I’ve been beavering away at this simulation lark for quite a while now. I think everyone who is interested in understanding an internal combustion engine ( Be quick ! they are bound to be replaced by supercap powered electromotors in the next decade I reckon) and does not have an allergy for computers and maths should get EngMod4T and a copy of publication SAE R-161. As far as engineering books go, it is quite readable, but a novella it is not. I have read the first few hundred pages, but let face it, it is quite hard to read it like a story. The Fun part (for me at least) is to fiddle with a model, then stumble upon one of many oddities that arise then refer back to the book and re read it. Sometimes you are just as confused as before, but every once in a while you start to get it.
What bugs me is people telling me that you need to spend $5000000 to get a simulation that works. The argument mostly goes like this : OEM’s spend millions on software, how can a $99 or $500 piece of software be any good.
OEM’s will spend more on the design of a doorknob than a mid level race team on engine development.
What OEM’s want, is to cut development time on a 800 million euro platform development, they are not really bothered by a 10K a month licence fee, if it does what they want it to do, fast. Something like AVL Fire or GT power can and has to, cost a lot because the market can bear it and because it costs a lot to develop software that can tell you what happens when you move a panel a bit like this, the pipe xyz gets more like that and how is that going to affect the emissions and the temperature of the glove compartment. Then only sell a few thousand seats at best.
But still it is a simulation, not reality. Just as the picture of the pipe above is not actually a pipe, the simulation is not actually an engine and in the case of a 1D model is it not even a mathematical 3d model of 3d physical model as you seen in many of colourful CFD picture and animations, but a more simple linear mathematical model. The maths and the ”laws” of physics are also a model of reality. Newtons laws of motion give a very good description of what happens when you chuck an anvil from your window. So for that situation the model ( f=m*a) is very accurate even though it does not take into account a whole lot of variables (air resistance, micro fluctuations in the earths gravity, eddies in the space time continuum, etc.). However if you chuck the anvil down a black hole it probably is not accurate at all. This is a round about way of saying is, that you can have a very capable model in common conditions that omits a fair bit of variables.
Gripe number two. People tend to think that when using simulations you all end up in the same place. ‘CFD is not the answer otherwise all F1 cars would look the same’ and more in this vein.
This is utter tripe. Just as a file/hammer/violin does primarily produce worthwhile work in the hands of a skilled operator. It is a tool.
IBM Pollyanna Principle: Machines should work. People should think.
There is no program that will design something from a blank sheet of paper.
Ergo you still have to come up with a starting point that is near some kind of optimum. All the program does is solve the ever so entertaining maths, that would not be practical to do by hand ( if would take you a 100 years and you would not have any friends left). There are systems that allow parametric optimization of for instance a intake port using CFD. That is clever stuff and can find the ”optimum” port shape but only within the constraints you gave beforehand.
It is a really good method to weed out the bad options from the probably quite good options so that you are in turn closer to an optimum point when you actually do the physical testing.
This are two plots I nicked/borrowed from the GT power/ Gamma technology website . It is promotional material for top level software so I’d expect it to be , well not the worst data you could expect.
How well does EngMod4T fare ( with my model ! If I make a mistake in the input , it of course gives a wrong answer)
One of the simulations I did.
the engine is an a-series for lower ‘budget’ race class
- stock bottom end 998cc
- A stock head big bore head (actual flow data of the head was used but the data was pretty generic for a good but bone stock 12g940 head).
- An LCB based on measurements on a Minispares Evolution Medium bore LCB I have on the shelf.
- The prescribed cam for the class (S96 files where used that can be used in a cam grinder so they are pretty good) valve motion was with 1.3 ratio rockers.
- Intake is a Weber on a long manifold.
- Burn model is not tweaked/forced to make it follow the dips and rises of the curve. You would get a better match but it would no longer produce accurate predictions for different set-ups.
The power values are left blank as it is a race engine and it is not public domain (although this is just the baseline before it gets some trick new parts) . The torque and bhp values do no cross over at 5250rpm btw as it is in KW and NM (It’s that new thing called The SI system)
All in all not half bad for a program that probably does cost less that half of the monthly licence fee of GTpower ( granted you get a few other bits and bobs).