View Single Post
Old 05-16-2014, 02:59 PM   #1 (permalink)
oldbeaver
EcoModding Apprentice
 
Join Date: Oct 2008
Location: Chile
Posts: 223

Mercedes 89 D - '89 Mercedes 300 E
90 day: 33.86 mpg (US)

Skodie - '09 Skoda Octavia TDI PD
90 day: 38.84 mpg (US)

1993 Mercedes 300D Turbo - '93 Mercedes Benz 300D Turbo W124
90 day: 26.19 mpg (US)

Crossie - '16 Subaru XV Crosstreak
90 day: 9.61 mpg (US)

Crossie - '16 Subaru XV Crosstreak
90 day: 33.34 mpg (US)
Thanks: 15
Thanked 9 Times in 7 Posts
Real time fuel vs rpm vs tps for dinosaur cars

Hello,

From long time I am trying to find an easy way to retrofit real time fuel consumption estimation for old cars, which are not OBDII and which don't have electronic injection signals. This is the case of many of them.

MPGuino needs two signals only:

VSS = km/h or miles/h
Fuel = km/l or miles/gallon

Many cars have the first but not the second, which is my case, by the way.

To obtain an estimation of the fuel consumed / injected at any time, without the use of two flow sensors which are complicated to handle, or last short or are too expensive, I am trying another method which is to estimate real time fuel consumption, regardless if it is gasoline or diesel as follows:

Fuel used per unit time = f( rpm, tps)

Where rpm is well known and always available or easy to make available.
Tps which is throttle position, preferable close to injection pump or carburator, not at the pedal (for precision).

Tps is a signal sometimes available, sometimes not. But you can instal a linear sensor position to almost any car.

The problem was to have a math model to make the relation between the independent variables rpm and tps and the dependent variable Fuel with precission. Not easy.

But some fellows from Korea did it, (Science and Technology Vol. 4 Nr 4, December, 2011).

The model developed took measurements of fuel used from onboard computer. They were rigorous in using stats to measure the accuracy of the model.

I am attaching a spreadsheet with the model and a surface graph of it. Want to open discussion on it, but more than that, I will implement in my car, using MPGuino as final onboard computer, using the model output as the second signal.

The model requires to be calibrated to one's particular car, for what I added one zero constant that can shift the complete model up and down, and a scale contante which can change the "slope" of the model in a simplified manner.

One should produce some points (fuel, rpm, tps) with the car using a small fuel tank or a complete tank and calibrate the model.

I know it will be not a precise model, but at least will serve as a relative model, to measure mods made to the car or engine, driving strategies and more. The more one calibrate it, the more it will work close to reality.

Anyhow, all onboard computers need callibration.

Any thoughts?

What I need now is a small IC processor to evaluate the polynomial expression in real time, before sending the result to MPguino together with VSS.

Any contributions? I am not electronic, so please easy, practical solutions.

Hope this will be of use to many.

Oldbeaver

Attached Thumbnails
Click image for larger version

Name:	Fuel spent vs rpm vs tps.jpg
Views:	71
Size:	134.2 KB
ID:	15009  
Attached Files
File Type: xls Consumo de combustible vs rpm y tps.xls (60.5 KB, 91 views)
__________________
Mercedes 300 D turbo 1993

Last edited by oldbeaver; 05-16-2014 at 03:04 PM.. Reason: spell
  Reply With Quote