Optimizing a Tail for Low Drag: Part 3
Initial
Testing
To start
the process of designing a drag-reducing tail, I threw together some boards
that would give me some adjustability of their angle relative to the rearmost
surfaces of my car. Then I headed out on the road for initial testing.
Before you
start testing, it’s important to recognize potential shortcomings and what your
tests can actually show you (as well as what they can’t). In this case,
separate boards at the top, bottom, and side of my car do not replicate
a complete, solid tail; rather, I’m using this test to get an idea of what
might be appropriate dimensions and taper angles to start my investigations of
a full tail. I’ll use these data to try and predict the drag changes from the
various angles and then use those as a jumping off point to design the real
tail rather than just guess at a shape.
You will
see people misunderstand this all the time online. Commonly, someone will make
a change to their car and then use one tank’s measured gas mileage as “proof”
it worked. Don’t fall for this! Often, the claims made from this sort of “test”
are unbelievably large in comparison to the change made, and the test can’t
actually show what the tester thinks it can. For instance, before I start in on
this tail I added outboard diffuser extensions to my car (outermost boards/vanes here):
The first
tank with these on returned a full 5 mpg higher than the previous tank—a
nearly 10% increase. Success, right? Wrong! Attributing that large a change in
gas mileage to this small (and as yet unknown) change to the car’s aerodynamics
is bananas. It also does not account for the fact that the previous tank was
almost entirely short trips to and from school and work (a distance of
approximately 3 miles), whereas this last tank included 200+ miles of freeway
driving. Of course the car got better mileage: the driving profile was
completely different and better suited to economical driving.
Keeping in
mind what I want to learn from these tests and how I want to use that
information, I headed out on the road.
Pressure
Measurement
First up,
I measured aerodynamic pressure on each board
toward its trailing edge as I varied its orientation. The way these boards are
cobbled together didn’t really allow for fine adjustments, so the angle changes
ended up being quite large—but the results were still informative.
I tested
on a N-S road on a day with winds out of the SSW at ~10 mph. So, the yaw angle
and freestream speed are different for each direction here—differences that
show up in the results.
Top board:
Bottom
board:
Side
board:
One
curiosity: the spike in negative pressure at 10° inward slant (where 5° is
approximately aligned with the bumper cover side). Concerned this might be an
error, I ran that test again in both directions and got the same result. That
sticks out as an angle I probably want to avoid in my tail design.
In
addition to the above, I measured the base pressure on the bare car (in the
center of the license plate) and found that it was -40 Pa southbound and 0 Pa
northbound:
Comparing
this to the measured pressures on the boards in each direction, this means that
the bottom and side boards were essentially supporting base pressure in all
positions—which might indicate that they are in separated flow regardless of
angle. The top board, on the other hand, developed higher than base pressure southbound, even up to a steep angle.
In the
meantime, I can conclude the following from these tests:
-drag reduction from pressure recovery on this tail will likely come from its upper surface more than from the bottom or sides
-any pressure recovery is likely to be a lot smaller than “theory” suggests
Angle from Horizontal |
South (Pa) |
North (Pa) |
Average (Pa) |
-13° |
0 |
-10 |
-10 |
-24° |
-10 |
0 |
-10 |
-34° |
-20 |
0 |
-10 |
Angle from Horizontal |
South (Pa) |
North (Pa) |
Average (Pa) |
+2° |
-40 |
-10 |
-30 |
+16° |
-30 |
-20 |
-30 |
+34° |
-40 |
-10 |
-30 |
+44° |
-40 |
-10 |
-30 |
Angle from Centerline |
South (Pa) |
North (Pa) |
Average (Pa) |
0° |
-30 |
-40 |
-40 |
-5° |
-40 |
-20 |
-30 |
-10° |
-60 |
-20 |
-40 |
-18° |
-20 |
-40 |
-30 |
-25° |
-40 |
-40 |
-40 |
-31° |
-40 |
-40 |
-40 |
You would be surprised how many people ask me what OPEC means. |
-drag reduction from pressure recovery on this tail will likely come from its upper surface more than from the bottom or sides
-any pressure recovery is likely to be a lot smaller than “theory” suggests
Unfortunately,
without more pressure recovery this tail won’t even come close to my drag
reduction goal. If the full mockup doesn’t develop higher pressures, I’ll have
to investigate shape changes to try and increase them.
Tuft
Testing
Second, I
went out on another day and had a friend drive my car while I took photographs
of tufts taped to the top, bottom, and side boards at various angles.
There was
a fairly gusty wind that day, 15-20 mph out of the NW. Despite the yawed flow,
the tufts showed attached flow in more configurations than I expected.
I couldn’t
see until I reviewed the burst images at home later, but the bottom board had
attached flow in its lowest position (it sagged a bit during testing). That
flow separated if the board was angled upward at all:
The side
board, which I had suspected was in separated flow in all positions, turned out
to have attached flow even in the crosswind, in both directions, if the board
was colinear with the body side (which is tapered inward about 5°). Slanting
the board further toward the centerline of the car, the tufts started to show
separation:
On the
top, parallel(ish) to the rear window the board showed attached flow. Angling
it further down (where my camera refused to focus; you can just barely make out
the tufts in the image), it still had attached flow on the verge of separation,
and further still resulted in completely detached flow:
I had
debated waiting for a less windy day but decided to go ahead since it’s harder
to coordinate testing with two people’s schedules. This turned out to be fine,
as these conditions are more representative of typical use and it is perhaps
more important that my final tail design perform well in conditions such as
this (gusty crosswinds) than calm, perfect days—which are a rarity. Ultimately,
reality—the real car, in the real world, in real conditions (which are almost
never “perfect”)—must be the arbiter of a design: it doesn’t matter how a
specific shape performs in CFD or a wind tunnel if those don’t reflect real
conditions. And aerodynamic designs must be tested in the real world.
Current CFD can only fully simulate very low speed, geometrically simple flows
(and it takes a supercomputer to do even that); all other CFD car simulations, even
those by OEMs and especially the generic ones you’ll see on Youtube videos and
blogs, depend on (sometimes massive) assumptions and approximations that remove
them several steps from reality. Ultimately, it does not matter if your design
is perfectly optimized in a digital environment if it turns out not
to perform the same way in the real world. So, go ahead and test on windy days.
Analysis
Now that I
have some data to work with, I can try and predict how a tail will perform and
use that as a starting point for my next test buck. To do this, I wrote
a Python program to calculate total base pressure drag as a function of top
board angle, top board measured pressure, bottom board angle, bottom board
measured pressure, side board angle, and side board measured pressure
(approximating pressure as constant over the entire surface area of each part
of the tail). Pressure drag is the sum of each pressure multiplied by the
surface area of each part normal to the x-direction i.e. in the yz-plane,
giving the x-component of the force developed by the pressures acting on the
slanted areas (I won’t bother here, but you can prove this relationship with
simple trigonometry). Each of those areas changes, as well as the remaining
vertical base area, with change in the boards’ orientations, and of course the
measured pressures change as well—which means this is a lot easier to let a
computer sort out and spit out an answer.
The
greatest drag reduction predicted by this algorithm is about 3% better than the
car with no tail—nowhere near my 15% goal. To achieve that, I will have to
essentially eliminate base pressure drag. So, for the next mockup I’ll test
some geometries, such as panel curvature, that might help to increase pressure
over a tail extension.
Keep in
mind, too, that these are only general estimates; I’ve made a lot of
assumptions (and ignored things like shear stress drag) to try and roughly predict
what a real tail will do. But that’s fine; I’ve only begun testing at this
point, and as I continue to test and measure, I will refine my parameters for a
final design. These initial tests are just a starting point.
Previous: Optimizing a Tail for Low Drag: Part 2
Comments
Post a Comment