You are on page 1of 5

Aerosol Science 38 (2007) 131 – 135

www.elsevier.com/locate/jaerosci

Technical note
Lift and drag forces on a sphere attached to a wall in a Blasius
boundary layer
L.G. Sweeney, W.H. Finlay∗
Department of Mechanical Engineering, University of Alberta, Edmonton, Alta., Canada T6G 2G8

Received 11 September 2006; accepted 20 September 2006

Abstract
Lift and drag forces on a sphere attached to a planar wall, over which a laminar flat plat boundary layer flows, are examined
numerically in this study. Particle Reynolds number ranged from 0.1–250, which represents steady, laminar flow about the sphere,
and the plate Reynolds number was held constant at 32 400. A finite-volume computational fluid dynamics program was utilised.
Simulation results were validated against analytical results for drag and lift in creeping flow and against experimental results
available in the literature for lift at higher particle Reynolds number. The model results were curve-fitted and interpolating drag and
lift coefficient functions are reported. The lift and drag results are shown to be weakly dependent upon plate Reynolds number. The
resulting correlations are expected to be useful in the development of particle impending motion and aerosol entrainment predictions
of particles adhering to planar walls.
䉷 2006 Elsevier Ltd. All rights reserved.

Keywords: Sphere; Particle entrainment; Reynolds number; Coefficient of lift; Coefficient of drag; Boundary layer

A relatively common mode of aerosol creation is the resuspension of particles or droplets from a surface over
which fluid flows. Examples of such phenomenon include pollen particles in an atmospheric boundary layer, dust
in the ventilation system of a building, and dry powder drug delivery in pharmaceutical applications (Finlay, 2001).
Determining the flow conditions under which particles will detach and move into the flow is a difficult task. Central
to this task are impending motion models, which are typically created by considering force and moment balances on
a particle (Dey, 1999; Ziskind, Fichman, & Gutfinger, 1995). To have predictive ability from the models, the forces
acting on the particles must be well characterized. In the case of a spherical particle attached to a planar wall, analytical
results for the hydrodynamic forces are known only for highly specific flow arrangements, the most well-known being
shear flow at low particle Reynolds number (O’Neill, 1968). Forces acting on particles attached to a wall in a turbulent
boundary layer have been examined experimentally and a few force correlations have been proposed (Hall, 1998;
Mollinger & Nieuwstadt, 1996). However, very limited research exists that focuses on particles in a laminar boundary
layer. Extending the understanding of hydrodynamic forces acting on spherical particles to different flow regimes
extends the predictive capability of such models. In the present study we examine lift and drag forces on a sphere
attached to a planar wall over which a laminar boundary layer flows, a geometry of interest in the above context.

∗ Corresponding author. Tel.: +1 780 492 4707; fax: +1 780 492 2200.
E-mail address: warren.finlay@ualberta.ca (W.H. Finlay).

0021-8502/$ - see front matter 䉷 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jaerosci.2006.09.006
132 L.G. Sweeney, W.H. Finlay / Aerosol Science 38 (2007) 131 – 135

In a general non-similar boundary layer, such as the developing portion of a Blasius boundary layer (Rex < 1000),
the coefficients of lift and drag of a sphere attached to a wall are functions of the Reynolds number of the sphere as
well as some measure of boundary layer thickness. However, in a Blasius boundary layer a dimensionless measure of
boundary layer thickness is the ratio of boundary layer thickness to sphere diameter, or alternatively, the plate Reynolds
number. As a result, we can write

CL = f (Resphere , Rex ),

CD = g(Resphere , Rex ). (1)

The Reynolds number of the sphere is defined using the velocity at the midpoint of the sphere

Resphere = Ud/2 d/. (2)

The plate Reynolds number is defined with respect to the position, x, at which the particle is located
Ux
Rex = . (3)

Here, d is the sphere diameter,  is the kinematic viscosity, and U is the freestream velocity.
A commercially available computational fluid dynamics package, ANSYS CFX, was utilized to solve the steady,
Navier–Stokes equations

∇ · (uu) = −∇p + ∇ 2 u − gk. (4)

The flow was assumed to have constant density and viscosity, in addition to satisfying the constraint of incompressible
flow

∇ · u = 0. (5)

The software uses an unstructured grid and the finite volume technique to solve Eqs. (4) and (5). The program uses a
collocated unstaggered grid, and the advection scheme is upwind differencing with an additive corrective flux.
The computational domain is a rectangular box that uses an xy stream-wise symmetry plane that splits the sphere
in half. The geometry was modified to remove the singular point where the sphere contacts the wall by inserting a
vertical cylinder that intersects the sphere and the wall. The no-slip boundary condition was used for the sphere and wall.
A specified fully developed (Rex =32 400) Blasius velocity profile was used for the inlet, and the remaining boundaries
were treated as openings with specified pressure and the velocity derivative normal to the boundary set to zero.
The upper bound Reynolds number of the study was set as the point where the flow about the sphere becomes
unsteady. Mochizuki (1960) examined the influence a spherical particle in a laminar boundary layer has on transition
to turbulence. From their flow visualization at various free stream velocities, the transition to unsteady flow was found
to occur above a sphere Reynolds number of 250.
The accuracy of the model was confirmed using convergence testing and validation. Discretization error was tested
extensively at a sphere Reynolds number of 1, and convergence was deemed acceptable when lift and drag coefficients
change less than 3% with subsequent grid refinement. The effect of boundary placement was examined at sphere
Reynolds numbers of 0.1, and 250 by simultaneously moving each boundary. The influence of the top boundary
placement on the lift and drag calculations was tested independently at a sphere Reynolds number of 250. The remaining
convergence test determined the influence of geometry alterations at the base of the sphere on the force coefficients. At a
sphere Reynolds number of 250, geometry-independent results were demonstrated at a sphere surface area truncation of
0.25%. A summary of verification results is presented in Table 1. An estimate of total error arising from discretization,
inexact boundary conditions, and geometry approximation is determined by assuming the convergence study results
to be approximations of error. The total verification and curve-fit error at a particle Reynolds number of 0.1 is 4.9%
and 5.0%, for coefficient of lift and drag, respectively. At a particle Reynolds number of 250, it is 5.0% and 5.3%,
respectively.
In the sphere Reynolds number range 0.1–250, a total of 23 simulations were used to resolve the trends in the
coefficients of lift and drag. The simulation results for coefficients of lift and drag are presented in Fig. 1. The model
results were validated against experimental and analytical findings. In creeping flow conditions, shear flow about a
L.G. Sweeney, W.H. Finlay / Aerosol Science 38 (2007) 131 – 135 133

Table 1
A summary of verification and validation errors in the simulation at particle Reynolds number of 0.1 and 250

Lift coefficient error (%) Drag coefficient error (%)

Re = 0.1 Re = 250 Re = 0.1 Re = 250

Grid convergence 0.8 0.04


Boundary convergence 1.3 1.8 < 0.01 1.8
Top boundary 1.9 3.3
Geometric convergence 0.1 0.4
Validation 7.9 > 100 0.3 N/A
Curve-fitting 4.3 3.7
with validation 9.3 > 100 5.0 N/A
Total error
without validation 4.9 5.1 5.0 5.3

1000 7

6
Simulation - Drag
100
Empirical Fit - Drag 5
Coefficient of Drag

Analytical Result - Drag

Coefficient of Lift
Simulation - Lift 4
10 Analytical Result - Lift
Empirical Fit - Lift
3

2
1

0.1 0
0.1 1 10 100 1000
Sphere Reynolds Number

Fig. 1. Results are shown for coefficient of lift and drag vs. sphere Reynolds number. The solid lines represent empirical fits to the simulation data
and the dashed lines represent the low Reynolds number analytical solutions.

sphere has been examined analytically (Leighton & Acrivos, 1985; O’Neill, 1968) and results for coefficient of lift and
drag have been obtained as given in Eq. (6). The lift coefficient was obtained through a first-order Reynolds number
perturbation to the creeping flow solution for drag force.
24
CD = 1.7009 ,
Resphere
CL = 5.8696. (6)
An important consideration here is to determine whether a Blasius boundary layer reduces at sufficiently low sphere
Reynolds number to the linear shear flow needed for Eq. (6) to be valid. Assuming a series solution in the vertical
distance from the wall, y, and substituting into the Blasius equation, one can obtain the following expressions for the
stream-wise (u) and normal (v) velocity components
u = Uf  (0) + O(4 ),
f  (0)2
v= √ + O(4 ), (7)
4 Rex
134 L.G. Sweeney, W.H. Finlay / Aerosol Science 38 (2007) 131 – 135

Table 2
A comparison of experimentally measured coefficient of lift values (Willets & Naddeh, 1987) to values predicted in the numerical simulation

Reynolds number range CL —Experimenta CL —Simulation

43–100 0.4 ± 0.08 0.67–0.47


83–140 0.05 ± 0.02 0.51–0.39
140–230 0.1 ± 0.01 0.39–0.27

The simulation lift coefficients are calculated using an empirical fit to the simulation data (Eq. (8)).
a Error results are estimated from the experimental data by assuming the resolution of the force measuring device to be its accuracy.


where  = y U/x. The Blasius solution to the boundary layer equations is valid only for high Rex (> 1000), and
consequently, for small y, the vertical velocity is small and the stream-wise component assumes a linear profile. The
analytical relationships in Eq. (6) are thus applicable for model validation at low sphere Reynolds number.
At the lowest sphere Reynolds number considered in the simulation, Resphere = 0.1, the value of coefficient of lift
was calculated to be 5.4017, which is 7.79% lower than the analytical value. However, as seen in Fig. 1, the trend of
the lift coefficient at low Reynolds number is asymptotically approaching the analytical value. The difference between
the analytical and computational value is likely physical, and the computational curve is expected to approach the
analytical value as sphere Reynolds number decreases below 0.1.
The coefficient of drag was calculated to be 409.5 at a sphere Reynolds number of 0.1, and the analytical value is
408.2, which agrees within 0.3%. In addition, over the sphere Reynolds number range 0.1–0.71, the 6 simulation points
were fitted with a power function and the resulting function was 1.71 times the creeping flow solution for uniform flow
past a sphere where no surface is present (24/Resphere ), which is 0.6% higher than the analytical coefficient in Eq. (6).
Thus, the simulation predicts the analytical behavior of the problem at low sphere Reynolds number quite well.
An important result of this work is determining at what sphere Reynolds number the creeping flow analytical results
become inaccurate. For coefficient of lift, the analytical solution is valid up to a sphere Reynolds number of 0.11 at
which point the error is ∼ 10%. For coefficient of drag and a similar 10% error, the analytical solution is valid up to a
sphere Reynolds number of 1.6.
For the geometry under consideration, experimental results exist for the coefficient of lift in the sphere Reynolds
number range 43–230 (Willets & Naddeh, 1987). The values are presented in Table 2 along with a comparison with
computational model predictions. The experimental value for the Reynolds number range 43–100 matches the compu-
tational model well, with an experimentally measured lift coefficient of 0.4 falling just outside the simulation predicted
range of lift coefficients 0.67–0.47 for the stated particle Reynolds number range. The remaining two measurements
predict positive lift forces; however, the magnitudes do not correspond well with the simulation values, with differences
in excess of 100%. The simulation predicts a monotonically decreasing trend in lift coefficient, similar to uniform flow
about a sphere, which the experiment does not. However, limited confidence should be placed in the experimental
results due to the difficulty in measuring such small forces reliably. Indeed, it is unexpected that the experimentally
measured coefficient of lift would reduce by nearly an order of magnitude with such a slight increase in Reynolds
number range. If the experimental verification is included in the error prediction, then the correlation developed for
lift coefficient becomes far less useful, with error greater than 100% at high Reynolds number. We suggest that further
experimental investigation is needed to examine the results of our simulations at higher Reynolds number in the range
considered.
The results are strictly valid at a plate Reynolds number of 32 400. However, the constraint of steady, laminar flow
about the sphere requires that the particle be relatively deep in the boundary layer. As shown in Eq. (6), the velocity
profile becomes linear at distances near the wall, and the coefficient of lift and drag functions, Eq. (1), become solely
functions of particle Reynolds number. Consequently, as particle Reynolds number decreases, the velocity profiles for
various plate Reynolds number collapse into a single, linear, profile. At the highest particle Reynolds number considered
in this research, Resphere = 250, the velocity profiles begin to deviate from linearity at the top of the sphere for various
plate Reynolds number. From this consideration, it is clear that the coefficient of lift and drag results generated in this
work should be weakly dependent on plate Reynolds number, and this dependence should increase as particle Reynolds
number increases. This effect was tested at a particle Reynolds number of Resphere = 250, for which the effect is largest
over the range considered. Coefficient of lift and drag results were calculated at a plate Reynolds number of 500 000,
and the lift and drag results increased by 4.5% and 17.4%, respectively, compared to the results at Rex = 32 400. This is
L.G. Sweeney, W.H. Finlay / Aerosol Science 38 (2007) 131 – 135 135

evidence that the results are indeed weakly dependent on the plate Reynolds number for the particle Reynolds number
range considered, and that for large particle Reynolds number the empirical relationships are only approximately valid
at plate Reynolds numbers different from 32 400.
The results of our simulations were empirically fitted with exponential curves. The method of nonlinear regression
was used in addition to a Gauss–Newton advancement method. The analytical low Reynolds number asymptotic
behavior of both the coefficient of lift and drag was enforced in the curve-fitting process. The curves are presented in
Eqs. (8) and (9) and are shown in Fig. 1.
 
40.812 0.2817
CD = 1− arcsin(0.238Re) , (8)
Re Re0.0826
CL = 5.811 − 4.339 Re0.0429 tanh(0.9395 Re0.3531 − 0.2966) + 0.0589 tanh(−0.1137 Re + 2.5386). (9)

The variable Re in Eqs. (8) and (9) represents the sphere Reynolds number defined in Eq. (2). The equations deviate
from the simulation results by less than 4.3% for the coefficient of drag and 3.7% for the coefficient of lift over the
range Resphere = 0.1–250. Including curve-fit and verification error, the simulation accuracy is shown, as in Table 1, to
be 5.1% for coefficient of lift and 5.3% for coefficient of drag.
In summary, numerical solution of the Navier–Stokes equations for incompressible flow has provided insight into the
forces on a sphere attached to a wall in a fully developed Blasius boundary layer. Convergence testing and validation
against low Reynolds number analytical solutions and high Reynolds number lift measurements provide confidence in
the model results. The coefficients of lift and drag were characterized at a plate Reynolds number of 32 400 and over
the sphere Reynolds number range 0.1–250, and empirical fits, valid in the sphere Reynolds number range 0–250, to
the data were supplied. The coefficient of lift and drag were shown to be only weakly dependent on the plate Reynolds
number. Consequently, the empirical equations can be used at various plate Reynolds numbers without serious loss in
accuracy. The results of this study lack extensive experimental validation; however, design and execution of a verifying
experiment is difficult due to the low magnitude of the forces involved. The empirical relationships developed in this
study can be used to aid in the development of an optimized and sensitive experimental apparatus. The lift and drag
correlations presented here are also anticipated to be useful in predicting resuspension and entrainment of aerosol
particles that have deposited on a flat surface.
The authors gratefully acknowledge support from Alberta Ingenuity and the National Science and Engineering
Research Council of Canada.

References

Dey, S. (1999). Sediment Threshold. Applied Mathematical Modelling, 23.


Finlay, W. H. (2001). The mechanics of inhaled pharmaceutical aerosols. London: Academic Press.
Hall, D. (1998). Measurements of the mean force on a particle near a boundary in turbulent flow. Journal of Fluid Mechanics, 187, 451–466.
Leighton, D., & Acrivos, A. (1985). The lift on a small sphere touching a plane in the presence of a simple shear flow. Journal of Applied Mathematics
and Physics, 36, 174–178.
Mochizuki, M. (1960). Smoke observation on boundary layer transition caused by a spherical roughness element. Journal of the Physical Society of
Japan, 16, 995–1008.
Mollinger, A. M., & Nieuwstadt, F. T. M. (1996). Measurement of the lift on a particle fixed to the wall in the viscous sublayer of a fully developed
turbulent boundary layer. Journal of Fluid Mechanics, 316, 285–306.
O’Neill, M. E. (1968). A sphere in contact with a plane wall in a slow linear shear flow. Chemical Engineering Science, 23, 1293–1297.
Willets, B. B., & Naddeh, K. F. (1987). Measurements of lift on spheres fixed in low Reynolds number flow. Journal of Hydraulic Research, 24,
425–435.
Ziskind, G., Fichman, M., & Gutfinger, C. (1995). Resuspension of particulates from surfaces to turbulent flows—review and analysis. Journal of
Aerosol Science, 26, 613–644.

You might also like