Professional Documents
Culture Documents
1 Introduction
The viscous flow past a two-dimensional bluff body and the industry and has been the focus of several fundamental studies
resulting recirculation region behind it have been the focus of on transition to turbulence and vortex generation.
numerous investigations. These studies have been motivated In this study we investigate the two-dimensional, unsteady
by the desire to understand the fundamental physics of such flow past a bluff body of square cross-section and the effects
flows as well as their practical importance in industry. This on this flow due to an adjacent wall which is either fixed or
study shows that, as in the case of a fixed wall, the presence sliding at the bulk fluid velocity. A sketch of the three cases
of a sliding wall in contact with a bluff body has a pronounced of interest is given in Fig. 1.
influence on the stability of the resulting flow and its dynamical
characteristics. 1.1 Previous Work. The literature relevant to the present
From the practical point of view, the flow past a bluff body study can be broadly divided into two groups: one concerning
near a sliding wall has been of interest for several years in the bluff-body flows far from any bounding walls and one con-
automotive industry, particularly to reduce drag for more fuel- cerning flows past bluff bodies near a plane wall. The latter
efficient vehicles. More recently, a strong interest in such flows group can be further subdivided depending on whether the
has arisen in relation to magnetic disc-storage devices in the plane wall is fixed or sliding with respect to the bluff body.
computer-peripherals industry. The flow past a bluff body For those studies focussing on bluff-body flows far from
fixed on a solid wall is also of interest to the electronics cooling any solid wall boundaries the unsteady features of the vortex-
shedding process are of particular interest. The review by Ber-
ger and Wille (1972) provides much of the information and
background for the current understanding of such flows. Sev-
Contributed by the Fluids Engineering Division for publication in the JOURNAL
eral additional detailed studies of the vortex-shedding process
OF FLUIDS ENGINEERING. Manuscript received by the Fluids Engineering Division behind a bluff body have appeared since the publication of
July 31, 1989. that review.
du da dP <d2u d2u\ Because of the elliptic nature of the flow fields investigated,
(uu) + —(uv)--
+ + -Li (2) the outlet boundary condition will have some influence on the
Jt dx dy -Tx yd^dy2)
Re ' development of the upstream flow. Two alternative conditions
dv d d dp <d2v d2v\ were used at the outlet plane in the present study. The first
(uv) + —(vv)-- (3) was to set the normal gradient of all dependent variables at
'dx dy Re \
the outlet plane equal to a constant ( ^ 0 ) . This is equivalent
where the following dimensionless variables are defined: to requiring that the second derivative of any dependent vari-
X _tlh able be zero and results in less artificial smoothing than the
X fully-developed flow condition. Using this condition the nu-
H -£•'-' H
merical procedure was stable and convergent for those flows
u V P UbH which eventually reached a steady-state result. However, for
'u =Tf P = ?> Rt
V the periodic flows which are of primary interest here, this
b
condition often led to instabilities in the numerical procedure
The tilde denotes a dimensional variable and the u and v ve-
and yielded diverging results. In addition, at the outlet plane
locity components correspond to the x- and j-directions of the
it also caused distortion of vortices generated by the obstacle
Cartesian coordinate system, shown in Fig. 1. The kinematic
in the flow domain.
viscosity of the fluid is denoted by v, and the dynamic pressure
by P. The various terms in the equations are nondimension- For these reasons, the open boundary condition described
alized using the fluid density, p, the height of the square rib, by Han et al. (1983) was used for the unsteady calculations.
H, and Ub, the uniform bulk velocity upstream of the rib. The This condition allows the shear-induced flow structures created
time, f has been normalized with the convective time scale, in the flow domain to pass through the exit plane without
T=H/Ub. Relative to the viscous time scale, T„ = H2/v, the reflecting waves at that boundary. The open-boundary con-
convective time scale was always at least two orders of mag- dition is based on the Sommerfeld radiation condition:
nitude smaller for the flows of interest here (Re> 100).
^ + Ao (5)
2.2 The Boundary Conditions. The recirculating viscous dt dx
flow past a bluff body poses an elliptic problem requiring where <j> is any dependent variable and c is the local wave speed
boundary conditions at all the boundaries of the calculation of the <j> wave (c>0), calculated from the values of 4> imme-
domain. In addition, since the flows of concern are known to diately ahead of the exit boundary. Preliminary tests were
be time dependent beyond a critical Reynolds number, an initial conducted to determine the influence of the outlet condition
condition is required. One of the following two initial con- on the flow field development. The same flow (Re = 250,
ditions was employed in the present calculations: (i) a uniform sliding-wall case) was calculated twice with the outlet plane
velocity field with an impulsive start at t = 0; or, (ii) the results located 15i/and 25H downstream of the rib, respectively. The
of a previous calculation at a different Reynolds number, which results showed that the outlet condition did not influence time-
often shortened the time required to reach a steady state or dependent quantities, such as the vortex-shedding frequency,
steady periodic flow behavior. or the spatial distribution of waves passing through the outlet
plane. (The little distortion that did arise resulted from ignoring
At the inlet plane of the flow domain, located five rib-heights
diffusive effects in this region, but it did not extend more than
(5H) upstream of the rib, the velocity was prescribed for all
one rib height upstream of the exit plane.)
three cases studied. This choice of entry length was fixed by
the fact that larger entry lengths did not significantly change
the character of the flow a distance 5H upstream of the rib. 3 Numerical Procedure
For all the results presented here, a uniform-flow profile was
The calculation procedure used in this study is essentially
fixed at the inlet:
the primitive variable formulation described in LeQuere et al.
u=Ub, v=0 (4a) (1981), modified as explained below. Following Harlow and
Welch (1965), a staggered-grid system is employed in which
T h e b o u n d a r y conditions used at a solid wall surface were the main grid nodes are the storage locations for pressure.
those of impermeability and no-slip: Velocity components are stored at points midway between the
v±=0, v, = vmn (4b) main grid nodes. The finite difference forms of the governing
equations are obtained by integrating the partial differential
where v± and vi are the velocities perpendicular and parallel equations over elemental rectangular control volumes or cells.
to the wall, respectively, and uWaii is the speed at which the The resulting set of algebraic difference equations is then solved
wall slides along the x axis in its own plane. These conditions iteratively, using the modified, strongly-implicit (MSI) pro-
apply to the wall adjacent to the rib in the wall-bounded flows, cedure of Schneider and Zedan (1981). A single iteration con-
as well as to the surfaces of the rib itself. vwaH is set to zero sists of the sequential solution of: (i) a Poisson-type equation
on the surfaces of the rib for all cases considered. For the for pressure, using the velocity field from the previous itera-
sliding-wall case t>wan = Ub at the lower bounding wall. For tion; (ii) the u,v momentum equations using the new pressure
the fixed-wall case vwaii = 0 at the lower b o u n d i n g wall down- field; and (iii) the pressure correction (continuity) equation,
stream of the rib and yWaii = 0 or Ub at this wall upstream of used to update the velocity field so that continuity is satisfied.
the rib. Whether the lower wall b o u n d a r y condition upstream This iterative scheme goes by the acronym SIMPLER (SIM-
of the rib is sliding or fixed has a significant impact on the PLE-Revised) and is due to Patankar (1981). For the flows of
resulting flow downstream of the rib. F u r t h e r discussion of interest to this study, the method for determining the pressure
this point is given in the results. field directly in the SIMPLER algorithm has represented a
For the freestream case, zero shear boundaries were specified significant improvement over SIMPLE in terms of both com-
6 rib heights above and below the rib centerline. F o r the wall- putational cost and numerical stability. This is because the
b o u n d e d flows the same condition was specified along the effectiveness of the SIMPLER solution procedure does not
flow (e.g., ywan = Ub), whereas in the other case (Case 2B)
the upstream wall was fixed. At the lowest Reynolds number
investigated (Re = 100), a steady solution was obtained for both
geometries by marching in time until the transient behavior
died away. Following this development period, no changes in
the flow field variables were observed. Figures 5(a) and 5(b)
0.3 0.4 0.5
show streamline plots of the results for Case 2A and 2B, re-
Strouhal Number spectively. In Fig. 5(a), the flow field is dominated by a single
Fig. 3(b) large eddy behind the rib, which is 30.8 rib heights in length.
Fig 3 Case 1: Flow past a rib in a freestream. Spectra for lift and drag
Figure 5(b) reveals a much smaller recirculation region for
-'--'— ,(a)
time histories, J
"Re
- 500, (b) Re = 1000. Case 2B. In comparing the two flow fields, it is also apparent
that the upstream boundary layer separation and resulting
wavelength of three shedding cycles. Another interesting fea- recirculation bubble upstream of the rib in Case 2B is absent
ture is that the behavior of the time records starting at t = 20 from the flow field in Case 2A. The absence of the upstream
appears to be repeated starting at t = 165. The correspondence separation in Case 2A results in increased upward momentum
is not exact, and from the present data it is not possible to say of the fluid as it travels over the rib and induces the larger
whether the entire history between t = 20 and 165 will be downstream recirculation shown in Fig. 5(a). An additional,
repeated or whether similarity will break down. Of course, in very small, counter-clockwise rotating eddy was also predicted
a real flow situation the vortex shedding process may be af- near the downstream corner of the rib for both cases. This
fected by three-dimensionality at this Reynolds number as well eddy is induced by the larger one which forms the main re-
as free-stream turbulence which might be present. Neither of circulation zone behind the rib.
these effects can be reproduced in the present two-dimensional At Re = 500 calculations were again performed with both
calculations. sliding and fixed walls upstream of the rib. The steady solutions
with Re = 100 were used as initial conditions. Marching in time
4.2 Case 2. Rib on a Fixed Wall. The geometry which did not produce steady solutions for either case at this Reynolds
corresponds to this case is shown in Fig. 1(b). All results pre- number. Instead, in both cases a time-varying flow developed
sented here were obtained with the wall fixed downstream of consisting of structures periodically shed from behind the rib
the rib (e.g., uwan = 0). Upstream of the rib, two different and which moved downstream. While this was true for both
boundary conditions were used at the wall. In one case (Case Case 2A and 2B, further examination of the data revealed that
2A) the upstream wall was allowed to slide with the incoming the unsteadiness was very weak for Case 2B and only minor
Fig. 5(b) Case 2B: Flow past a rib on a wall which is fixed upstream
and downstream. Streamlines for steady flow at Re = 100.
Fig. 7 Case 2: Flow past a rib on a fixed wall. Drag coefficient and
cross-stream velocity fluctuations, Re = 500.
>
Fig. 9(6) Case 3: Flow past a rib on a sliding wall. Vorticity contours
for quasi-steady flow at Re = 100
0.7-
0.5-
0.3-
0.1-
• 1
0.9-
Fig. 10 Case 3: Flow past a rib on a sliding wall. Instantaneous vorticity
contours for the unsteady flow case, Re = 500
0.7-
0.5-
0.3-
0.1-
1 , .
-0.1-
0.2 0.3 0.4
Strouhal Number
Fig. 12 Case 3: Flow past a rib on a sliding wall. Spectra for lift and
drag time histories, Re = 500.
between the flow past a rib near a sliding wall and the flow
0.05-
past an obstacle in a freestream at an angle of attack. The
latter flow geometry was one of the cases considered in the 0 Freestream
study by Davis and Moore (1982). In both flows the front D Fixed wall
stagnation region on the obstacle moves downward off center, A Sliding wall
while the rear stagnation region moves upwards. This is because o.oo-l , 1 1 1
more of the fluid impinging on the front is forced over the 0.0 250.0 500.0 750.0 1000.0
top of the obstacle. The rib near a sliding wall is in fact a Re
Fig. 16 Strouhal number variation with flow Reynolds number for the
limiting case, where all the fluid is forced over the top. This three flow configurations under consideration
results in a pair of vortices forming downstream which do not
behave in a symmetrical way. In both cases the behavior of Strouhal number, is plotted as a function of Reynolds number
the lift and drag coefficients is modified accordingly, to ac- for the three cases in Fig. 16. The frequency used to determine
count for the unequal behavior of the two vortices. The shed- the Strouhal number is the frequency at which a pair of vortices
ding of the negative vortex leads to a small local minimum in is shed from behind the rib. In all three cases, the Strouhal
the drag coefficient while the shedding of the positive vortex number is only a weak function of the Reynolds number. This
leads to a large decrease in the drag. Further distinctions be- relative insensitivity of the shedding frequency to Reynolds
tween the freestream and the wall-bounded flows are discussed number has been observed in other studies and holds over a
in the next section. large Reynolds number range extending well into the turbulent
The mean drag coefficient for the three cases is plotted as flow regime; see, for example, Durao et al. (1988) and Davis
a function of Reynolds number in Fig. 15. In all three cases, and Moore (1982). Although the Strouhal number varies only
the mean CD increases with increasing Re over the Reynolds slightly with Reynolds number, it is quite different for Case
number range studied. The fixed-wall case (Case 2) shows the 1, where the rib is not adjacent to a wall. The wall-bounded
smallest dependence on Reynolds number variations. The mean flows have Strouhal numbers approximately 50 percent lower
drag in Case 1 (freestream) was always less than the wall- than the values for the freestream case where vortex shedding
bounded flows of Cases 2 and 3. At the lowest Reynolds num- occurs. This distinction is discussed in the following section
ber (Re = 100), the drag is approximately the same for Case 2 along with some questions regarding the stability of bluff body
and 3. This indicates that for the wall-bounded flows, where flows and the influence of the neighboring wall on such flows.
no periodic vortex shedding was observed, the sliding wall has
no significant influence on the drag. For higher values of Re, 5 Discussion
where shedding does occur, the effect of the sliding wall is to Although it was not the intention of this study to accurately
increase the mean drag as well as the amplitude of the fluc- determine the conditions marking transition from steady to
tuations in CD. For example, at Re = 500, for Case 2 the range unsteady flow for the three cases investigated, some comments
of the drag coefficient was 2.04<CD<2.69 while for the slid- can be made regarding their relative stability. In this section
ing-wall case (Case 3) the range was 1.58<C£)<3.78. we discuss this point and various issues raised in the previous
The variation in the nondimensional shedding frequency, or section.
/
\ u
\y _ 1 d u u
profiles in Cases 1 and 3 are possibly unstable. Case 2, which
has no inflection points in the profile, should be stable to
inviscid perturbations. It should be noted that downstream of
Frecstream Fixed Wall • Sliding Wall the recirculation region in Case 2 the velocity profile develops
quickly into a boundary-layer-like flow, and it is well known
Fig. 17 Sketches of the streamwise velocity profiles downstream of
the rib for the three cases under consideration, (a) Case 1, rib in a that the Blasius boundary layer is stable to inviscid disturb-
freestream; (b) Case 2, rib on a fixed wall; (c) Case 3, rib on a sliding ances. The similarity between the profiles of Cases 1 and 3 is
wall. more noticeable if the freestream profile below the symmetry
plane is included for comparison as drawn in the figure.
We first summarize the results of the previous section re- Thus, based on careful observation and appealing to inviscid
garding the stability of the three flows. For Case 1, the rib in stability theory, we conclude that the flow past a rib near a
a freestream, all three Reynolds numbers (Re = 100, 500, 1000) fixed wall (Case 2) is the most stable of the three flow con-
yielded unsteady solutions in which vortex shedding occurs. It figurations. This conclusion is supported by the findings of
is well known that the transition from steady recirculating flow the present study which determined unsteady solutions for the
to a flow with periodic vortex shedding occurs at approximately freestream (Case 1) and sliding-wall (Case 3) configurations
Re = 50 for a bluff body in a freestream. For Case 2, the rib at lower Reynolds numbers than for the fixed-wall geometry.
on a fixed wall, a steady state solution was obtained at Re = 100 We turn now to a comparison of the freestream and sliding-
while at the higher Reynolds numbers (Re = 500,1000) periodic wall flows and note that in the freestream case the transition
vortex shedding occurred. For Case 3, the rib on a sliding wall, Reynolds number is about half that for the sliding-wall case.
a steady solution was obtained at a Reynolds number of Re = 50. In comparing the two flows when they are steady, there are
For the flows discussed in the present study, all three Reynolds two features which are distinctly different. First, the stagnation
numbers yielded unsteady solutions. However, we described point on the upstream side of the rib is close to the wall for
the flow at Re = 100 for Case 3 as quasi-steady since, although Case 3 since all the fluid has to pass over the top of the rib.
waviness occurred in the wake, no vortex shedding developed. In Case 1 the stagnation point is close to the midpoint of the
At higher Reynolds numbers (Re > 250) the flow was unsteady rib and only half of the fluid impinging on the rib passes over
and periodic with regular vortex shedding. From this brief the top, the remaining half of the fluid passing underneath.
summary it is apparent that the transition to a periodic shed- The second feature distinguishing the two flows is the dif-
ding flow occurs at the lowest Reynolds number (Re = 50) for ference in the conditions governing the locations of their re-
Case 1; at an intermediate Reynolds number (50 < Re < 100) spective stagnation points upstream of the rib. When the flow
for Case 3 and at the highest Reynolds number (Re> 100) for goes unsteady in the sliding-wall case, the position of the stag-
Case 2. We now attempt to understand why the freestream nation point remains fixed at the lower upstream corner of the
and sliding-wall cases are more unstable than the fixed-wall rib. For the freestream flow the stagnation point is not con-
flow. strained to remain on the geometric symmetry plane but is free
Consider the steady solutions which develop at low Reynolds to move. In the freestream flow the velocity normal to the
numbers for all three cases. Intuitively, one would expect a geometrical symmetry plane (v-component) is not constrained
sliding wall near a bluff body to be a destabilizing influence to be zero. Indeed, after vortex shedding develops, the cross-
on the wake region. In this configuration a mechanism is avail- stream velocity is often not zero. However, for the sliding-
able for removing relatively stagnant fluid from behind the wall flow, the cross-stream velocity is constrained to be zero
rib. This mechanism is not available when the wall is fixed. at the stagnation point and all along the wall upstream of the
The moving wall also provides a mechanism for the generation rib and downstream as well. In contrast, note that the stream-
of a strong counter-rotating (positive) vortex. This vortex in- wise velocity conditions on the symmetry plane in Case 1 and
teracts with the negative vortex which develops as a result of on the sliding wall in Case 3 are not so different. In the free-
forcing the oncoming flow over the rib. The structure of the stream case the streamwise velocity is not constrained on the
flow behind the rib in Case 3 for which the wall is sliding, is symmetry plane, while in the sliding-wall case it is forced to
very similar to the flow behind the rib in a freestream. In both acquire the wall velocity by the no-slip condition.
cases, two vortices of opposite sign and roughly equal strength It seems reasonable to assume that the freestream flow be-
are present behind the rib. Also, in both cases, the absence of comes unstable at a lower Reynolds number than in the sliding-
a fixed wall downstream of the rib makes flow reattachment wall geometry because it is susceptible to perturbations which
impossible. Instead, the zero velocity point is free to move in break the symmetry of the steady flow. This allows non-zero
the wake of the rib. cross-stream velocities on the symmetry plane and permits the
This argument, based on observation, can be strengthened upstream stagnation point to oscillate about a mean position.
by referring to Fig. 17 which presents sketches of the stream- Similar perturbations are not permitted in the region near the
wise velocity profiles downstream of the rib for the three cases, upstream stagnation point in the sliding-wall flow because the
when the flow is steady. We evaluate these profiles in light of cross-stream velocities are forced to zero by the boundary
the inviscid stability of nearly parallel flows. According to conditions imposed at the wall. As a result, the stagnation
Rayleigh's point-of-inflection theorem: point is not free to oscillate about a mean position but is,
instead, constrained to remain fixed.
A necessary (but not sufficient) condition for inviscid in-
stability is that the basic velocity profile has a point of in- .Examination of the time histories presented in the previous
flection. section indicates that the sliding wall of Case 3 also has a
A related theorem which gives a slightly stronger criterion substantial influence on the type of unsteady behavior which
for stability is Fjortoft's theorem: is observed. For this case the unsteady simulations always
If y0 is the position of a point of inflection in the basic resulted in a short transient followed by a perfectly periodic
profile, u(y),and u0 = u(y0), then a necessary but not suf- shedding cycle. This was true not only at Re = 500 and 1000,
ficient condition for inviscid instability is that but also at Re = 1500 and 2000, for which we have not presented
time histories. At Re = 500 the calculation was started impul-
d2u sively and fully periodic shedding was observed at approxi-
-rj(u-u0)<0
mately 80 time scales, (Fig. 11). The initial condition for the
Contents (continued)
Technical Briefs
509 Oscillatory Fluid Flow Through a Porous Medium Channel Bounded by Two
Impermeable Parallel Plates
J. M. Khodadadi
511 Visualization of the Wing-Tip Vortex in Temporal and Spatial Pressure Gradients
X. Liang and B. R. Ramaprian