Aerodynamics discussion

[Dimman]

Originally Posted by plucas

I am EarlQ partner. I am the aero guy who does the aero development and CFD. Also looking at purchasing an FRS....



PM me if you are interested in making a solid model. I can give you a good starting point to work from. I just cannot distribute the model we are currently using. We use Creo/ProE but also have a lot of experience in SolidWorks. Glad to see some other engineers on here with interest in racing cars!



We are using OpenFOAM. Basically we use creo/proE for the cad model, then OpenFOAM for the pre-processing and solver, and then paraview for post-processing. We run this on a linux workstation. Depending on the analysis and its requirements, runs can take anywhere from 12-36 hours.

Correct that cfd is not "real life." It is an experiment with conditions that do not change to get repeatable results. It is in no way 100% and that is a big misconception. However, it can definitely drive quality design changes that drastically improve performance. Most automotive cfd models are very simplified because of computing time and it won't drive design changes on a big scale. How it is set up makes a big difference. I have validated our program and setup using automotive test of ahmed bluff body and a class 8 heavy vehicle (GCM). Both have data from university for validation of setup and program.

Also windtunnels are not "real life" either. That is also a big misconception. It is also just an experiment to get repeatable results that represent what is going on in real life (just like cfd). Both are doing the same thing, just in different ways.




Not all the models are of the same quality. However, the models still need to be reworked anyways to make them watertight for cfd purposes. If you have any questions on computer hardware or cfd in general, let me know in this thread or pm. I am happy to help :happy0180:

I'm starting to research hardware and Linux, after over a decade being out of the loop for what is cutting edge.

I was looking at starting with an Intel i7, but saw on your blog you are using Intel Xeon. What should I focus the build around? Which resources should be the priority (cpu, core number, cache, ram, video card)? Is the CFD mostly cpu/ram intensive, and ParaView graphics card intensive?

Also, how much of a difference in performance between a regular Linux install (Ubuntu) and going through VirtualBox on Windows?

Thanks for any help you can offer.

[EarlQHan]

Originally Posted by Dimman

I'm starting to research hardware and Linux, after over a decade being out of the loop for what is cutting edge.

I was looking at starting with an Intel i7, but saw on your blog you are using Intel Xeon. What should I focus the build around? Which resources should be the priority (cpu, core number, cache, ram, video card)? Is the CFD mostly cpu/ram intensive, and ParaView graphics card intensive?

Also, how much of a difference in performance between a regular Linux install (Ubuntu) and going through VirtualBox on Windows?

Thanks for any help you can offer.

Paul told me it's CPU/RAM intensive. I believe he has dual i7's with 16GB and a full car model takes around 18-24 hrs to process. With our workstation, it can be done in 2-4 hrs. Not sure of the performance difference but we use Ubuntu through VirtualBox. I don't know much about computers though, so I'm sure he can tell you more.

[EarlQHan]

Originally Posted by plucas

Correct that cfd is not "real life." It is an experiment with conditions that do not change to get repeatable results. It is in no way 100% and that is a big misconception. However, it can definitely drive quality design changes that drastically improve performance.

Also windtunnels are not "real life" either. That is also a big misconception. It is also just an experiment to get repeatable results that represent what is going on in real life (just like cfd). Both are doing the same thing, just in different ways.

THIS! CFD testing and wind tunnel testing is a means to drive the design, speed up design iterations, and reduce costs. You still want to correlate your results from either method with real world results.

Don't mistake wind tunnel tested for wind tunnel optimized. They are two different things. This is the equivalent of thinking someone is a lawyer just because he or she has taken the bar exam. It's easy to get too wrapped up in pictures of smoke streams or wool tufts.

And anything you put in the air stream will have an aero effect. Whether it is a beneficial effect or not is a different story.

[Dimman]

@EarlQHan

Speaking of wind-tunnels not being real world, either: are spinning tires and moving ground available in OpenFoam?

[jdzumwalt]

I consider my self pretty good with most things but this is way over my head. Please keep us all informed because I too want to go faster.

[OrbitalEllipses]

Originally Posted by industrial

Does anyone have more pictures of the TRD Griffon underbody work? I'd like to try and replicate their work since it's probably the most extensively tested stuff for our cars so far.

Race cars and street cars are very different. The TRD Griffon car was BUILT to go around Tsukuba. Race cars are use flat bottoms for aero; however, I don't see a flat bottom working well with a daily driver that's sitting at lights and traffic. Why? Flat bottoms generally have issues with cooling that require smart engineering to fix. You introduce one variable, it changes another. @Turn in Concepts can attest to this: they put a flat bottom belly pan on their race car several years ago and subsequently ran into trouble with transmission and diff overheating during races because these parts didn't get enough airflow presumably. A street car might not car get as hot, but it certainly has even less of a chance to have effective cooling.

As far as the diffuser goes, there's been talk of the stock muffler aiding in reducing lift at the rear (posters and I believe someone tracked down an official comment). I didn't get a chance to drive the car much before switching out the muffler for a different unit, but it seems to me that the rear of the car has become less stable at highway speeds. It could to winter tires and a bad headwind, but that's what I feel. I'd like to see some definitive results on this.

[CSG Mike]

Originally Posted by Dimman

I'm starting to research hardware and Linux, after over a decade being out of the loop for what is cutting edge.

I was looking at starting with an Intel i7, but saw on your blog you are using Intel Xeon. What should I focus the build around? Which resources should be the priority (cpu, core number, cache, ram, video card)? Is the CFD mostly cpu/ram intensive, and ParaView graphics card intensive?

Also, how much of a difference in performance between a regular Linux install (Ubuntu) and going through VirtualBox on Windows?

Thanks for any help you can offer.

PM me if you need help making a box

[jamal]

There are places on the hood where the pressure differential wants to draw air in, and places where it wants to go out. Toward the front is usually negative pressure and then at the windshield there will be a lot of pressure that would force air into the engine bay. If someone wanted to test they could put manometers in different spots of the hood.

I always thought it was interesting that the fake hood vents on the old GC impreza are exactly where the radiator and intercooler duct outlets were on the WRC car, so there was something to it.

[Shankenstein]

Originally Posted by OrbitalEllipses

As far as the diffuser goes, there's been talk of the stock muffler aiding in reducing lift at the rear (posters and I believe someone tracked down an official comment).

Absolutely, it reduces lift! Like a golf ball, therefore woooosh! \s

The dimples will initiate a turbulent boundary layer, which will hug the diffuser further up the rear. The higher it goes, the less wake cross-section. I expect the contribution of the dimples is very low.

The main way the muffler reduces lift isby providing a smooth transition between the pumpkin and the bumper. They could have left it level, but angling it will prevent the flow from getting too caught up in the rear suspension area. I'd hope that the aftermarket exhaust kits maintain that slope, but most welders are more concerned with fitment than aero.

[plucas]

Originally Posted by Dimman

I'm starting to research hardware and Linux, after over a decade being out of the loop for what is cutting edge.

I was looking at starting with an Intel i7, but saw on your blog you are using Intel Xeon. What should I focus the build around? Which resources should be the priority (cpu, core number, cache, ram, video card)? Is the CFD mostly cpu/ram intensive, and ParaView graphics card intensive?

Also, how much of a difference in performance between a regular Linux install (Ubuntu) and going through VirtualBox on Windows?

Thanks for any help you can offer.

I have an i7 for my personal computer with 32gb of ram. Basically meshing is very ram intensive and computing is cpu intensive. This however does not mean that the other isn't important also. Paraview is very graphics card intensive. I would personally do it around an i7 to keep cost down. Then the quickest and most ram you can run. Then for graphics card, I am not sure what you would need. I am not a computer expert and I had a friend help with specing out and building my computer.

We now however have a nice workstation with xeons. I will run the same simulation on my personal computer vs the workstation and compare results. I will also then let you know the complete specs of both.

Originally Posted by EarlQHan

Paul told me it's CPU/RAM intensive. I believe he has dual i7's with 16GB and a full car model takes around 18-24 hrs to process. With our workstation, it can be done in 2-4 hrs. Not sure of the performance difference but we use Ubuntu through VirtualBox. I don't know much about computers though, so I'm sure he can tell you more.

This basically. I will run an analysis to compare.

Originally Posted by Dimman

@EarlQHan

Speaking of wind-tunnels not being real world, either: are spinning tires and moving ground available in OpenFoam?

Yes both are available but not called the same. You just have to set the boundary conditions. Really almost anything is possible with OpenFOAM since you can write your own code.

Originally Posted by OrbitalEllipses

Race cars and street cars are very different. The TRD Griffon car was BUILT to go around Tsukuba. Race cars are use flat bottoms for aero; however, I don't see a flat bottom working well with a daily driver that's sitting at lights and traffic. Why? Flat bottoms generally have issues with cooling that require smart engineering to fix. You introduce one variable, it changes another. @Turn in Concepts can attest to this: they put a flat bottom belly pan on their race car several years ago and subsequently ran into trouble with transmission and diff overheating during races because these parts didn't get enough airflow presumably. A street car might not car get as hot, but it certainly has even less of a chance to have effective cooling.

As far as the diffuser goes, there's been talk of the stock muffler aiding in reducing lift at the rear (posters and I believe someone tracked down an official comment). I didn't get a chance to drive the car much before switching out the muffler for a different unit, but it seems to me that the rear of the car has become less stable at highway speeds. It could to winter tires and a bad headwind, but that's what I feel. I'd like to see some definitive results on this.

Flat bottoms can work and work well on street cars. It just has to be engineered. It will not work well if you just slap a cover on the bottom of the car and do not change anything else. The whole cooling of the car has to be changed. Engine cooling, trans cooling, and diff cooling would all have to change. Cars are a system and need to be engineered as a system.

[ayau]

Originally Posted by Shankenstein

Absolutely, it reduces lift! Like a golf ball, therefore woooosh! \s

The dimples will initiate a turbulent boundary layer, which will hug the diffuser further up the rear. The higher it goes, the less wake cross-section. I expect the contribution of the dimples is very low.

The main way the muffler reduces lift isby providing a smooth transition between the pumpkin and the bumper. They could have left it level, but angling it will prevent the flow from getting too caught up in the rear suspension area. I'd hope that the aftermarket exhaust kits maintain that slope, but most welders are more concerned with fitment than aero.

Can you explain the text in bold in layman's terms? I've always wondered what those dimples were used for.

[OrbitalEllipses]

Originally Posted by ayau

Can you explain the text in bold in layman's terms? I've always wondered what those dimples were used for.

http://entertainment.howstuffworks.c...question37.htm

[Captain Insano]

Originally Posted by OrbitalEllipses

Race cars and street cars are very different. The TRD Griffon car was BUILT to go around Tsukuba. Race cars are use flat bottoms for aero; however, I don't see a flat bottom working well with a daily driver that's sitting at lights and traffic. Why? Flat bottoms generally have issues with cooling that require smart engineering to fix. You introduce one variable, it changes another. @Turn in Concepts can attest to this: they put a flat bottom belly pan on their race car several years ago and subsequently ran into trouble with transmission and diff overheating during races because these parts didn't get enough airflow presumably. A street car might not car get as hot, but it certainly has even less of a chance to have effective cooling.

As far as the diffuser goes, there's been talk of the stock muffler aiding in reducing lift at the rear (posters and I believe someone tracked down an official comment). I didn't get a chance to drive the car much before switching out the muffler for a different unit, but it seems to me that the rear of the car has become less stable at highway speeds. It could to winter tires and a bad headwind, but that's what I feel. I'd like to see some definitive results on this.

I have not switched out my exhaust yet, I'm getting ready to... BUT, that part I put in bold, winter tires DEFINITELY affected the feel of my car on the highway even at speeds as low as 65-70 mph. The car feels less stable on the highway for sure with my winter tires and I put them on when it was absolutely bone stock so I know it's the tires and no other modification. There is a good chance you are feeling your winter tires more than the aero differences caused by swapping your OEM muffler.

[wbradley]

My TSX had a CD of 0.26

As long as the drag coefficient is not excesssive, say below 0.30, fuel economy shouldn't take too negative a hit. Creating downforce usually creates drag. How much downforce does this type of car need? Not a lot, as long as it is down force, not lift.

In this day and age a passenger vehicle should have low drag. It might not impact city driving as much, but at highway speeds better mileage and reduced noise are attainable/desireable compared to a decade to two ago..