I was just curious if there is any work being done on a system that would self tune itself. I remember reading some information over a year ago with regards to the subject, but was wondering if there was any new information?

Is the reason its so hard to do is accurately judging the power/systems computation/AFR or something else. I would think the routines to make a program would be fairly straight forward. We have CNC machines that adjust for wear on the cutting tools, it would be sorta along those line.

I realize it would be much more difficult than I am making it sound, but surely there are more complicated systems out there that have return loop feedback/adjustments.

Again I was just curious.

 

Are you thinking of VEMaster? There is also another one, that I can't remember. But they were more for tuning the VE Tables for an SD based vehicle as I recall.

 

No, I mean something that would conitually modify the tables to find optimum settings for spark, fuel etc. Sorta what we all do, but it would do it real time, so we don't. Then it would remember the setting for future. Kinda eliminating what we have to do.

Don't bash me about that is why we are in the hobby, etc etc.

 

Nope, I won't do that. However, it's been my experience that computers are great for doing computations quickly but they lack intuition or intellegence.

On some GM ECMs, they have "Low Octane" routines (SD) does. The intent was to run the highest spark possible, but if you got "bad gas", the ECM would lower the spark advacne to compensate for the lower grade.

Like Communism, it's a wonderful theory. But in actual practice, I found it to be a pain in the butt. I found it better to disable the Low Octane Routine and just control the spark myself. My SD car runs far more consistent with the Low Octane disabled.

I llike to think that I know more than a computer. Maybe that is why I still do my VE Tables myself, instead of through a program. I can make a determination if a particular piece of datum should be part of the "set" or excluded due to other factors that don't make it truly representative.

 

And when it's all done, lets call it Block Learn Memory

Just kidding. I assume you're more interested in WOT. You would need some feedback like a W/B O2.

 

Speed Pro, and some of the other aftermarket ECMs do have the capability for WB-O2 closed loop, including WOT.

If you got crazy enough i'm sure you could write code to make a GM ECM do the same.

 

I'm sure you were kidding, Right.

I mean something that would constantly be adjusting the timing, fuel etc while monitoring the power output somehow, then remembering that rpm, spark, fuel, load etc for the next time. It would have to be pretty complex to be able to do it, but back in the 60s I'm sure if you would have asked tuners if this was possible they would have laughed.

Just thought it would be interesting to see what is going on..or if anyone has some inside info on new late breaking systems.

 

If I remember right, autotronics was the company that was working on that....

 

Ski,

IMO, this is the crux of the "problem". to do something like this you would have to have both inteligence to run this routine & a standard by which to measure gain/loss. I dont think that the intelegence would be that big of an issue. I think that the issue would be the measurement of gain or loss.

if all you wanted to do was to dial in VE's, then I think that a "calibration" mode could simply be activated & the ECM could run this routine & dial in the VE's automatically, the "gold standard" would have to be in this case the O2 sensor. IMO, this is something that folks ought to look into (aftermarket, Holley, Accell, FAST, etc, etc). this would make the dialing in for the street driven cars much simpler. I dont think that it ought to be something that is continually ongoing, thats what a BLM & INT are for. I think that they do a good job of "thier job". the issue here is what happens as the O2 degrades? it kills the tune & it should not be.

I also think that those folks that support a WBO2, ought to dial in a WOT setting but I would think that for 2 reasons this would be touchy @ best. the first is that they are relying on the basic tuneup being close & the 2nd is that they are relying on the WBO2 being both "right" & fast enough to self compensate. I think that both assumptions are not safe ones to make. I know that this is offered, & folks use it, but from the folks that I talk with they dial it in & then turn this feature off.

to take this the other direction though, I am willing to bet that the big 3 already have this, they just dont issue it out with the auto's that they sell. It seems to me that its a logical jump from dialing in all the VE tables manually to doing it automatically & you could cut the tune time on a new car/motor/cam/head/exh/....... by a BUNCH with an automated tool like this. With flash proms & real time ECM's that ought to be a simple thing to do. then they can fine tune the drivability stuff automatically.

I bet that some enterprising young person with the smarts to write simple routines would be able to pull this off in a GMECM, say an old MAP setup like mine or a MAF like yours. cut all the smog stuff out of it & add a routine that checks an input bit for the "go tune it" bit & then run the routine. it seems to me though that with the VE master (I have only worked on my SD setup) that its taken most of the hard labor out of the fine tuning.

thats my 7 1/2 cents.

BW

 

The fuel is the easy part. It's the spark that makes self tuning a PITA. I think it's very possible. The code would need to recognize slow smooth throttle positions (already does this with highway mode) and then advance the timing until knock is detected, then store the highest value minus 2 degrees (safety net) in that cell. I would hope that it would have some better logic built in that would be able to interpolate values. It would need to understand that at 100kpa there should always be less timing than at the same engine speed but lower MAP. If it couldn't do that then things would get messy.
The wideband o2 sensor would be used for fuel trim. I don't think it could ever get the pump shot right no matter how fancy your coding skills are. Other than that I think it could easily learn your Vol Eff tables. The problem with all of this is that a computer is only as smart as it's programmer. In this case you'd be the programmer so why not just tell it exactly what you want.
For our formula team we're working on some self tuning software for when the engine is on the dyno. Having a dyno is a lot easier for self tuning programming since the dyno can measure torque, hold an rpm and control throttle position.
An easy way to use a wideband for vol eff is to run in closed loop and output inj pulse width. Log the data and you'll have a vol eff table in no time.

 

Just need a torque measuring devise.
ie driveshaft deflection, in chamber pressure transducer.
Or a WB, K/S and possibly EGT.

Probably see it in OBDIII.
Or something real close.

 

We have the wideband o2, we have the torque measuring load cell, we don't have a k/s, we have EGT for each cylinder and we're getting in chamber pressure transducers . The problem is setup and what cars come with all of these sensors? I know of a few that have a couple of the above but none with all. Like you said, probably OBD3.

 

That is what I was thinking of.......some way to measure cylinder pressure would be the cat's butt, along with a knock sensor....man that would be awesome.