FIGURE 10-73
‘Nondimensional axial velocity (bine
curve) and pressure coefficient (green
curve) along the floor below a vacuum
leaner modeled as an irrotational
region of flow.
‘Masia!
FIGURE 10-74
Based on an irrotational flow
approximation, the maximum speed
along the floor beneath a vacuum
cleaner nozzle occurs at x= tb. A
stagnation point occurs directly below
the nozzle.
aaoo 1 2 3
[Normalized disiencealeag the Boor, x*
a
Ee eee ae
a tatenis)
Maximson speed along the floor.
‘We expect that the vacuum cleaner is most effective at sucking up ditt from
the floce when the speed along the floor is greatest and the pressive along, the
floor is lowest. Thus, contrary to what you may have thought, the best perfor
nance is wot directly below the suction ints, but rather at x= b, as illustrated
in Fig. 10-75.
Discussion Notice that we never used the wid of the vacuum nozzle in our
analysis, since a line sink kas no length scale. You can convince yourself that a
‘vacuum cleaner works best at x a 4b by performing a simple experiment with a
‘acum cleaner and some small granular material (ike sugar or salt) om a aed
floor. It wus oot that tbe irotational approximation is quite realise for flow into
the inlet ofa vacuum cleaver everywhere except very close (othe floor, because the
flow is rotational there.
250m/s (16)
‘We conclude this section by emphasizing that although the irrotational
flow approximation is mathematically simple, and velocity and pressure
fields are easy to obtain, we must be very careful where we apply it, The
irrotational flow approximation breaks down in regions of non-negligible
vorticity, especially near solid walls, where fluid particles rotate because of
viscous stresses caused by the no-slip condition at the wall. This leads us
to the final section in this chapter (Section 10-6) in which we discuss the
boundary layer approximation,
40-6 = THE BOUNDARY LAYER
APPROXIMATION
‘As discussed in Sections 10-4 and 10-5, there are at least two flow situ-
ations in which the viscous term in the Navier-Stokes equation can be
neglected. The first occurs in high Reynolds number regions of flow where
net viscous forces are known to be negligible compared to inertial and/orpressure forces; we call these inviscid regions of flow. The second situation
occurs when the vorticity is negligibly small; we call these irrotational or
potential regions of flow. In cither case, removal of the viscous terms from
the Navier-Stokes equation yields the Euler equation (Eg. 10-13 and also
Eq, 10-25). While the math is greatly simplified by dropping the viscous
terms, there are some serious deficiencies associated with application of the
Euler equation to practical engineering flow problems. High on the list of
deficiencies is the inability to specify the no-slip condition at solid walls.
This leads to unphysical results such as zero viscous shear forces on solid
walls and zero aerodynamic drag on bodies immersed in a free stream, We
can therefore think of the Euler equation and the Navier-Stokes equation as
two mountains separated by a huge chasm (Fig. 10-75a). We make the fol-
lowing statement about the boundary layer approximation:
The boundary layer approximation bridges the gap between the Euler
equation and the Navier-Stokes equation, and between the slip condition
and the no-slip condition at solid walls (Fig. 10-756).
From a historical perspective, by the mid-1800s, the Navier-Stokes
equation was known, but couldn't be solved except for flows of very sim
ple geometries. Meanwhile, mathematicians were able to obtain beautiful
analytical solutions of the Euler equation and of the potential flow equa-
tions for flows of complex geometry, but their results were often physi-
cally meaningless. Hence, the only reliable way to study fluid flows was
empirically, ic., with experiments. A major breakthrough in fluid mechan-
ies occurred in 1904 when Ludwig Prandtl (1875-1953) introduced the
boundary layer approximation. Prandtl’s idea was to divide the flow into
two regions: an outer flow region that is inviscid and/or irrotational, and
an inner flow region called a boundary layer—a very thin region of flow
near a solid wall where viscous forces and rotationality cannot be ignored
(Fig. 10-76). In the outer flow region, we use the continuity and Euler
equations to obtain the outer flow velocity field, and the Bernoulli equa-
tion to obtain the pressure field. Alternatively, if the outer flow region is
irrotational, we may use the potential flow techniques discussed in Section
10-5 (eg., superposition) to obtain the outer flow velocity field. In either
case, we solve for the outer flow region first, and then fit in a thin boundary
layer in regions where rotationality and viscous forces cannot be neglected.
Within the boundary layer we solve the boundary layer equations, to be
discussed shortly. (Note that the boundary layer equations are themselves
approximations of the full Navier-Stokes equation, as we will see.)
The boundary layer approximation corrects some of the major deficien-
cies of the Euler equation by providing a way to enforce the no-slip coi
dition at sofid walls. Hence, viscous shear forces can exist along walls, bodies
immersed in a free stream can experience aerodynamic drag, and flow
separation in regions of adverse pressure gradient can be predicted more
accurately. The boundary layer concept therefore became the workhorse of
engineering fluid mechanics throughout most of the 1900s. However, the
advent of fast, inexpensive computers and computational fluid dynamics
(CFD) software in the latter part of the twentieth century enabled numeri-
cal solution of the Navier-Stokes equation for flows of complex geome-
try. Today, therefore, it is no longer necessary to split the flow into outer
flow regions and boundary layer regions—we can use CFD to solve the
559
Eu hp :.
FIGURE 10-75
(a) A lmge gap exists between the
Baler equation (which allows slip at
‘walls) and the Navier-Stokes equation
(which supports the no-slip condition);
(8) the boundary layer approximation
bridges that gap.
Outer How (inviscid andlor
inouational region of flow)
Boundary layer (tational with,
on-negligitle vizcoue force)
WA
FIGURE 10-76
Prandil’s boundary layer concept
splits the flow into an outer flow
region and a thin boundary layer
region (not to scale.560
zee
Flow of a uniform stream parallel
to a flat plate (deavings not to scale).
(a)Re, ~ 10, (6) Re, ~ 10". The,
larger the Reynolds nitober, the thinner
the boundary layer along the plate st a
given x-location
FIGURE 10-78
Flow visualization of a laminar
fat plete boundary layer profile,
Photograph taken by F. X. Wortmann
in 1953 as visvelized with the tellurionn
method. Flow is from left to right, snd
the leading edge of the flat plate is far
to the left of the field of view,
Wormnain PX, 1077 AGARD Cont, Prono
mapper 12
full set of equations of motion (continuity plus Navier-Stokes) throughout
the whole flow field. Nevertheless, boundary layer theory is still useful in
some engineering applications, since it takes much less time to arrive at a
solution. In addition, there is a lot we can learn about the behavior of flow-
ing fluids by studying boundary layers. We siress again that boundary layer
solutions are only approximations of full Navier-Stokes solutions, and we
must be careful where we apply this ot any approximation,
The key to successful application of the boundary layer approximation is
the assumption that the boundary layer is very thin, The classic example is a
uniform stream flowing parallel to a long flat plate aligned with the x-axis.
Boundary layer thickness 6 at some location z along the plate is sketched
in Fig. 16-77. By convention, 6 is usually defined as the distance away from
the wall at which the velocity component parallel to the wall is 99 percent
of the fluid speed outside the boundary layer. It turns out that for a given
fluid and plate, the higher the free-stream speed V, the thinner the boundary
layer (Fig, 10-77), In noncimensional terms, we define the Reynolds number
based on distance x along the wall,
Reynolds number along a flat plate: (10-60)
Hence,
‘At s given x-location, the higher the Reynolds number the thinner the
boundary layer
Tn other words, the higher the Reynolds number, all else being equal, the more
reliable the boundary layer epproximation. We are confident that the bound-
ary layer is thin when 6 Re, .). Itis also com-
‘mon in heat transfer to use this value as the critical Re; in fact, relations
for average friction and heat transfer coefficients are derived by assuming
the flow to be laminar for Re, lower than Re, .«, and turbulent otherwise.
‘The logic here is to ignore transition by treating the first part of transition
as laminar and the remaining part as turbulent. We follow this convention
throughout the rest of the book unless noted otherwise.
The transition process is unsteady as well and is difficult to predict, even
with modern CFD codes. In some cases, engineers install rough sandpaper
or wires called trip wires along the surface, in order to force transition at a
desired location (Fig. 10-83). The eddies from the trip wire cause enhanced
local mixing and create disturbances that very quickly lead to a turbulent
boundary layer. Again, the vertical scale in Fig. 10-83 is greatly exaggerated
for illustrative purposes.SS a eT Se
An aluminum canoe moves horizontally along the surface of a lake at 3.5 mifh
Fig. 10-84). The temperature of the lake water is 50°F. The bottom of the
1x cance is 20 ft Tong and is flat, Is the boundary layer on the canoe bottom laminar
pr turbulent?
SOLUTION We are to assess whether the boundary layer on the bottom of a
canoe is laminar or turbulent.
Assumptions 1 The flow is steady and incompressible. 2 Ridges, dings, and
ther nononiformities in the bottom of the canoe are ignored—the bottom is
assumed to be « smooth flat plate aligned exactly with the direction of flow. 3 From
the frame of reference of the canoe, the water below the boundary layer under the
feanoe moves at uniforin speed V = 3.5 map,
Properties The kinematic viscosity of water at T= SOF is v= 1407 x 107
i,
Analysis First, we caleulate the Reynolds number at the stern of the cance,
Vx __G.5 mifh}20 fy) (= 8) th )
Fy 1407 x 105 RG Imi /\36008,
Since Re, is much greater than Re, (5 x 10°), and is even greater than
Re, iin (50 10°), the boundary layer is defintely turbulent by the back of
the canoe,
Discussion Since the canoe bottom is nether pecfeetly senacth nor perfeotly
fist, and since we expect some disturbances in the lake water due to waves, the
paddles, swimming fish, ete, transition (0 turbulence is expected to occur much
eaulier and more rapidly than illustrated forthe ideal case in Big. 10-81. Hence we
ae even more confident that this boundary layer is turbulent.
= 7.30 x 10°
The Boundary Layer Equations
Now that we have a physical feel for boundary layers, we need the
equations of motion to be used in boundary layer calculations—the
boundary layer equations, For simplicity we consider only steady, two-
dimensional flow in the x-plane in Cartesian coordinates. The methodology
used here can be extended, however, to axisymmetric boundary layers or
to three-dimensional boundary layers in any coordinate system. We neglect
gravity since we are not dealing with free surfaces or with buoyancy- 6, the pressure
gradient term is orders of magnitude greater than the advective terms on
the left side of the equation. Thus, the only term left in Eq. 10-64 is the
pressure term, Since no other term in the equation can balance that term,
we have no choice but to set it equal to zero. Thus, the nondimensional
j-momientum equation reduces to
apt
ay
or, in terms of the physical variables,
ter Bo
Bound yer SS ow
Soma pesan gio srghatomdaytons ad yy
In words, although pressure may vary along the wall (in the x-direction),
there is negligible change in pressure in the direction normal to the wall.
This is illustrated in Fig. 10-88, At x = x,, P= P, at all values of y across
the boundary layer from the wall to the outer flow. At some other location,
X= xy the pressure may have changed, but P = P, at all values of y across
that portion of the boundary layer
The pressure actoss @ boundary layer (y-direction) is nearly constant.
FIGURE 10-88
Pressure may change along a
boundary layer (x-direction), but the
Physically, because the boundary layer is so thin, streamfines within the change in pressure across a boundary
boundary layer have negligible curvature when observed at the scale of the layer direction) is negligible.
boundary layer thickness. Curved streamlines require a centripetal accelera-
sion, which comes from a pressure gradient along the radius of curvature.
Since the streamlines are not significantly curved in a thin boundary layer,
there is no significant pressure gradient across the boundary layer.566
‘Outer flow, \
Boundary ayer
FIGURE 10-89
‘The pressure in the irrotational region
of flow outside of a boundary layer
ccan be measured by static pressure
taps in the surface of the wall. Two
such pressure taps ae sketched.
7 ta)=v
=| wer
ae
FIGURE 10-90
‘An order-of-magnitude analysis of
the laminar boundary layer equations
along a flat plate reveals that 5 grows
like x (not to scale).
One immediate consequence of Eq. 10-65 and the statement just presented
is that at any x-location along the wall, the pressure at the outer edge of the
boundary layer (y & 6) is the same as that at the wall (y = 0). This leads to
2 tremendous practical applicstion; namely, the pressure at the outer edge
of a boundary layer can he measured experimentally by a static pressure tap
at the wall directly beneath the boundary layer (Fig. 10-89), Experimental-
ists routinely take advantage of this fortunsie situation, and countless airfoil
shapes for airplane wings and turbomachinery blades were tested with such
pressure taps over the past century.
‘The experimental pressure data shown in Fig. 10-64 for flow over a cir
cular cylinder were measured with pressure taps at the cylinder’s surface,
yet they are used to compare with the pressure calculated by the irrotational
‘outer flow approximation. Such a comparison is valid, because the pressure
obtained outside of the boundary layer (from the Euler equation or poten-
tial flow analysis coupled with the Bernoulli equation) applies all the way
through the boundary layer to the wall.
Returning to the development of the boundary layer equations, we use
Eq, 10-65 to greatly simplify the x-component of the momentum equation.
Specifically, since P is not a function of y, we replace dP/ax by dPldx,
where P is the value of pressure calculated from our outer flow approxima-
tion (using either continuity plus Euler, or the potential flow equations plus
Bernoulli). The x-component of the Navier-Stokes equation becomes
au ou Lap a Pe
gatz a t Ya t “ey
ew “Dosey 13 Pw
After some algebra, and after multiplying each term by L/L, we get
oH yd ae (ode ey out
wet Senn aat (ar) amt lar)ls) je
Comparing terms in Eq. 10-66, the middle term on the right side is clearly
orders of magnitude smaller than the terms on the left side, since Rez =
ULiv > 1. What about the last term on the right? If we neglect this term,
we throw out all the viscous terms and are back to the Euler equatio
Clearly this term must remain. Furthermore, since all the remaining terms in
Eq. 10-66 are of order unity, the combination of parameters in parentheses
in the last term on the right side of Eq. 10-66 must also be of order unity,
iy
“Again recognizing that Re, = U/L/v, we see immediately that
61
EO VRE,
This confirms our previous statement that at a given streamwise location
along the wall, the larger the Reynolds number, the thinner the boundary
layer. If we substitute x for Z, in Fg. 10-67, we also conclude that for a lami-
nar boundary layer on a flat plate, where U(x) constant, 5 grows like
the square root of x (Fig. 10-90).
(10-67)In terms of the original (physical) variables, Eq, 10-66 is written as
Se ee
at ay (10-68)
Note that the last term in Eq. 10-68 is not negligible in the boundary layer,
since the y-derivative of velocity gradient du/dy is sufficiently large to
offset the (typically small) value of kinematic viscosity v. Finally, since we
Know from our y-momentum equation analysis that the pressure across the
boundary layer is the same as that outside the boundary layer (Eq, 10-65),
wwe apply the Bernoulli equation to the outer flow region. Differentiating
with respect to x we get
Phere constant + LP ey (0.
poe = pa =)
where we note that both P and U are functions of x only, as illustrated in
Fig. 10-91. Substitution of Eq. 10-69 into Eq. 10-68 yields
ou ae, aU Ou
wet ay Uae tos
(10-70)
and we have eliminated pressure from the boundary layer equations.
We summarize the set of equations of motion for a steady, incompress-
‘ble, laminar boundary layer in the xj-plane without significant gravitational
effects,
ae, de
ax * oy
eg ye yy SH
ae ay ae 7 OF
0
Boundary layer equations: (10-79
Mathematically, the full Navier-Stokes equation is elliptic in space, which
means that boundary conditions are required over the entire boundary of the
flow domain. Physically, flow information is passed in all directions, both
upstream and downstream. On the other hand, the a-momentam boundary
layer equation (the second equation of Eq. 10-71) is parabolie, This means
that we need to specify boundary conditions on oaly three sides of the (two-
dimensional) flow domain. Physically, flow information is not passed in the
direction opposite to the flow (from downstream). This fact greatly reduces
the level of difficulty in solving the boundary layer equations. Specifically,
we don't need to specify boundary conditions downstream, only upstream
and on the top and bottom of the flow domain (Fig. 10-92). For a typical
boundary layer problem slong a wall, we specify the no-slip condition at the
wall (w= 0 = 0 at y = 0), the outer flow condition st the edge of the bound-
ary layer and beyond [u = U(x) as y + oo}, and a starting profile at some
upstream location [1 = tjaing()) At X= Spies WHEE paring MAY OF MAY
not be zero]. With these boundary conditions, we simply march downstream
in the x-direction, solving the boundary layer equations as we go. This is
particularly attractive for numerical boundary layer computations, because
‘once we know the profile at one x-location (:}, we can march to the next
x-location (x;,,), and then use this newly calculated profile as the starting
profile to march to the next x-location (2;,.), ete
367
P=PX,U=00)
ay Y,
Oo) Pay ae
Pas
FIGURE 10-91
Outer flow speed parallel to the wall
is UG) and is obtained from the outer
flow pressure, P(x). This speed appears
in the x-component of the boundary
layer momentum equation, Eq. 10-70.
[No bousdary conditions on
dowasteat edge of fw dotmain
FIGURE 10-92
‘The boundary layer equation set is par-
abolic, 50 boundary conditions need to
be specified on only three sides of the
flow domain.