Page 4 of 5 FirstFirst 12345 LastLast
Results 61 to 80 of 83

Thread: NN tuning help

  1. #61
    Advanced Tuner
    Join Date
    Feb 2016
    Posts
    431
    Mmm... nah not really.
    Its pretty much just a strategy to deal with VVT and emissions et all.
    Advanced Neural Network is a scary cool name and all but it is not like we are dealing with the Matrix here.
    If it was so advanced, it would tune itself for maximum performance and best emissions and make us all look like dunces.
    Its a looooooong way from anything like that.


    Quote Originally Posted by Monzsta View Post
    There's so much going on in the pcm we can't see that would scare you. The fact we can do what we are allowed to is amazing. The NN is more than a nuisance among a bunch of tables. Among Ford, Chevy, and Dodge the Dodge pcm is almost sentient. The NN is as close to AI in an engine controller as it gets. Dodge has built a brain that looks at everything going on and makes it's own decisions, regardless of tables or data. Think of it as a mini Watson in the background.

  2. #62
    If you don't turn it off, does it not decide to ignore changes in tables when it feels like it? The thing THINKS, man! It's COMING FOR YOU!!! lol, j/k. But it's certainly no slave to the tables. That's pretty neat in my book. The manufacturer's goals and ours are very different. Running what we would call a safe aggressive tune a calibration engineer could point out a hundred things that got violated like piston temp, emissions output, crank bending loads, ect ect. Plus, with the EPA having a gun to their head that controller better have 10 different ways to counter a bad sensor and not pollute excessively.
    Last edited by Monzsta; 04-14-2017 at 11:11 AM.
    Level 3 Master Chrysler/Jeep/Dodge/Ram technician
    20 years experience

  3. #63
    Advanced Tuner
    Join Date
    Feb 2016
    Posts
    431
    I don't think a lot of tuners get the NN's real purpose.
    Lets say you hold the engine at 2000rpm and 80kpa of manifold pressure.
    Adjust relevant cell in VE table to get AF where you want it. Now advance the cam 5 degs, VE will probably need a trim
    to correct AF, now retard cam 12 degs, will probably need another trim. Not to mention that to maintain that manifold pressure
    the TB will probably have to move. Now throw emission requirements into the mix where you need to swing the cam 40 degs or more
    and you can see that the old VE table with one fuel value ain't gonna cut it anymore, the CL will be chasing its tail constantly.
    So the NN is just a different way of calculating how much fuel to put into the engine.
    It doesn't even use the VE table and there are plenty of other tables not really used until you turn the NN off.
    Also don't think for a second that any of the NN stuff is for maximum performance. It is for maximum emissions performance.
    So the engineers can swing the cam around whenever and wherever they like to get the emissions where they need to be and the fueling will
    stay correct. That is the purpose. The cam timing they use, does P/T performance no favours.
    That is why most modern engines are quite lame down low despite running high comp ratios.
    Most all manufacturers go about complying with emissions in similar fashion.
    If you retune cam timing with performance as the main focus, it moves around much, much less...
    And so then also simplifies the engines fuel requirements.
    Last edited by Hemituna; 08-14-2019 at 03:20 PM.

  4. #64
    So how would you go about the cam tuning side of things? Is there a cookie cutter approach to get it close say for a 5.7 stock cam?

  5. #65
    Senior Tuner
    Join Date
    May 2016
    Posts
    1,134
    Low Baro table or something like that is what it's called. You can tell from the log if you're changing the correct table. I haven't spent much time on it but just moved mine closer to the values in the WOT table and it is definitely more responsive driving it.

  6. #66
    Advanced Tuner Blue Bee's Avatar
    Join Date
    Mar 2017
    Location
    DFW
    Posts
    330
    Quote Originally Posted by Abdulaziz_pe View Post
    what i discovered these days i tune the car on the dyno on 12.4 AFR

    on the street it goes to 13, i rework it to 12.4 or 12.5. next week 11.9 then i rework it to 12.4 then it goes to 13 afterwards !!

    especially 6.4 across all its models. could NN cause this issue ??
    Quote Originally Posted by Abdulaziz_pe View Post
    i tune on 4th if the car on 3.06 or 3rd if it is 2.82 or less.
    What Dyno are you using? I think something to point out with tuning you want to be consistent with pulls on the dyno, however making a pull having the transmission not in it's 1:1 gear not only skews HP numbers (especially on an inertia dyno) but your logged data points for tuning happen faster and logs will not be as accurate. You should always make pulls with trans in 1:1, it does not matter what what gear is the rear end. This is why people get frustrated and talk smack on dyno shops when they get numbers from two different places even if they have the same dyno, a lot of it IS the operator. An example would be the usual run of the mill inertia dynojet, you can't tune on those no matter what anyone says, the car is not loaded the same on that dyno as it would be on the street, so of course your tune is going to be off. If you make a pull in 3rd then another in 4th (if 4th is 1:1), your torque numbers are going to be way different, because you have used the trans to multiply torque. With that being said, tune the car on a proper dyno, trans in 1:1, then on the street make a pull in the SAME gear<<--very important, before you decide the tune is off and needs reworked. If it's tuned in 4th and make a 2nd gear pull on the street, things happen faster and your wideband might not display what you think it should fast enough. Don't mean to change the thread topic but I think this needs to be factored in no matter how you tuning so you are not chasing your tail between the dyno and going for a drive.

  7. #67
    Advanced Tuner
    Join Date
    Feb 2016
    Posts
    431
    The slower the ramp rate the better the AF reading.
    Wideband will always show lean in the lower gears, also until the entire exhaust heats up etc. etc.
    I pay little attention to the reading till gear 4-5-6-7-8 on road and when on the dyno ramp em in as high a gear as the dyno max speed will allow.
    Otherwise in top gear WOT they always go rich.

  8. #68
    Senior Tuner
    Join Date
    Feb 2006
    Location
    Canada
    Posts
    1,957
    Tuning my first Dodge today and wanted to say thanks for the thread guys....I actually understood this,lol
    2000 Trans Am WS6

  9. #69
    Advanced Tuner Blue Bee's Avatar
    Join Date
    Mar 2017
    Location
    DFW
    Posts
    330
    Screenshot (9).png
    Has anyone noticed this? Above is my car, 2016 6.4 Hemi, pretty much nothing under the NN tab, but below is a 2013 Jeep wrangler and there's all kinds of fun stuff to change! Actually almost all the new Dodge stuff has all this except my car! Even another scat pack charger same year as my challenger has all this...why is that?
    Screenshot (8).png

    Another thing to note about what was said earlier by Hemituna,

    Hemituna
    I don't think a lot of tuners get the NN's real purpose.
    Lets say you hold the engine at 2000rpm and 80kpa of manifold pressure.
    Adjust relevant cell in VE table to get AF where you want it. Now advance the cam 5 degs, VE will probably need a trim
    to correct AF, now retard cam 12 degs, will probably need another trim. Not to mention that to maintain that manifold pressure
    the TB will probably have to move. Now throw emission requirements into the mix where you need to swing the cam 40 degs or more
    and you can see that the old VE table with one fuel value ain't gonna cut it anymore, the CL will be chasing its tail constantly.
    So the NN is just a different way of calculating how much fuel to put into the engine.
    Probably doesn't even use the VE table as it is fairly ugly looking stock.
    There are plenty of other tables that are bypassed until you turn the NN off so why not the VE table too.
    Also don't think for a second that any of the NN stuff is for maximum performance. It is for maximum emissions performance.
    So the engineers can swing the cam around whenever and wherever they like to get the emissions where they need to be and the fueling will
    stay correct. That is the purpose. The cam timing they use, does P/T performance no favours.
    That is why most modern engines are quite lame down low despite running high comp ratios.
    Most all manufacturers go about complying with emissions in similar fashion.
    If you retune cam timing with performance as the main focus, it moves around much, much less...
    And so then also simplifies the engines fuel requirements.
    Spot on, anyone ever look into some of the newer GM stuff? You have to go up to edit and select the "Virtual VE" table? My guess this is basically the same kind of system, while in GM's VVE table if you change the cam timing by 1 degree the entire table changes. I wonder if NN could be laid out the same way, making it more usable for heavily modded cars...? Because by the way I read, some guys are turning off NN and going VE for "rowdy" cams, but with only 1 VE table (2 Bank tables, yes I know), going straight VE then proper way to lay down a repeatable tune is to also completely disable cam timing. As stated above, as soon as the cam moves, VE table is out the window unless it grabs another VE table with that set cam timing.

  10. #70
    Senior Tuner
    Join Date
    May 2016
    Posts
    1,134
    Ford maps the cam profiles to different "Mapped Points " for those of you not familiar with their approach. Each mapped point has different Speed Density, Spark, and Torque models.

  11. #71
    Advanced Tuner Blue Bee's Avatar
    Join Date
    Mar 2017
    Location
    DFW
    Posts
    330
    Quote Originally Posted by ridenrunwv View Post
    Ford maps the cam profiles to different "Mapped Points " for those of you not familiar with their approach. Each mapped point has different Speed Density, Spark, and Torque models.
    Yea the Ford stuff gets crazy time consuming, but that's what I'm getting at, we are still kind of left behind I think for getting into all the dodge stuff, but maybe those tables simply don't exsist.

  12. #72
    Senior Tuner
    Join Date
    May 2016
    Posts
    1,134
    Neural Network could probably be a complete solution. I had to work with them in a computational intelligence class in school and it's crazy how accurate of results you can get if you train them properly.

    This leads me to the problem that we would have making use of a neural network in non OEM calibrations... Training a neural network requires a known set of outputs so the calculated output can be compared to the actual output and the weights adjusted. The big question for me is how is the Dodge neural network making the determination of what it's calculated error is to properly adjust these weights. I have not had much time to look at what have been mapped for the inputs and layers of the neural network in the tunes but if I ever do and the scanner contains the parameters needed to dive into testing I will try to write a program or something and spend some time on it.

  13. #73
    Tuner Ohioc5's Avatar
    Join Date
    Apr 2008
    Location
    Ohio
    Posts
    93
    Let's bump this up.
    Here's a pic of my LTFT fuel mass. '16 Hellcat
    Wanted to make sure I understand. Some of you use injector pulse width and some use fuel mass?
    Looking at my LTFT if you wanted a 2% reduction in fuel I could multiply the Injector pulse width cell by .98 and make the inverse table match.

    My question is, could I do the same with fuel mass? Multiply cell by .98 for 2% and change inverse cell this would add 2% fuel

    Would I set my target AFR and then trim fuel mass.

    Did I read this correctly?

    I think I got it now, I re-read post #5.

    Last edited by Ohioc5; 06-20-2017 at 07:27 PM.
    2016 Hellcat - 800 Miles - It's winter

  14. #74
    Advanced Tuner
    Join Date
    Apr 2014
    Posts
    704
    Quote Originally Posted by Ohioc5 View Post
    Let's bump this up.
    Here's a pic of my LTFT fuel mass. '16 Hellcat
    Wanted to make sure I understand. Some of you use injector pulse width and some use fuel mass?
    Looking at my LTFT if you wanted a 2% reduction in fuel I could multiply the Injector pulse width cell by .98 and make the inverse table match.

    My question is, could I do the same with fuel mass? Multiply cell by .98 for 2% and change inverse cell this would add 2% fuel

    Would I set my target AFR and then trim fuel mass.

    Did I read this correctly?

    I think I got it now, I re-read post #5.

    If you would like to alter fuel mass instead of PW to decrease fueling, you would need to multiply by the inverse of the amount you want to trim.

    So, if LTFT was showing -2% and you wanted to pull 2% fueling via fuel mass value alteration, you would need to find the inverse of .98 (1/.98 = 1.0204) and multiply the fuel mass values by that.

    It's backwards from altering pw values because the PCM does fuel mass calculations and then uses the calibration table to find the appropriate PW for said fuel mass.

  15. #75
    Tuner Ohioc5's Avatar
    Join Date
    Apr 2008
    Location
    Ohio
    Posts
    93
    Thank you for the info, I appreciate it. Great information in here.
    2016 Hellcat - 800 Miles - It's winter

  16. #76
    Tuner Ohioc5's Avatar
    Join Date
    Apr 2008
    Location
    Ohio
    Posts
    93
    I may be too picky for NN, but thank you for the thread everyone.

    Not excited about initial logs.
    Last edited by Ohioc5; 06-21-2017 at 12:56 PM.
    2016 Hellcat - 800 Miles - It's winter

  17. #77
    Tuner Ohioc5's Avatar
    Join Date
    Apr 2008
    Location
    Ohio
    Posts
    93
    Quick NN question.

    The numbers in the Power enrich vs. Pratio Vs. RPM don't seem logical. In the tune at 5200 rpm and 1.74 pressure ratio it's commanding .0722 (1/.0722 = AFR 13.85 stoich) + .0238 = .096, thus 1/.096=10.41 afr. Doesn't make sense to me. Yet I'm commanding .0110 in the PE aircharge table. .0110+.0722 = .0832, 1/.0832=12.019

    When does this table come into play during PE? My understanding is you set your target AFR in the Power Enrich vs. Aircharge vs. RPM table

    This is a '16 Hellcat with NN on, no mods.
    2016 Hellcat - 800 Miles - It's winter

  18. #78
    Advanced Tuner
    Join Date
    Jul 2015
    Posts
    925
    Make your life easier, start using Lambda. Forget about AFR

  19. #79
    Tuner in Training
    Join Date
    Jul 2013
    Posts
    24
    I noticed this the other day after a recent update as well. I'm not nearly as much of a math guru as some of you guys but just based on what I am reading and seeing it seems to me that if I would adjust the cam values based on the new cam profile the ANN should be able to compensate for it?? Or am I giving the ANN way more credit than what it can actually do?

    I spend today street tuning my 2013 Durango R/T with Headers, Intake and SRT catback in VE mode with NN turned off. I was using the STFT and LTFT math to come up with error at this time. I haven't put in the wideband yet. I got it pretty close but it still needs a little work but man that stock table was terrible. I turned the ANN back on and all my trims are 3% and lower except for the lower aircharge and RPM areas. Those areas are more like 5-7% lean but I'm not sure what to do to the ANN table to correctly account for the error.

  20. #80
    Potential Tuner
    Join Date
    May 2018
    Posts
    3
    Quote Originally Posted by Ohioc5 View Post
    Quick NN question.

    The numbers in the Power enrich vs. Pratio Vs. RPM don't seem logical. In the tune at 5200 rpm and 1.74 pressure ratio it's commanding .0722 (1/.0722 = AFR 13.85 stoich) + .0238 = .096, thus 1/.096=10.41 afr. Doesn't make sense to me. Yet I'm commanding .0110 in the PE aircharge table. .0110+.0722 = .0832, 1/.0832=12.019

    When does this table come into play during PE? My understanding is you set your target AFR in the Power Enrich vs. Aircharge vs. RPM table

    This is a '16 Hellcat with NN on, no mods.
    Do you have Cat Over Temp enrichment? Or knock enrichment?