SScam68

iTrader: (20)

don't need to be a student, use the internet


Here's some info on wiki on it

http://en.wikipedia.org/wiki/Tensor

gametech

iTrader: (1)

I'm glad some of you guys understand this. I had to spend too much time at work to do my calculus homework back in school. I ended up auditing the class to keep from having an F on my record. BTW, thanks to all who are working on this.

RedHardSupra

iTrader: (1)

i'm stuck on math, and i need to solve this:

a=x*(b+(c-b)*exp^(-k*x))

exp() is a function of raising the base of the natural log to some power.
please solve this for x in terms of k. the rest you can treat as a bunch of constants for now. and your reward will be..umm.. to be able to tune your car right, good enough?

cantdrv65

iTrader: (3)

If anyone needs a program that will solve multivariable equations for you like MatLab.....check out Mathcad. Its readily available for free if you look in the right places.

cantdrv65

iTrader: (3)

If I only had my laptop with Mathcad tonight.... If I remember I'll let Mathcad solve it for you tomorrow.

RedHardSupra

iTrader: (1)

that will be a good test, MatLab gave up on it, it could solve for k but not for x. havent used anything more symbolic math oriented, does anyone have mathematica?

beast69camaro

iTrader: (14)

I'm pretty sure that equation can't be solved explicitly for x in terms of k because of the presence of x in both an exponential and linear term. I don't know what you're doing with that equation,.....but it looks like x would need to be solved for numerically.

miami993c297

iTrader: (6)

Hi Marcin,

you've got an email.

Christian

SScam68

iTrader: (20)

yep, and if you know K plug in numbers into x until they match up to the value that you need.

I get

K=ln((c-b)/(a/x-b))/x

miami993c297

iTrader: (6)

Hi Marcin,

Did you get my email with the Excel attachment?

Does this solution about the equation works well enough?

This spreadsheet was done using Excel to generate a numerical solution.

On the equation side, I can give you the solution in Matlab code as well, if that would help out more...working on it.

Christian

joecar

Ugghhh, that's fugly...

rearranging: (a/x - b)*exp(k*x) = c - b

it's still fugly...

RedHardSupra

iTrader: (1)

Ok folks, I think I got the math for figuring out the BIAS table, which seems to be a major source of imprecision. I think that's one of the 'hidden variables' we didn't account for before at all, thus lack of 'convergence' of VE tuning. This is only a first step to a whole new way universal way of tuning that i've been working on for a while.
I hope you're not afraid of differencial equations, physics, and linear algebra, cause in 5 pages i got it all

http://www.marcintology.com/tuning/T...reModeling.doc

this is an early version, written mostly past 2am, so there's bound to be some wrong **** in there, so please let me know. More to come later, must graduate first.

Bink

Congrats! Very impressive.
I need to sit down and read your blog. Also, Very impressive.
What's your thesis on??
Graduating with what degree?
My hats off to you........

RedHardSupra

iTrader: (1)

blog's gonna get updated once proof of concepts get tested and automated until your grandma is able to run through the process
i'm getting a masters in computer science/computer security, with thesis doing a vulnerability assessment of a secure neighbor discovery protocol in ipv6.

so any bugs yet? i'm still not sure of it's correctness in few spots.

P Mack

iTrader: (6)

This is what I got from that paper:

Using wideband and and fuel mass you can get Airmass. Using airmass, MAP, and VE you can figure out what the PCM is calculating for Temp. Then once you know what the pcm has calculated, you can solve for r. Is that right? So my question is, are you trying to find out what the pcm is using for r, or are you trying to find what value you're going to program into the pcm for your tune?

Bink

Should Intake Time = t(sec) = 240/RPM??

RedHardSupra

iTrader: (1)

pmack: r is just a calibration. i want to tweak it in such a way that the calculations from the air side (the one with gmve in it) and the ones from fuel side (afr*ifr*ipw) agree. once we know r and somehow manage to translate it to corresponding MAF values (since the BIAS table is on the MAF scale) then we can just use it as a constant and use it to tweak gmve values. but that's later, for starters we gotta figure out the constants. but you're right, it's totally not obvious (as in i dont have a clue yet how i'm gonna approach it) to create proper airflow vs bias from rpm vs r we have now.

bink: why do you say it's 240/rpm? wouldnt then 240/6000rpm=0.040 secs? last i checked we got 20msec window at 6000rpm. what did i miss?

P Mack

iTrader: (6)

Ok, I have a problem with the whole premise of your r calculation. It is circular. You are calculating r based on what you get for TEMPexpected. TEMPexpected was calculated based on IPW, which was calculated by the PCM based on cylinder airmass, which is based on TEMP, which is based on r.

Your error was that you set TEMPexpected=TEMPobserved. If that is the case, then instead of calculating an actual r, you are calculating what the pcm thinks r is.

Also, as far as the 240/rpm thing goes, the number you use is meaningless. r was derived empirically, and 240 is just a coefficient in front of r. It doesn't matter. All that matters is that e is raised to (constant * time). The time the air is exposed to heat is not equal to the time between intake valve events anyway.

RedHardSupra

iTrader: (1)

of course it's circular, you got airflow that depends on temperature that depends on the amount of airflow that's the whole problem, and that's why differential equations start to show up (against my will )

yes, you are absolutely correct, we are not calculating actual r, we're just getting r to do what pcm it should do. for now it's a iterative method, just like the old ve tuning, gotta do it few times. however this time it should converge better. i do have another approach where i tune both r and ve at the same time, which is gonna be pretty much the way the e38/e67 stuff must be done, but that's pretty friggin involved. i got a spreadsheet with a very early idea how to do it if you want to take a look at it.

hmm...that rpm multiplier is kinda pointless, you're right, r is just going to 'absorb' it anyway...good thinking, this is exactly why i posted this early stuff, this **** is entirely too complex for my blonde self :/

the time the air is exposed to heat is kinda important, that's why at lazy rpm/low airflow it's much more ECT biased. but you're right, what kind of relationship it is, i have no clue. any ideas what else influences it? if i have few potential suspects, i can figure out some weighing system and see which one influences it the most.

so far, great feedback. as a pimp said to his ho's: keep 'em coming!

Bink

I'm sorry. My goof...I was thinking 1 stroke in 4 (4cycle). Stupid on me .
.
.
Carry on.