[bwilson4web]

I've started calibrating our ZVW30 indicated versus actual mileage. The first part is to define the standard, the actual mileage, and then look at the dependent variable, the indicated mileage. For our standard:

• same pump - #6 at the Shell station on South Parkway in Huntsville AL
• same fill-up protocol - manually listening for the rising tone when the gas backs up into the filler tube. Pause at least 5 seconds and then top to 'one click' manually
• same fuel - Shell 87 octane, E10
• quantity - at least 1 gallon since pump reports to 1/1000th of a gallon
• temperature - not recorded

There needs to be a model for how mileage is calculated and assuming the distance measurements are valid, there are two methods for measuring fuel consumption: (1) injector timing and summing, or (2) mass flow stoichiometric operation.

Injector timing, as Ken@Japan pointed out, is subject to mechanical delay and linear output outside of these two. When the injector is first turned on, it takes a finite amount of time until fuel begins to flow, the delay. This means, as Ken@Japan pointed out, a fixed overhead has to be 'taken out' of the fuel measurement.

The second approach is to use mass flow at the 14.7 air to 1 unit of gas ratio. This works when the engine operates in the closed-loop, feedback system range. But this range does not start working until the O{2} sensors are operational, which typically takes the first 45 seconds in my NHW11. At high power settings, the ZVW30 engine begins using cooled, exhaust gas to moderate exhaust temperatures. In contrast, the NHW11 used a richer mixture to avoid burning out the catalytic converter. In this case, one would expect the error to change as a function of power used, something proportional to vehicle speed.

To figure out the calibration curve, we need to know which mechanism, injector timing or mass flow, is being used. I don't know but errors in the data by looking at average speed, gives an indication of which mechanism is being used. I'm still gathering data and only have four samples.

PRELIMINARY INJECTOR COUNTER

Assuming injector timing is used to calculate fuel consumed, the correction curve should be a line with possibly a fixed offset:

The "14.467" constant in the linear trend-line would correspond to a fixed overhead, the delay that Ken@Japan reported for the 1.5L engine. With just four records, I don't see a speed related factor, which suggests mass flow is not being used.

After I get enough samples, the trend-line values will become more and more accurate. I can then solve the inverse function and convert the indicated MPG to true MPG. But I will only be able to do this for my particular vehicle. Others will need to gather their own calibration data. The inverse function:

my_true_MPG = (indicated_MPG - 14.467) / 0.8376
my_true_MPG = (indicated_MPG - 14.467) * 1.194

So if I use this correction formula on my 'on flash' measurements:

60.8 MPG -> 55.32 MPG
132.6 / 55.32 -> 2.40 gallons residual fuel versus 2.18 gallons

This would add ~.22 gallons to the tank absolute capacity or nearly 12.3 gallons.

If we were to gather enough data points, we could do a population study. This would allow us to test different hypothesis of the source of the error. But that is beyond the scope of this study as I'm just using this, my raw data:

pump MPG	indicated MPG	avg mph	gallons
54.4	61.0	33	2.28
54.8	60.2	33	1.444
49.8	56.4	35	9.65
48.5	55.0	32	9.8

.

Impact of Calibrated MPG

Once I'm sure of the calibration formula, I'll update my earlier mph vs MPG chart to include both the indicated and calibrated MPG. I'll also have to correct the hill climb data, again showing both calibrated and indicated fuel consumption. This will serve the purpose of showing folks why their indicated mileage can be wrong as well as a resource for those who claim to have a true MPG reading, a reference curve to use for their data.

Bob Wilson

[bwilson4web]

One late thought, if you would like me to analyze your indicated versus pump mileage data in the MyHybridCar database, please do the following:

• standardize the fill-up - changing up the fuel or other variables makes calibration ... risky
• use Trip A or B for tank - I like to use Trip A for tank to minimize the risk of resetting the wrong trip meter. For ZVW30 owners, please record all three values: distance, MPG and average speed
• let me know you have data

I'll analyze your data and post the spreadsheet. You can then pick it up and continue any further investigations.

Thanks,
Bob Wilson

[geodosch]

Bob,

On a related topic, when you were doing your 'running out of gas' tests, did you do any observations as to the accuracy of the 'Miles to empty' indicator?

[bwilson4web]

geodosch said:

. . .
On a related topic, when you were doing your 'running out of gas' tests, did you do any observations as to the accuracy of the 'Miles to empty' indicator?

I always exceeded the 'Miles to empty' by about a factor of two. I didn't write it down but I remember them showing ~50/60 miles when 'flash' started but I continued to drive for nearly 120 miles.

Bob Wilson

[bwilson4web]

With 109 samples, manually collected from the PriusChat community, we now have enough data. First the calibration:

The upper, outliers are still sparse but they seem to be around the trendline. This is not enough to 'bet the farm' but enough to not worry too much if extended.

The other interesting aspect is the double-humped, mileage distribution:

This closely agrees with the double-humped distribution from the small number of user mileage reports from the EPA database:


At this point, the ball is in Toyota's court. Absent some way to modify the software that calculates the indicated MPG, we just need to understand the problem and depending upon the requirements, there are these options:

• use pump MPG for accuracy and indicated for relative MPG performance
• use correction formula: actual = (.85 * indicated) + 5.3
• use 3d party MPG indicator: scangauge, supermid or others (mpguino)

Feel free to share this with other Prius groups and thank the PriusChat community for their data contributions. I had to manually collect the data from hundreds of postings.

Bob Wilson

[bwilson4web]

geodosch said:

. . .
On a related topic, when you were doing your 'running out of gas' tests, did you do any observations as to the accuracy of the 'Miles to empty' indicator?

Nothing formal. My impression was it showed ~50-60 miles remaining while in truth I was going closer to 120 miles.

Given the stealthy nature of running out of gas, I don't want to repeat those experiments until I have a trustable indication that the vehicle is really out of gas. This is a safety issue.

Bob Wilson

[geodosch]

bwilson4web said:

Nothing formal. My impression was it showed ~50-60 miles remaining while in truth I was going closer to 120 miles.

Bob,

Thanks, that's just what I was looking for. So the 'Miles to go' isn't miles to flash, which is what I had thought it might be.

I'm going to start tracking it, to see how linear the estimate is, and also how it varies based on overall MPG.

[bwilson4web]

We've increased the number of samples at PriusChat and this has improved the calibration formula:

MPG_actual ~= (0.86 * MPG_indicated) + 5.0

Current sample set:


Bob Wilson