Finding shape with lowest drag coefficient

[julien.decharentenay]

Hi,

I have been thinking of a simple little project for which I would require your participation. The project is to find the most aerodynamic shape that can hold at least one person (laying down is ok).

Some background: I am a cross between software development and scientist (expertise in Computational Fluid Dynamics). I recently quit my job to focus on my own projects. One of which is Aerodynamic on Demand - add https in front of aerodynamicondemand.appspot.com

Further details: I would like interested people to use the above webportal (using the Discovery option for free) to find the most aerodynamic shape. Aerodynamic is measured by the drag coefficient Cd (i.e not the drag area), the lower the Cd the more aerodynamic the design is.

The project closing date will be 9th October 2013.

To participate it should be relatively simple:

1. Create your geometry in a CAD package (I personally use SketchUp):
   • Geometry does not need to be a car or bike (i.e do not worry with wheels, etc) - do not spend more than a couple of hours unless you have a very specific shape in mind
   • Geometry is to be able to accommodate at least one occupant
   • Geometry is not taller than 5m
   • The lowest point of the geometry is to be no more than 0.5m above the ground
   • See the geometry requirements on the website
2. Export the geometry in STL format (using meters units);
3. Go on the website to the "Start a calculation" page;
4. Fill in your details and use the promo code ECO1023 - if you click on "update" it should report a AU$0 cost for the Discovery option. Use a speed
   between 50km/h (13.889m/s) and 100km/h (27.77m/s) with a zero yaw angle;
5. Select the Discovery option.

At the moment, there are 50 free sims available (happy to increase if there is more interest) to be used by the closing date. If the website redirect you to Paypal, DO NOT PAY, send me an email instead!!!!

Note: the Discovery option is a coarse simulation.

Report your results: It is probably best to give your results by responding to this thread.

What am I going to do? I am planning to organise the results by decreasing drag coefficient to create 30 seconds YouTube video(s) of the geometries - I won't share or take credit for the geometries beyond the video(s). I would like to credit the participants so it would be great if when posting your results, you give out any twitter, facebook name that I can add to the credits (otherwise, I will just use your EcoModder username).

This is only running on the EcoModder forum, so please do not disclose the promo code outside. I will include a mention of EcoModder to the video (if acceptable).

Please feel free to ask if you have any questions.

Julien

[mikeyjd]

Sounds cool. I don't have a cad program, but thanks for the idea.

[julien.decharentenay]

Thanks Mike. You can try SketchUp - there is a free version caller SketchUp maker and is reasonably easy to use.

To get things started:

- Cd = 0.8908

- Results and images at: https://aerodynamicondemand.appspot....1-4ef147317882
- Twitter @HibouSoftware

[gone-ot]

Wonder what information is available for all the different bullet shapes? Obviously, they've done LOTs of aerodrag computations and analyses on the many, many, different military & civilian ogive 'shapes'?!?

[aerohead]

*Since the ground is mentioned,this will be a road vehicle?
*If it is a road vehicle,must it contain a source of locomotion along with the occupant?
*It is to be operated only on a closed-course?
*Closed-course vehicles would have a chance to utilize laminar profiles if operating solo,on a dead calm track day.
*The 'laminar' profiles would never work on a real roadway.
*We might mention Reynolds number operating conditions as they affect boundary layer type and both pressure drag and surface friction drag.

[aerohead]

Originally Posted by julien.decharentenay

Thanks Mike. You can try SketchUp - there is a free version caller SketchUp maker and is reasonably easy to use.

To get things started:

- Cd = 0.8908

- Results and images at: https://aerodynamicondemand.appspot....1-4ef147317882
- Twitter @HibouSoftware

There is a vast amount of empirical data regarding the aerodynamic drag of differing geometric forms in the public domain dating back to before the turn of the 20th Century.
Some minimums have already been established for forms operating in ground proximity.Typically,streamline bodies of revolution and ellipsoids and their half-bodies return the lowest drag on a frontal area basis.
There are delayed-transition 'laminar' forms which can work with torpedos,submarine hulls,airships,wing sections,and fuselage,artillery projectiles and bullets,but they don't perform in ground effect.
Unless your CFD has 'Full Navier-Stokes' capability with the ancillary turbulence models built in,it will be unable to anticipate 3-dimensional wake forms behind road vehicle bodies,something critical to accurate CFD drag prediction.

[julien.decharentenay]

Originally Posted by aerohead

*Since the ground is mentioned,this will be a road vehicle?

Yes. To keep it in the spirit of the forum

Originally Posted by aerohead

*If it is a road vehicle,must it contain a source of locomotion along with the occupant?

No. It is intended as a fun and quick experiment with shape and analysis, not a vehicle design.

Originally Posted by aerohead

*It is to be operated only on a closed-course?*Closed-course vehicles would have a chance to utilize laminar profiles if operating solo,on a dead calm track day.
*The 'laminar' profiles would never work on a real roadway.

Thanks for explanation as I was not familiar with closed-course. You can assume that it is operated on a closed-course. Since it is modelling based, you need to be aware that the model uses a turbulence model. I would be interested to see the difference in model prediction between a profile optimized for laminar flows and one optimized for transitioning/turbulent flows. I am assuming that the model will indicate that the latter is better - another limit of modelling...

Originally Posted by aerohead

*We might mention Reynolds number operating conditions as they affect boundary layer type and both pressure drag and surface friction drag.

Sounds good. Re = rho x U x L / mu where
rho = 1.205 kg/m3 and mu = 1.8e-5 kg/m.s
U is the travelling speed selected for the analysis
What is L ??? It could be a characteristic length taken as the longer of the height, width or length or based on the frontal area (L = sqrt(frontal area))?

[julien.decharentenay]

Originally Posted by aerohead

There is a vast amount of empirical data regarding the aerodynamic drag of differing geometric forms in the public domain dating back to before the turn of the 20th Century.
Some minimums have already been established for forms operating in ground proximity.Typically,streamline bodies of revolution and ellipsoids and their half-bodies return the lowest drag on a frontal area basis.

I agree with you. My expertise is with software and CFD, not shape optimisation. I was thinking of trying a couple more shapes during the week, but I lack the background to go into it in a lot of details. Add I would love to have others testing my software.

Originally Posted by aerohead

Unless your CFD has 'Full Navier-Stokes' capability with the ancillary turbulence models built in,it will be unable to anticipate 3-dimensional wake forms behind road vehicle bodies,something critical to accurate CFD drag prediction.

The CFD is based on a steady-state RANS simulation package. The simulation uses a k-omega SST turbulence model - with is quite common, but lacks a number of capabilities (which most 2-equations turbulence models lacks): namely laminar to turbulent transition and anisotropy of the Reynolds stress tensors closure.

[aerohead]

Originally Posted by julien.decharentenay

Yes. To keep it in the spirit of the forum



No. It is intended as a fun and quick experiment with shape and analysis, not a vehicle design.



Thanks for explanation as I was not familiar with closed-course. You can assume that it is operated on a closed-course. Since it is modelling based, you need to be aware that the model uses a turbulence model. I would be interested to see the difference in model prediction between a profile optimized for laminar flows and one optimized for transitioning/turbulent flows. I am assuming that the model will indicate that the latter is better - another limit of modelling...



Sounds good. Re = rho x U x L / mu where
rho = 1.205 kg/m3 and mu = 1.8e-5 kg/m.s
U is the travelling speed selected for the analysis
What is L ??? It could be a characteristic length taken as the longer of the height, width or length or based on the frontal area (L = sqrt(frontal area))?

For road vehicles,the source flow is considered already fully turbulent and the vehicle totally immersed in a turbulent boundary layer.Which is good,as the TBL can feed kinetic energy in,in a hostile region of adverse pressure gradient over the aft-body,helping to prevent reverse-circulation and ensuing separation in this 3-D domain.
In a protected,closed-course environment,with a lone vehicle on the course at a time on a calm day,it might be able to get a bit of laminar boundary layer up to the first minimum pressure position,then transition over.But vehicles have such little surface area compared to say an aircraft,the skin friction is insignificant compared to pressure drag.
------------------------------------------------------------------------
With respect to the Rn,yes L=body length as we treat it with vehicles.

[aerohead]

Originally Posted by julien.decharentenay

I agree with you. My expertise is with software and CFD, not shape optimisation. I was thinking of trying a couple more shapes during the week, but I lack the background to go into it in a lot of details. Add I would love to have others testing my software.



The CFD is based on a steady-state RANS simulation package. The simulation uses a k-omega SST turbulence model - with is quite common, but lacks a number of capabilities (which most 2-equations turbulence models lacks): namely laminar to turbulent transition and anisotropy of the Reynolds stress tensors closure.

I peeked at some online CFD stuff,Imperial College,etc..
' looks like they recognize some contextual limitations with the models,but boy are they advancing the state-of-the-art!
When Wolf H.Hucho was with Volkswagen,they would put a car model in the wind tunnel,use visualization techniques to define the wake,along with velocity and pressure probe survey,then import the car-wake coordinates into the software as an aerodynamic singularity,panelizing the wake right along with the car within the wireframe virtual domain.
This gave them a priori knowledge of the actual 3-D wake with which to include in the numerical analysis for comparison to the empirical measurements found in the tunnel.
The CFD just gets better and better.