+ Reply to Thread
Page 2 of 4 FirstFirst 1 2 3 4 LastLast
Results 11 to 20 of 38

Thread: Testing, Testing 1,2 ...

  1. #11
    I chucked a couple of shifted TLLTD values through. Anything lower than about 50% TLLTD struggles to put power down at the higher speeds due to aero drag. A locked diff would help to power down, but as with a lot of things its implementation into my sim is on the horizon.

    The Roll by Ay plot has been brought to life, but it also doesn't play too nicely with the given transfer function. It looks about right to me, so am I just completely missing something?

    With the yaw velocity vs roll/Ay, I'm not too sure exactly how you can get a cross over point (at least with a FSAE car). When running at different speeds the roll by Ay plot doesn't really change much, but yaw velocity gain increases with speed. Unless I'm missing a unit conversion somewhere it seems like these two plots are a hop, skip and a jump away from being able to intersect. A guess at what you're getting at though is if a vehicle has its peak yaw velocity response at the same frequency (and magnitude) as its Roll by Ay, when driving over rough road/side wind gust/under acceleration (assuming torque steer is a player), these two effects will constructively interfere and make the vehicle get a progressively bigger and bigger tank slapper going on.
    Attached Images
    Michael Geist
    Monash Motorsport

  2. #12
    Senior Member
    Join Date
    Mar 2008
    Location
    Brighton, MI
    Posts
    618

    Yawn

    Sweet ! Roll just looks heavily damped. Wouldn't your car have some peak roll just from tires, alone, though?

    On the other question, roll probably doesn't have a lot of speed dependency while yawrate has quite a bit of. On some production cars, there can be a speed where the peak frequencies cross over. Above this speed roll steer signs invert: rear roll understeer behaves like roll oversteer (which ain't a real good feeling at 180 kph). Corvette is a good example. So, some manufacturere adopt a policy in which roll oversteer is designed into the rear "because it feels so good above 180 kph", other run some seriously stiff roll damping to attenuate the problem and others just can't figure it out.

    Problem with the rear roll oversteer theory is that your base cars suffer from slow and lazy responses, require extra stiff rear tire stiffness to get the rear cornering compliance back down and your signature cars acquire a reputation of loosing control by "speening out" and crashing.
    Last edited by BillCobb; 12-19-2015 at 03:32 PM. Reason: spilling air

  3. #13

    Scratch that

    Yep, my damping values were a bit off. I gave it another go with a few changes and it's looking a bit more believable. As for now, I think it's time for some eggnog.
    Attached Images
    Michael Geist
    Monash Motorsport

  4. #14
    I managed to get a simulation up and running. It is an 'idealized' model with no compliance effects and very primitive suspension. Regardless, lets give this a go and see what we get.

    The test conditions mirror what Geist ran - 100 km/h with 3 degrees of steered angle. For interest sake, I cranked up the chirp frequency up to 6 Hz just to see what happens at higher frequencies. The regression fits to the magnitude, but to double check I plot the phase as well.

    Here is what I got.

    EDIT: Re-uploaded plots to be reasonable sized
    Attached Images
    Last edited by turtle; 01-30-2016 at 06:16 PM.

  5. #15
    Senior Member
    Join Date
    Mar 2008
    Location
    Brighton, MI
    Posts
    618

    Looks Good !

    Use the "unwrap" function to remove the jump in phase plots to make them continuous. To include the effects of suspension compliance, incorporate their action as a calculated loss in tire cornering stiffness. Cornering compliances as equivalent tire stiffnesses, makes this analysis easy. It also reveals how the best tires in the world are easily ruined by having suspension compliances that wash them out.

    If you apply this technique to real world road testing, try to develop the finite width pulse step steer input technique instead of the chirp. Your driver will appreciate the shorter segment duration. The steer PSD is different, but as long as there is some power in it, your transfer functions will be revealing.

    Since you have produced a ~zero frequency gain value, calculate the car's understeer from it.

  6. #16
    Just for fun, I applied the same technique to some random data I have laying around. Formula North has some nice slaloms for this purpose. Near constant forward velocity, equal cone spacing and nearly equal amplitude left-to-right. Combined with a driver that likes to wrestle with the steering, I get a response that vaguely represents a pulse step steer if you squint.

    From the DAQ data, the forward velocity is 45 km/h with the steering oscillating between 7 degrees-ish of steering. The lowest logging frequency is 20 Hz, which in hindsight isn't ideal. The analysis suffers and only the response below 3 Hz is observed.

    The phase plots are questionable but the magnitude plots could be believable. Lets say the steering sensitivity of the car is 0.135 g/deg just inspection of the steering sensitivity plot. Cracking open a trusty copy of Milliken & Milliken, I can find the DC understeer. The final computed value: 1.5 deg/g.

    Some quick observations: Both the steering sensitivity and yaw-by-steer DC gains are a lot lower than what the simulation predicts. The yaw-by-steer response is heavily damped and the roll-off occurs much earlier in reality than in simulation. The data is sketchy at best. This is a car that has been accused of "understeering like a pig".

    EDIT: And only now do I realize I am comparing two datasets with different operating conditions. Oops.
    Attached Images
    Last edited by turtle; 02-01-2016 at 12:45 AM.

  7. #17

    Revival

    I'm surprised this thread and its counterpart over in the 'Static Events' section haven't gained more traction, considering Bill is serving it all up on a silver platter. Anyhow, I thought I'd chuck some more plots in the mix to get the ball rolling again.

    Building upon the frequency response plots, I've thrown together the lateral acceleration and yaw rate phases for both 60 km/h and 100 km/h. The Mean Phase Difference can then be computed (yaw - lat accel). You can then churn out the 0 Hz gradient of the Ray function to come up with a rear cornering compliance values of 1.94 and 1.71 deg/g for 60 and 100 km/h respectively. It's a bit of an academic exercise considering it's using tire data and a vehicle simulation instead of real on-track stuff, but it's a bit of fun nonetheless. It seems like the forum isn't allowing me to upload images no matter what I do, so here are some links.

    Phase plot - 60 km/h - http://i.imgur.com/2xVAu0C.jpg
    Phase plot - 100 km/h - http://i.imgur.com/IGLt72U.jpg

    Shifting focus slightly, I've temporarily dropped the non-linear tire model for some nice constant cornering compliance values. The result is a carpet plot of vehicle yaw damping and lateral acceleration response time for various rear cornering compliance (DR) and understeer (K) values. The yaw velocity peak to steady state ratio is from the same old frequency response stuff and the lateral acceleration response time is from a step steer input (calculated from 50% steer to 90% steady state lateral acceleration). The results are pretty much what you would expect, being that while you're cooking up a new vehicle and choose the wrong recipe, you (or the driver) are gonna have a bad time.

    Carpet plot - http://i.imgur.com/XdGIUCW.jpg
    Michael Geist
    Monash Motorsport

  8. #18
    Senior Member
    Join Date
    Mar 2008
    Location
    Brighton, MI
    Posts
    618

    Re-Revival

    Not really surprized by the null conversation about these synthesis tools. That's because, really, there are more mechanics than engineers involved in this pendeavor, and the engineers appear to be more fluent in CAD than CAE. That essentially makes their cars 'kit' versions, because a design process to meet specifications alludes them. They just hope their cars don't fail in mass, durability, and performance. ("How much does my skin made up of little triangles and squares weigh ?")

    What makes this more difficult is the lack of competitive performance analysis (Called "fingerprinting" in the trade.) You measure the BOB and WOW characteristics (Best of the Best and Worst of the Worst) of several or many different versions of a vehicle and weigh the merits of each design feature based on performance. Then you assign values to your 'winning' choice and flow down the key attributes from the metrics into sub-system and component specifications (I.e Rear cornering compliance and aligning moment compliance steer given a tire cornering coefficient or stiffness. Your (Excellant BTW) carpet plot of performance surface features is esentially the beginning of a good vehicle design. Cost, mass, manufacturability, strength and specifications then assign a pin-point on this map.

    Yaw velocity peak to steady state ratio at some (high) speed is a desireable objective metric because it is directly obtained from a step or frequency response test data processing program. But, just in case you are stuck with a controls engineer at the helm, it is easily transcribed into 'Yaw damping', yaw dampin', or yaw dampink' by the following function:

    function y=po2zeta(po)
    y= sqrt((log(po/100)^2)/(pi^2 +(log(po/100)^2)));

    so 4.32% overshoot is a dam pink value of 0.707. Sound familiar ? (I thought so...) At Lateral Acceleration Response time is a good counter measure because it help constrain the cornering compliance specifications. You can actualy make it too fast (as in there's too slow, but don't over-do it because once your driver can't sense it, that's short enough).

    C'est la vie.

  9. #19
    Senior Member
    Join Date
    Feb 2011
    Location
    North Carolina
    Posts
    111
    I think there also isn't many posts here because on most teams you usually have 1 to maybe 3, if you're lucky, team members working on vehicle dynamics projects (plus building the car and school work), so for this time of year you're probably not gonna hear much from them. I know when I was in college from January to May I would just occasionally check the forums here. I also think a lot of people don't necessarily want to post on FSAE.com despite reading it because of the high likelihood of being called an idiot, lazy, etc... So you have a lot of people who just view and learn without actually joining the fray.

    Hopefully I can play around with this here soon since I have two off weekends this month.
    Trent Strunk
    University of Kansas
    Jayhawk Motorsports
    2010-2014

    Now in NASCAR land. Boogity.
    Opinions Are My Own

  10. #20
    I would love to post something here and contribute a meaningful amount, but I imagine like others; so far I have only been a student of this thread. I have been learning Matlab for a relatively short time and within my team I'm currently (and have been for a few years) the sole student on kinematics/dynamic behaviour.

    I enjoy this thread and check up on it regularly, so don't feel your posts are ignored. I for one appreciate the insight given and look forward to more.

+ Reply to Thread
Page 2 of 4 FirstFirst 1 2 3 4 LastLast

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts