Help me evaluate CFD results
I have been playing around with a model of my car and putting it in a cfd program. The model is not that great, however it is pretty accurate (shape wise) from the center line of the car. Based on this I have been messing with the angle of the hood to windshield and the nose. I have a limited understanding of how the numbers should be interpreted and would like more opinions on the subject. Rather than posting a bunch of screen shots directly here I have inserted them in a google document.
(quick sample) https://ecomodder.com/forum/member-m...st-current.png https://ecomodder.com/forum/member-m...current-55.png https://docs.google.com/document/d/1...it?usp=sharing The program I used for modeling is Autodesk Inventor and the CFD is Autodesk Flow Design. The wind speed is set to 10 m/s (I forgot to increase it) unless otherwise noted. The 2d plane the pressure is mapped on is in the same spot for all screen shots.
Thank you in advance, Matt |
•The car is not sitting on the ground plane, would this effect how to interpret the front and top of the car?
Is the ground plane moving? Are the wheels represented? •I had two options for mapping, pressure or velocity. Is one better to look at or is it a combination of the two that makes the data interpretable? They interoperate, or something. There should be some algorithm to combine the results. •Is there anything that you guys suggest I try and mock up and test? Is this analyzing a 2D plane though the centerline of the vehicle? If so it's meaningless. The entire 3D form is necessary. •The first thing I wanted to test was the simple hood height (because of visibility and simplicity), the only differences I see is a lower high pressure (on the scale) and a more localized section of low pressure where the windshield and the roof meet. The only difference I see is the speed and pressure. What hood height difference is there? Once again, the 3D form is significant. A Jeep and a VW Beetle have flat windshields at a similar angle. But the Jeep's is square while the Beetle has a teardrop that has been truncated at the front. It's body is widest at the C-pillar. edit: I looked around a bit. You may find answers at www.cfd-online.com/Forums/autodesk-simulation-cfd/ |
Quote:
1.5 Would that algorithm be the cD number or similar? 2. Can you expand on that? Other threads look at 2d profiles, which would be basically the same as a smoke overlay (like in the first post of the wind tunnel thread) and we still can make assumptions based on that data such as how small the vortex at the back of the car is. I just want to try and figure out how much difference it makes and have a plan before trying to make something on the car. https://ecomodder.com/forum/showthre...ead-26678.html 3. Its relatively small. Like 3-4 inches up the windshield. I only put it in the center of the model because thats where the plane is. Jeep and beetles have different approaches though. Jeeps hoods are pretty much parallel to the ground. Ignoring the part at the base of the windshield the beetles is more obtuse. I will check that out, thanks. Wind speed set to 25 m/s. velocity instead of pressure. https://ecomodder.com/forum/member-m...nes-tracer.png Wind speed at 25 m/s velocity instead of pressure. It also seems like flow stays attached longer at the back. https://ecomodder.com/forum/member-m...ght-tracer.png |
It appears you have a 3D shape that's being analyzed in a single plane. Do you have an isometric view? The single plane reduces the computation needed. How long does it take to generate a single view?
Further into the Wind Tunnel and Smoke thread: http://theansweris27.com/wp-content/...02-768x466.png Permalink #817 IIRC aerohead found a blister on the hood had no appreciable effect on his T-100 pickup. |
From memory the software is not that slow. And it does it in 3d.
It is sometimes less confusing to view it in 2d. |
CFD model
I have low confidence in all desktop CFD simulations.Optimization of shape involves details which can only be evaluated in 3-D,in ground contact,at 'full-scale,' above critical Reynolds number,on software we can't afford,running on a supercomputer we also can't afford.
From your model,the nose is okay.The nose slope would be at saturation )no additional steeper angle would affect drag),and the windshield angle would also be near saturation.So far so good! The aft-body ought to follow the 'template.'The last portion of your roof-line cannot support attached flow. The diffuser angle should be relaxed down to 2.8-4.0 degrees otherwise the flow won't follow that either. The VW XL1 would be a good example to follow.It's dead-nuts on as far as shape goes. |
Quote:
a single view in the CFD is fairly quick. less than a minute for the 2d plane. About 45 seconds for the 3d and then time to settle down I guess. The reason I am avoiding a 3d evaluation is because my model is lacking. I created it based on pictures of the car (from a distance zoomed in), but I don't have one from above to get the taper of the tail and such. I can post one if you think it would provide more information. I wanted to test the hood to windshield because the air appeared to be separating at the back of the roof. I was hoping that it would indirectly keep flow more attached at the back by changing how it transitioned from the front roofline. Current car @ 25 m/s https://ecomodder.com/forum/member-m...ero-tracer.png After letting it sit and run it changed to this: https://ecomodder.com/forum/member-m...o-tracer-2.png The red box is where the flow lines are originating from. https://ecomodder.com/forum/member-m...o-tracer-3.png Quote:
At the very end where it takes a sharp turn or where the roof ends and meets the tail? I had to compromise on the bottom of the car. Do you think the CFD can be used as a data point? Not necessarily accurate for numbers, but just the flow pattern around the car. |
CFD
Quote:
The bumper would be okay if it had adequate corner radii,but as depicted in plan-view,it has none at all,and guaranteed separation.Front fenders Dittoo.A-pillars ditto. Nothing you can do at the front can solve the issue you have at the back.The present roofline is too aggressive towards the tail,and will absolutely trigger separation.No where can you exceed 23-degrees downslope. I understand the diffuser issue,although there's no reason an active diffuser,which could deploy downwards on the highway wouldn't solve the issue.GM did it with their 'Epcot' concept of the early 1980s. I think you'd be better off just looking at the wind tunnel flow images of real cars here on the Aero Forum. I'm okay with CFD,just not the 'toy' kind.If you want to get serious,you better bee a millionaire! |
Do with this what you will: The Ford 3D Store
|
Quote:
Are you eyeballing the slopes or calculating? I have been eyeballing. When I overlay the template (I understand has limitations) the tail pretty closely matches the upper shape. I would like to do some tuft testing, but don't really have the stuff required to record it. https://ecomodder.com/forum/member-m...er-overlay.png Quote:
Do you think a gurney flap at the tip of the tail would help flow? I remembered I have 2 wide angle dash cams I might be able to use to record from a chase vehicle. I will try to get something this weekend. |
eyeballing or..................
Quote:
He dedicates an entire section to forebody design,drawing on Jaray,Fachsenfeld,Kamm,Waters,Pawlowski,Carr,Ahmed ,Hoerner,Emmelmann,Janssen,and Hucho. From the visual schematics,I can just look at your rendering,and know that you\'re already at saturation.There\'ll be no problem with attached flow over the roof peak and onto the rear slope. The important part is the slope/contour back there.That\'s where your streamlining potential is.The \'template\' contour is derived from the lowest-drag half-streamline-body of revolution.My model which used it,measured Cd 0.12 just as predicted.You\'ll never find a shape with lower drag potential.That\'s why I shared it. |
I appreciate the advice from you guys.
Can you expand on what you mean by saturation? Are you saying the profile is as good as its going to get and I won\'t see much improvement modifying the boat tail anymore? Therefore, I need to focus on the rest of the car. Also, I finished the bottom section of the boat tail I just hadn\'t when I took that picture. https://ecomodder.com/forum/member-m...214-133908.jpg Edit: Assuming I am correct in what you mean by saturation I think I have some useful information. On a particular road in town that was recently repaved and has a 55 mph speed limit the tail makes absolutely no noise. If there was a lot of turbulence the unsupported sections would flap around right? The road is decently protected from crosswinds on both sides by trees and buildings. |
Thanks for posting that. Just yesterday in another thread I was lamenting the fact we haven't seen a full boat tail (on a Metro) lately.
Do you have construction pics? That show the internal structure (if any?). What do you do about tail light and the license plate? I'd be curious what it is like living with it. Parking and & etc. Quote:
As for 'as good as its going to get', consider truncating it less than a foot. See the Peter Brock Cobra Coupe or Dave Cloud's Dolphin. http://img.photobucket.com/albums/v7...phin/DD006.jpg https://ecomodder.com/forum/showthre...tml#post559726 |
saturation
Quote:
The tail looks delicious! Thanks for doing that!:D |
Quote:
Quote:
|
4% @ 70
Quote:
|
Quote:
Holy moly! They let you ride around in that? Gotta love the USA. |
Quote:
0.34=>0.30 sounds about right. Now go after the wheelwells. Quote:
The result may vary by jurisdiction. |
Check out the software that freebeard linked a couple of months back. Appears to be slightly more capable that the Autodesk stuff.
|
Permalink #4 was just for purposes of illustration.
The key to low-compute-power CFD IMHO is OpenVDB: Quote:
https://i.ytimg.com/vi/R8Y9VqU_YTI/maxresdefault.jpg https://i.ytimg.com/vi/R8Y9VqU_YTI/maxresdefault.jpg Instead of computing every voxel in the virtual wind tunnel, it looks more diligently closer to the aeroform. It's above my pay grade. :( |
Quote:
Quote:
Quote:
Quote:
|
You've added a crumple zone. Worst case and heaven forfend, but if you're rear-ended the fool behind you will be safer. :thumbup:
You should strongly consider the old thread from 2013, Military theme for fun ...and function Bumblebee stripes and some cryptic Zombie Response Vehicle markings would not be out of place. Quote:
That kind of work involves a tool-chain. OpenVDB is a visual database. It can be queried by anything that can structure a query. It's currently employed by the movie industry, Maya and (apparently*) Blender. I took a deep dive into the [pseudo]code structure a few years back. I recognized a lot of it from when I was trained to edit Mac OS HFS structures (I worked for Symantec for 9 years). B-trees allow you to get to any point in the point cloud in a discrete number of steps. Shallow trees have fewer steps to get to an arbitrary point. There was a brilliant way of encoding things like pressure or a motion vector for each point but I don't recall and prolly couldn't explain if I did. So there's that. I'm like a scarecrow in a field pointing thataway. *I struggle with Blender itself, let alone where the API for the C++ libraries are. Have you looked at Blender? I'm not doing any render farming on my laptop. |
Gurney flap
So far,no really low drag car has used a Gurney flap as a palliative for flow separation.Hucho and the others,who've explored really minimum drag all recommend that we stay on the streamlined contour,then if length is an issue,just chop the body as Koenig-Kamm did at the FKFS,next door to Daimler-Benz.
|
The Gurney flap aka wickerbill was a field expedient. As was the Gurney Bubble.
https://www.gt40s.com/media/mkiv-gur...l?d=1532040050 https://www.gt40s.com/media/mkiv-gurney-bubble.901/ |
I came across something and this thread seems most relevant:
https://www.youtube.com/watch?v=N4YMYGDEcPM This speaks to the integration of OpenVDB with Blender. It is about importing data sets into the program, where for CFD the internal physics engine would be used to generate the data. But it shows that the future isn't quite here yet. |
All times are GMT -4. The time now is 03:20 PM. |
Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2024, vBulletin Solutions Inc.
Content Relevant URLs by vBSEO 3.5.2
All content copyright EcoModder.com