Hi Robert,
I always enjoy reading informed comments:
Quote:
Originally Posted by RobertSmalls
I do love seeing mountains of data, but I have a few questions about, and issues with, your plots.
Is the "actual" data the mpg according to the pump and odometer? There's error in pump data too, but it will be random, not biased, and it'll average out in the long run.
Most of your outliers in the 54-65 mpg range do not look like outliers. They fit the trend pretty well. Also, all of the outliers are on the same side of the trendline, and by excluding them, you move the trendline up. I would actually include them all. If you think one of them represents an incompletely filled tank, you need to throw out that data point and one for the tank that follows it.
|
We are using the pump and trip meter for calculated and the vehicle indicated MPG.
Throwing out the outliers is always risky and the first passes, I tossed out the upper 5% and lower 5%. My theory being that errors at the extremes have an inordinate effect on the line. But as the number of samples increased, I noticed the trend line was approaching a slope of 0.85 with a Y axis intercept approaching 5.
pri_2010_630.jpg
pri_2010_630a.jpg
pri_2010_630b.jpg
pri_2010_630c.jpg
pri_2010_630d.jpg
pri_2010_630e.jpg
Fortunately, Ken@Japan reported a conversation with a Toyota engineer that the error was 5%. Suddenly it made sense. The Y-axis offset, 5, was approaching 10% of the mid-point value, ~50 MPG. So the offset was tossing the slope the wrong way. I quickly plotted the distribution of errors and sure enough, the upper 10% started at a slope of 45 degrees, the "knee in the curve." So I changed the outlier definition and you see that curious grouping of outliers.
So with Ken's report and the current chart showing ~5% error, you reach a point of diminishing returns. As Dick Hicks, one my GE managers once said, "90% isn't good enough, it is perfect!"
Quote:
Originally Posted by RobertSmalls
That's a pretty scandalous conclusion: The Prius' onboard instrumentation lies about MPG. Is Toyota aware that the community is aware of this? I would write them a letter.
|
We believe they are. I've not written mine since I'm not sure there is a standard for MPG accuracy.
Quote:
Originally Posted by RobertSmalls
If the transmission had discrete gears, plot 3 would be simpler, and I'd slap you for fitting a high-order polynomial to it. In fact, I think I will anyway. Gears or no gears, work (y) required to go speed (x) is a parabola, and your trendline should be as well. Also, what's up with the guy at 50mph, 70mpg? Is he P&G'ing, or aeromodding, or lying, or what?
|
All I can do is use the data that is available. It isn't as if there a pile of money riding on the results. <grins>
Quote:
Originally Posted by RobertSmalls
When I make mpg vs mph plots, I definitely use gal/mi vs mph. I believe they make more intuitive sense that way.
|
I know and my BSFC charts are that way. But when dealing with folks who have an "MPG" meter, well I try to communicate without getting bogged down in a lesson on physics or thermal dynamics.
Quote:
Originally Posted by RobertSmalls
Plot four: it's good to see the numbers confirm what I had been hoping: The Insight outperforms EPA estimates, and while it doesn't beat the Prius, it's not a total embarassment to Honda.
|
This is something that bothers me, the absence of an 'mph vs MPG' chart in the owner's manual. Knowing the 'knee in the curve,' I choose to drive 65 mph on the highway and 70,000 miles later, 52 MPG. But I had to make my own chart to see this relationship. I would rather see a graph than those EPA numbers.
For the Insight owners, they too can choose peak efficiency speeds on the interstates and highways ... if they know what it is. In the City, it looks like they have no problems.
Quote:
Originally Posted by RobertSmalls
Plot five: What happens when you compare your hill-climb data to mpg vs mph on flat land? At what speed to you most efficiently convert gasoline to gravitational potential energy, and how dramatic is the difference across the range of speeds you would consider climbing hills? And what's up with the second line where the guy gets better mpg at 88mph than at 83mph?
|
Those were two different octanes, 89, the top curve, and 87, the bottom. What happens is 87 octane appears to have more energy per unit so mileage is best in ordinary driving with 87. But 89 octane allows the engine to tune to the 'knock' sensor. So going up the hill at high speeds, the Prius does not have to draw on traction battery energy. If you draw too much energy from the traction battery, the car slows down ... it still climbs but not like before.
If I'm going to drive in the mountains, I'll switch to a higher octane so I won't put as much stress on the traction battery. But once I get to the 'flat land,' I'm back to 87 octane.
Quote:
Originally Posted by RobertSmalls
What are you using to gather data to the 0.0001 of a gallon? That's pretty impressive. Me, I'd like to hook up an ABS wheel speed sensor to a data acquisition card and measure the energy consumption of a pothole.
|
The fuel pump reports .001 gallons on the receipt. You're right that the distance accuracy could cause a loss of resolution but I've got GPS, trip meter, and map readings of the distance. Toss a coin and I could have gone either way. But there is a basis of estimate, the pump receipts.
Bob Wilson