[s-cars] Fuzz logic vs adaptive learning

[QSHIPQ at aol.com]

Kirby:
Comments inserted
In a message dated 1/24/2004 1:25:03 PM Central Standard Time, 
kirby.a.smith at verizon.net writes:
Two points:

>a) The A'pexi learns what WGFV duty cycle to use to achieve its
>commanded boost pressure (gauge) vs. rpm and gear, and throttle.  When
>conditions change, it has to tweak what it knows.  Overboost occurs
>temporarily under colder conditions, where it is more easily tolerated. 
>Underboost occurs temporarily when it warms from cold conditions.

I only speak to the difference between Adaptive and Fuzzy logic.  Adaptive 
can only plot in the known points, if you have 16 options to program in each 
gear for instance, it can only "learn" 16 boost values (based on desires vs 
inputs).  Fuzzy logic is capable of much more.  You can program 16 'desires', and 
FL cannot only learn and assign values to those 16, it can take unnassigned 
values between the 16 and assign a boost profile to them.  

>b) If we had a motronic that controlled the top of the WG, and could be
>programmed by gear to make up for Audi's little embarrassing mistake in
>tranny design, then I would agree with you wholeheartedly, rather than
>just conceding that motronic is lower risk to the engine, but not to the
>tranny unless OEM programmed.

Again, not a hardware problem.  Mac 11 used a top feeder AND Mac 11/14 used a 
first gear lower boost (the wire is snipped if it's a 5spd).  I would also 
venture to say that no FV boost intervention in first gear is probably better 
than *any* control.  Let the turbo ramp itself, even a k24 won't reach close to 
a maximum boost map profile in first gear without FV assistance IME.  Want a 
top feeder valve?  You are darn close to just inverting the values assigned to 
a bottom feeder.  Again a software issue.

Scott J