I have been trying to use road test data to ‘rediscover’ the drag formula for my model Prius, 0.42*(V**2) + 190 N. My road testing includes multiple “neutral” rolling tests in both directions and the results have been mixed. Short of a vehicle dynometer and wind tunnel testing, I’m trying to figure out how a layman might get a usable, vehicle drag formula from the road. Right now, I agree that there are too many variables.

By integrating the drag formula over distance and adding the ‘overhead’ energy, one comes up with a pretty useful, theoretical MPG vs MPH curve. Differences between road tests and this drag defined, limit, identifies where improvements are needed or better than expected performance, ‘sweet spots’ can be found.

Within tolerable error, the road test MPG vs MPH has been reproducible and revealed or validated major factors: (1) engine oil level needs to be 75% full and never over F, (2) tire pressure closer to side wall maximum, (3) importance of low-power, rolling warm-up, (4) auto-off, control law boundary at 67 kph, and (5) transaxle MG1 rpm speed limit at 112 kph.

I’m not giving up on road testing but was looking for ideas on how it might be used to validate or derive the drag formula coefficients.

Thanks,
Bob Wilson

Obviously a public road has more variables than a racetrack, and if the road testing is done over a long period, even more variables will come into play.

In short, I cannot actually see an outcome – but for what it’s worth, the road would be better than the track!

I’ve been road testing my 2003 Prius, NHW11, since I bought it nearly two years ago. My primary interest is mapping the fuel consumption as a function of speed and other variables. So far, this has identified some sweet speeds, 60 kph (38 mph) city and 105 kph (66 mph), gasoline brands, wheel alignment, warm-up practices, route planning, hill climbing, and transaxle lubricants. Every test succeeded even when they came back with “don’t do that again, ever!” But I’m still having difficulty getting a stable, math model that integrates all of the data:
http://home.hiwaay.net/~bzwilson/prius/calculated_MPG_Rev_B.jpg

Controlling and tracking all of the variables is not trivial and I’m beginning to wonder if track testing suffers from the same problems I’m running into on the road . . . too many variables resulting in different results depending upon preparation, weather and track conditions. How do you handle making sure your results are reproducible on the track and subsequently the road?

I am leaning towards recording everything, all trips and as much data a possible and then trying to do a multi-variable analysis. With enough data, hundreds of megabytes with as many variables as possible, it should be possible to have software look for correlations. In short, I’m thinking of trying to harvest an engineering model from a huge mass of data points. This is driven my by frustration that every trip is potentially a test but I need to reduce the labor associated with reducing the data. Have you run into similar testing approaches?

Thanks,
Bob Wilson