Hi folks,

Here is my latest data on ICE specific fuel consumption:


There may be some problems with the mixture ratio at rpms under 1,600 rpm but I'm pretty comfortable with everything above 1,600 rpm.

Bob Wilson

 

You sure those units are right? I would have expected kWh/g or else power vs. mass flow rate of fuel. As it stands you're dividing power by energy. I'm guessing this is just a typo.

 

OPPS! It should be "ICE W/g".

Also, I widened the data window and added 0.65 ms. to bring the fuel-air ratio into a more linear relationship. We are hobbled by the absence of calibration constants and formula for both the fuel injector timing and air mass flow sensor. A similar correction factor for the air mass flow sensor might accomplish the same effect.

Bob Wilson

 

I think I'm still confused, but it could just be me. A fuel gram will produce xx joules of thermal energy, producing so many watts of power if released continuously over a set number of seconds. So your Y axis unit is relating a unit of power to a unit of mass, which we can convert into a unit of energy. To me that says "1/s". Is it ICE J/g (ICE Ws/g)?

 

What I do is use this formula to get the ICE watts:

ICE Watts = (ICE_Nm * ICE_rpm) / 9.549

The ICE watts are over a sample interval, ~1.084 sec.

To calculate the fuel burned during this interval, I use:

grams = (2*(ICE_rpm/60)*dT*(Inj_ms+0.65)) * 2.5046 g/ms

2 - fuel injection pulses per rotation
0.65 ms - correction factor to make fuel_air ratio remain linear (*)
2.5046 g/ms - experimentally derived ratio using MFD calculated fuel burn
dT - sample interval time in seconds

The spreadsheet is a little over 3 MB and I have no problem sharing it. This includes the raw data and all formulas and processing.

Bob Wilson

(*) calibration of both the injector and air mass flow are unknown. The raw data is reported by the engine ECU but the air-to-fuel ratio suggests both may have non-linear characteristics. Ken uses directly measured injector time and has already reported 0.5 ms of dead time. But two of us have data suggesting the reported injector time has 0.65 ms of unreported time. Needless to say, this begs going back and comparing the two as well as taking a closer look at mass air flow data.

 

Ahh, alright. I think I see what's going on. The only piece missing is that you need to take your calculation a step further by multiplying ICE_W by your sample interval (1.084s) to get the total energy produced (ICE_J) during said sample interval. Then you can divide by your estimate for fuel burned during that interval to get ICE_J/fuel gram. Otherwise it doesn't really mean anything because ICE_W is just a measure of how fast the engine is producing mechanical energy, not how much energy is produced.

 

Let me think about it a little longer but I think I'll wind up doing this. It will increase the energy by about 8.4%.

One of the problems I still have is dealing with the units I'm getting versus the Toyota reported values. Unlike the Toyota paper, these are what I'm seeing from the Prius in my driveway and I can trace back to fuel receipts and clocks.

If you haven't guessed, I'm headed for an open source, Prius energy model, probably excel based. Eventually, I be able to develop a set of formulas and constants that model the vehicle. I want a model whose data fidelity is such at anyone entering any given profile can predict the performance and see the results on their next fill-up. Then we can investigate and do some truly exciting things.

Bob Wilson

 

I look forward to seeing the results. I can see where you are going with this, which is why I suggest the change that puts joules on that Y axis. (Again, multiplying by the time step only serves to give youthe quantity you need. You would have to do the same thing if your time interval was exactly 1s, except only the unit would change in that case.) Once you do that it becomes trivial to multiply fuel grams by energy content (~47kJ/g) to get chemical energy input, and then divide energy output by energy input for each time step and get ICE efficiency. From a quick hand calc on this data it looks like this engine is peaking somewhere in the 33% range. Pretty darned good!

In a couple of weeks I'll be able to start providing NHW20 data using the CAN datalogging utility from Yoshi (ken@japan) and my laptop. All I'm waiting for is the CANUSB dongle and the parts needed to solder up the OBDII cable with the right connector.

Another question: Does the miniscanner have access to a measurement of torque, or are you solving backward from velocity so that you can use F=ma, then P=Fv? I ask because it looks like I'll be stuck trying to estimate power based on acceleration, which presents all kinds of limitations and uncertainties. For example I need a dead-flat test track, no wind, known mass, and I can't go very fast because above maybe 25mph aerodynamic drag starts to become significant. Clearly, the value of that data would be minimal without some fairly extensive calibration.

 

AHHH!!! That makes sense. I had calculated ICE efficiency on a tank from DC to AL at 31%. But if I correct the time interval to 1.08 sec that 31% becomes 33%.

Excellent.

It reports MG1 and MG2 torque. With MG1 torque, I can calculate ICE torque. The respective torques should be proportional to the current. You can get ICE rpm either directly from injector timing or indirectly from MG1 and MG2 rpm.

When you get the scanner, PM and we'll talk.

Bob Wilson

 

That makes a lot of sense. Behold, the beauty of electric power transmission! I don't know if Yoshi's program can get MG1 torque, but the driver for the CAN dongle ought to be able to isolate pretty much anything in the datastream.

I'll definitely drop a line once it's up and running.