Initial ANN Trainer VE Settings and Axis range Setup

[HemiStude]

?14 Jeep JK 5 speed NAG 3.6L. ON3 Turbo kit. Stock rear back-half exhaust, gearing & tire size. Larger injectors. High energy coils. Rear O2's removed and associated MIL's disabled. Mail order tuning was efficiently done by ON3?s resident tuner.

It starts and runs very well with no CEL?s in 8 months daily driving since the tuning was finalized. 20 mpg highway. I?m getting up to 8 psi boost and it really rips under WOT. ANN is enabled but no ANN tuning was done.

I?m quite satisfied with most everything but it has slightly uneven, shaky acceleration at gentle acceleration and on cruise control. I also have a ?14 Grand Cherokee (3.6 - 8 speed) and it is silky-smooth under the same light acceleration. I want to get that same OEM smoothness from my turbo?d Wrangler.

I?ve taken the HP Academy online Webinar and Dodge ANN training and I?m ready to try my hand at using the HP Tuners ANN Training tool. I?ve set up all the histograms in Scanner (a huge amount of tedious work.)

I?m now wondering if the upper range of the pressure ratio of 1.06 in the twenty-five factory VE tables will support the boost I?m making. I know I can change the table values, but are ANN VE table characteristics besides the table values flexible like they are in the conventional VE tables? For example, can I change the scale of the Pressure Ratio axis in the 25 VE tables as well as the PRatio Max and Max VE PRatio Max in the training values, or are these exact values required by the ECU?

The top Pressure Ratio value in my OEM VE table was 1.0. During tuning this upper value was changed to 1.3 to account for the boost. I think I would need to increase the value in each of the 25 ANN VE tables to from 1.06 to 1.3 as well as increase the PRatio Max and Max PRatio Max in the training values to 1.3 before even starting ANN Training.

Is there anyone who is experienced in ANN Training on a 3.6 who is willing to provide some guidance for me? My Layout and graphs are attached.

One more question: A GUI question... I created 50 histograms in one layout based on the HP Academy training (25 for closed loop and 25 for Power Enrichment.) I am only able to view/access about half of the histograms. They aren't visible. Any thoughts on how to handle this situation?

Thanks in Advance,

Dave

[Stoopalini]

You can't change the Axis of the 25 VE tables in the trainer. To set the limit higher, go to the Airflow - Neural Network tab, and look for PRatio Max. This will most likely be set to something like 1.059 ... if you change it to 1.6, or 1.5, or whatever you want ... then produce a new training data set and run it through the NN Trainer, it will then have the 25 virtual VE tables maxing out at this value.

[HemiStude]

I'll do that next and re-run the training. Thanks for that info.

Did you open my layout to see if my histogram setup looks workable? I'd appreciate feedback since the function filtering is very new (and a little uncomfortable ) for me. Thanks!

[Stoopalini]

I did not open it, as I have a 2023 with a 5.7L hemi with a GPEC2A controller and I'm not sure how it would be different from your 3.6L and different controller.

I also have not taken the Tuning School Courses, so would not be able to see if you did what they recommend correctly.

I can show you what I did though? I setup an LTFT+STFT chart with the same axis as the virtual VE tables, then added a filter for exhaust cam position. On the Hemi, there is a single cam, so the exhaust and intake lobes are locked in sync. So I used the Exhaust drop down in the trainer and wrote down the 5 values. Then I setup 5 charts in Scanner, with each chart using a filter to populate +5 and -5 from the ANN exhaust drop down values.

Here's an example for the 117.25 exhaust cam position chart:

CAM_ANN_Chart.JPG

So I ended up with 5 charts, each showing the fuel trims when the exhaust cam position matches the virtual VE tables values for the exhaust position, plus/minus 5 degrees for each.

Does the 3.6L have dual cams, and thus you need the 25 virtual tables?

[DodgeBoy_8HP]

Could you post the tune and the NN file after the first training?

[HemiStude]

Could you post the tune and the NN file after the first training?

I will post as soon as I get a log. I just retrained with 1.6 PRatio Max and Max PRatio Max on the advice of Stoopalini and changed my row axis values in my scanner histograms to match the new ANN Result file.

I will make a road run a little later. I'm just going to drive very easy, trying not to get into power enrichment. I hope I get some data. If I set it up right, maybe I will. I'm kinda new to tuning so any advice you can give relative to HOW to drive while logging for ANN would be helpful.

Thanks DodgeBoy!

[DodgeBoy_8HP]

To start, definitely drive easy.

Do long log, with throttle sweeps. Once you get more confident, increase throttle percentage (throttle sweeps still).

Idea is to hit difference CAM angles (as many as combinations you can) with PRatios.

[HemiStude]

DodgeBoy-

I'm including a few logs. I forgot to plug my WB in so that data is missing. Not really a problem for these tenderfoot runs though. I didn't get much data. I'm not sure why, but I did make short logs just to get some data with which I could evaluate my Histogram/Graph setups. I think I have them pretty well set now. I'll make a series of longer logs tomorrow with a fuller range of R's and pressure ratios... or at least that's my aim. The NAG1 transmission has a mind of it's own so trying to put a steady sweep of the rpm range is difficult. An Eddy dyno would be much more time efficient. It downshifts constantly and unexpectedly and I don't know how to stop it. Keeping it from upshifting is easy.

Maybe seeing my logs and my histogram scheme will give you an idea of my concept. I'm wide open for suggestions. The amount of data is almost overwhelming. I probably need to just get used to looking at the key references but right now, I don't have a firm enough grasp yet. I'll get there.

ANN Tuning.Layout.xmlANN with 1.6 Pressure Ratio result.htt.pdf3.hpl2.hpl1.hplANN Layout Closed Loop Tables1.Graphs.xml

On the file named ANN with 1.6 Pressure Ratio result.htt.pdf you will need to cut the .pdf off the end of the filename. I couldnt get it to upload as a .htt.

I hope my uploads are OK. It's the first time I've done much file attaching in HP Tuners.

Thanks for whatever time you put in helping me to learn the process. I'm a budding 71 year old Hot Rodder and Dirt Biker with a ton of enthusiasm and just a few technical skills. My first EFI was a 900cfm Holley Projection on my numbers matching 70 Charger R/T in 2001. The Jeep is my first turbo.

Dave

[HaasExp]

?
I?m quite satisfied with most everything but it has slightly uneven, shaky acceleration at gentle acceleration and on cruise control.

The top Pressure Ratio value in my OEM VE table was 1.0. During tuning this upper value was changed to 1.3 to account for the boost.

Hi Dave,

just out of curiosity, what has led you to think that it's the ANN causing the listed issues under acceleration and cruise control?

I just had a brief look at one of your datalogs and the fueling is way off literally everywhere (eg +25% trims in some cases).
To me it seems like your injector data is wayy off because even though you have a turbo kit, your airflow and fueling requirements at idle, light throttle, gentle acceleration and cruising should be almost the same as stock.
This means, that even on the stock tune, your trims should be much closer than they currently are provided the injector data is reasonably close. This is what I've experienced on many boosted 3.6 pentastars.
Without the correct injector data, you are just trying to mask the inaccuracies by trying to banaid over them by "tuning" the ANN. You will end up going around in circles here.
On a descent tune, the vehicle should virtually drive like factory out of boost (because nothing has really changed there) and become an animal on boost.

?
The top Pressure Ratio value in my OEM VE table was 1.0. During tuning this upper value was changed to 1.3 to account for the boost.

If you are getting approx. 8psi max boost then your maximum pressure ratio would be (14.7+8)/14.7 = 1.54 so rescale up to around 1.6.

[HemiStude]

HaasExp - Thanks for joining this thread.

"Without the correct injector data, you are just trying to mask the inaccuracies "

I thought the same thing about the injector data... that is, if I put in the mfg supplied injector data (which I did) that it should run exactly the same as before the injector change. I replaced the injectors as the first step in the turbo build, weeks ahead of adding the turbo. I noticed that it didn't run as well out of the gate with the new injectors. Andre, from the HP Academy says in the online tuning class I bought, that injector changes should be transparent (unless insanely large injectors were installed.) I double checked with Tech at the Injector Mfg. to be sure I was using the correct table and entry data. They assured me I was. That could be my basic problem. The guy who tuned it may have tweaked them too. I need to review those two tables.

That makes me go back to a thought that I had initially regarding injector size. Since the injectors installed in my 2014 Wrangler's 3.6 from the factory are identical (same P/N) to those installed in my wife's E85-capable Grand Cherokee, why did I need to go to bigger injectors for gasoline-only operation? If stoich is ~10 for E85 and 14.7 for gasoline, then it takes ~45% more fuel volume with E85. In my simple thinking, that is definitely enough extra duty cycle to get me to 8 psi on gasoline with the factory injectors.

Do you have experience with boosted 3.6's on factory injectors? It might be a good test for me to go back to the factory injectors and the factory Grand Cherokee injector data. I'll never go past 8 psi and I never stay in boost for more than 10-12 seconds anyway with the auto tranny.

Do you see a flaw in this line of thinking?

[DodgeBoy_8HP]

DodgeBoy-

I'm including a few logs. I forgot to plug my WB in so that data is missing. Not really a problem for these tenderfoot runs though. I didn't get much data. I'm not sure why, but I did make short logs just to get some data with which I could evaluate my Histogram/Graph setups. I think I have them pretty well set now. I'll make a series of longer logs tomorrow with a fuller range of R's and pressure ratios... or at least that's my aim. The NAG1 transmission has a mind of it's own so trying to put a steady sweep of the rpm range is difficult. An Eddy dyno would be much more time efficient. It downshifts constantly and unexpectedly and I don't know how to stop it. Keeping it from upshifting is easy.

Maybe seeing my logs and my histogram scheme will give you an idea of my concept. I'm wide open for suggestions. The amount of data is almost overwhelming. I probably need to just get used to looking at the key references but right now, I don't have a firm enough grasp yet. I'll get there.

ANN Tuning.Layout.xmlANN with 1.6 Pressure Ratio result.htt.pdf3.hpl2.hpl1.hplANN Layout Closed Loop Tables1.Graphs.xml

On the file named ANN with 1.6 Pressure Ratio result.htt.pdf you will need to cut the .pdf off the end of the filename. I couldnt get it to upload as a .htt.

I hope my uploads are OK. It's the first time I've done much file attaching in HP Tuners.

Thanks for whatever time you put in helping me to learn the process. I'm a budding 71 year old Hot Rodder and Dirt Biker with a ton of enthusiasm and just a few technical skills. My first EFI was a 900cfm Holley Projection on my numbers matching 70 Charger R/T in 2001. The Jeep is my first turbo.

Dave

I will take a look at the logs tomorrow morning. Busy at work...

[HaasExp]

Do you have experience with boosted 3.6's on factory injectors? It might be a good test for me to go back to the factory injectors and the factory Grand Cherokee injector data. I'll never go past 8 psi and I never stay in boost for more than 10-12 seconds anyway with the auto tranny.

Do you see a flaw in this line of thinking?

I assume you are talking about the factory 8-hole injectors from the PUG 3.6 out of the GC.
These do not flow anymore than the gen 1 Pentastars, they just have fine atomization for improved idle, transient and emission performance (supposedly economy improves too).
Stock Pentastar injectors are maxing out at around 75% DC on pure gasoline with the stock tune which doesn't leave much headroom at all. The factory flex fuel 3.6's still use the same injectors and they definitely max out on E85, however the factory tune limits the IPW and thus starts pulling the throttle as a stop gap to avoid WOT lean outs. This is obviously perfectly acceptable for the performance of a daily driver which 95% of these vehicles are used for.

At 8psi boost (approx 0.5bar) you will theoretically need around 50% more fuel flow to maintain reasonable AFRs. That will put you at around 112% "effective" DC. This is obviously not possible, and you will lean out once the injector goes static which is probably around 95% DC (eg when the injector is held fully open and cannot physically flow anymore fuel).

Fuel injector clinic (FIC) 660cc injectors are the typical go to as a drop-in replacement injector for boosted Pentastars. They work pretty well and are quite reliable.
While they do supply data with their injectors, you will probably still find that you need to tweak this data slightly for your specific vehicle if you stick with the ANN tuning route and begin with the factory tune.
The other tuning alternative is VE tuning which does work very satisfactory provided you avoid swinging the cams all over the place like the stock tune does.

[DodgeBoy_8HP]

I am having issues opening the .htt files...

Could you send me via email?
[email protected]

Thanks.

[HemiStude]

I think that explains it pretty well. Thanks

I have the 660 cc Fuel Injector Clinic Injectors.

PW vs Fuel Mass
0.000359375 0.456
0.0008125 0.536
0.001703125 0.6
0.003046875 0.64
0.003109375 0.672
0.0034375 0.752
0.003890625 0.832
0.004484375 0.912
0.00521875 0.984
0.006546875 1.112
0.007953125 1.256
0.016765625 2.064
0.293390625 35
0.829015625 200

Fuel Mass v. PW
0.456 0.000359375
0.536 0.0008125
0.6 0.001703125
0.64 0.003046875
0.672 0.003109375
0.752 0.0034375
0.832 0.003890625
0.912 0.004484375
0.984 0.00521875
1.112 0.006546875
1.256 0.007953125
2.064 0.016765625
35 0.293390625
200 0.829015625

Startup Injector Scalar - 9
Injector PW .440

Pulsewidth offset
XXX 7.99 9.49 10.99 11.99 12.74 13.49 14.00 14.76 15.50
-80 2.368 2.072 1.768 1.528 1.216 0.848 0.608 0.464 0.376
-40 2.56 2.232 1.896 1.616 1.304 0.888 0.632 0.488 0.392
0.313 2.73 2.368 2.016 1.712 1.368 0.92 0.648 0.504 0.4
40.47 2.92 2.52 2.136 1.808 1.44 0.968 0.688 0.544 0.416
80 3.112 2.664 2.248 1.904 1.488 1.016 0.712 0.552 0.432

[HaasExp]

I have the 660 cc Fuel Injector Clinic Injectors.

If those are the logs from the FIC660cc injectors based on the ON3's tune, then it is wayyy off and you really need to get it retuned properly.
Do you have a wideband O2 sensor setup for tuning?
Feel free to send me a PM if you want to discuss further.