2yan

I've just been having fun analyzing data so I figured I'd share my analysis with you guys:
I've attached it to this forum post.

Oh I also got another Turbo Miata, I'm a bit of a data guy so I don't mind fixing someone else's tuning and I wanted to start with a semi built car.

Deets:
Mega squirt 3, Miata 1997 (Need to figure something out before emissions)

Goals: Learning to tune and be super dope.


Edit: Also if someone is down to send me their Tuner studio Logs i'd love to pour through them and see what I can figure out. I get data and you get a bunch of pretty charts out of it.

Socals14

Interesting. Python, eh? Thought about doing this in R, myself. Especially with maps...wondering if the smoothing maps could be better implemented.

2yan

I've been working on it a bit extra lately actually. I'm really fond of my auto tune algorithm right now. My standard deviation from my target afr is about .2 and I figured out how to account for throttle enrichment. I'm still tweaking things to perfect it. It's not a one shot algo yet. But it's smart enough in that It takes about 2-3 drives to go from a terrible table to a really good one.
On the note of R vs Python. With the scientific python toolkits it mirrors R's data frame and charting functionality really well but I don't think you can go wrong with either.
I found it easier to write custom code to load up ve tables/ export VE tables with python though.

2yan

Update, now tuned a friends car with the algo and it works pretty well. Running correlations to figure out where issues are is also pretty awesome. ( We found that his coolant temperatures had an interesting hysteresis and that was connected with his IAC values). Figured it out my sorting by correlation coefficient of different variables to the coolant temp.

jetskibruce

Hi Ryan, I've been tuning and looking for ways to better analyze the logs and I came across your post here. I'm a bit of python programmer but like most things, I find there is always a lot more to learn. I've been looking for a reason to get into jupyter notebooks too, so this is great. I think I've entered the code correctly to dup what you posted initially, at first look into my logs I wasn't capturing Load and to get through I changed it to MAP but the results look better with Load. I added Load and the engine states to my sdcard logs today. Then logged a drive and just went back to the code and I 'think' I changed it back to using Load ok, had you used MAP
Looking at this early work of yours, I'll bet your latest tuning code is good! Your work looks good to me and I immediately see the value in working in a notebook, it just adds to the python ability to quickly code and test.

Attached is a pass in the code from a drive today - today I was also running the auto tune to keep tweaking my setup. I've got a few ideas I want to try to see if I can get some code to help.
Even saving the jupyter project as html was new to me, and I didn't see how to toggle that ability to 'hide the code' in the HTML ms-analyze.html
I was thinking of coding up a website to let users upload their own log and see the results. ( Not sure if I'm that ambitious or have the time )