CFD analywis on aerodynamics

[taero]

Im now analizing flow around formula car, especially on under the vehicle often called floor, undertray or underbody.

I got a result showing negativeforce?downforce?on floor and as a hole vehicle. But the analisyswasdonewith k-?.

Turned on k-?, then I got positive force as hole vehicle, almost 0 force on floor. I can under stand flow separation occurs stronger on k-?, then downforce will be decreased.

One model with k-? shows similar downforce to experimantal data, but when with k-?, shows lift force!

Another with k-? shows too much excessive downforce, but with k-?, shows similar to experimantal data.

All the surface under the body is meshed as prism layer, and others are tetra. Is this a problem?

Or are there other problems?

I can't resolve this problem for a long time!

Please help me. Please.

[taero]

Im now analizing flow around formula car, especially on under the vehicle often called floor, undertray or underbody.

I got a result showing negativeforce?downforce?on floor and as a hole vehicle. But the analisyswasdonewith k-?.

Turned on k-?, then I got positive force as hole vehicle, almost 0 force on floor. I can under stand flow separation occurs stronger on k-?, then downforce will be decreased.

One model with k-? shows similar downforce to experimantal data, but when with k-?, shows lift force!

Another with k-? shows too much excessive downforce, but with k-?, shows similar to experimantal data.

All the surface under the body is meshed as prism layer, and others are tetra. Is this a problem?

Or are there other problems?

I can't resolve this problem for a long time!

Please help me. Please.

[taero]

Sorry, there were some mistakes.
This is a correct below.


Im now analizing flow around formula sae car, especially on under the vehicle often called floor, undertray or underbody.

I got a result showing negativeforce?downforce?on floor and as a hole vehicle. But the analisys was done with k-?.

Turned on k-?, then I got positive force as hole vehicle, almost 0 force on floor. I can understand flow separation occurs stronger on k-?, then downforce will be decreased.

One model with k-? shows similar downforce to experimantal data, but when with k-?, shows lift force!

Another with k-? shows too much excessive downforce, but with k-?, shows similar to experimantal data.

All the surface under the body is meshed as prism layer, and others are tetra. Is this a problem?

Or are there other problems?

I can't resolve thisproblem fora long time!

Please help me. Please.

[Steve_Chung]

Hey Taero,

There are a lot more factors than simply which model you use that will affect the final result of the simulation.

Here at Maryland we have used K-e, K-w, and Reynolds Stress models. All three models have resulted in accurate simulations. The key here is they are all methods of solving the RANS equations, and as a result are sensitive to different parameters depending on the simulation.

The gesture you are making with k-w vs. k-e predicting separation earlier is incorrect; they respond to different meshes and conditions which result in different trends. By playing around with your turbulence intensity ratio and/or initial guess, I can make the trend you're seeing go the other way around.

Anyhoo looks like the problem lies with your misunderstanding of how these models solve RANS. Figure that out and see where the shortcomings of your setup and simulation lie. Remember, the numbers never lie. Your own presupposition of these numbers always will.

Also, remember to go through ALL your initial conditions and make sure they make sense for your application.

-Steve
University of Maryland FSAE
Aero guy and workhorse

[taero]

Thank you for your kindfull replying. I understand that you meantha initial value like turbulence, and initialconditions are really important.

If you don't mind, could you tell me especially on which you set the value?
Or we should do it step by step with parameters changed?

Looking forward to your assistance.

[Xeilos]

More information is required on a whole for you to evaluate whether or not your simulation is valid. I would look at your mesh and the work your way down from there. Most people do not understand the importance of meshing and how critical it is to achieving an overall satisactory result. You should have spend 40-60% of your total time on meshing alone.

The questions you should answer (to yourself) first are:

a) What is my y+ value. I hope you know what y+ is...
b) Is my mesh size appropriate for the size of simulation I am doing?
c)Does my y+ value within the recommended values for my turbulence models and wall functions that I am using.
d) Am I meeting (at the very least generally) acceptable levels of mesh quality indices for my simulation?

These are all fairly basic things that should be checked; otherwise what you are doing is simply nothing more than CFD = Colorful Flow Diagrams OR CFD = Counterfeit Fraud and Deciept.

Now with the parameters of the k-e model(if a k-e model is appropriate here???) are usually correlated with experimental data (as should all CFD anyway). If you do not have experimental data, then do a literature search for equivalent or similar CFD simulations to the case you are looking at for which values they used. If the above two are not applicable then use the standard ones with your software package. If the last suggestion is not acceptable, I suppose you could ask someone else on a forum for them

[taero]

Xeilos

Thank you for your kind message.
I didn't know that k-epsilon is much accurate than k-omega.
Is that because of wide use of k-epsilon so many research has been done for k-epsilon?

I had thought that the region is all covered with boundery layer, k-omega works well(if the mesh is appropriate)

How's about that?

I'm now rebuild mesh considering y+.
k-omega might work worth with thinner prism layers.

[Xeilos]

taero:

I never said that k-epsilon is more accurate than k-omega. Any turbulence model is only as accurate as you make it. There is no one-size fits all and will give you accurate solutions out of the box. The entire premise behind RANS is that it is one big approximation for one of nature's most complex phenomenon (turbulence). Each model has its limitation regardless of how much you tune it.

You should know these limitations before you even start simulating otherwise your results have a huge chance of being bogus. For example, k-epsilon is not so great for determining the exact location of seperation. However with proper tuning of the equations themselves, the model can be made to approximate the effects of seperation without actually displaying the seperation.

If you are getting the sense that CFD is not something that you can just press GO and simulate, then you are learning. On a personal note, I am becoming increasingly prejudiced against FEA and CFD software packages due to their ease of use. The majority of people who are now using these software packages do not understand the mechanics that drive them. I may not pretend to know these packages inside and out but as I learn, I am beginning to understand the limitations of them. And, always, always, always verify your results against some sort of basic calculations or correlations.

There tends to be more papers on k-epsilon and the effects of tuning with it as it was one of the 'original' turbulence models when D.B.S. published it. It is also quite stable (converges easily enough) and solves decently quickly (it is only two equations).

Your surfaces of where a boundary layer forms should have a inflation layer as a general rule.

Also keep in mind that y+ is often an iterative process and for k-omega thinner is not always better. As I said before there is a range to which the y+ works well in certain flow conditions. Please look these up in a technical manual, text book or hell, use the help for your CFD package.

[Mike Cook]

<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by Steve_Chung:
Hey Taero,


-Steve
University of Maryland FSAE
Aero guy and workhorse </div></BLOCKQUOTE>

Don't you have racecar to be tending to.

[Steve_Chung]

<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by Mike Cook:

Don't you have racecar to be tending to. </div></BLOCKQUOTE>

Well...... it's technically doesn't exist http://fsae.com/groupee_common/emoticons/icon_wink.gif

-Steve

[taero]

Guys, what kind of mesh is mostly used in your team? Tri , tetra, hexa or others?

[Adambomb]

<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by Xeilos:
These are all fairly basic things that should be checked; otherwise what you are doing is simply nothing more than CFD = Colorful Flow Diagrams OR CFD = Counterfeit Fraud and Deciept.
</div></BLOCKQUOTE>

That is the most awesome thing I've heard all year.

<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by taero:
Guys, what kind of mesh is mostly used in your team? Tri , tetra, hexa or others? </div></BLOCKQUOTE>

1. Most teams are really good at making Colorful Flow Diagrams (stolen and used already :P)

2. The most appropriate mesh element to use depends on the application. I'd suggest getting this information from a textbook.