Great website and interesting analysis! It seems that we are opponent not only in the KVRC, even if CAEdevice is more focused in mechanical design than in CFD simulation
Thanks, it is always good to hear some nice words about a small project like a homepage.
It would be cool to see some more competition by small or larger engineering companies. I think KVRC is a great platform to test and show CFD and other capabilities. Too much of our work stays behind closed doors.
A quick hi. In case you are wondering about the late results. I am back in Australia to organise for a number of item on a personal level. I flew over on Monday and had the pleasure of preparing some of the car from HK airport. This has delayed me a lot - particularly as I have to focus on a my short two weeks here to make sure that I get everything done. My internet connection is also very inconsistent and I typically have to download files a couple of times before getting them in full. I am going through the simulations slowly but surely. The emails will pop-up during the week when the results are available...
Meanwhile i'll share the data i've just received, being this the race weekend (stop reading if you don't want the surprise ruined).
Well, once againg i'm not happy with the numbers i've received.
During the first two races my private CFD simulations gave hugely different results from the official simulations, which is quite understandable, given the different setup i've used compared to the official ones.
In particular:
Nurburgring Round 1: Drag=-19%, DF tot=-30%, DF front=-15%, DF rear=-42%
Monaco: Drag=-15%, DF tot=-27%, DF front=-8%, DF rear=-39%
However, this round the discrepancy is much, much bigger:
Nurburgring Round 2: Drag=-22%, DF tot=-39%, DF front=-17%, DF rear=-53%
If the discrepancy had been similar to Monaco, my car would have had these numbers:
Drag=2256N, DF tot=9746N, DF front=4720N, DF rear=5026N
Instead, i've received the followings:
Drag=2067N, DF tot=8146N, DF front=4304N, DF rear=3842N
There are two reasons why i'm posting this.
The first one is to ask wheter that's normal or not... Is anyone else experiencing similar discrepancies? Has the official solver been modified so much? What could lead to these numbers?
(For clarity, i'm using this setup: Khamsin package, half car in a 3m wide/5m tall/13m long wind tunnel, 5mm hexahedral mesh on the car, 100mm max mesh for the environment, 44.4m/s air velocity, 10m/s floor velocity, turbulent flow k-epsilon. No rotating wheels (which could be quite an issue). Usually my simulations take 6 hours on a 3.3GHz computer.)
The second reason why i'm posting this is to say that i may not be able to complete the championship. That's because i can't keep on basing the design of my car on unreliable results, and at the same time i've got no time to spend after the optimizations of the CFD parameters. Also, the regulation changes impose me to design a brand new car for the next round, which is something i could barely afford in normal conditions, and something that i cannot afford to do at all given the CFD problems. (when i mention the lack of time i mean that i will be physically away from my computer...)
...i don't know...i'll try, but i don't guarantee my presence.
My advice is to use the OFFICIAL framework. I did this choice at the beginning of the championship: I had problems with the first two races, but I'm sure to move in the right direction. I'm confident about the last three races (the software is still not stable, but the improvements I see on my workstation are confirmed on the virtual track).
I agree considering the development of the numerical simulation a (the) key point for the future of the championship.
Anyway, as I wrote to another partecipant who decided to don't take part to the second race, it's a game!
If anyone would like to run a test with the local release of OCCFD, I'd be happy to make workstation available.
Generally speaking: last year, to have a decent correlation between the software I was using (k-eps based) I had to set a time dependent simulation with a large time step (about 0,01s): I'm quite sure that a steady state solution is very sensitive to the vortex model (a part of the energy of the flow is not considered). A switch to a dynamic solution would help: as a sponsor I can consider to help with the cloud services if the staff would agree (for the 2016 edition).
RicME85 wrote:Would it not be better to just do a run or two with AWS? My results with AWS were pretty much exactly the same as the official results.
Interesting... Similar results every round? In that case AWS would be the best choice.
CAEdevice wrote:My advice is to use the OFFICIAL framework.
As you underlined, the official one (run on "local") hasn't been particularly stable...
Now it's too late for me to start playing around with new softwares: i've got no time.
The questions remain: why are my results so different this round in particular? Has OCCFD changed so radically with the last update? What's setup difference could justify the discrepancy?
CAEdevice wrote:Anyway, as I wrote to another partecipant who decided to don't take part to the second race, it's a game!
I know, I know, but not getting to the podium is not an option for me. Nor it makes sense developing a brand new car completely blinded: i wouldn't learn anything.
I do not understand the issues with own CFD simulations and KVRC results. I also run using a symmetry plane (do not do that on normal cars!) and get results that are very close to KVRC.
variante wrote: 10m/s floor velocity, turbulent flow k-epsilon. No rotating wheels (which could be quite an issue)
This is something that hardly can get into my brain.
Why 10m/s floor velocity? Why should you expect correct results?
k-epsilon? Why? KVRC uses k-Omega-SST.
No rotating wheels. Again, why?
You are using a very different CFD setup and complain that you cannot reproduce the results.
It is not even hard to change these parameters in most CFD software.
I suggest you change these things and have a look at what happens.
No, there were not big changes about the reasults between the different releases of OCCFD (less than 5% in df and drag).
About the stability. OCCFD still have important weak points during meshing (but the Linux version works significantly better). Anyway I prefer to launch the same analysis 5 times (luckily if it has to fail, it happens in 15 minutes) than to run it one in a not official boundary box, with different vorticity laws.
About the time: I had to pause the development of my car a week before the second race and untill the end of July beacuase of job and family issues. Not a big problem, I had fun to imagine new solutions without a workstation for two months So take it easy!
Hope you are not really going to leave the KVRC, I'm preparing a car with a big surprise for the 4th race, you will not want to miss it?
LVDH wrote:This is something that hardly can get into my brain.
Why 10m/s floor velocity? Why should you expect correct results?
k-epsilon? Why? KVRC uses k-Omega-SST.
No rotating wheels. Again, why?
You are using a very different CFD setup and complain that you cannot reproduce the results.
It is not even hard to change these parameters in most CFD software.
I suggest you change these things and have a look at what happens.
10m/s floor --> increasing floor velocity should decrease boundary layer under the floor and in general increase air velocity under there. This would increase rear downforce. However the greatest discrepancy from the official results is focused in that area...therefore i try to be "pessimist" with my own results decreasing floor velocity. Do you see why i'm doing that now?
k-epsilon is all i can have from Khamsin. I thought OCCFD used epsilon anyway...especially judging from the post processing photos previously published...
No rotating wheels because Khamsin doesn't get that rotating wheel plugin for free.
Anyway, the problem is not the result difference itself...as long as that that difference remains pretty much constant, like it has happened during the first two races. Why isn't that happening this third round??
CAEdevice wrote:No, there were not big changes about the reasults between the different releases of OCCFD (less than 5% in df and drag).
Understood. Thank you.
CAEdevice wrote:Hope you are not really going to leave the KVRC, I'm preparing a car with a big surprise for the 4th race, you will not want to miss it?
mmmh intriguing! Hope i'm not leaving too! Depends wheter i can find the time and solve solver problems
I repeat that i'll race only if i'll be convinced to have produced a good machine.
LVDH wrote:How much % do you have to redesign because of rules and how much because your car is too draggy for the next races?
I can keep the airfoils and the canopy. And that's it.
(drag doesn't worry me that much: that car would be flexible enough to get around that problem)
I know, I know, but not getting to the podium is not an option for me. Nor it makes sense developing a brand new car completely blinded: i wouldn't learn anything.
