As it pertains to the modeling a base calculation of rod/stroke ratio should be part of the calculation. That determines piston decel, dwell & accel rates. Decel & accel rates will vary as the crank goes through it's circular range of motion based on the rod/stroke ratio.

 

Rod ratio also plays a role in whether the force being created is going into pushing the piston down/turning the crank/being converted into work, or if it's being wasted by pushing the piston sideways against the cylinder walls and just turning into heat/friction.

 

I quoted this post instead of both. Nothing to apologize for. I like nerding out. But DavidBoren is correct. It gets way blown out if we don't narrow the scope first to prove it out.

OK, so for the non-physics folks who might be reading this, when we say "adiabatic", what we are saying is that the total heat in the system is constant other than what is added or extracted on purpose (short version to keep it relatively simple). In reality, there is something called an adiabatic curve that and engine could theoretically follow if it were perfect. A quick example is the compression stroke. Imagine compressing the air, and all the pressure and temperature changes stayed in the air with no bleed off, and no heat loss. Not gonna happen, but that's the basic concept.

There are a lot of things I'll never be able to accurately calculate, but I don't think it will turn out to be necessary once they are at least accounted for. For example, during the power stroke, a ton of energy goes through the heads straight to the coolant and does no work. I may never get that number, but I might still be able to account for it by first getting the data points that I can account for.

Which is why I started with parasitic acceleration. That can be calculated pretty close. Where I'm still reworking is that there is a dual wave function going on, and I'm not sure yet what function to assign the second wave. Using the rod/stroke ratio is only part of it, because the angle makes a tremendous difference as well. I need to figure out the correct function and then integrate both over crank angle rotation and then divide by 2Pi to get the average acceleration. It was easy with sinusoidal only function, but I'm still working this one out.

I've also started working on fueling, and I'm actually very comfortable with how this is coming out. I started with the VE's and after a ton of unit conversion - using a straight 12.8 AFR (which was my dyno session) - I got 70% duty cycle at 7200 RPM. On my HPT logs, 70% is right about where I seem to max out on repeated runs. So, the first go on fueling looks to be within a few percent, which gives some raw energy inputs to start working with. I want to look more at the logs and compare before posting these equations.

My method was to take the VE and cylinder volume, RPMs, and air density to get grams of air per second. Then use AFR to get grams per second of fuel. Then use the heat of combustion of gasoline to get the total heat input to the system - assuming perfect combustion to CO2 and H2O. Then, convert KJ/S to KW (1:1) to HP to get into the same units.

I'm netting 1095 HP of energy input to extract 438 HP out according to the model at peak power - 6800 rpm.

I'm netting 787 HP of energy input to extract 379 HP out according to the model at peak TQ - 4800 rpm.

 

I can say this much...

No matter how I try to run the calculations, air and fuel definitely continue to rise after peak power. I have to modify the VE's to fall off the face of the earth to get air and fuel to decrease immediately after peak power, but then I start to get efficiencies that are way too high and the power output falls far faster than the dyno curve actually does. In other words, I have to force the model using known erroneous/false data to make the airflow and fuel fall off after peak power

It is almost certainly internal resistances adding up.

 

Your fuel calculations look very promising. And closely support your calculations for the horsepower required to spin the rotating assembly at those rpms. So at least the work you have done so far seems to be checking out. Either you're getting close to correct, or both sides are equally wrong together, either way you are consistent.

 

Now you are talking about the resolution of error I was talking about. The parasitic loss you're trying to equate for, or are seeing/ not seeing, is due to ,as I said, the error in the sum of the whole as far as its scalar variance from a theoretical adiabetic function.

The farther from optimal the model becomes the larger the exponent of calculated error will be. What the actual error is can range from a squared to a cubed function across the range of BSFC.

To further add to your comments on adiabatic function for those not in the know, they can start here https://en.m.wikipedia.org/wiki/Adiabatic_invariant

Furthermore, The model of adiabatic function is a square. Time a linear measurement, energy a vertical measurement. In other words the function of energy rise & drop would happen in zero time & the enthalpy would sustain, over time, as a constant. The Conclusion is that adiabetic function can only exist outside of space & time. The net result of complete loss of enthalpy would return to the energy level before the energy reaction begins, or ambient temp. That is 100% adiabetic, or .000 BSFC.

To compare the non-adiabatic curve of an ICE, by overlaying it BSFC map, in comparison to the hypothetical "square" of adiabatic operation shows how far off it is.

As it pertains here, we only need concern ourselves with as close to adiabatic as we can get, within the constraints of energy loss over time & in mechanical transfer ability of the engines design, & factoring in remaining enthalpy.

There has to be energy retention in the system for the ICE to continue to make power. This has to do with the fact we can't operate outside of space & time. The very upper limit, from the testing I have been educated on, that is around 800*F EGT. That is in that 5hp/ cu in power range, under stoich burn conditions, at a RPM equivalent to a hypothetical Compression ratio.

 

this is like restricting your theory and question to a single engine. Why can't you compare all engines on the planet? Use data from every combustion engine to shape your theory. Its like trying to learn about DNA by only comparing human Genomes. You MUST compare every available orgamism's genome to get the more accurate model. You MUST use data from all engine platforms, in all engine scenarios, from airplane to boat perspectives, with every available part possible. Science does this for DNA data- its called bioinformatics. We do not have such free easy facility for engineformatics (is that what it would be called?) where part data would add up to scientific literature.



Next, a performance engine near peak torque frequently experiences VE above 100%. If you are using 100% in the old fashioned equation it would put you at a 10% loss near redline because your initial reference was using 100 when it was actually 110%. So from your mathematical perspective you have been missing 10% and maybe also calling it "parasitic loss".

Many engines employ variable camshaft timing now, even while you drive some can adjust electronically for a flat VE curve. These engines, like the Honda S2000 engine, have a nearly flat VE curve across the board to redline. If the engine is prepared for the application there should not be a "magic brick wall" so much related to the weight or friction of the design, more likely there will be a valvetrain related or piston speed limit related problem first. In other words, a real racing engine has every element of friction reduced as much as feasible to maximize horsepower, in general, so in these applications the engine is prepared properly, and such "steadily increasing friction component forces" are going to be reduced per the application, as opposed to using too tight of clearances as in OEM engines, and causing OILING/FRICTION issues because of OIL FLOW related failures.

 

I almost spit my drink out when I read that! Too funny. Like in Caddyshack 2. "Just aim way over that way and play the slice - you'll never correct it anyway."

I'm starting to really think it's getting close to correct. I put my previous cam's stuff - VE's, DCR, etc, and it was ground 9 degrees off, which is why I swapped it out. It accurately predicts a double hump torque curve, an early torque peak at 3600, and an early power peak. It is not correctly predicting the peak power RPM. the motor peaked at 5600, but it's calculating 6200.

 

Yes. And so many internet calculators expect you to just key in your BSFC as if you just sort of know it, so you can play with the number to make it give you what you want. I'm trying to actually back into my BSFC.

There comes a point, though, when you have to just use a number that gets you close. In the GM tune, there is a table that tells the computer the estimated friction losses at various RPM and MAP. So, GM got to that point as well, though admittedly their fudge factor will be pretty damned accurate compared to mine...

 

You should be able to calculate a close BSFC off of known fuel mass vs airmass. If you have a quantitative value for HP on the log of air mass then you have a start to a scalar map.

For simplification purposes a modern EFI motor will probably not exceed, in almost every normal application, .5BSFC. That would be the start to the model I would see as getting as close as you can to a hypothetical.

The base model I am talking about would be in the .3-.35 BSFC range. That is about our known limit at the moment. Hyper-scram Jet engines even exceed that number

 

First question - answer is that I don't have access to the data on the level I would need. If I did, I would simply run a neural net regression on a very wide data set, wait a month for the computer to crunch the numbers, and I'd have it. The only engine I have access to with the amount of data I need to begin building a model is my own. but, I can log my built LS1, my stock 5.3, use logs from my stock LS1 and previous build on the same LS1, logs from a stock-ish 4.8L I have and a stock-ish 6.0. That said, it's still very LS-specific, but it's what I have access to and can read via HPT.

Second statement - I'm using my actual VE's, not estimates. They do exceed 100% from 4400 to 6400 RPM. Based on the AFR on the dyno at 12.6-12.9 AND the PE multiplier of 1.15, I'm getting exactly the wideband readings off the dyno I should get with the VE values in my table. My stock VE's are no good, because the car was never tuned on the stock cam. The VE table looks like stir-fried a$$ in the stock tune. Curious as to how GM ended up with those numbers.

Lastly, It's starting to look like friction is overall a linear function. I've got some more work to do on it, but it actually can be somewhat calculated. The main variable is the oil viscosity changes with temperature. Everything else in the friction system appears to be relatively constant. The side loading of the pistons will be tricky, because engine load will determine piston side loading. Tricky, because the angle on the crank is not equal to the angle on the rod. I don't think it will be flat like "60 all the way across" or something easy like that, but I do think it will simplify to a multivariate line.

 

Well, I got the BSFC numbers, and they didn't make me hold my sides laughing or rush back to the drawing board.

I'm not going to count the low end, as I have this area deliberately lean - holding about 0.35 from idle to 3000. Starting at 3200 RPM, I'm at 0.4000 Lb/HpHr. From there, it is basically a steady rise to 0.508 Lb/HpHr at 6000 rpm. Then, it sort of jumps to 0.53 at 6400, 0.55 at 6800, and 0.58 at 7200.

TO me, that is a classic example of the law of diminishing returns - it is taking more effort to increase the engine speed a similar interval as the speed increases.

 

Can I see your raw numbers?

 

Thanks for that. Took over an hour to watch due to pausing and note taking. It does answer one of the burning questions as to why race engines will peak at 8500 vs 6300, and that is the reduction in parasitic losses on the mechanical side.

Also, breaking it down to mean effective pressure might simplify a lot of this for me.

Good to see you posting, Martin!

 

the BSFCs or all of them? I'm fine with whatever, but I just want to know what you want to see.

 

@GTFoxy - is this what you were looking to see?

OK, I found a mistake since my previous post - I had used my calculated TQ/HP values instead of my actuals when I calculated BFSC. I redid it with the actuals, and it makes a lot more sense now:

RPM......HP......BFSC
2400...152......0.456
2800...184......0.448
3200...227......0.43
3600...267......0.43
4000...309......0.436
4400...345......0.452
4800...380......0.468
5200...412......0.475
5600...436......0.48
6000...456......0.482
6400...471......0.487
6800...471......0.509

But, you can see that in the data, peak HP was crossed between 6400 and 6800, but it requires more fuel to obtain 471 at 6800 vs 6400, indicating that adding energy is not increasing engine output.

The above is calculated BSFC vs actual HP. based on injector duty cycle logs, BFSC is calculating correctly.

 

You can clearly see that BSFC is worse at 2400rpm, it gets better, then it gets worse again. Whatever made it worse at 2400 is the same thing that is making it worse at 6800, and it has nothing or little to do with friction or parasitic losses.

 

That's a pretty common pattern from what I can tell. You would think the low point would correlate with peak TQ, because that is generally considered peak efficiency, but it doesn't. Some cursory research I did was saying to pretty much ignore BSFC at or near idle, because the idle power output is so far down, it wrecks the denominator.

Another thought I had about this is that the VE is quite low at and near idle due to overlap. This came up earlier in this thread - losses out the pipe due to overlap. i think this is what is presenting on the low end. The cam really comes into itself around 3K and then pulls like a freight train. Martin discussed a similar phenomenon on his LSA thread. In this case, I think that around 3K, the overlap event is short enough in duration that the air and fuel quit short circuiting out the exhaust and the additional contribution to energy production begins. Past peak HP, I don't think that the overlap is magically getting longer in duration on the same cam, so I'm pretty sure the increasing losses are responsible at the high end.

The video Martin posted actually helped quite a bit to generalize things. They take the average effects for two revolutions and calculate the Mean Effective Pressure. That made it much easier to work with some of the numbers. I was able to somewhat generalize Brake Mean Effective Pressure, and there are two very distinct and obvious curves - one before peak TQ and one after peak TQ. When you see it, it becomes very obvious that the mechanisms affecting low RPM performance are not the same as those affecting high RPM performance. It is quite stark. I'll post a picture when my computer starts cooperating.

 

Red is after peak Tq, blue is before. X axis is VE and Y axis is brake mean effective pressure.

 

I was actually hoping to see the raw numbers, at peak torque & HP, & Injector size at static running pressure, pulse width & duty cycle as well as air mass rates at those points.