neural network and/or fuzzy logic

[Phil Lamovie]

Hi All,

Jens wrote "my major interest is to use a neural network
                      and/or fuzzy logic "

You made reference to fuzzy logic allowing you to dispense
with the dyno so I'll assume you mean for closed loop fuel & spark
control.

My first question is do you plan to measure engine torque
with a rotating and transmitting load sensor ?

If not then fuzzy is not the way to go. the only algorithm that
comes to mind is based on torque peak i.e. MBT per cylinder.

You could feel up one of the headbolts but that will give you
a pressure peak reading that will always be higher  when the
engine is knocking.

Have you considered using a GPS these have quite good
feed back or perhaps  a little bit of code to run FFTs on
the doppler shift from your home built radar array.

Then again you could fit an accelerometer to a perfectly good
barometer and just tune using accel runs at diff throttle settings.

My personal choice is still a speed density system that
has a lookup table calibrated on a dyno.

It seems to me that if you would consider running an " unmapped "
engine you would also prefer the Dentist to dispense with the boring
x-ray and just drill away. I predict the results will be similar.

If your engine mapping is only 10% out you will burn at least 10 % if
not more fuel. If you do the sums I'll think you find that the cost in

wasted fuel is much higher than an hour or so on a chassis dyno.

phil