Why mpg testing is difficult and variable...

[ZugyNA]

There are many variations on the approaches to engine and fuel management design by engineers from many different cultures...the only common ground in the US being the fuel used and the EPA emissions regs.

Also each vehicle might have a different total mileage and state of tune....and possibly other modifications from stock.

So a modification that might increase mpg on one vehicle...won't necessarily have the same effect on another.

So this might be why using an A-B-A testing method with statistical analysis might be a kind of overkill. Though with a good mpg readout method and a good test route...that might not be so relative to one particular car.

Main point is: even if you do careful testing of a mod on one car and find it doesn't increase mpg...that doesn't mean this mod might not work on another car.

1) you might not have set it up right...the devil is in the details?

2) this kind of mod just might not work on this particular car?

So...to try and extrapolate from testing an mpg mod on one car to all cars...is NOT statistically acceptable?

[Mighty Mira]

Originally Posted by ZugyNA

So...to try and extrapolate from testing an mpg mod on one car to all cars...is NOT statistically acceptable?

Yes, it is more complex than simply saying x mod increases mpg by y percentage. It's because fuel economy is not a simple function.

Fuel is wasted in several different areas:

Load at highway speed:
1) In heating up the air through turbulence (high CdA)
2) In heating up the tyres and the road (high Crr tyres)

Engine/transmission efficiency:
3) In an engine that is in too high an rpm range for the load
4) In an engine that is too large for the load.
5) In too many, inefficient parasitic loads on the engine (alternator, pumps, etc)

In braking:
6) Through an overly heavy car
7) Because we heat up brake pads/discs/drums instead of returning kinetic energy to a battery.

2, 6 and 7 mainly apply in city driving. The rest apply to highway driving.

FE in a car can be visualized as being constrained by three bottlenecks. Attack one bottleneck, and eventually you approach diminishing returns. For example, start off with a car built like a brick and geared for city driving. Do some aero mods, and the highway performance will improve. Do more mods and it may not improve the FE much, because the gearing will be much too high for the steady state load.

Note that in each of those different areas, it is possible to measure the variables and so compute a function that will give a good approximation as to expected FE. However, it's more complex than saying "Do this and get X increase".

Now, since most driving is done in the city, it makes sense for car companies to attack the braking issue, and the only way to do that is with electric technology - you can't turn kinetic energy back into gasoline again. Hence the hybrid car - one with both a gas and electric motor.

However, if you can solve the other problems (i.e. decreasing the load on the engine), you can do away with the gasoline engine entirely and go solely to electrical, as range is a major factor in EVs not catching on.

I would say that as gs.org gets more of an understanding, people will learn how to measure and understand the effect of the different variables in their car, in the same way as people currently understand the interplay between different variables in producing a fast quarter mile time etc.

[95metro]

Originally Posted by ZugyNA

So a modification that might increase mpg on one vehicle...won't necessarily have the same effect on another.

So this might be why using an A-B-A testing method with statistical analysis might be a kind of overkill.

I agree that FE results from modifications will vary from vehicle to vehicle. Some vehicles may see a lot, some just a little.

But all results need to be repeatable and that's where A-B-A testing comes in. To even say a device works on one specific vehicle needs to be able to be proven over and over - at any time on any day in any season.

With something as highly variable as FE I don't find statistical analysis to be overkill at all. It gives you the ability to make educated guesstimates much closer to the truth than evidence with unestablished baselines will.

Since FE is already variable and testing adds more variables you have to have accurate data to build on to ascertain baselines, margins of error, and to be able to separate real results from flukes and assumptions.

[Mighty Mira]

Nice, 95Metro.

If you don't do proper testing, then you open yourself to trusting all sorts of things from snakeoil to duct tape.

[MetroMPG]

I'm with you 100%, 95Metro - defender of A-B-A and otherwise controlled-testing!

[95metro]

MetroMPG - The Great Testing Guru! You've proven over and over again just how important proper testing is.

Between your site (which you need to update! ) and Tony's fuelsaving site I've come to understand how important it is. At least I won't be performing anymore acetone "tests" (see links) like I originally did when I was new to the world of FE...

http://www3.telus.net/metro/log/nov0305.htm
http://www3.telus.net/metro/log/nov1805.htm

[cheapybob]

Originally Posted by Mighty Mira

Yes, it is more complex than simply saying x mod increases mpg by y percentage. It's because fuel economy is not a simple function.

Fuel is wasted in several different areas:

Load at highway speed:
1) In heating up the air through turbulence (high CdA)
2) In heating up the tyres and the road (high Crr tyres)

Engine/transmission efficiency:
3) In an engine that is in too high an rpm range for the load
4) In an engine that is too large for the load.
5) In too many, inefficient parasitic loads on the engine (alternator, pumps, etc)

In braking:
6) Through an overly heavy car
7) Because we heat up brake pads/discs/drums instead of returning kinetic energy to a battery.

2, 6 and 7 mainly apply in city driving. The rest apply to highway driving.

FE in a car can be visualized as being constrained by three bottlenecks. Attack one bottleneck, and eventually you approach diminishing returns. For example, start off with a car built like a brick and geared for city driving. Do some aero mods, and the highway performance will improve. Do more mods and it may not improve the FE much, because the gearing will be much too high for the steady state load.

Note that in each of those different areas, it is possible to measure the variables and so compute a function that will give a good approximation as to expected FE. However, it's more complex than saying "Do this and get X increase".

Now, since most driving is done in the city, it makes sense for car companies to attack the braking issue, and the only way to do that is with electric technology - you can't turn kinetic energy back into gasoline again. Hence the hybrid car - one with both a gas and electric motor.

However, if you can solve the other problems (i.e. decreasing the load on the engine), you can do away with the gasoline engine entirely and go solely to electrical, as range is a major factor in EVs not catching on.

I would say that as gs.org gets more of an understanding, people will learn how to measure and understand the effect of the different variables in their car, in the same way as people currently understand the interplay between different variables in producing a fast quarter mile time etc.

All those things are variables, and the more variables you have the less likely that consistant results will result. Skip the attempt at "city" testing, IMO. Its hopeless to get non-variable conditions. Even on a highway test its difficult.

[Matt Timion]

Originally Posted by cheapybob

All those things are variables, and the more variables you have the less likely that consistant results will result. Skip the attempt at "city" testing, IMO. Its hopeless to get non-variable conditions. Even on a highway test its difficult.

Agreed... besides, when you only test highway you get better bragging rights :P

Seriously though, an extra stop in the city could cause you your new MPG record. When on the freeway/interstate you can easily use the same route, same lane, etc. The variables you'll have (temperature, weather, etc.) can be quantified and can be taken out of the equation.

[JanGeo]

20/20 last night they interviewed the Turbonator inventor - apparently it does NOT work but they have sold 150,000 units and say that it does. Also intervieved a platium gas additive that works in MOST cars to improve mileage again another failure when tested by EPA however they rely on car vibration / motion to add the chemical and EPA runs on a dyno.

[ZugyNA]

Here is a chart showing the results of tank to tank testing.

Notice the mpg difference between the dark blue line (2000) and the light orange line...same period in 2006...showing mpg gains. Avg on this line in 2001 is about 21 mpg...in 2006 it's about 26.5 mpg...around a 26% gain.

Notice the green arrow showing where I started testing acetone and an FA2000 in 2004. The poor results in early 2004 were due partially to a failing AC control which ran the AC too much.

The FA2000 didn't contribute much due to my thinking it was open too far, while the real problem was a lack of a shraeder valve to keep the raw gas from overloading the O2. So actually it was shut off most of the time.

Easy to see the time of the year (ambient temps) vs mpg.

There are gaps where the car wasn't driven. No records for 2001 thru 2003.