RESEARCH ARTICLE | JUNE 29 2022

Drag, lift, and torque coefficients for various geometrical

configurations of elliptic cylinder under Stokes to laminar
flow regimes
Geunwoo Oh; Hyunwook Park; Jung-Il Choi 

AIP Advances 12, 065228 (2022)

https://doi.org/10.1063/5.0097916

CrossMark

 
View Export
Online Citation

Articles You May Be Interested In

Evaluating causes of foreign accent in English sentences spoken by native speakers of Italian differing in
age of arrival (AOA) in Canada
J Acoust Soc Am (April 2003)

English vowel production by native Mandarin speakers: Influences of AoA, LoR, education, perception, and
orthography
J Acoust Soc Am (September 2005)

07 October 2023 09:16:46

 AIP Advances ARTICLE scitation.org/journal/adv

Drag, lift, and torque coefficients for various

geometrical configurations of elliptic cylinder
under Stokes to laminar flow regimes
Cite as: AIP Advances 12, 065228 (2022); doi: 10.1063/5.0097916
Submitted: 3 May 2022 • Accepted: 6 June 2022 •
Published Online: 29 June 2022

Geunwoo Oh,1 Hyunwook Park,1 and Jung-Il Choi2,a)

AFFILIATIONS
1
Department of Computational Science and Engineering, Yonsei University, Seoul 03722, Republic of Korea
2
School of Mathematics and Computing (Computational Science and Engineering), Yonsei University, Seoul 03722,
Republic of Korea

a)
Author to whom correspondence should be addressed: jic@yonsei.ac.kr

ABSTRACT
In this study, the drag, lift, and torque coefficients are derived as a function of the axis ratio (AR) and angle of attack (AOA) for elliptic

07 October 2023 09:16:46

cylinders with simple and straightforward correlations in flow regimes ranging from Stokes to laminar flow. An immersed boundary method,
based on an implicit direct forcing scheme to impose a more accurate no-slip condition for viscous flow, is utilized to compute the hydro-
dynamic forces on the elliptic cylinder. Numerical simulations are performed for two-dimensional flow around an elliptic cylinder. The ARs
and AOAs of the elliptic cylinder are within 0.3–1.0 and 0○ –90○ , respectively. A critical Reynolds number (Recrt ) map is obtained, indicat-
ing a transition between steady and unsteady flows as a function of the AR and AOA. Based on Recrt for the circular cylinder, it is observed
that a more prolonged elliptic cylinder delays Recrt to a higher Re value at low AOAs but causes an unsteady transition at a lower Re value
when the AOA exceeds ∼30○ . The correlations are selected as simple functions, such as power laws and trigonometric functions, based on the
hydrodynamic force behaviors according to variations in the AR, AOA, and Re. The prediction accuracy of the proposed correlations assessed
in terms of mean relative errors is ∼1.5%, 8.9%, and 11.2% for drag, lift, and torque, respectively. This comparison demonstrates that the
proposed correlations are suitable for accurately predicting hydrodynamic forces in Stokes to laminar flow regimes, even when using simple
basic forms.
© 2022 Author(s). All article content, except where otherwise noted, is licensed under a Creative Commons Attribution (CC BY) license
(http://creativecommons.org/licenses/by/4.0/). https://doi.org/10.1063/5.0097916

I. INTRODUCTION Subsequently, the particle motion was calculated using the predicted
drag force and Newton’s second law.6 Practical particles, such as air
An accurate particle transport prediction in fluid flow is vital in pollutants, dust, and coal powder, have a variety of non-spherical
many environmental and industrial engineering applications, such irregular shapes.7 Owing to the effect of the particle shape on the
as pollutant and particle matter dispersion in the urban bound- flow patterns around the particles, the hydrodynamic forces acting
ary layer, dust filtering with a porous bed, combustion in coal on the particle, such as drag, lift, and torque, change as a function
power plants, and in fields dealing with pulverized particle trans- of the particle shape, orientation, and Re, leading to variations in
port. Early numerical particle transport simulations in fluid flow transport prediction. To consider various non-spherical particles in
were performed under the assumption of spherical particles.1,2 The particulate flow simulations, the effects of the particle shape and
drag force exerted on the sphere was accurately predicted as a orientation should be considered.8
function of the Reynolds number (Re) using Stokes’ law, which Among regular shapes, an elliptic cylinder can represent a vari-
arises from the Navier–Stokes equations under the assumption of ety of particle shapes by varying its axis ratio (AR).7 An elliptic
creeping flow.3 Because the Re value of the particles of interest cylinder can be altered to a spherical shape when AR = 1 and to
exceeds the Stokes flow regime, more accurate corrections have been a fiber-like shape as the AR approaches zero. The angle of attack
proposed,4,5 which cover an Re range up to the subcritical regime. (AOA) of an elliptic cylinder, along with its AR, which also has a

AIP Advances 12, 065228 (2022); doi: 10.1063/5.0097916 12, 065228-1

© Author(s) 2022
 AIP Advances ARTICLE scitation.org/journal/adv

significant effect on drag, introduces an asymmetric wake pattern ratios of 1–32. Sanjeevi et al.18 expanded the Re limitation of the cor-
and induces lift and rotational forces. Because their effects have been relations to 2000 and observed good agreement with the DNS data
an interesting phenomenon in fluid mechanics, researchers have for three different ellipsoids. Recently, Ouchene et al.19 derived the
proposed drag, lift, and torque formulations considering the par- correlations for oblate particles valid for aspect ratios of 0.2–1 and
ticle shape and orientation, non-Newtonian fluid effects, and heat Re values of up to 100.
transfer.7,9–24 Leith10 derived the shape factor, which is the drag To the best of our knowledge, although Zastawny et al.15
force modification factor for a non-spherical particle based on a and Sanjeevi et al.18 carried out a comprehensive parametric study
Stokes’ law extension. The derived shape factor was compared with according to Re and orientation, only ellipsoids with two aspect
measured literature data, exhibiting good agreement. Mittal and Bal- ratios were considered. Ouchene et al.17,19 provided accurate CD , CL ,
achandar11 simulated the flow around an elliptic cylinder with a and CT correlations for prolate and oblate ellipsoids based on a com-
body-fitted mesh in elliptic-cylindrical coordinates. The drag and plete parameter set, including various aspect ratios. They proposed
lift coefficients (CD and CL , respectively) were observed for fixed non-trivial correlation forms as an extension of previous analytic
parameters and various boundary treatments. They found that the research.14,15,27 However, the variation trend according to the aspect
Strouhal number (St) and CD obtained from their numerical study ratio, orientation, and Re is not immediately visible. In addition,
agreed well with the experimental data. Faruquee13 investigated the previous studies conducted steady-state simulations without distin-
effects of the AR on the flow over an elliptic cylinder with AR = 0.3–1 guishing between steady and unsteady flows, which induce different
at Re = 40 and observed the variations in CD and wake length behind flow pattern phases.28
the cylinder. The study also investigated the contributions of the In this study, we performed unsteady numerical simulations of
pressure and friction drag coefficients (CDp and CDf , respectively) the flow around an elliptic cylinder using full-parameter simulations
to the total drag coefficient and provided correlations between each for 0.3 ≤ AR ≤ 1 and 0○ ≤ AOA ≤ 90○ and derived a complete set of
drag coefficient and AR as polynomial functions of the AR. Raman CD , CL , and CT correlations valid in a subcritical Re regime. We
et al.16 considered the unsteady flow around an elliptic cylinder. obtained a subcritical regime in which all elliptic cylinders exhibit
In particular, they observed variations in time-averaged CD and St steady flows by observing the evolution of CL with increasing Re
according to the AR for moderate Re. Additionally, they constructed for all ARs and AOAs. Subsequently, we conducted additional sim-
functional relationships between CD and AR for different Re values. ulations in the subcritical regime for 0.1 ≤ Re ≤ 10. The proposed
The aforementioned studies investigated the effect of the AR for dif- correlations consist of simple functions to explicitly represent the
ferent Re values in a viscous flow regime; however, the effect of the hydrodynamic force variations according to AR, AOA, and Re. The

07 October 2023 09:16:46

orientation was not considered. novelty of this study lies in providing an Recrt map, indicating a
Orientation-induced asymmetry in flow patterns around transition between steady and unsteady flows and simple correla-
elliptical particles causes a lift force and torque. Holzer and tion functions for accurately predicting hydrodynamic forces under
Sommerfeld14 considered the orientation by introducing the cross- Stokes to laminar flow.
wise sphericity (ϕ– ), which is defined as the ratio between the The remainder of this paper is organized as follows: Sec. II
cross-sectional area of a sphere and the projected area of the volume defines the problem and describes the numerical methods and val-
equivalent ellipsoidal particle perpendicular to the streamwise direc- idation cases. Section III is devoted to the subcritical flow regime
tion. Furthermore, sphericity (ϕ), which is defined as the ratio of the identification and analysis of the flow pattern, hydrodynamic coef-
surface area of a sphere to the volume equivalent ellipsoidal parti- ficients, and proposed correlations. Finally, Sec. IV offers several
cle,25 was introduced to construct the correlation. In these studies, concluding remarks.
the orientation and shape effects exhibit variations in CD in terms
of ϕ– and ϕ. Zastawny et al.15 constructed correlation functions for
an ellipsoidal particle based on the numerical data obtained from II. NUMERICAL METHODS AND PROBLEM SETUP
immersed boundary-based simulations. Their results revealed low
accuracy predictions for CL and the torque coefficient (CT ), whereas A. Projection and immersed boundary methods
CD showed good agreement with the predictions. Consequently, The conventional approach in computational fluid dynamics
they proposed new equations for CD , CL , and CT , which agreed well (CFD) for simulating the flow around an object begins by gener-
with the simulation results for four different non-spherical particles. ating body-fitted grids on the solid surface. The flow field grids
Ouchene et al.7 performed direct numerical simulations (DNS) to are radiated from the surface grids with a proper growth rate,
provide CD , CL , and CT for three different ellipsoidal particles for and a no-slip boundary condition is imposed on the surface grids.
Re = 0.1–290 with a body-fitted mesh. They compared their corre- As our research object is an elliptic cylinder, this study requires
lation predictions with previous studies14,15,26 and showed that the body-fitted grids with elliptic-cylindrical coordinates.11 However,
correlation functions of Zastawny et al.15 provided the best predic- generating a grid for every AR and AOA set is time-consuming
tions of CD , CL , and CT compared with the DNS results, with relative and unnecessary. To avoid the uncertainty associated with various
deviations of up to 20%. Despite these superior predictions, they grid sets, the immersed boundary method (IBM) is utilized, which
found that the correlations of Zastawny et al.15 show significant devi- can express the presence of the object via a forcing added to the
ations outside of the Stokes flow regime. Ouchene et al.17 derived a Navier–Stokes equations on Cartesian grids. An elliptic cylinder
set of CD , CL , and CT correlations for prolate ellipsoids immersed with various ARs and AOAs can be expressed on the domain with-
in an uniform flow considering the aspect ratio, incident angle, and out generating numerous body-fitted grids. The governing equa-
Re as parameters. The correlations covered a complete set of para- tions of incompressible fluid flow with an immersed object are as
meters for prolate ellipsoids in an Re range of up to 240 and aspect follows:

AIP Advances 12, 065228 (2022); doi: 10.1063/5.0097916 12, 065228-2

© Author(s) 2022
 AIP Advances ARTICLE scitation.org/journal/adv

∂u 1 non-dimensional drag, lift, and torque coefficients are, respectively,

+ (u ⋅ ∇)u = −∇p + ∇2 u + f , (1)
∂t Re defined as follows:
∣FD ∣ ∣FL ∣ ∣T∣
CD = 1 2 D
, CL = 1 2 D
, CT = 1 , (4)
∇ ⋅ u = 0, (2) 2
ρu ∞ h 2
ρu ∞ h 2
ρu2∞ D2h

where u is the fluid velocity vector field, p is the pressure, and f is where ρ is the fluid density, u∞ is the reference velocity, and Dh is the
the forcing term. The fluid flow lies on the Eulerian grid (Cartesian hydraulic diameter of the elliptical cylinder. F D and F L are the drag
coordinate system), and the surface of the solid body is formed as a and lift forces, respectively, which are calculated by summing the
set of Lagrangian points. The volume forcing f , an additional term forces on the Lagrangian points along the cylinder surface as follows:
in Eq. (1), imposes a no-slip boundary condition on the immersed L L
boundary surface. It lies on the Eulerian grid and is calculated using FD = ∑ Fx (Xl ), FL = ∑ Fy (Xl ), (5)
the interaction between the fluid flow and the immersed surface. l=1 l=1
The interaction is achieved via interpolation, which is the discrete
delta function of a three-point continuous function used by Roma where L is the number of Lagrangian points on the immersed surface
et al.29 The immersed boundary forcing F on the Lagrangian points and X l denotes the locations of the Lagrangian points. The torque
was obtained by interpolating the flow field velocity and the desired acting on a particle is calculated using the surface forces and relative
velocity. The calculated forcing F is extrapolated to the Eulerian position vector as follows:
grid as volume forcing f . As calculating a suitable F value using a
naïve treatment does not yield accurate results for low-Re flow cases, T= ∮ r × {(−pI + τ w ) ⋅ n}dS, (6)
∂S
Park et al.30 proposed an IBM based on the implicit direct forcing
scheme to improve the accuracy for such cases by using an iterative where I is the Kronecker delta stress, τ w is the shear stress, and n is
procedure to estimate F as follows: the normal vector on the surface element on the object.

Fk = Fk−1 + (Ud − BI u∗,k−1 ), (3) C. Problem definition

The computational domain and related boundary conditions
where Ud is the desired velocity on the Lagrangian points, k is the are illustrated in Fig. 1(a). Sen et al.35 found that a blockage ratio (B)

07 October 2023 09:16:46

iterative step of the implicit direct forcing, BI is the interpolation of 0.01 is sufficiently low to avoid the blockage effect; B is defined
operator, and u is the flow velocity on Eulerian grids. as the ratio of Dh to the length of the domain perpendicular to
The governing equations were solved using the projection the streamwise direction. Accordingly, the size of the computational
method.31 The incompressibility constraint was first ignored, and an domain used in this study was 100Dh × 100Dh to avoid the blockage
intermediate velocity field was obtained. Subsequently, the velocity effect. A uniform flow (u/U ∞ = 1, v = 0) in the streamwise direc-
field of the next time step was obtained by projecting the inter- tion was imposed at the inlet of the domain. A Neumann condition
mediate velocity field onto the divergence-free field to obtain the was applied at the top and bottom boundaries to allow the flow to
incompressibility condition. The above equations are discretized properly exit the outlet. For the outlet, a convective boundary con-
via finite-difference formulations of the second-order central dif- dition was applied to allow the flow and vortices to smoothly exit
ference scheme for spatial derivatives. To increase the numerical the domain. For an efficient computation, a non-uniform grid clus-
stability of the calculations, all terms are advanced in time with the tered near the cylinder was used. A uniform grid was embedded
Crank–Nicolson scheme. Block LU factorization was used to decou- only near the cylinder to utilize the discrete delta function for cal-
ple the velocity and pressure. Additionally, the coupled intermediate culating the surface forces. The size of the uniform zone was 3Dh
velocity components were decoupled via approximate factorization. × 3Dh , which covered the elliptic cylinder expressed as Lagrangian
Both factorizations retain second-order accuracy in space and time. points.
Details of the projection method and IBM algorithms can be found The distance between a Lagrangian point and its neighbor-
in studies by Pan et al.32 and Park et al.,30 respectively. ing points must be the same as the grid size of the uniform grid
zone. Thus, a simple strategy is required to form evenly distributed
B. Drag, lift, and torque coefficients Lagrangian points, as shown in Fig. 1(b). The well-known ellipse
The hydrodynamic forces acting on the particles are defined equation is as follows: (x/a)2 + (y/b)2 = 1, where a and b are the
by the drag, lift force, and torque. The drag and lift forces are the semi-major and semi-minor axis lengths, respectively. The first
components of the total force with respect to the streamwise direc- Lagrangian point, L1 , is laid on (a, 0). A circle with radius DS and
tion (flow direction) and perpendicular to the streamwise direction, center L1 can provide two intersection points using the ellipse equa-
respectively. The lift force and torque are the result of flow asym- tion. One solution is defined as the second Lagrangian point L2 . In
metry caused by non-aligned particles in the streamwise direction. addition, L2 becomes the center of a new circle, which also provides
According to Newton’s second law, a particle’s translational motion two intersection points using the ellipse equation. Because one of
is induced by the drag and lift forces, which corresponds to the equa- the intersection points coincides with L1 , the other point is defined
tion of motion.1,33,34 Similarly, rotational motion is induced by the as L3 . The same process along the surface of the ellipse generates
torque acting on the particle.7,15,17,19 a set of evenly distributed Lagrangian points. The process contin-
The forces acting on the cylinder, such as the drag, lift, and ues until the distance between the newly generated points Li and L1
torque, can be characterized by non-dimensional coefficients. The is smaller than DS. The intersection points are easily calculated as

AIP Advances 12, 065228 (2022); doi: 10.1063/5.0097916 12, 065228-3

© Author(s) 2022
 AIP Advances ARTICLE scitation.org/journal/adv

FIG. 1. (a) Computational domain for

the flow around an elliptic cylinder and
related boundary conditions. (b) Build-
ing up Lagrangian points using a fixed
distance DS.

solutions to the fixed-point iteration method. The inclination of the and maximum deviations from them. The CD results for flow over an
ellipse was determined using vector rotation. The rotated position elliptic cylinder for AOA = 0 (no incident angle) are compared with
(x′ , y′ ) originated at point (x, y) is obtained as x′ = x cos θ − y sin θ, a previous study in the low-Re regime, as shown in Table III. These
y′ = x sin θ + y cos θ. The AR of an elliptic cylinder is defined as the comparisons show good agreement between the results obtained in
minor-to-major axis length ratio (i.e., AR = b/a). The AR values used this study and those in the literature in the low to moderate Re range.
in this study were 0.3, 0.4, 0.5, 0.6, 0.8, and 1. The incident angle is
denoted as the AOA and was varied from 0○ to 90○ at 10○ intervals. III. RESULTS AND DISCUSSION
All elliptic cylinders had the same Dh value of unity.
A. Critical Reynolds number and subcritical
Re regime
D. Flow solver validation
Fluid flow characteristics around a circular cylinder according
We performed a grid convergence test for flow around a circu-

07 October 2023 09:16:46

to Re have been extensively documented for decades via experimen-
lar cylinder at Re = 40. Four uniformly spaced grids in the uniform
tal and numerical approaches.28,35,38,39,41,42 The fully attached flow
zone (3Dh × 3Dh ) were considered using grid spacings of h = Dh /20,
for low-Re flow exhibits a steady separation bubble behind the cylin-
Dh /40, Dh /80, and Dh /160. Considering the CFL condition, the
der that grows in the streamwise direction as Re increases. Flow
computational time step Δt was defined as Δt = h/U ∞ . Table I shows
instability occurs as Re increases, and periodic flow patterns, which
CD and its relative error (ϵr ) obtained from the simulations with
are called von Kármán vortex streets, are observed. Particularly, Re
five different grid resolutions. Here, the relative error is defined as
corresponding to the unsteadiness threshold is defined as the critical
ϵr = ∣CD,h − CD,h f ∣/CD,h f , where CD,h and CD,h f denote CD with h and
Reynolds number (Recrt ).
with the finest grid spacing (hf /Dh = 1/160), respectively. The refer-
ence value for the grid test was set to CD,h f . The results indicated
that the grid resolution with h/Dh = 80 in the uniform zone pro-
vided reasonably accurate predictions of hydrodynamic forces acting TABLE II. CD , CL , and St for flow over a circular cylinder.
on the cylinder. Thus, for the following simulations for the elliptic Re = 1 Re = 40 Re = 100
cylinder cases, a non-uniform grid with 361 × 361 points was used
in the entire domain (100Dh × 100Dh ), while a uniform grid with CD CD CD CL St
241 × 241 points was used in the uniform zone (3Dh × 3Dh ).
The implemented flow solver was validated by comparing the Choi et al.36 ⋅⋅⋅ 1.52 1.34 ± 0.011 ±0.315 0.164
aerodynamic coefficients for flow over stationary circular and elliptic Liu et al.37 ⋅⋅⋅ ⋅⋅⋅ 1.35 ± 0.012 ±0.339 0.164
cylinders in the low to moderate Re range. The predicted aerody- Russel and Wang38 ⋅⋅⋅ 1.60 1.38 ± 0.007 ±0.300 0.169
namic coefficients for flow over the circular cylinder are compared Tritton39 11.70 1.48 ⋅⋅⋅ ⋅⋅⋅
to those of previous studies, as shown in Table II. In this table, CD Present 10.99 1.54 1.35 ± 0.009 ±0.331 0.164
and CL for the Re = 100 case indicate the time-averaged quantities

TABLE III. CD for flow over an elliptic cylinder for AOA = 0.

TABLE I. Effect of grid spacing h in the uniform zone on CD for flow around a circular
cylinder at Re = 40. AR = 0.5 AR = 2.0
h/Dh 1/20 1/40 1/80 1/160 Re 15 20 25 15 20 25
CD 1.5797 1.5502 1.5382 1.5383 Chen and Yen40 2.414 2.099 1.882 2.317 2.055 1.882
ϵr 2.69 × 10−2 7.74 × 10−3 6.50 × 10−5 ⋅⋅⋅ Present 2.433 2.108 1.892 2.327 2.067 1.893

AIP Advances 12, 065228 (2022); doi: 10.1063/5.0097916 12, 065228-4

© Author(s) 2022
 AIP Advances ARTICLE scitation.org/journal/adv

Because the different elliptic cylinder geometric configurations The flow patterns around an elliptic cylinder with AR = 0.4 and
induce variations in the flow field near the cylinder, the value of Recrt AOA = 0○ , 30○ , 45○ , 60○ , and 90○ for Re = 10, 40, and 100 are shown
varies, and CD and CL vary accordingly. Raman et al.16 and Paul in Fig. 2. As a fundamental case, Paul et al.28 identified Recrt = 48 for
et al.28 studied Recrt variations according to the shape of an ellip- the circular cylinder case. In our study, when Re = 10, a steady flow
tic cylinder, which was oriented with and inclined with respect to was observed for all elliptic cylinder cases. For Re = 40, unsteadi-
the streamwise direction. These previous studies confirmed that the ness did not occur at AOA = 0○ or 30○ . As Re increased to 100, an
shape of the cylinder significantly influences Recrt . Consequently, it elliptic cylinder with AOA = 0○ induced a steady flow. It is observed
is necessary to observe the effects of the AR and AOA on Recrt and that the shape of an elliptic cylinder induces lower and delayed Recrt
divide the Re regime accordingly. Paul et al.28 identified Recrt by compared to that of the circular cylinder case. Consequently, Recrt
observing the evolution of CL , which is one of the intuitive methods for various ARs and AOAs was identified by increasing Re in inter-
among the diverse methods described in that study. For stable flow, vals of 1, as shown in Fig. 3. The subcritical regime is achieved
the transient oscillatory behavior of CL with time decreases to zero. when Re < 16, which means that all elliptic cylinder cases induce
However, the amplitude of the periodic behavior of CL increases a stable flow. The fluid flow exhibits an unsteady behavior for all
and has a saturated value when the flow is unsteady and fully elliptic cylinder cases at Re > 165. The Re value for a particle in an
periodic. actual situation is low and is in the steady flow regime. Therefore,

07 October 2023 09:16:46

FIG. 2. Vorticity contours for the flow
around an elliptic cylinder (AR = 0.4)
for various Re and AOA values; the con-
tour color indicates the vorticity sign (red:
positive and blue: negative).

AIP Advances 12, 065228 (2022); doi: 10.1063/5.0097916 12, 065228-5

© Author(s) 2022
 AIP Advances ARTICLE scitation.org/journal/adv

high-pressure region is the widest, while the high-vorticity region is

the smallest for all AOA cases.

C. Hydrodynamic coefficients and the proposed

correlations
The correlations reported in recent numerical studies7,15,17–19
were proposed as extension forms based on the analytic correlations
given by Happel and Brenner,27

CD = CD,ϕ=0○ + (CD,ϕ=90○ − CD,ϕ=0○ )sin2 ϕ, (7)

CL = (CD,ϕ=90○ − CD,ϕ=0○ ) sin ϕ cos ϕ, (8)
FIG. 3. Recrt of an elliptic cylinder according to different ARs and AOAs. where ϕ is the orientation angle. CD was expressed as a sine-squared
function in previous studies, except for the work of Mando and
Rosendahl,43 which used a cubic sine function. Sanjeevi et al.18 con-
firmed that the functional form of Eq. (7) is valid for high Re, even as
it is reasonable that the observations and resultant correlations are high as 2000. In addition, it has been consistently confirmed that the
performed for Re under 15, which corresponds to the steady flow functional form of sin ϕ cos ϕ in Eq. (8) can represent the behavior
regime. of CL . Furthermore, Zastawny et al.15 scaled CL with the functional
form of the power of sin ϕ and cos ϕ.
B. Flow field near an elliptic cylinder under Recrt The drag coefficients of the extremum cases (i.e., CD,ϕ=0○ and
The hydrodynamic force variations were determined using the CD,ϕ=90○ ) were modeled using various functional forms. Zastawny
pressure and vorticity on the surface of the cylinder. The pressure et al.15 used a simple power form of Re, while Ouchene et al.17,19 pro-
and vorticity contours around an elliptic cylinder for various AOAs posed a combination of the power forms of the aspect ratio and Re.
are shown in Fig. 4. We considered the flow around the elliptic cylin- Sanjeevi et al.18 introduced exponential Re terms into the functional
der with AR = 0.4 and Re = 10 as a representative case because other form proposed by Zastawny et al.15 Ouchene et al.19 confirmed

07 October 2023 09:16:46

cases show similar variations. It is observed that the high-pressure that the mean relative deviations (MRDs), ∣CD,correlation − CD,CFD ∣
and -vorticity region shifts as the elliptic cylinder is rotated. For non- × 100/CD,CFD (%), of their correlations for CD,ϕ=0○ and CD,ϕ=90○ are
zero AOAs, the high-pressure region is confined to the stagnation 4% and 10.9%, respectively. They also confirmed that the MRDs
point, whereas the high-vorticity region spreads broadly on the sur- of the correlations by Zastawny et al.15 were ∼12.5% and 26.5%,
face of the elliptic cylinder. As the AOA increases, the stagnation respectively.
point moves to the hill of the elliptic cylinder, and the high-pressure
region spreads widely. In addition, high-negative-pressure regions 1. Drag coefficient
formed on the tips of the downside of the elliptic cylinder. The high- The drag coefficients obtained via numerical simulations for an
vorticity region was maintained on the front of the elliptic cylinder elliptic cylinder with different ARs and AOAs for the three Re cases
and diminished on the back as the AOA increased. As the AOA are shown as symbols in Figs. 6(a)–6(c). It is observed that the over-
reached 90○ , the pressure and vorticity fields near the elliptic cylin- all magnitude of CD decreased with increasing Re for all ARs and
der exhibited symmetric patterns similar to the AOA = 0○ case. The AOAs. This is consistent with numerous previous studies35,39,41,42

FIG. 4. Pressure contours (upper row) and vorticity contours (bottom row) near an elliptical cylinder (AR = 0.4 and Re = 10) with the increasing AOA.

AIP Advances 12, 065228 (2022); doi: 10.1063/5.0097916 12, 065228-6

© Author(s) 2022
 AIP Advances ARTICLE scitation.org/journal/adv

well with the simulation results, with an MRD of ∼0.049%. It can

also be observed from Fig. 5 that Eq. (9) is in good agreement with
the results of previous studies.
Regarding the rotation of the elliptic cylinder, the amplitude
of CD exhibits a larger increase for an elongated elliptic cylinder as
it rotates from AOA = 0○ to 90○ . Additionally, the amplitude of CD
is also dependent on the AR. Similar trends were observed for all
Re cases. Furthermore, the variation in magnitude according to the
AR when AOA = 0○ for each Re case was larger for lower Re. The
drag coefficient of an elliptic cylinder reflecting the aforementioned
effects can be expressed as follows:
b3
b1 1 b4
CD = CD,AOA=0 + ( − 1) ( b − cos(2AOA)). (10)
FIG. 5. CD of flow around a circular cylinder under low-Re regime flow. The drag Reb2 AR Re 5
coefficient calculated using the correlation in Eq. (9) is represented by the solid
black line. The model coefficients were found to be b1 = 0.77, b2 = 0.37,
b3 = 1.03, b4 = 1.09, and b5 = 0.12. The overall magnitude of CD is
computed using the first term, which is given by Eq. (9). The ampli-
and is established as a correlation. Sen et al.35 expressed the CD − Re tude variation for each elliptic cylinder as it rotates is dependent
relationship based on previous studies, which is given as on Re and AR, which are described as power forms in the second
term. The limiting behavior of the trigonometric term (second term)
a1
CD,AOA=0 = + a3. (9) approaching zero when the AR approaches unity (circular cylinder)
Rea2 is reflected as 1/AR − 1, and the correlations revert to those of the
By using a non-linear least-squares curve fitting, the model coef- circular cylinder case. The trigonometric function is assumed to be
ficients were found to be a1 = 9.9, a2 = 0.75, and a3 = 1.09 for 0.1 b4 /Reb5 − cos(2AOA) to reflect the Re-dependency at AOA = 0○ , as
≤ Re ≤ 10. Qualitatively, the results obtained with Eq. (9) agree very discussed above. It is worth noting that this function is reasonable

07 October 2023 09:16:46

FIG. 6. CD of the elliptic cylinder as a function of the AR and AOA: (a) Re = 0.1, (b) 1.0, and (c) 10. (d) CD comparison between the CFD results and the correlation function.

AIP Advances 12, 065228 (2022); doi: 10.1063/5.0097916 12, 065228-7

© Author(s) 2022
 AIP Advances ARTICLE scitation.org/journal/adv

because it has a 1 − cos(2AOA) form and is the converted form of summation of two opposite contributions and decreases as the AOA
the sine-squared function utilized in previous studies. The proposed approaches 90○ .
correlations show good agreement with the simulation results, as The behavior of CL corresponds to trigonometric functions, as
shown in Figs. 6(a)–6(c). Moreover, Fig. 6(d) confirms that the pro- shown in Eq. (8). The variations in the CL magnitude according to
posed correlation for CD in Eq. (10) is qualitatively trustworthy, with AR are similar to those in the CD cases for the same reason that the
an MRD of ∼1.5%. Note that the correlations for CD proposed by amplitude of the trigonometric function induced by the elongation
Sanjeevi et al.18 and Ouchene et al.19 show that the overall MRD was diminished as the AR approaches unity. The maximum values for
∼2% and less than 10%, respectively. each Re case were achieved at 40○ ≤ AOA ≤ 50○ . It is worth noting
that the AOA exhibiting the maximum value shifts to AOA = 45○
2. Lift coefficient and the curves become symmetric as Re decreases. This observation
The AR- and Re-dependency of CL is shown in Figs. 7(a)–7(c). was also reported by Zastawny et al.15 and Ouchene et al.7 Similar
It is well known that a lift force is induced by the pressure dif- in context as the functional form of CD , the proposed correlation for
ference between the upper and lower sides of an object. The cases CL is as follows:
where AOA = 0○ and 90○ lead to a symmetric flow pattern, as illus- c1 1 c3
trated in Fig. 4, and CL for these cases is zero. As the elliptic CL = ( − 1) sin(AOA) cos(AOA), (11)
Rec2 AR
cylinder is inclined, the fluid flow exhibits an asymmetric pattern.
In Fig. 4, the pressure contour for AOA = 30○ shows that a high- where the model coefficients were found to be c1 = 0.74, c2 = 0.27,
pressure region is formed around the head of the elliptic cylinder. and c3 = 0.78. The magnitude terms also reflect the Re-dependency
It is observed that the vertical components of the positive and neg- of CL and reduce the trigonometric functions to zero as the AR
ative pressure values in that region are in the negative y-direction. approaches unity. The proposed equation (11) has the same trigono-
This contributed to the total lift force. As AOA increased to 60○ , metric function as the correlation proposed by Zastawny et al.15
another negative pressure region appeared at the tail of the ellip- and Ouchene et al.,17,19 scaling CL with sin ϕ cos ϕ. Sanjeevi et al.18
tic cylinder. Its vertical component is in the opposite direction reported a correlation for CL with an overall MRD of ∼3%–4%.
compared to the previous contribution, which is in the negative Ouchene et al.19 reported that the MRD is in the order of less than
y-direction. Subsequently, the total lift force decreases due to the 10%. The CL correlation proposed in this study is presented as lines

07 October 2023 09:16:46

FIG. 7. CL of the elliptic cylinder as a function of the AR and AOA: (a) Re = 0.1, (b) 1.0, and (c) 10. (d) CL comparison between the CFD results and the correlation function.

AIP Advances 12, 065228 (2022); doi: 10.1063/5.0097916 12, 065228-8

© Author(s) 2022
 AIP Advances ARTICLE scitation.org/journal/adv

07 October 2023 09:16:46

FIG. 8. CT of the elliptic cylinder as a function of the AR and AOA: (a) Re = 0.1, (b) 1.0, and (c) 10. (d) CT comparison between the CFD results and the correlation function.

in Figs. 7(a)–7(c), with an MRD of ∼8.9%. Figure 7(d) compares the The correlations for CT also show good agreement with the simu-
simulation results and the correlation for CL , demonstrating that the lation results, as shown by the lines in Figs. 8(a)–8(c). The calculated
proposed correlation is reliable compared with previous studies. MRD was ∼11.2%. The MRD of the CT correlations proposed by
Sanjeevi et al.18 and Ouchene et al.19 is ∼3.5% (or less) and less than
3. Torque coefficient 8%, respectively. Therefore, it is also qualitatively confirmed that the
A torque acts on the elliptic cylinder for the same reason as proposed correlation for CT in Eq. (12) can represent the behavior of
in the case of the lift force. The torque is zero when the direction CT , as shown in Fig. 8(d). However, it fails to capture the movement
of the total force acting on the elliptic cylinder is in the streamwise of the AOA, showing a maximum at Re = 10 for both the CL and CT
direction. The asymmetry caused by the incident angle induces non- correlations.
zero lift force values and rotates the direction of the total force. This
asymmetry also causes a disagreement between the center of pres-
sure and the center of gravity. The total force acts on the center of
IV. CONCLUSIONS
pressure and induces torque on the elliptic cylinder.26
The torque is also the result of asymmetrical flow and exhibits a Numerical simulations of the flow around an elliptic cylinder
behavior similar to that of the CL case, as shown in Fig. 8. Thus, the with various ARs and AOAs in the low-Re regime were performed
CT correlation function has the same structure as the CL case. The to improve the hydrodynamic coefficient prediction. The projection
CT correlation function is proposed as follows: method combined with the immersed boundary method was vali-
dated for various Re values. An iterative procedure for calculating
d3 the forcing term was applied to obtain more accurate no-slip bound-
d1 1
CT = d
( − 1) sin(AOA) cos(AOA), (12) ary conditions for the immersed surface. The numerical results
Re 2 AR
showed that the geometrical conditions have a significant influence
where the model coefficients were found to be d1 = 0.085, d2 = 0.068, on shifting the Recrt value and on CD , CL , and CT variations. For
and d3 = 1.17. CT was zero when AR = 0○ and 90○ . The variation in AOA = 0○ , Recrt was delayed to high values when the elliptic cylinder
amplitude according to Re is not significant for the CT case, as shown became elongated in the streamwise direction. The incident angle of
in Figs. 8(a)–8(c), explaining the very small values of d1 and d2 . the elongated cylinder induced earlier unsteady flow. The subcritical

AIP Advances 12, 065228 (2022); doi: 10.1063/5.0097916 12, 065228-9

© Author(s) 2022
 AIP Advances ARTICLE scitation.org/journal/adv

regime was obtained for Re < 16 for all elliptic cylinders and AOA 4
C. W. Oseen, Newer Methods and Results in Hydrodynamics (Akademische
cases. Verlagsgesellschaft, Leipzig, 1927).
5
Detailed numerical simulations in the subcritical regime were L. Schiller and A. Naumann, “Uber die grundegenden berch-nungen bei der
performed for 0.3 ≤ AR ≤ 1.0, 0○ ≤ AOA ≤ 90○ , and Re = 0.1, 1, and schwerkraftaufbereitung,” Z. Ver. Dtsch. Ing. 77, 318–320 (1933).
6
C. Yin, L. Rosendahl, S. Knudsen Kær, and H. Sørensen, “Modelling the motion
10. The functional forms of the correlations were carefully chosen
of cylindrical particles in a nonuniform flow,” Chem. Eng. Sci. 58, 3489–3498
with simple power and trigonometric functions as an extension of (2003).
those reported in previous studies. The effects of Re and AR were 7
R. Ouchene, M. Khalij, A. Tanière, and B. Arcen, “Drag, lift and torque coeffi-
expressed as amplitude terms when the elliptic cylinder became cir- cients for ellipsoidal particles: From low to moderate particle Reynolds numbers,”
cular. The amplitude of the trigonometric functions is reduced to Comput. Fluids 113, 53–64 (2015).
8
zero as the AR approaches unity to reflect the limiting behavior. B. Arcen, R. Ouchene, M. Khalij, and A. Tanière, “Prolate spheroidal particles’
The effect of Re was significant for CD and minor for CT . The con- behavior in a vertical wall-bounded turbulent flow,” Phys. Fluids 29(9), 093301
vex curves of CL and CT became symmetric, and the maximum (2017).
9
G. B. Jeffery, “The motion of ellipsoidal particles immersed in a viscous fluid,”
values shifted to approximately AOA = 45○ as Re decreased. The
Proc. R. Soc. London, Ser. A 102, 161–179 (1922).
accuracy of the proposed correlations was proven to be trustworthy 10
D. Leith, “Drag on nonspherical object,” Aerosol Sci. Technol. 6, 153–161
and in agreement with the simulation results, with the mean relative (1987).
deviations of 1.5%, 8.9%, and 11.2% for the drag, lift, and torque 11
R. Mittal and S. Balachandar, “Direct numerical simulation of flow past elliptical
coefficients, respectively. The proposed correlations demonstrate cylinders,” J. Comput. Phys. 124, 351–367 (1996).
12
the ability to achieve accurate predictions for Euler–Lagrangian S. A. Johnson, M. C. Thompson, and K. Hourigan, “Flow past elliptical cylinder
particulate flow simulations and fixed-bed porous simulations. at low Reynolds numbers,” in Australasian Fluid Mechanics Conference (Elsevier,
Adelaide, Australia, 2001).
13
ACKNOWLEDGMENTS Z. Faruquee, D. S.-K. Ting, A. Fartaj, R. M. Barron, and R. Carriveau, “The
effects of axis ratio on laminar fluid flow around an elliptical cylinder,” Int. J. Heat
This work was supported by the National Research Foun- Fluid Flow 28, 1178–1189 (2007).
14
dation of Korea (NRF) (Grant Nos. 2017R1E1A1A03070161 and A. Hölzer and M. Sommerfeld, “New simple correlation formula for the drag
2022R1A2C2003643) funded by the Korean government (Ministry coefficient of non-spherical particles,” Powder Technol. 184, 361–365 (2008).
15
of Science and ICT). M. Zastawny, G. Mallouppas, F. Zhao, and B. van Wachem, “The derivation of
drag and lift force and torque coefficients for non-spherical particles in flows,” Int.
J. Multiphase Flow 39, 227–239 (2012).

07 October 2023 09:16:46

AUTHOR DECLARATIONS 16
S. K. Raman, K. A. Prakash, and S. Vengadesan, “Effect of axis ratio on fluid flow
Conflict of Interest around an elliptic cylinder - a numerical study,” J. Fluids Eng. 135, 111201 (2013).
17
R. Ouchene, M. Khalij, B. Arcen, and A. Tanière, “A new set of correlations of
The authors have no conflicts to disclose. drag, lift and torque coefficients for non-spherical particles and large Reynolds
numbers,” Powder Technol. 303, 33–43 (2016).
18
S. K. P. Sanjeevi, J. A. M. Kuipers, and J. T. Padding, “Drag, lift and torque cor-
Author Contributions relations for non-spherical particles from Stokes limit to high Reynolds numbers,”
Int. J. Multiphase Flow 106, 325–337 (2018).
Geunwoo Oh: Conceptualization (equal); Investigation (equal); 19
R. Ouchene, “Numerical simulation and modeling of the hydrodynamic forces
Methodology (equal); Software (equal); Validation (equal); Visual- and torque acting on individual oblate spheroids,” Phys. Fluids 32(7), 073303
ization (equal); Writing – original draft (equal). Hyunwook Park: (2020).
Conceptualization (equal); Methodology (equal); Software (equal). 20
Z. Cao, D. K. Tafti, and M. Shahnam, “Modeling drag force in ellipsoidal particle
Jung-Il Choi: Conceptualization (lead); Investigation (equal); suspensions with preferential orientation,” Powder Technol. 378, 274–287 (2021).
21
Project administration (lead); Supervision (equal); Writing – review R. P. Bharti, R. P. Ram, and A. K. Dhiman, “Computational analysis of cross-
& editing (lead). flow of power-law fluids through a periodic square array of circular cylinders,”
Asia-Pac. J. Chem. Eng. 17(2), e2748 (2022).
22
R. P. Ram, R. P. Bharti, and A. K. Dhiman, “Forced convection flow and heat
DATA AVAILABILITY transfer across an in-line bank of circular cylinders,” Can. J. Chem. Eng. 94(7),
1381–1395 (2016).
The data that support the findings of this study are available 23
L. He and D. Tafti, “Variation of drag, lift and torque in a suspension of
from the corresponding author upon reasonable request. ellipsoidal particles,” Powder Technol. 335, 409–426 (2018).
24
X. Li, M. Jiang, Z. Huang, and Q. Zhou, “Effect of particle orientation on the
drag force in random arrays of prolate ellipsoids in low-Reynolds-number flows,”
REFERENCES AIChE J. 65(8), e16621 (2019).
25
H. Wadell, “The coefficient of resistance as a function of Reynolds number for
1
C. Marchioli, A. Soldati, J. G. M. Kuerten, B. Arcen, A. Tanière, G. Goldensoph, solids of various shapes,” J. Franklin Inst. 217, 459–490 (1934).
26
K. D. Squires, M. F. Cargnelutti, and L. M. Portela, “Statistics of particle dispersion L. Rosendahl, “Using a multi-parameter particle shape description to predict the
in direct numerical simulations of wall-bounded turbulence,” Int. J. Multiphase motion of non-spherical particle shapes in swirling flow,” Appl. Math. Model. 24,
Flow 34, 879–893 (2008). 11–25 (2000).
27
2
A. Vié, H. Pouransari, R. Zamansky, and A. Mani, “Particle-laden flows forced by J. Happel and H. Brenner, Low Reynolds Number Hydrodynamics: With Special
Applications to Particulate Media (Springer Science and Business Media, 2012),
the disperse phase: Comparison between Lagrangian and Eulerian simulations,”
Vol. 1.
Int. J. Multiphase Flow 79, 144–158 (2016). 28
3
I. Paul, K. A. Prakash, and S. Vengadesan, “Onset of laminar separation and vor-
C. T. Crowe, J. D. Schwarzkopf, M. Sommerfeld, and Y. Tsuji, Multiphase Flows tex shedding in flow past unconfined elliptic cylinders,” Phys. Fluids 26, 023601
with Droplet and Particles (CRC Press, 2012). (2014).

AIP Advances 12, 065228 (2022); doi: 10.1063/5.0097916 12, 065228-10

© Author(s) 2022
 AIP Advances ARTICLE scitation.org/journal/adv

29 36
A. M. Roma, C. S. Peskin, and M. J. Berger, “An adaptive version of the J.-I. Choi, R. C. Oberoi, J. R. Edwards, and J. A. Rosati, “An immersed boundary
immersed boundary method,” J. Comput. Phys. 153, 509–534 (1999). method for complex incompressible flows,” J. Comput. Phys. 224, 757–784 (2007).
30
H. Park, X. Pan, C. Lee, and J.-I. Choi, “A pre-conditioned implicit direct forcing 37
C. Liu, X. Zheng, and C. H. Sung, “Preconditioned multigrid methods for
based immersed boundary method for incompressible viscous flows,” J. Comput. unsteady incompressible flows,” J. Comput. Phys. 139, 35–57 (1998).
Phys. 314, 774–799 (2016). 38
31 D. Russel and Z. J. Wang, “A Cartesian grid method for modeling multiple mov-
K. Kim, S.-J. Baek, and H. J. Sung, “An implicit velocity decoupling procedure
for the incompressible Navier-Stokes equations,” Int. J. Numer. Methods Fluids ing objects in 2D incompressible viscous flow,” J. Comput. Phys. 191, 177–205
38, 125–138 (2002). (2003).
32 39
X. Pan, C. Lee, K. Kim, and J.-I. Choi, “Analysis of velocity-components D. J. Tritton, “Experiments on the flow past a circular cylinder at low Reynolds
decoupled projection method for the incompressible Navier–Stokes equations,” numbers,” J. Fluid Mech. 6, 547–567 (1959).
Comput. Math. Appl. 71, 1722–1743 (2016). 40
S.-S. Chen and R.-H. Yen, “Resonant phenomenon of elliptical cylinder flows in
33
H. Zhang, G. Ahmadi, F.-G. Fan, and J. B. McLaughlin, “Ellipsoidal particles a subcritical regime,” Phys. Fluids 23, 114105 (2011).
transport and deposition in turbulent channel flows,” Int. J. Multiphase Flow 27, 41
C. Wieselberg, “New data on the laws of fluid resistance,” in National Advisory
971–1009 (2001).
34 Committee For Aeronautics (Technical note, 1922), p. 9.
K. T. Shanely and G. Ahmadi, “A numerical model for simulating the motions 42
of ellipsoidal fibers suspended in low Reynolds number shear flows,” Aerosol Sci. A. E. Hamielec and J. D. Raal, “Numerical studies of viscous flow around
Technol. 45, 838–848 (2011). circular cylinder,” Phys. Fluids 12, 11–17 (1969).
35 43
S. Sen, S. Mittal, and G. Biswas, “Steady separated flow past a circular cylinder M. Mandø and L. Rosendahl, “On the motion of non-spherical particles at high
at low Reynolds numbers,” J. Fluid Mech. 620, 89–119 (2009). Reynolds number,” Powder Technol. 202(1-3), 1–13 (2010).

07 October 2023 09:16:46

AIP Advances 12, 065228 (2022); doi: 10.1063/5.0097916 12, 065228-11

© Author(s) 2022