Dan Ruggirello wrote:
Paul,
Profile R5 is what I ran last night. I doubled the radiator vanes
and I
thought I had an open
area of 1/3 but after I ran the model I rechecked it and found that
I
mistakenly made it 25%. I
found that according to the Stewart website
(http://www.stewartcomponents.com/html/tech_support/advanced_cooling_system_
basics.asp) radiators
have typically 1/3 open area. Not sure if this is true for all
radiators? This model took about
5.5 hours to run and has ~41k nodes.
R6 is the profile you sent to run. The velocities through the duct
seem
greater than previous
runs, perhaps due to a relatively large open area (89% open area).
In
general it looks like
modifying the top of the wedge with a slight downward curve as you
did
seems to push the velocity
distribution towards the left. This model took about 1.2 hours to
run
and has ~25k nodes.
What do you think about an enclosed exhaust vent with the cowl flaps
inside? It may help keep the
exhaust stream closer to the plane thereby reducing drag.
Dan
25% is the bottom limit I would guess. 33% might be slightly higher
drag.
I think it is a good idea but hard to implement with a liquid cooling
system.
One could also put a butterfly valve in the intake.
IMHO there is something a bit wrong on the velocity data. How can the
velocity
in the duct in front of the rad be much much higher than the velocity
just
on the other side of the rad. V4_R6c looks more like pressure data.
Velocity through the rad should be a function of the pressure
difference
across it.
Paul Lamar
Here is my thought on this. The wedge seems to act as a diffuser and a
venturi. As the flow aproaches the wedge vertex, it must accelarate
since
the vertex is farther away. This also corresponds to the lower pressure
the
closer you get to the wedge vertex. The wedge also expands the flow out
as
a diffuser, the closer you get to turning into the radiator vanes.
Through
the vanes, the average velocity should be lower than the average
velocity of
the duct near the inlet (wherever the velocity in most uniform) since
the
radiator open area is greater than the duct cross-sectional area. (In
this
case, you have 89% open area, so 16*.89= 14.2 inches vs 4 inches. )
Immedately after the radiator, the flow expands further, reducing the
velocity even more until the flow begins to recombine downstream.
Dan
Indeed the wedge is acting as a diffuser. The avg. velocity of the air
going
in the scoop is reduced by the ratio of rad area over scoop area at the
face
of the rad.
The mass (weight) of air per minute going in the duct must equal the mass
(weight)
of air per minute coming out of the rad as we are not adding heat in this
simulation. If we pressurize air we increase the density or weight per
unit volume.
The pressure is higher on the the front face of the rad than it is over
the rear
face of the rad. Therefor the average velocity of the air over the front
face of
the rad must be lower than the average velocity of the air coming out the
back
side of the rad. This is why I think the illustrations are mislabeled. Or
it
is one over these velocity numbers.
Vance what do you think about this?
That would be true (depending on the pressure values) if density changes
significantly. Gas flows with negligible heat transfer can be treated as
incompressible as long as the velocities are small relative to the speed of
sound. (had to crack open ye old fluids book) For Mach numbers < .30, the
max density variation is less than 5%. So conservation of mass is
rho1*V1*A1 = rh2o*V2*A2 and becomes ~ V1*A1=V2*A2. (I hope Vance agrees
;-)
This Navier-Stokes solver is only valid for incompressible fluids. BTW, I
believe that the boundary layers (BL) were excessively large for the
previous runs. Am told this is the case when using large Reynolds numbers
and large elements relative to the true BL thickness. The solver can
account for this by allowing the BL to slip (or become freestream velocity)
beyond a specified amount. I calculated a rough upper bound boundary layer
thickness of about 0.20" to use for slip. So far, the models are looking
more reasonable although I still see the same overall behavior of the wedge.
I will post more results tommorow. In retrospect, I probably should have
worked out all the issues before posting. But I got too excited ;-)
Dan Ruggirello
The AirCraft Rotary Engine NewsLetter. Powered by Linux.
ACRE NL web site. http://home.earthlink.net/~rotaryeng/
Copyright 1998-2002 All world wide rights reserved.