[chrs]

For constant-speed, level road driving, the real-time MPG indictor tells you exactly what you need to know. You can compare using differnet gears and throttle positions and know which is the best choice.

But for accelerating and hill climbing, it can be hard to tell what's best. If you are accelerating hard, the bar is down at 25 mpg or so, which seems bad, but in fact the engine may be operating very efficiently--using a lot of energy but storing the result as kinetic energy of the car. You may be better off opening the throttle pretty wide (where the engine is efficient), accelerating quickly, and then suddenly leveling off your speed, rather than gently sneaking up on the desired speed, even though the latter looks better on the indicator.

My purpose isn't to start at debate about acceration techniques (and if you want to have one, I suggest starting, or picking up a thread in the MPG Issues forum). Rather, I want to suggest an indicator that is corrected for acceleration and slopes. Then, rather than debating techniques, we can learn what works best by watching the indicator.

So how do we sense and correct for slopes and acceleration? Actually, an accelerometer can't tell the difference between accelaration and slope (nor can an inclinometer), which sounds like it might be a problem, except that it turns out that for the purpose of what I'm suggesting here, we don't need to know which it is. Either way, we are storing energy. What I want to know is whether the bar has dipped to 25 mpg because I'm using the engine in a very inefficient regime, or because I'm storing most of the energy coming out of it.

Here's the algorithm to do it: From the acceleration and the speed, one can easily calculate the rate at which energy is being stored--the power going into energy storage. The hard part is figuring out how to correct the raw mpg number given that number. What I'd suggest is to calculate how many more miles you could theoretically go (at the present speed) with the extra energy you are storing.

The IMA energy storage should also factor in, but let's set that aside for now.

To be mathematically specific, the indicator now displays

MPG = velocity (mph)/(gas flow rate (gallons per hour))

That can be re-written as

MPG = (velocity/power)*(power/gas flow) = (velocity/power)*(efficiency)

What I'd like to display instead is

MPG_effective
= ( velocity/(power for steady speed) )*(efficiency of engine now)

= ( velocity/(power for steady speed) )*(power now/gas flow)

To get power for steady speed, I suggest that data be stored in a table. That data could come from a roll-down test--record speed as a function of time on a level road, while coasting in neutral. Or it could come from calculations using the published Cd and A, and an estimate of tire rolling resitance. If it comes from the roll-down test, power (from F = ma) is m*a*v where a is velocity and a is acceleration.

Using the fact that total power now is the power going into accelertion, plus the power needed to maintain speed:

MPG_effective
= ( v/(m*|a_rolldown|*v) )*(m*(a_now+|a_rolldown|)*v)/gas flow)
= [(a_now+|a_rolldown)/(|a_rolldown|)] * (v)/(gas flow)
= [(a_now+|a_rolldown|)/(|a_rolldown|)] * (actual present MPG)

That's a nice simple result.

So now now can that be implemented? How can one get the actual present MPG data? In fact, how does the FCD sense the fuel consumption? Can one tap into that sensor? Is that data would be available on OBDII? I got a palm-pilot based OBDII scanner (Auterra Dyno-Scan) that is supposed to be able to display real-time or trip MPG for people who don't have a FCD, but that feature doesn't seem to work on an Insight.

If all else fails, an option is optical sensing of the FCD bargraph, or just making a unit that would display the factor [(a_now+a_rolldown)/(a_rolldown)], and the driver could do mental arithmetic to calculate the result--that way, if you are driving your Insight instead of walking or riding a bike, you'd still get mental exercise, even if not any physical exercise!

One cool thing about this formula is that the result goes negative when you brake. That's a nice way to show the negative effect of braking on mpg--something else that doesn't show up in the present system. (I'm again ignoring the effect of IMA and regenerative braking)

Charlie

 

[b1shmu63]

Whoa, Nice concept!

 

[Mike Dabrowski 2000]

Hi CHRS
Dont mean to jump on your thread, but think you may have uncovered another thing MIMA can do.

But for accelerating and hill climbing, it can be hard to tell what's best. If you are accelerating hard, the bar is down at 25 mpg or so, which seems bad, but in fact the engine may be operating very efficiently--using a lot of energy but storing the result as kinetic energy of the car. You may be better off opening the throttle pretty wide (where the engine is efficient), accelerating quickly, and then suddenly leveling off your speed, rather than gently sneaking up on the desired speed, even though the latter looks better on the indicator.

I follow your concept, see if I have it right.
The thing not considered in the present(efficiency) indication is the stored energy in the cars inertia. The energy to change that stored inertia while accelerating, and the amount of gas to acheve that acceleration is The true efficiency, not what it says on the MPG indicator.

MIMA will have VSS to determine acceleration, MAP to show engine efficency, and throttle to let it know what you want to do.
I could see it displaying an efficiency graph on the MIMA bargraph.

 

[chrs]

Hi CHRS
Dont mean to jump on your thread, but think you may have uncovered another thing MIMA can do.

Awesome!

But to get fuel consumption from MAP, you need to know the mixture, and that depends at least on whether you are in lean burn or not. And of course the RPM but that's easy. Will MIMA know the mixture or some proxy for it?

People have talked about how hard it is to make an optical sensor for the instantaneous MPG bar graph. I think it would be possible to make that work with an adaptive threshold based on the signal from another optical sensor, but that's messy and complicated.

Using VSS for acceleration makes a lot of sense--better than my idea of using an accelrometer, especially since speed is needed for my calculation anyway.


Bottom line: if you have a way to get a signal that is proportional to fuel use, and it works in or out of lean burn, I'd love to help you set up the calculations for this.

Charlie

 

[Mike Dabrowski 2000]

Chrs
Will have to see what kind of acceleration resolution I can get from the VSS, to know how good we can compute acceleration from the signal, I did a quick look at the VSS signal with my scope, and it goes from 0 HZ when the car is stopped to about 55HZ at 40MPH, and it is a linear change.
The map is the mixture determining signal that the ECM uses to control the fuel injectors, and is very sensitive indicator of engine load, and from what I can see, has a signal that goes from 0.79V at idle with no load to about 1.2V at 100MPG steady state, to 0.65V when the engine is being back driven during engine braking. It maxes out at about 2.6V at full throttle. It should be a good indicator of engine efficincy, but not MPG, but as you stated, the MPG is not the best indicator of efficiency.
The simplest hardware to determine what the MPG bargraph is showing was suggested in the past, just capture the video image,from one of those mini tv cameras like I used for my rear MPG display, which has AGC to compensate for lighting changes. Do a simple OCR scan of the captured frame to decode the MPG numbers. May be able to do a progressive frame grab with a PIC microcontroller,kind of like the Snappy frame grabbers of several years ago. I happen to have 2 of those in my parts bin..

:wink:

 

[chrs]

Will have to see what kind of acceleration resolution I can get from the VSS, to know how good we can compute acceleration from the signal, I did a quick look at the VSS signal with my scope, and it goes from 0 HZ when the car is stopped to about 55HZ at 40MPH, and it is a linear change.

That's a good data point. It might be that we have to time the interval between pulses, rather than counting pulses in a fixed time, in order to get enough resolution on the velocity signal to also calculate a good acceleration signal. Is that within the capability of your hardware, or should I start thinking about an extra circuit to do that function?

The map is the mixture determining signal that the ECM uses to control the fuel injectors, and is very sensitive indicator of engine load....

Right--the amount of fuel injected should be linearly related to the MAP signal, as long as the fuel/air ratio is held constant. But when you enter lean-burn, the mixture is suddenly drastically changed. So there's an unknown nonlinear relationship between fuel per cycle and MAP. Furthermore, there are other factors that affect whether you go into lean burn, at least including what gear you are in. If all I wanted was a general indication of doing well or poorly on MPG, it would be fine. But I actually want something truly proportional to fuel use, because what I want to do is compare gently accelerating in lean burn to open-throttle acceleration.

It should be a good indicator of engine efficincy...

It certainly doesn't directly indicate efficinecy--it gives a rough nonlinear indication of fuel consumption per revolution. But for efficiency we need fuel consumption per power out. That's where my adjustments factor in--to take numbers on fuel consumption and change to efficiency. But I'm not sure it's worth correcting for accleration if we can't get the fuel consumption number more accurately to start.

Does anyone know whether the mixure is roughty constant until you enter lean burn, and roughly constant in lean burn, or whether there is a lot of variation within each mode? If it's just two constants we might be able to detect lean burn and switch the factor.

One of the NREL papers included an engine map giving efficiency vs. rpm and torqure--but apparently it was only for non-lean-burn mode. It showed best efficiency for the throttle nearly wide open, which is what you'd expect without lean burn, but doesn't help answer the interesting question of what happens with lean burn.

The simplest hardware to determine what the MPG bargraph is showing was suggested in the past, just capture the video image,from one of those mini tv cameras like I used for my rear MPG display, which has AGC to compensate for lighting changes. Do a simple OCR scan of the captured frame to decode the MPG numbers. May be able to do a progressive frame grab with a PIC microcontroller,kind of like the Snappy frame grabbers of several years ago. I happen to have 2 of those in my parts bin..

As primarily an analog electronics guy, I'm amused when "simple hardware" involves multiple chips each with millions of transistors, but of course from a practical point of view that is simpler than an analog circuit with a few transistors that needs pots adjusted for every 10 degree temperature change....

Maybe I need to test different acceleration methods simply by accelerating different ways, going a few hundred meters, and resetting the FCD after each run. Of course, once I have MIMA installed, that will be easier, since I can turn off the IMA for the tests and get a better indication of how the engine performs.

Charlie

 

[Mike Dabrowski 2000]

I managed early on to get a good reading of the whole bargraph with a single photodiode. The only problem was that to make it work in bright sun in the hole ambient to total darkness needed me to cover the whole bargraph with opaque material. The next step would have been to be to make a repeater bargraph that would replace the hidden one, and run off of the analog level.
Timing the interval between rising edges of the VSS should be straighforward to measure with a lot of accuracy, since the processor can do 10MIPs
I have not actually worked on that part of the program yet, so cannot give more info about resolution, but would expect resolution much better than the 1MPH that we have on the speedometer. and since we will do our own calculations, we could correct it to a GPS so it is more accurate than the car.
I think the fuel consumption is computed by counting the duration of opening and the frequency of opening of the actual Fuel Injectors. At a fixed pressure (fuel Pump) and a known temperature fuel, you could with some calibration compute it independantly by capturing those signals. I can see another PIC microcontroller that could measure the fuel injectors and VSS, and compute a corrected actual efficiency for a hill climb.