It is currently Tue Sep 25, 2018 7:27 am



Post new topic Reply to topic  [ 7 posts ] 
Author Message
 Post subject: Self learning
PostPosted: Wed Jun 19, 2013 10:00 am 
Offline
User avatar

Joined: Sat May 11, 2013 9:45 am
Posts: 729
Location: PA, USA
In a flash rework email discussion a couple comments have popped up making me think a wider discussion might be needed.

The questions were around how the FW itself alters the flash info for self learning...and my first thought is that should NEVER EVER happen under ANY circumstances because it's WAY to dangerous to implement in generic type open source ECU.....but I tend to be the nervous type so I want to put this out to the group.

We currently can use the VeAnalyze feature in TS which will look at a lambda table and O2 readings to correct the fuel table, then it allows the user to decide whether or not to accept the changes....but obviously the tuner needs to be connected to do this.

We can in theory and probably should add a closed loop control feature to the ecu itself but normally this doesn't alter flash it just applies a correction to the final pulse width.

This is where I thought o5e would end and any kind of self learning in the ECU was beyond the of scope of the project due to the complexities I think it brings. I think we have about 2-4 years of work still ahead of us as it is before we have a well tested, full feature ecu that can do what it's told to do.

My guess is it would be another 2-10 years on top of that to get it to where it can do what's right regardless of what the user told it. Maybe I'm making this harder than it needs to be but I think I see a minefield of issues/risks that need to be resolved here.

so...thoughts?


Top
 Profile  
 
 Post subject: Re: Self learning
PostPosted: Wed Jun 19, 2013 10:57 am 
Offline
User avatar

Joined: Wed May 15, 2013 2:57 pm
Posts: 33
I do not think that self learning without computers is that helpful for the short term. Datalogging and off line calculations is a better way of doing the fuel calibration.

Long term drift is another topic where this might be helpful.

Closed loop lambda control is very helpful and needed (at the very least for the ones using catalysts) and quite good for street mapping on the first drives..

I think that a self learning table based on the lambda feedback values might be worthwile. It is easy to cap the values with a min/max table (safety). Then fueling does not have to wait for the lambda controller to converge. I think that a plausability check for the lambda sensor might be good also.

This self learning feature is however final polish.. So short term drop it.


Top
 Profile  
 
 Post subject: Re: Self learning
PostPosted: Wed Jun 19, 2013 12:29 pm 
Offline
User avatar

Joined: Sat May 11, 2013 9:52 am
Posts: 304
Location: Over here, doing 'over here' things.
The only thing I would consider using self-learning / auto-tuning for is establishing baseline fueling tables at first start-up and no-load / light-load conditions and O2 sensing would be a must for this. Something where at first start the ECU / tuning software has you go run the engine through its RPM range at 250-500 RPM increments and records the necessary fueling to achieve stoichiometric at each point. Beyond that, fueling would need to be hand-tuned for things like boost, acceleration enrichment, et cetera.

Regarding firmware updating itself, as in the values in flash for the running engine, I don't see a problem IF something like the variables / data stored in flash are copied to RAM and RAM is the working set so that if RAM is updated it can be copied back into flash without disrupting engine operation.

Glenn's mention of long-term deviations is a point worth making too. Many OEM vehicles incorporate fuel trim compensations, where if the ECU finds itself consistently either enriching or leaning the mixture outside of the pre-programmed values, over a prolonged period, the ECU stores an offset in a semi-permanent portion of memory so the calculations can be expedited. In essence, there is short-term trim which is in effect when the engine is running, and a long-term trim that is an average of short-term trims over a period of time.


Either way, I see it as a feature that can wait until there is operational firmware, running an engine and not blowing it up. ;)

_________________
/me goes off to the corner feeling like Jerry Springer with a mullet.

My O5E candidate: 1982 Honda CX500TC motorcycle.


Top
 Profile  
 
 Post subject: Re: Self learning
PostPosted: Wed Jun 19, 2013 2:41 pm 
Offline
User avatar

Joined: Sat May 11, 2013 9:45 am
Posts: 729
Location: PA, USA
glenn wrote:
I do not think that self learning without computers is that helpful for the short term.

.........

This self learning feature is however final polish.. So short term drop it.


It's not helpful at all at the moment, that's for sure. What I'm wondering is if it's would ever be helpful to o5e...and I'm not sure it would.


As I said the question arose from the flash re-work project and ground word that was being laid setting up memory in a way that could allow it. To me this seemed like an expansion of the scope of the flash re-work in a way that probably doesn't provide any value......although I could see using it for error codes I guess.

The self tuning part of the question in my head is would this ever make sense on o5e or is it more trouble than it's worth...my 1st thought is it's more trouble than its worth because we are not OEMs trying to pass EPA testing by getting to optimal closed look mixture as fast as possible, or allowing you to not detonate your engine when you run regular instead of premium at 38psi boost.

I'm just afraid this path leads to a mountain of code and settings that are very application specific and can ever really be tested well enough to trust an I'd like to see it OFF the table...which brought me back to the first part of the question and whether or not there was value in the extra work to configure the FW to have access to permanent storage.


Top
 Profile  
 
 Post subject: Re: Self learning
PostPosted: Wed Jun 19, 2013 3:40 pm 
Offline
User avatar

Joined: Wed May 15, 2013 2:57 pm
Posts: 33
First off how many write cycles can the flash memory take?

If this number is large being able to store data permanently is a good idea.

Thinking engine total run time. Maximum boost, rpm etc.


Top
 Profile  
 
 Post subject: Re: Self learning
PostPosted: Wed Jun 19, 2013 3:46 pm 
Offline
User avatar

Joined: Sat May 11, 2013 9:45 am
Posts: 729
Location: PA, USA
glenn wrote:
First off how many write cycles can the flash memory take?

If this number is large being able to store data permanently is a good idea.

Thinking engine total run time. Maximum boost, rpm etc.


10k-100k I think I read but I'll need to check.

One thought is store in the battery backed up sram section so it's saved through normal power cycling but not truly permanently.


Top
 Profile  
 
 Post subject: Re: Self learning
PostPosted: Fri Jun 21, 2013 8:41 am 
Offline
User avatar

Joined: Sat May 11, 2013 9:45 am
Posts: 729
Location: PA, USA
mk e wrote:

One thought is store in the battery backed up sram section so it's saved through normal power cycling but not truly permanently.


It's looking like the battery back-up of the an sram section will make the HW spec list.

We aren't going to do anything with it just yet, but long term storage will be available for future use......you know for after we actually have engines running and generating long term data :)


Top
 Profile  
 
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 7 posts ] 

Who is online

Users browsing this forum: No registered users and 1 guest


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to:  
cron
Theme designed by stylerbb.net © 2008
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group
All times are UTC - 5 hours [ DST ]