http://listenonrepeat.com/ watch/?v=X2Q_udvSakg#Col - - PowerPoint PPT Presentation

http listenonrepeat com watch v x2q udvsakg col umbia
SMART_READER_LITE
LIVE PREVIEW

http://listenonrepeat.com/ watch/?v=X2Q_udvSakg#Col - - PowerPoint PPT Presentation

http://listenonrepeat.com/ watch/?v=X2Q_udvSakg#Col umbia_FSAE_Metlife_Stadiu m_Autocross Making R acecars Go Faster with R Matthew Sheridan Overview 1. Vehicle Dynamics 2. Some Stats 3. Results! Ways to Make a Racecar Fast Big Engine


slide-1
SLIDE 1

http://listenonrepeat.com/ watch/?v=X2Q_udvSakg#Col umbia_FSAE_Metlife_Stadiu m_Autocross

slide-2
SLIDE 2

Making Racecars Go Faster with R

Matthew Sheridan

slide-3
SLIDE 3

Overview

  • 1. Vehicle Dynamics
  • 2. Some Stats
  • 3. Results!
slide-4
SLIDE 4
slide-5
SLIDE 5
slide-6
SLIDE 6
slide-7
SLIDE 7
slide-8
SLIDE 8
slide-9
SLIDE 9
slide-10
SLIDE 10

Ways to Make a Racecar Fast

slide-11
SLIDE 11

Big Engine

slide-12
SLIDE 12

Downforce

slide-13
SLIDE 13

Really Good Downforce

slide-14
SLIDE 14

Downforce Gone Bad

slide-15
SLIDE 15

High Power Low Drag

slide-16
SLIDE 16

Lower Drag Even Lower Power

slide-17
SLIDE 17

Bolt is not grip limited right now…

slide-18
SLIDE 18

Sliding = Slow

slide-19
SLIDE 19

Cool, but Slow

slide-20
SLIDE 20

Formula SAE Course

slide-21
SLIDE 21

We care most about Lateral Grip

  • Tune to Win
slide-22
SLIDE 22

What about weight?

No Area Term?

Coefficient of Static Friction varies with weight, pressure, temperature, and camber angle

slide-23
SLIDE 23

Camber Angle

slide-24
SLIDE 24

Want max grip while also having drivable car

  • Tune to Win
slide-25
SLIDE 25

I Don’t Trust Software that’s not R

slide-26
SLIDE 26
slide-27
SLIDE 27

Printer + Pen + Scanner + MS Paint

slide-28
SLIDE 28

Magic Formula

y = sy+Dsin(C atan(B(x+sx)-E(B(x+sx)-atan(B(x+sx)))))

  • y = side force output, x = slip angle input
  • Sy = vertical shiff, sx = horizontal shift
  • B = Slope from origin
  • C = horizontal bounds of sin wave
  • D = peak value
  • E = curvature factor
slide-29
SLIDE 29

But in a billion dimensions nah….

slide-30
SLIDE 30

KNN Regression

slide-31
SLIDE 31

Local Regression

slide-32
SLIDE 32

Thin Plate Smoothing Spline

  • Actual MAGIC!
  • Minimizes

cross-validated residuals while also minimizing number knots in splines

  • MGCV Package

Simon Wood

slide-33
SLIDE 33

Nonlinear Least Squares on Magic Formula

slide-34
SLIDE 34

Trying GAM

Y = s(SA) + fz * SA^n

slide-35
SLIDE 35

Loess, Locfit

slide-36
SLIDE 36

More Locfit

Turn nearest neighbor value to n = 0.1

slide-37
SLIDE 37

Play with NN

slide-38
SLIDE 38

KNN Regression

slide-39
SLIDE 39

Locfit

slide-40
SLIDE 40

GAM

slide-41
SLIDE 41
slide-42
SLIDE 42
slide-43
SLIDE 43

Comparison of Tires

slide-44
SLIDE 44

Camber Interpolation

slide-45
SLIDE 45

Why don’t you try interpolating FZ too?

slide-46
SLIDE 46

Scale the Z

slide-47
SLIDE 47
slide-48
SLIDE 48

Questions?

A.matthew@Columbia.edu