Results by using a polynomial curve fit to help calibrate MAF?

[evolmotorsprt]

I've been working on tuning my MAF and noticed that fitting a curve to the data may be a good way to refine the results from logging. I can't be the first one to think of this, so has anyone else had good results by curve fitting the data and writing the results to the PCM?

I'm am going to try this tomorrow and report back, but I'd like to hear if anyone has done this. Thanks

Curve Fit.PNG

[cobaltssoverbooster]

Been doing it for years. I like the way it fits data to uncharted zones. Fit predicts higher load cells when I'm trying to tune in cells but had to stop shy because the error got to an undesired limit

[Pwiles]

I've been working on tuning my MAF and noticed that fitting a curve to the data may be a good way to refine the results from logging. I can't be the first one to think of this, so has anyone else had good results by curve fitting the data and writing the results to the PCM?

I'm am going to try this tomorrow and report back, but I'd like to hear if anyone has done this. Thanks

Curve Fit.PNG
How do you use fit curve?

[cobaltssoverbooster]

Do some reading on exponential regression

[evolmotorsprt]

Been doing it for years. I like the way it fits data to uncharted zones. Fit predicts higher load cells when I'm trying to tune in cells but had to stop shy because the error got to an undesired limit

Thanks for the response. I liked the fit of the data to populate the top extremes of the MAF curve, but I didn't like the fit below ~3000Hz. My car wanted a smidge more fuel down there, YMMV. My advice to those that want to experiment with this is first know what data to expect, break the MAF curve into 2-3 sections, and use your head to know what is reasonable. Test it again with a wideband after cleaning everything up. I was well within 2% of my target above 3000Hz.

Playing with this is pretty easy in excel, and being able to find the equation of a line is a nice thing to know.

[cobaltssoverbooster]

there always is error because you are driving in a wide zone to collect data. the data typically isn't as refined as say steady state which means it already has a dynamic error. not a huge deal its usually only a few percent at most.
this process can help you produce an accurate maf curve very quickly assuming you have the correct base airmass calibration settings, base fuel component settings, and perform testing under solid state or a simulated solid state using custom filter options.

remember the more aggressive the camshaft the more unstable the idle air flow is which makes that data harder to predict. there are also a slew of idle airflow model factors coming into play down there to help better predict the behavior at lower rpm. if any of them are off they can also induce an error as well.

if you can program excel to do this regression analysis and prediction for you, then you can make adjustments that allow you to progress further into wot with every adjustment. for those guys that start from scratch and work pull by pull until they get to a full pull, this method can reduce the amount of required tuning pulls from between the initial run and a full pull. nice time saver when you get it right.