Designing a tail extension for the first-gen Insight

[RobertSmalls]

Ah, yes, there it is.

Here's what happens when you follow the template literally:


Here's the same, with the bottom curving upward along a spline that crosses a 7° construction line about 2/3 of the way along its length.



The rear truncation is just slightly taller than a US license plate. Looks like you will have room on both sides for reverse lights and a license plate light.

Also, as you can see, I was a little sloppy where the tail attaches to the car, leaving a gap.

EDIT: I should add, this model is 4.89m long, 95cm longer than stock. That's much better than the 5.42m that I predicted in post #1 of this thread.

[Cd]

Robert, what format does your model need to be in to work in the CFD program that you have ?

Perhaps this might be useful :

Honda Insight 3d model | Free 3D Models for Maya and 3DS MAX


1999 Honda Insight by ZXT - Google 3D Warehouse

[RobertSmalls]

Ooh, those models look nice. They are just surface models, but I could convert them to solids and hollow out an engine bay, add a radiator, etc.

I'd need to open them in Pro/E to add the tail, then run them through a CFD simulation. Pro/E is supposed to be able to import .DXF, but the free .3ds -> .dxf converter I tried didn't produce anything I could open.

[3-Wheeler]

The AeroHead Airflow Template has been regenerated for a better fit with the template graphic.

A table of data was then generated, using a small CAD program to get X-Y values for the curvature-of-departure for the template.

A second order polynomial curve fit is then generated, which allows anyone wanting to make such a template, to easily do so.

Below is the data...



And here is what the data looks like in graphical format...

You can see from the graph, that the Honda Insight rear deck curvature is slightly steeper than the template.

When designing the curvature of the new tail extension, I will try to match the Aero Template over the curvature on the car.



The data clearly shows that for anyone designing an extension to the tail of the car, the biggest-bang-for-the-buck is to make the tail as long as possible, since the Y-Axis curvature is increasing with increasing length.

Trying to make the shape too short, causes the effort to be wasted because the departure angle is relatively small at short lengths.

The polynomial expression for the curve is...

y = 7.91943079843437E-04 * x^2 + .134609376182944 * x - .737888871593933

Jim

[RobertSmalls]

What's the R² on your curve fit to the template? Can you fit a parabola to the Insight curvature data, too?

[3-Wheeler]

Originally Posted by RobertSmalls

What's the R² on your curve fit to the template? Can you fit a parabola to the Insight curvature data, too?

Hi Robert,

I don't see an R^2, but are you referring to the X^2?

In the equation, you place your X-Axis distance from the back of the car into X in the equation, and your answer for the Y-Axis offset is given in Y.

Does that answer your question?

As far as fitting a parabola, I'm not quite sure why you would want to simply the equation further, since it is really easy to setup in a spreadsheet and calculate Y for any value of X.

Actually, when it comes time to make a "master template" of this shape for the Insight, I'm going to calculate X-Y moves with the equation and translate that data to a full size template in Styrofoam.

That will allow one to glue blocks and/or sheets of foam to the desired curvature to match the template by AeroHead. This template can be used on all three sides of the car surface, when construction begins.

Jim.

[RobertSmalls]

You can get Excel to fit a parabola to anything, but the R² (the correlation coefficient) tells you how tightly your curve fits the data. 0 is no correlation, and 1 is a perfect correlation. In Excel, go to Add Trendline - Options - Display R-squared on chart.

Since the template was hand-drawn, it probably is a parabola, so I expect an R² around 0.99. It would be interesting to know if the Honda's tail is also a perfect parabola, or only a looser approximation of one.


Actually, you'll need a second foam template for the sides. The first picture in post #41 shows one template in the vertical plane, and the same template on a different scale in the ground plane.

You want your tail to have the same height and angle where it takes over from the Honda tail, so you'll need to scale your equation so the y and dy/dx of the template match those of the Honda, at the point where the Honda ends. Or just eyeball it as I did in the second picture in post #41.

[3-Wheeler]

Hi Robert,

OK, now I know what you meant by R^2!!

I typically don't rely on the R^2 value for curve fit comparisons, since it favors numbers that are larger in value.

When I did the coast down testing analysis, I noticed that the example at the Instructables used the R^2 technique, and again, it puts more emphasis on larger velocities during the coast down test, than the smaller one's close to zero miles per hour.

Rather, I like to calculate the difference between the curve fit and raw data in percent. This technique gives equal footing between the smaller numbers in the data set and the larger one's. So if I'm trying to come up with a good curve fit for some raw data, the percentage method gives the entire curve the same emphasis, and not skew the deviation towards the larger numbers at higher speeds.

I have included this approach in the table below, so you can see how accurate the curve fit is for our template in question.

Note: one should not be alarmed by the large error at the smaller X-Axis distances. It should be easy to either curve fit the lower data over a smaller range, or make a template in foam, then smooth the transitions using the eye-ball method.

[3-Wheeler]

Originally Posted by RobertSmalls

...Actually, you'll need a second foam template for the sides. The first picture in post #41 shows one template in the vertical plane, and the same template on a different scale in the ground plane.

You want your tail to have the same height and angle where it takes over from the Honda tail, so you'll need to scale your equation so the y and dy/dx of the template match those of the Honda, at the point where the Honda ends. Or just eyeball it as I did in the second picture in post #41.

Hi Robert,

Actually, what you mentioned above is just what I had in mind for the top and sides. The "template" will simply have to be trimmed to fit at various positions on the car, depending one the shape in the surrounding area.

Here is the Honda Insight data in case you are interested. I tried to be very careful in picking the shape of the car, but if one were to calculate the delta-X and delta-Y of the raw data, you would probably see some variance, and be able to correct for it.

I have not tried this, since these dimensions are close enough for me, and when the foam template is made, the working edge of the foam will be "artfully" sanded for a shape that looks good to the eye. It's similar to an artist who paints!!



By the way, the scaling number of 3.555 is shown here as well, and the CAD numbers in inches were converted to the full scale value of 94.4.. inches, which is the wheelbase dimension, so the shape should be very close.

Jim.

[jime57]

Fascinating thread. I been looking at the rear of my own Insight and rolling some of this over in a very rough fashion. From a practical viewpoint, it looks like one would want to remove the rear bumper facade, the tail lights, and the hatch minispoiler to get a "clean" starting point and attachment point. That would also shorten the tail extension slightly. It also looks like the sides of any extension get rather complex since the rear wheel area is wider than the upper sides. The complex side must add some aerodynamic complexity to the basic equations, but I can envision how the air might flow, if this shape was extended into the tail. Anyway, enjoying the discussion among the heavyweights!