicted in Fig. 3. All geometries in this study are discretized in underlying implementation of the numerical scheme can only be exam-
the same way; in particular, the wall resolution and domain size are ined to a limited extent. For that reason it is necessary to perform an ex-
identical. In dependence on the particle shape, the number of finite vol- tensive validation for the flow and heat transfer [22]. In [15] the authors
ume cells differs and is in the range of 1.3 to 1.9 million cells, see Table 1. validated the drag coefficient, Nusselt number, Strouhal number and
It should be noted that for the simulations of different angles of attack recirculation length for spherical particles against published data. The
the meshes in the inner and outer domains are not modified, and only authors also studied the influence of the computational domain on
the inner domain is rotated. For that reason the particle resolution and heat and flow characteristics. Wittig et al. applied an in-house Finite
Volume solver together with a Continuous-Forcing Immersed Boundary
approach to study heat transfer and fluid flow [16]. The authors com-
pared their results for cd and Nu against data taken from [15] for a spher-
5 ical particle and demonstrated an excellent agreement between the two
different flow solvers. On the other hand, the general validity of the slid-
ing interface approach was carefully discussed by the authors in [21]. To
4
extend the validation process to include non-spherical particles, the
drag coefficient of a cubic particle was estimated numerically as func-
3 tion of the Reynolds number. The main difference compared to the
cd

validation for a spherical particle is that the cubic particle features geo-
2 metrical singularities, i.e. sharp edges, which can affect the heat and

1
Table 4
Regression formulae for the Nusselt number.
1 25 50 75 100 125 150 175 200
Re Particle shape Correlation
1 pffiffiffiffiffiffi 1 2
Spherical Nusphere ¼ 0:568Pr 3 pRe ffi 0:0104Pr 1Re 2 þ 1:70
ffiffiffiffiffiþ
3 3
Fig. 5. Drag coefficient for different particle shapes based on new regression formulas (see 1
Ellipsoidal Nuellipsoid ¼ 0:640Pr 3 Re −0:0390Pr 3 Re3 þ 1:59
Table 3). The particles are oriented in the direction of flow. Numerical data are taken from 1 pffiffiffiffiffiffi 1 2
Cubic Nucube ¼ 0:781Pr 3 Re−0:115Pr 3 Re3 þ 1:14
[15]. Symbols: numerical results, solid lines: regression curves.
 A. Richter, P.A. Nikrityuk / Powder Technology 249 (2013) 463–474 467

Details about the flow physics of such particles without an angle of

attack are given in [15].
For the tracking in Euler–Lagrange approaches the coefficients for
heat and fluid flow of non-spherical particles are typically estimated
with the diameter d and the project area A of the volume-equivalent
sphere. The present work follows this approach: all coefficients were re-
lated to the geometrical parameters for the volume equivalent sphere.
In accordance with previous works, all particles feature an identical vol-
ume (1 mm3). With this definition the reference values d and A (used
for drag, lift and torque coefficients and for the Nusselt number) are
equal for all particle shapes that were considered here. The Reynolds
number, too, is based on the diameter of the volume-equivalent sphere
and reads
!
ρj u jd
Re ¼ : ð2Þ
Fig. 7. Definition of angles ψ and φ. μ

The drag coefficient is defined as

fluid flow around the particle. Reference values were given e.g. by Fx
cd ¼ ð3Þ
Haider and Levenspiel [3]. The authors provided a closure correlation 0:5ρju→
∞
j2 A
for the drag coefficient of a cubic particle, which precisely fits the exper- !
imental data published by Pettyjohn and Christiansen [23]. In addition, with F x ¼ nx  F as the force in x direction and the force vector
numerical results for a cube provided by Saha [24] were also taken into
!    
! ! T !
account. As illustrated in Fig. 4, the agreement between the numerical F ¼ ∮S −p n þ μ ∇ u þ ∇!
u  n dS: ð4Þ
results is perfect. However, it should be noted that for larger values of
Re the numerical and experimental results slightly differ. The mean de- For the drag coefficient of a sphere placed in a laminar, homoge-
viation between the experimental and numerical results is 4.7%. neous flow several correlations between cd and Re are given in the liter-
ature. Based on a generic form for the drag coefficient [6]

4. Results a b
cd ¼ þ pffiffiffiffiffiffi þ const ð5Þ
Re Re
4.1. Heat and fluid flow of particles oriented in the flow direction
individual correlations were developed for spherical, ellipsoidal and
In [15] the authors studied in detail the heat and fluid flow past ellip- cubic particles in order to quantify the correlation between cd and Re.
soidal and cubic particles that are oriented in the flow direction, and de- To do so, statistical software was applied (for more details see [27]).
veloped closure correlations for cd and Nu in dependence on different The resultant correlations are summarized in Table 3. The maximum
shape factors. In this section individual correlations are developed for deviation between the correlations and numerical data is lower than
the drag coefficient and Nusselt number of ellipsoidal and cubic parti- 0.6%, and the average deviation is 0.2%. Fig. 5 illustrates the validity of
cles without an angle of attack; in the next chapters results are added the given correlations by comparing them with numerical data.
for different angles of attack. The ellipsoidal particle features an aspect From Table 3 it can be noted that the coefficients a and b are similar
ratio of 2. Qualitatively, the particle shapes follow SEM measurements for all particle shapes. The constant value c on the other hand is much
of different coal types [1,25,26]. It is worth noting that these assump- higher for the cube in comparison with the other particles. In depen-
tions do not hold true e.g. for biomass applications with significantly dence on the Reynolds number the drag coefficient of an ellipsoidal par-
larger aspect ratios. In such cases the given correlations have to be ex- ticle is reduced by 9.5–27% (Re10–200), and for a cubic particle it is
tended to incorporate the aspect ratio as an independent variable. increased by about 12–14%, compared to a sphere.

(a) ψ = 0 deg (b) ψ = 0 deg (c) ψ = 45 deg (d) ψ = 90 deg

(e) ψ = 0 deg (f) ψ = 15 deg (g) ψ = 30 deg (h) ψ = 45 deg

Fig. 8. Streamlines for (a) spherical and (b–h) non-spherical particles at Re = 100.
468 A. Richter, P.A. Nikrityuk / Powder Technology 249 (2013) 463–474

(a) ψ = 0 deg (b) ψ = 0 deg (c) ψ = 45 deg (d) ψ = 90 deg

(e) ψ = 0 deg (f) ψ = 15 deg (g) ψ = 30 deg (h) ψ = 45 deg

Fig. 9. Non-dimensional temperature field Tn along the xy plane for Re = 100.

As shown in Fig. 6, the situation changes if the heat transfer is con- sufficient approximation for the influence of the Prandtl number
sidered. The spherical particle features the maximum heat transfer, [12,28]. The final correlations are listed in Table 4.
while for the ellipsoid the Nusselt number is decreased by 3–8%, and In contrast to the findings for the drag coefficient, the constant
for the cube by 13–19%. It should be noted that the Nusselt number is part of the Nusselt number is comparable for all particle shapes. The
an area-averaged value, so coefficient a also remains comparable for all particle shapes, but the
coefficient b is negative for the non-spherical particles. The given corre-
αd 1 −λ∂T lations reflect the numerical data base very well, with a maximum devi-
Nu ¼ ; α ¼ ∮S ∂n
dS ð6Þ
λ S T part −T ∞ ation of 0.6%, and an average deviation of 0.2%. This is also demonstrated
in Fig. 6 by a comparison of data taken from numerical simulations and
with α as the heat transfer coefficient at the particle surface. The total from the given regression curves.
heat transfer for both the cube and the ellipsoid is slightly above that
for a volume-equivalent sphere, but the total surface area is higher for
non-spherical particles. For that reason the area-averaged Nusselt num- 4.2. Flow characteristics of particles at different angles of attack
ber is below that for the sphere.
Starting with correlations given by Ranz and Marshall [28] and This section studies the influence of the particle orientation on heat
Whitaker [12] and fluid flow past non-spherical particles. In the literature, the shape of
pffiffiffiffiffiffi the particle is often prescribed via the sphericity, but this value provides
1 1 2
Nu ¼ aPr 3
Re þ bPr Re þ const
3 3
ð7Þ no information about the orientation of the particle. In the literature
(see e.g. [9]), additional characteristics such as the crosswise and
new correlations were developed for the Nusselt number as a function lengthwise sphericity were introduced in order to overcome this prob-
of Re and Pr. The influence of the Prandtl number was not investigated lem. When such values are used, additional computational work is
1
explicitly, but from the literature it is known that the term Pr is a 3
required to track particles, since the projected surfaces need to be

2.0 Re = 50
Re = 100
1.8

1.6
cd

1.4

1.2

1.0

0.8
-90 -60 -30 0 0 15 30 45
ψ / deg ψ / deg
(a) ellipsoid (b) cube
Fig. 10. Drag coefficient for cubic and ellipsoidal particles in dependence on ψ and Re.
 A. Richter, P.A. Nikrityuk / Powder Technology 249 (2013) 463–474 469

Table 5 recirculation domain is comparable to that for a rigid sphere, and rises
e
Regression terms for cd as a function of ψ. slightly, if the angle of attack is increased.
 
Particle shape e
f cd ψ
  4.3. Influence of particle orientation on particle forces and heat transfer
Ellipsoidal 2:21 2
sin e
ψ
Re0:303
 
Cubic 0:059 2
sin 2ψ e
Re0:292 4.3.1. Drag forces
The relation between the drag coefficient and angle of attack is
highlighted in Fig. 10 for different particle shapes and different Reynolds
calculated each time the drag coefficient and other coefficients are eval- numbers. For all particle shapes and Reynolds numbers the drag in-
uated. As an alternative the particle orientation relative to the flow field creases along with the angle of attack. The influence of ψ on cd is at
can be taken into account directly. It should be noted that the number of most 80% for the ellipsoid at Re = 200. In this process the main effect
degrees of freedom that describe the particle's orientation in the flow for the ellipsoid is that the pressure forces rise due to the increased
field depends on the particle shape. In this work the angle ψ denotes projected surface area, while the friction forces remain nearly constant.
the angle of attack, which is the angle between the local flow vector In dependence on Re and ψ the flow can separate from the particle's sur-
(x direction) at the barycenter of the particle and one axis of the particle. face. At ψ = 90 deg, for example, this is true for ellipsoidal particles and
For the ellipsoid it is the axis of rotational symmetry. Due to symmetry for Reynolds numbers above 10. With an increasing angle of attack the
properties the orientation of an ellipsoid in the flow field is fully defined separation region moves from the backward to the forward stagnation
by the angle of attack. For a cube, a second angle φ becomes relevant; region. In the present data set no rapid change in pressure of friction
this angle defines the rotation around the particle axis. Both angles ψ forces could be identified depending on the occurrence or movement
and φ are linearly independent of each other. Alternatively it is possible of the separation region. For the ellipsoid a strong relation between cd
to describe the orientation as the rotation around each coordinate axis and Re was found up to a Reynolds number of 200. It is a well-known
in a global coordinate system. In that case not two but three angles are fact that for spherical particles the drag force becomes independent of
necessary to describe the orientation, and the angles are not linearly inde- Re for Reynolds numbers between 800 and 3.8 · 105 [29], thus for the
pendent of each other, which complicates the estimation of semi- ellipsoidal particle a similar Reynolds number range can be expected,
empirical correlations. The definition of angles ψ and φ is illustrated in at which drag forces are independent of Re.
Fig. 7. In this work only the influence of ψ was considered. Deeper insights If a cubic particle is rotated, on the other hand, the projected surface
into heat and fluid flow past a cube at different φ are provided in [16]. area is changed only slightly, which causes a minor increase in the pres-
To illustrate the effect of a particle's orientation, for Re = 100 Figs. 8 sure forces. However, the friction forces rise significantly, which is con-
and 9 show the heat and fluid flow past spherical and non-spherical par- trary to the ellipsoidal particle. Flow separation could be identified even
ticles at different angles of attack. The temperature field is illustrated via at Re = 10. As the Reynolds number increases the separation region
the non-dimensionalized temperature moves from the bottom to the top of the particle. An increase in ψ, on
the other hand, leads to a larger recirculation domain, but has only
 
minor influence on the position of the separation region. In agreement
T n ¼ ðT−T ∞ Þ= T part −T ∞ ð8Þ
with results for ellipsoidal particles the drag was not seen to be inde-
pendent of the Reynolds number due to flow separation. This fact indi-
with T∞ the free-stream temperature and Tpart the surface temperature cates a possible range at which the drag does not depend on Re up to
of the particle. From Figs. 8 and 9 it becomes obvious that the orienta- Reynolds numbers significantly larger than 200.
tion of the particle in the flow field significantly changes the flow char- The maximum impact of ψ on the drag coefficient for the cube is
acteristics and hence the heat transfer. It is a well-known fact that at approximately 32%. A statistical analysis of the flow characteristics
Re = 100 a symmetric recirculation domain exists in the wake of the showed a distinct relation between the drag coefficient and the squared
sphere. Similar wake structures are present for an ellipsoid at ψ = sine function of ψ for the ellipsoid, and between cd and the squared sine
0 deg and ψ = 90 deg, but for 0 deg the recirculation domain is much function of 2ψ for the cube. The qualitative dependence on sin2(ψ) or
smaller, and for 90 deg it is significantly enhanced. At 45 deg the flow sin2(2ψ), respectively, is independent of the Reynolds number, as
field exhibits an asymmetry normal to the flow field. This asymmetry shown in Fig. 10.
results in an asymmetric pressure distribution along the particle and Next the influence of the angle of attack on the drag coefficient and
hence to additional lift forces and torques that act on the particle. By heat transfer was quantified by deriving closure relations, which can
contrast the flow field past a cube is not axis-symmetric, and the flow also be used in sub-models for CFD calculations. Since the symmetry
fields at ψ = 0 deg and ψ = 90 deg are identical. The length of the properties are different for ellipsoids and cubes, the most promising

5 5
ψ=0 deg ψ=0 deg
ψ=45 deg ψ=22.5 deg
4 ψ=90 deg 4 ψ=45 deg

3 3
cd
cd

2 2

1 1

0 50 100 150 200 0 50 100 150 200

Re Re
(a) ellipsoid (b) cube
Fig. 11. Comparison between predicted and numerically calculated drag coefficients at different Re and ψ. (a) ellipsoid, (b) cube. Lines denote semi-empirical correlations, symbols stand
for numerical data.
470 A. Richter, P.A. Nikrityuk / Powder Technology 249 (2013) 463–474

7.5

7.0

6.5

Nu 6.0

5.5

5.0 Re=50
Re=100

4.5
-90 -60 -30 0 0 15 30 45
ψ / deg ψ / deg
(a) ellipsoid (b) cube
Fig. 12. Nusselt number as a function of ψ. The Reynolds number equals 50 and 100. (a) ellipsoid, (b) cube.

approach to taking into account the angle of attack is to develop individ- 4.3.2. Heat transfer
ual models for each particle shape. Due to the symmetry of the particles For the Nusselt number the influence of the angle of attack is lower
it is possible to transform the angle of attack in the form compared to that for cd, which is illustrated in Fig. 12. At a Reynolds
number of 200 the influence ranges from 7% for the cube to 12% for
e
ψ ellipsoid ¼ jððψ þ 90Þ mod180Þ−180j ð9Þ the ellipsoid. Similar to the findings for the drag coefficient, the Nusselt
number for the ellipsoid varies as a function of sin2ψ. By contrast, the
e
ψ cube ¼ jððψ þ 45Þ mod 90Þ−90j: ð10Þ Nusselt number relation for the cube is not straightforward. Observing
Fig. 12b it becomes evident that the Nusselt number drops slightly for
e in closure relations for cd, the generic form
In order to include ψ   for ψ ≤ 15 deg, and rises for larger angles of attack. Although ψ modifies
the drag coefficient (5) was extended by adding a new term f cd ψ e , so the Nusselt number significantly at Re = 100, the influence of ψ on Nu
is only minor for Re = 50.
a b  
cd ¼ e :
þ pffiffiffiffiffiffi þ const þ f cd ψ ð11Þ The relation for the Nusselt number (7) was extended in a similar
Re Re way compared to cd, so

This approach offers the advantage that existing correlations for pffiffiffiffiffiffi  
Nu ¼ aPr
1
3
1 2
e :
Re þ bPr Re þ const þ f Nu ψ
3 3
ð12Þ
non-spherical particles oriented in  flow direction can be easily ex-
 the
tended by the additional term f cd ψ e . Alternatively, the angle of attack
can be included in the drag correlations by modifying the coefficients a
The resulting term fNu is listed in Table 6 and can also be added to the
and b. However, preliminary studies have showed that in that case the
correlations given in Table 4. The mean error of the semi-empirical corre-
quality of the approximation was always lower compared to the chosen
lation is between 0.2 (ellipsoid) and 1.4% (cube). The relation for the heat
approach.
transfer is not the same as the findings for cd. In particular, the heat trans-
Table 5 provides the additional term fcd as a function of the    
e . These terms can be directly added to fer of a cubic particle depends on two terms, i.e. onsin 2ψ e andsin2 2ψ e ,
transformed angle of attack ψ
the correlations that are summarized in Table 3. The mean deviation be- but with different signs. The generally good prediction of Nu for non-
tween the numerical results and the regression model for cd is in the spherical particles at different angles is demonstrated in Fig. 13.
range of 0.4 (cube)–0.6% (ellipsoid). From Table 5 it can be seen that
the Reynolds number dependency for cd is comparable for all particle 4.3.3. Lift forces
shapes, but the leading coefficient is much lower for the cubic particle. Due to the symmetry characteristics of their shape, ellipsoidal and
This fact reflects the findings discussed above that the impact of ψ on cubic particles do not experience additional lift forces if they are orient-
cd is more distinct for an ellipsoid than for a cube. In order to demon- ed in a streamwise direction. For ψ N 0 deg, the flow field on the upper
strate the validity of the derived models, Fig. 11 compares the numerical and lower sides of the particle is different. In regions with accelerated
results against those from the derived correlations and demonstrates flow the pressure drops, and rises in the vicinity of stagnation points.
the good prediction of cd for ellipsoidal and cubic particles. The resulting asymmetric pressure distribution induces additional lift
forces and torques. The lift forces for ellipsoidal and cubic particles are
shown in Fig. 14. In this illustration a sinusoidal dependence on 2ψ for
Table 6 the ellipsoid, and on 4ψ for the cube can be identified. Generally the
Orientation-dependent terms for Nusselt number.
lift forces for the ellipsoidal particle are four times larger than those
 
Particle shape e
f Nu ψ
for the cubic particle. At high Reynolds number flows around two-
 
dimensional geometries the lift coefficient primarily depends on the in-
Ellipsoidal 0:0153Pr 3 Re0:807 sin2 ψ
1
e duced circulation. For the Reynolds numbers considered in the present
   
Cubic 1
−0:22Pr Re
3
−0:0825 2
sin 2ψ e þ 0:0456Pr 13 Re0:48 sin 2ψ
e work, the pressure and viscous force fractions of the lift feature the
same order of magnitude. Additionally, the flow fields have distinct
 A. Richter, P.A. Nikrityuk / Powder Technology 249 (2013) 463–474 471

10
8 ψ=0 deg
9 ψ=22.5 deg
7 ψ=45 deg
8
7 6

Nu

Nu
6 5
5 ψ=0 deg 4
ψ=45 deg
4 ψ=90 deg
3
3
0 50 100 150 200 0 50 100 150 200
Re Re
(a) ellipsoid (b) cube
Fig. 13. Nusselt number correlations (lines) and numerical results (symbols) for different particle shapes and Reynolds numbers.

0.2
Re=50
0.1 Re=100

0.0
c1

-0.1

-0.2

-0.3

-0.4
-90 -60 -30 0 0 15 30 45
ψ / deg ψ / deg
(a) ellipsoid (b) cube
Fig. 14. Lift coefficient for (a) ellipsoid, and (b) cube. Reynolds numbers equal to 50 and 100.

0.6
ψ=15 deg 0.1
0.5 ψ=45 deg
ψ=60 deg
0.08
0.4
c1

c1

0.06
0.3
0.04
0.2 ψ=7.5 deg
ψ=22.5 deg
0.02 ψ=30 deg
0.1
0 50 100 150 200 0 50 100 150 200
Re Re
(a) ellipsoid (b) cube
Fig. 15. Lift coefficient for ellipsoidal and cubic particles as a function of Re; (lines) correlation, (symbols) numerical results.

!
three-dimensional distributions. For these reasons it is not feasible to with the lift force F y ¼ ny  F as the force in y direction (see Eq. (4)).
consider the circulation for interpreting the lift forces. Contrary to the drag forces and the heat transfer, the influence of the
The lift coefficient is defined as Reynolds number on cl is not evident, and finding a general rule for
the interaction of Reynolds number dependence and the influence of
ψ is not a straightforward task. For the ellipsoidal particle a non-linear
Fy
cl ¼ ð13Þ dependence on Re can be detected, and for all ψ the lift coefficient
0:5ρju→
∞ j A
2
increases slightly in dependence on Re.
472 A. Richter, P.A. Nikrityuk / Powder Technology 249 (2013) 463–474

Table 7 Table 8
e
Lift coefficient in dependence on ψ. Semi-empirical correlations for cm.

Particle shape cl Particle shape cm

     
Ellipsoidal e − 0:0262
sin 2ψ
0:97 e
cos 2ψ Ellipsoidal e 0:408 þ 0:369
− sin 2ψ
Re0:25 Re0:25 Re Re0:126
Cubic 0:0814 sin 4ψe Cubic 0:00129Re0:697 sin4ψe

The problem becomes more evident if cubic particles are investigat- of pressure and friction forces at the particle wall, which induced a
ed. For example the lift coefficient at ψ = 30 deg (see Fig. 15b) in- torque that acts on the particle. As an example, the gas flow at the
creases in the Reynolds number range between 10 and 50, than front side is accelerated for ellipsoidal particles at ψ N 0 deg. Close to
slightly falls at Re = 100, and finally it remains constant for Re N 100. the tip a pressure rise is incurred on the particle's underside and a pres-
A detailed consideration of streamlines and the pressure distribution sure drop on the upper side. At the opposite end, the flow on the under-
of the flow past a cube at ψ = 30 deg (not shown here) revealed that side is accelerated and provokes an additional pressure drop (see
for Re ≤ 50 the separation region at the rear side of the particle moves Fig. 8c). Due to the unsymmetrical pressure distribution the ellipsoidal
from the bottom to the upper side of the particle. For Re ≥ 50 the particle experiences a torque, which rotates the particle against its
flow regime does not changes significantly. The recirculation domain original orientation. Friction forces additionally amplify this effect. For
is located close to the upper rear corner of the particle, and the center the cubic particle the situation is different. At ψ N 0 deg the pressure
of the recirculation domain moves away from the particle with increas- peak in the forward stagnation region is moved to the leading edge of
ing Re. This effect leads to pressure forces that are independent of the the particle. On the upper and back side the flow field forms a wide
Reynolds number for Re N 50. It explains the discontinuity at Re = recirculation domain, which induces a pressure drop in this region
100 displayed in Fig. 15. For ellipsoidal particles a similar behavior (cf. Fig. 8f–g). The torques that come from the unsymmetrical pressure
could not be identified. The qualitative behavior at ψ = 7.5 deg and distribution partially compensate for the friction forces. For that reason
22.5 deg is comparable, but the range at which cl gains is shifted to larg- the torques acting on the ellipsoidal particle are approximately eight
er Re numbers. At ψ e ¼ 45 deg, the flow field remains symmetric along times greater than those for the cubic particle. In contrast to the ellip-
the xy plane, thus no lift forces are present if the Reynolds numbers soid the torques act in the opposite direction, which means a cubic
are below 150. If the Reynolds number is further increased, the flow particle tends to come back to its original orientation.
!
field loses its symmetric characteristic and induces additional lift forces. The torque coefficient is calculated using the torque Mz ¼ nz M , so
Table 7 lists the lift coefficients for both non-spherical particles esti-
Mz ! h    i
mated by regression analysis. In addition, Fig. 15 shows the lift coeffi- ! ! ! T !
cm ¼ ; M ¼ ∮S r  −p n þ μ ∇ u þ ∇!
u  n dS:
cients for different ψ e and as a function of Re. For the ellipsoidal particle → 2
0:5ρju∞ j A d
the lift coefficient resulting from regression analysis has a mean devia- ð14Þ
tion of 7.3% from the numerical data. Due to the irregularities discussed
above, for a cubic particle it is impossible to identify a general relation !
The vector r defines the vector between the specific wall cell and
between cl, ψ and Re. For that reason the regression curve listed in the center point. Fig. 16 displays the torque coefficient cm for different
Table 7 gives only a rough estimation for the cubic particle, also illustrat- particle shapes as function of ψ. This coefficient denotes the rotation
ed in Fig. 15. It is worth noting that in spite of a discontinuity in the cl–Re about the z-axis (cf. Fig. 7). Since only the influence of the angle of attack
relation no rapid drop in cl (stall) can be identified in Fig. 15. ψ was considered in this work, no additional torques around different
axes occur for regularly-shaped particles such as ellipsoids or cubes.
4.3.4. Torques From Fig. 16 it becomes obvious that cm is a function of sin(2ψ) for ellip-
As discussed above, an unsymmetrical flow field around non- soidal particles and a function of sin(4ψ) for cubic particles. With these
spherical particles at ψ N 0 deg leads to an unsymmetrical distribution relations in mind, semi-empirical relations for cm as a function of ψ were

0.2

0.1
cm

0.0

-0.1

Re=50
-0.2
Re=100

-90 -60 -30 0 0 15 30 45

ψ / deg ψ / deg
(a) ellipsoid (b) cube
Fig. 16. Torque coefficient for different particle shapes as a function of ψ.
 A. Richter, P.A. Nikrityuk / Powder Technology 249 (2013) 463–474 473

-0.05 0.06
ψ=7.5 deg
-0.1 0.05 ψ=22.5 deg
ψ=30 deg
-0.15 0.04
cm

cm
-0.2 0.03

-0.25 0.02
ψ=15 deg
-0.3 ψ=45 deg 0.01
ψ=60 deg
-0.35 0
0 50 100 150 200 0 50 100 150 200
Re Re
(a) ellipsoid (b) cube
Fig. 17. Semi-empirical correlations (lines) and numerical data (symbols) for cm.

derived. Table 8 compares the correlation functions for ellipsoidal and and 80% for the ellipsoid. The influence of ψ on Nu is moderate; here,
cubic particles. The mean deviations of the correlations from the numer- the maximum is between 7 and 12%. Due to the non-symmetric flow
ical results are 1.4% for the ellipsoidal and mean 8.8% for the cubic at different angles of attack, additional lift forces and torques are in-
particle. The numerical results and the correlations are compared in duced, which may have a significant influence on the trajectory of
Fig. 17, which demonstrates the accuracy of the model. non-spherical particles. Closure relations have been derived for the
drag coefficient, the Nusselt number, the lift force and the torque
5. Discussion around the z axis as functions of Re, ψ and Pr (for the Nusselt number
only). The focus of this work was to provide a comprehensive set of
In particle-laden flows, non-spherical particles are not oriented in a closure relations for the heat and fluid flow past non-spherical particles.
stream-wise direction. This can be caused by an initial rotational com- These closure relations can directly be used in particulate flows for
ponent due to the feeding system, by shear stresses on the flow field tracking non-spherical particles, and to model the heat transfer of
or by particle–particle or particle–wall interactions. As discussed such particle shapes.
above, the lift forces can exceed 50% of the drag forces, ignoring the
lift force that results from additional shear stresses (Saffman force). It Acknowledgments
becomes obvious that the lift forces significantly influence the trajectory
of non-spherical particles, and hence they have an impact on the parti- This research has been funded by the Saxon Ministry of Science and
cle distribution in technical systems. At the same time the particles ex- Fine Arts in the framework of Virtuhcon (project number 4-7531.50-02-
perience torques, which can compensate the effect partially. Tracking 0390-09/1).
non-spherical particles necessitates the extension of the equations of
motion for the Lagrangian particles by three equations for the torques References
acting on the particle. It is clear that this approach results in increased
[1] J. Brix, P.A. Jensen, A.D. Jensen, Modeling char conversion under suspension fired
computational work, but that may be necessary if the particle rotation conditions in O2/N2 and O2/CO2 atmospheres, Fuel 90 (2011) 2224–2239.
becomes dominant. In that case, relations for the Magnus force for [2] J.D. Hottovy, N.D. Sylvester, Drag coefficients for irregularly shaped particles, Ind.
non-spherical particles are also required. The development of such Eng. Chem. Process. Des. Dev. 18 (1979) 433–436.
[3] A. Haider, O. Levenspiel, Drag coefficient and terminal velocity of spherical and
models and the investigation of forces that act on non-spherical parti-
nonspherical particles, Powder Technol. 58 (1989) 63–70.
cles embedded in a shear-stress field are far more challenging and still [4] G.H. Ganser, A rational approach to drag prediction of spherical and nonspherical
largely unresolved. particles, Powder Technol. 77 (1993) 143–152.
In this work, only one angle was considered. The influence of the sec- [5] S. Tran-Cong, M. Gay, E.E. Michaelides, Drag coefficients of irregularly shaped parti-
cles, Powder Technol. 139 (2004) 21–32.
ond angle φ, which describes the rotation around the main body axis, [6] H.N. Yow, M.J. Pitt, A.D. Salman, Drag correlations for particles of regular shape, Adv.
can be neglected for an ellipsoid, but not for a cube. Due to the minor Powder Technol. 16 (2005) 363–372.
impact of φ on the heat and fluid flow around a cubic particle the influ- [7] J. Gabitto, C. Tsouris, Drag coefficient and settling velocity for particles of cylindrical
shape, Powder Technol. 183 (2008) 314–322.
ence of φ was not considered in the present work, but it was referred to [8] E. Loth, Drag of non-spherical solid particles of regular and irregular shape, Powder
the work done by Wittig et al. [16], where the flow past a cube was stud- Technol. 182 (2008) 342–353.
ied more in detail. Since ellipsoidal and cubic objects feature different [9] A. Hölzer, M. Sommerfeld, New simple correlation formula for the drag coefficient of
non-spherical particles, Powder Technol. 184 (2008) 361–365.
symmetry properties, it remains questionable whether an extension of [10] L. Rosendahl, Using a multi-parameter particle shape description to predict the mo-
the derived correlations to more general particle shapes would be useful tion of non-spherical particle shapes in swirling flow, Appl. Math. Model. 24 (2000)
or not. For that reason correlations were developed for each individual 11–25.
[11] A. Hölzer, M. Sommerfeld, Lattice Boltzmann simulations to determine drag, lift and
particle. On the other hand, an extension to different aspect ratios torque acting on non-spherical particles, Comput. Fluids 38 (2009) 572–589.
should be possible, but has yet to be implemented. [12] S. Whitaker, Forced convection heat transfer correlations for flow in pipes, past flat
plates, single cylinders, single spheres, and for flow in packed beds and tube bun-
dles, AICHE J. 18 (1972) 361–371.
6. Conclusion [13] E.M. Sparrow, J.P. Abraham, J.C.K. Tong, Archival correlations for average heat trans-
fer coefficients for non-circular and circular cylinders and for spheres in cross-flow,
Int. J. Heat Mass Transfer 47 (2004) 5285–5296.
The study presented is focused on the characterization of heat and
[14] A.K. Saha, Three-dimensional numerical study of flow and heat transfer from a cube
fluid flow past ellipsoidal and cubic particles at different angles of placed in a uniform flow, Int. J. Heat Fluid Flow 26 (2006) 80–94.
attack. To achieve this, three-dimensional calculations of ellipsoidal [15] A. Richter, P.A. Nikrityuk, Drag forces and heat transfer coefficients for spherical, cu-
and cubic particles were carried out at Reynolds numbers between 10 boidal and ellipsoidal particles in cross flow at sub-critical Reynolds numbers, Int. J.
Heat Mass Transfer 55 (2012) 1343–1354.
and 200. The numerical results showed that the angle of attack has a [16] K. Wittig, A. Richter, P.A. Nikrityuk, Numerical study of heat and fluid flow past a cu-
major impact on the drag force, with a maximum of 23% for the cube bical particle at sub-critical Reynolds numbers, Comp. Therm. Sci. J. 4 (2012) 1–14.
474 A. Richter, P.A. Nikrityuk / Powder Technology 249 (2013) 463–474

[17] ANSYS Inc., ANSYS-FLUENT™ V 14.0—commercially available CFD software package Dr.-Ing. Andreas Richter is the head of the research group
based on the Finite Volume method, 2012. (Southpointe, 275 Technology Drive, Interphase Phenomena within the Center for Innovation Com-
Canonsburg, PA 15317, U.S.A., www.ansys.com). petence VIRTUHCON at the TU Bergakademie Freiberg. He
[18] T.A. Johnson, V.C. Patel, Flow past a sphere up to a Reynolds number of 300, J. Fluid studied mechanical engineering at the Dresden University
Mech. 378 (1999) 19–70. of Technology, and finished his PhD in Dresden in the field
[19] B.P. Leonard, A stable and accurate convective modeling procedure based on qua- of aero-acoustical simulations.
dratic upstream interpolation, Comput. Methods Appl. Mech. Eng. 19 (1979) 59–98.
[20] J.H. Ferziger, M. Peric, Computational Methods for Fluid Dynamics, 3rd edition
Springer, Berlin, 2002.
[21] A. Richter, P. Nikrityuk, Heat and fluid flow around a sphere with cylindrical bore,
ASME J. Heat Transfer 134 (2012) 071704.
[22] M. Andrews, Guidelines for use of commercial software and diagnostics in articles
for the journal of fluid engineering, J. Fluid. Eng. 133 (2011) 010201.
[23] E.S. Pettyjohn, E.R. Christiansen, Effect of particle shape on free-settling rates of
isometric particles, Chem. Eng. Prog. 44 (1948) 157–172.
[24] A.K. Saha, Three-dimensional numerical simulations of the transition of flow past a Dr.-Ing. habil. Petr Nikrityuk is an associate professor at the
cube, Phys. Fluids 16 (2004) 1630–1646. University of Alberta (UoA). Before taking his current posi-
[25] M.J.G. Alonso, A.G. Borrego, D. Alvarez, R. Menéndez, Pyrolysis behaviour of tion at UoA, Petr Nikrityuk was the head of the research
pulverised coals at different temperatures, Fuel 78 (1999) 1501–1513. group Interphase Phenomena within the Center for Innova-
[26] G.D. Micco, A. Nasjleti, A.E. Bohé, Pyrolysis behaviour of pulverised coals at different tion Competence VIRTUHCON at the TU Bergakademie
temperatures, Fuel 95 (2012) 537–543. Freiberg. He studied mechanical engineering at the Moscow
[27] StatPoint Technologies Inc., STATGRAPHICS Centurion XVI ™—commercially Aviation Institute—MAI, where he obtained his PhD on the
available software package for statistical analysis, 2011. (Warrenton, Virginia, topic of mathematical modeling of thermal processes. Before
U.S.A., www.statlets.com). taking his current position in Freiberg, Petr Nikrityuk worked
[28] W.E. Ranz, W.R. Marshall, Evaporation from drops, Chem. Eng. Process. 48 (1952) as software developer in the field of computational fluid dy-
173–180. namics and as a post doc fellow in the Institute for Aerospace
[29] D.A. Jones, D.B. Clarke, Simulation of flow past a sphere using the FLUENT code, Engineering at the Dresden University of Technology.
Tech. rep, Austral. Gov., Dep. of Def., Def. Sc. and Techn. Org, 2008.