SimScale CAE Forum

Optimizing mesh quality and Y+ layer formation on an FSAE car


#357

Yes, MRF… If you want to keep it private, just share the private one with me…


#358

here it is i just made it public … shouldnt be a problem. Second to last run is our final run


#359

COOL, I can play with that :slight_smile:
Since you guys only saved your forces results every 5 iterations last year, I had to use ORSI 100i30mai to make it give similar results to every iteration saves:

This brings up a point, for ORSI, only ONE forces (preferably a coefficients) results item of all faces needs 1 iteration saves… For any others, there is no problem making them save every 5 or more (unless you want to ORSI them accurately, which you probably do, so you can ignore this note that could save core hours…), which would save core hours during the sim run…

EDIT: What FSAE team are you with and how did you do last year :question:

Also, you really should try my MRF idea…


#360

So based on the ORSI of the last years car, we did have 1% stability in the results… thats good to know

Yes for the coefficients i will do 1 iteration saves. I do have a decent amount of core hours, I dont plan on using a lot as i need to do maybe 10 -15 full car simulations total.

I am with the FSAE team Raceyard from Kiel Germany, The account is Raceyard53 as you have probably seen. We did ok last year, there are always problems with the car working properly/being completed on time. The team has a lot of advanced systems but isnt very good at time management or keeping deadlines. This caused the 2018 car to not run at any events and last years car only started working towards the end of the season. Its too bad because we loose a lot of points at the design events due to having nothing on the car validated.

I would love to try your MRF idea. I read also the Curved wind tunnel thread Cornering Aerodynamics with Simscale

And the PDF within this thread Here

In the PDF it shows the following graphs towards the end. While i would love to experiment with the model in a curved wind tunnel, I have a few reasons NOT to peruse this test which are explained at the end.


With this summary at the end;

Summary
The cornering condition can have a significant effect on aerodynamic
performance of vehicles but is unable to be represented
experimentally. Experimental solutions developed for dynamic
motion of aircraft are not readily adaptable for automotive bodies.
The close ground proximity, high blockage ratio and the specific type
of motion all add complexity which increases the difficulty of
achieving the required flow conditions.
Fortunately numerical simulation permits investigation into the
condition. However aerodynamicists must remain aware of the
change affected by this condition. Numerical simulation must
accommodate curved flow occurring within a non-inertial reference
frame, and this requires additional considerations when constructing
such models. Due to the motion, drag begins to act in a curved path
and variation in Re occurs within the domain. Results highlight the
importance of adopting the correct analysis techniques when
evaluating aerodynamic performance for cornering vehicles.

Reasons not to do Curved wind tunnel test

  1. This difference in data, to me, seems fairly minimal in comparison to a straight wind tunnel.
  2. These tests on the Ahmed body are at 25 m/s. This is quite fast for what our car will be seeing on the track. My tests would include-

a. Official skid pad event inner radius at 7.625 meters

b. Low speed skid pad test at around 4-5 meters
c. High speed skid pad test at around 10-12 meters

  1. (cont.) Therefore, outside of the higher speed skid pad, i do not see our cornering speed exceeding 20-22 m/s as most of the turns on the FSG track are slower speed turns. I have been using 18.06m/s for previous inlet velocity tests because this is the average speed for all turns on the track. Optimizing the car by simulating low and high speed cornering situations was not possible last year because i was the only one in the Aero team for the design phase. I simply didnt have enough time.

  2. Speaking of not having enough time, my last reason is just that. If we can justify that a straight on simulation ( including the pitch, yaw, roll, and steering input changes for each skid pad trial) are enough to have accurate results within reason. I would vote to leave it as a straight wind tunnel simulating the car in those pre-set cornering angles.

Basically - is this worth the trouble? Are the results that much more accurate that we NEED to do this?


#361

Let me start a simple project to explain and test the MRF setup that I have suggested for cornering flow analysis…

At that point can answer your questions…

Then I will start a new topic on that where we can have another record length discussion on it :wink:


#362

I mean if YOU wish to proceed with this test, while i continue with the same cornering simulations in a straight wind tunnel then we will have an excellent data set.

I just dont want to commit to this extra work when i still have the rest of my research for vehicle dynamics to do, get the car prepped for testing, and start writing the damn thing. I really dont want to make this stressful at the end.

P.S. I still really do want to do the curved tunnel

EDIT: I should really send you the full solidworks car model. I designed it with adjustable suspension so changes to geometry can be made quickly The tutorial for what i did is here


#363

No problem, parallel is good…

EDIT: Rhino imports Solidworks fairly well, or I can steal your 2020 wheels turned geometry at some point from a SimScale project if it is step format imported…


#364

Ok whatever you want. If you have the original file you can change the pitch, roll, yaw, ride height, and steering angles to whatever you want. Im not sure if an imported STEP can do that. Either way ill be importing the cornering geometries for each case in the next couple weeks.


#365

Give me an approximate turn radius you want please…


#366

Official FSG skid pad size for 2019 is:

7.625 meters inner radius
3m track width

so

Approximate center of car radius is

9.125m turn radius


#367

Wow, small, what speed are you hoping to drive around that with all this downforce you have (without drifting or do you drift too)?


#368

haha no drifting, although that would be sweet. Im not exactly sure what speed yet i have to talk to who was in charge of last years vehicle dynamics development. This is where i have to decide to do real world testing first, then use that data to as inputs to the CFD or vice-versa. I can get approximate data for this but the goal is improving our times/lateral acceleration not just comparing results


#369

Just a rough speed for now is OK…

WOW here are the 01234 U plots, very interesting (I will add layer lines soon, they take time as I have to draw them):


.

.

.

.


#370

wow this is awesome. Three things i notice.

  1. Looking at the underside of the wing, the separation location is about the same for 4,3,2, layer sims but in the 1 layer sim it stays attached longer.

  2. Looking at the outbound wake direction (far right of the picture) i notice that 3 layer and 1 layer are turned more upwards, while 4 and 2 layer are straight or downward facing.

  3. At the bottom of the wing, where the Dark red area is around the gray area, in the 4,3,2 layer tests this dark red is about the same size while in the 1 layer it grows bigger

Any thoughts on this or would you consider this just variations in flow?


#371

I am not sure but the 0 layer obviously has to be significantly different because of the major difference in CD and CL with ANY layered mesh…

I am showing the full solutions sets at the iteration that closely matches ORSI values, so that removes that issue (some even match to 4 significant digits :slight_smile: )

The CD and CL of all layered meshes are pretty close to each other and I don’t know if we are seeing true differences between # of layers in the U mapping or flow patterns yet… I hope we are, as that would be awesome…

What I find really interesting is the fact that 0 layer flow is almost not separated at all!!! (just a little in the camber area)


#372

Yea i forgot to say that your ORSI% difference values are spot on!!

Im not sure if its flow patterns or not, but for me… three simulations have one point of separation, and one has a later point. Unless this is a fluke, this points to something. Maybe running the 0,1,2,3,4 test again with the wing at a different AOA? See if the pattern of later separation for the 1 layer simulation is still there?


#373

AHA, more reason to do a dual test on existing car, cornering speed and do some INVALUABLE tuft grids at fixed speed here too :wink:


#374

yes the tuft test will most likely be done first at a straight line run for a basic setting. Then at each skidpad radius and comparing the best overall setting. We can change the AOAs for each event though.


#375

Maybe if the 2,3 and 4 layers pointed to the fact that they are giving better CL and CD values that could be related to having a slightly further forward separation point on that surface, but they do not seem to point that way with the way CD and CL vary over 2,3,4. Also 3 and 4 layers get that 2nd eddy developed in the camber region, so double whammy for predictions of what is happening here…

Did you not tuft test those wing packages when they were put on to begin with, that largest chord wing just looked wrong incidence angle to me before I saw CFD results :question: Tuft testing will not tell you that you have the wings set for most downforce but I think for sure they won’t be at most downforce incidence with separations like that.


#376

I would have liked to do a tuft test but literally all of our testing time from last year was just getting the car to work. Then came suspension setup and torque vectoring. In terms of a driving car, aero is last on the list

This wing profile was from the 2018 season, developed by the previous developer. Same for the front wing. Knowing i had no help for aero design, i had to make decisions on where to invest my time, so i left these alone. I focused on a turning simulation with many different geometry setups to improve overall performance.

You have the 2019 cornering simulation project now. I had about 20-25 different geometry changes to test how the aero packet worked and influenced the packet as a whole.

The current aero developer for the 2020 car has already noticed and changed this wing profile