12 replies to this topic

[Canuck]

18 years ago...ah man we've been around here a long time...anyway, 18 years ago, this thread and McGuire's post (and Greg's) touched on CFD use for cylinder head development before it went off the rails into a 2V vs. 4V debate.

 

And 18 years later we have vastly more powerful computers but searching recently for cylinder head CFD hasn't been any more successful now than it was then. Software with a cylinder head flavour is aimed at OEMs and priced accordingly (WAVE, ANSYS, Converge, ETC). openFOAM is perhaps better than it ever has been (and more approachable), but there are no cylinder head / ICE modules in the public domain, and I get the distinct impression that even if there were, I might need to go to university before I can figure out how to meaningfully use and interpret it. The scanning stuff is a piece of cake, I own more than one. They are more like a Linux environment than an iOS device, which is to say they work well but they remain the domain of people willing to tinker and grind to figure it all out.

 

I suppose this is just a big whine about the state of affairs in this niche area. In reality, we're living so far into the future I am absolutely amazed when I pause long enough to think about it but in this area it would seem there's been precious little movement. Perhaps that opportunity remains.

[Greg Locock]

Bear in mind that for the last 5 years IC development has slowed to a crawl in the OEM world. We got rid of our powertrain people 3 years ago, the dynos are now shuttered.

[desmo]

Chasing hard-fought incremental ICE powertrain gains at this point has to be throwing good money after bad.

[Canuck]

Bear in mind that for the last 5 years IC development has slowed to a crawl in the OEM world. We got rid of our powertrain people 3 years ago, the dynos are now shuttered.

Fair enough, but I'm whining about the non-OEM level, pinheads like me. Racing hasn't stopped, tuning hasn't stopped, hotrods are still being built and while you'd swear that everything from Audis to Zastavas are running the same turbocharged LS swap there are still those of us (I assume the us as opposed to me) that want to understand cylinder heads better than we do. There's nothing wrong with just buying whatever everyone else is running I suppose but I squirm under the thumb of the notion that everything that could be done, has been done by someone already. 

[PJGD]

Agreed; for example outward opening valves (as opposed to conventional inward opening poppet valves) have not been given enough attention.  Here lies improved air flow.

[Greg Locock]

Trouble is cfd on combustion (not flow through valves) is way beyond cutting edge - you've got turbulent flow, rapidly changing gas properties and a whole lot else.

[Canuck]

Fair enough, there are reasonable limits I suppose. There remains, to my knowledge, a (large) gap between hobby / prosumer level software that appears to use reverse-engineered experience and fairly basic math to predict port flow, and software that uses more complex inputs and calculations to more accurately model it with a visual representation.

 

Here's my specific, personal hang-up. Why does the 3.5 litre straight 6 in the (growing number of) 80s BMWs in my driveway, in a 3200lb car get nearly the same mileage as the 7000 lb, 400 hp V8 I haul the family with? Some of that no doubt is friction and heat management, but I have to believe that the vast majority of it is airflow management. Why are a cam, a $3500 cylinder head overhaul/port and a set of headers worth a pathetic 15%-20% gain on an engine barely squeaking out 60 hp/litre? Or I suppose I could ignore all of it, toss the lumps on the scrapheap of history and be YALSS - yet another LS swap (to be fair to the LS platform, there's a reason they're so ubiquitous I would suppose).

[Greg Locock]

Engines in the 80s were just getting to grips with EFI and emissions regs and cats .Even the first generation of EFI was better than carbs for fuel efficiency, but your modern engine has had the benefit of 30 years of quite intense competition. Vehicle mass has an effect on fuel consumption, but if you are mostly highway driving, less than you'd expect. Also of course you now have a DCT trans, better tires, and so on.

 

My1973 Escort wagon (carby 1.3 litre), probably 850 kg kerb weight, got exactly the same fuel consumption as my 2.2 tonne 4wd diesel 2.0 litre biturbo SUV gets now.

[Canuck]

Which I suppose is sort of my point. What advances can I apply retroactively to this old lump to improve its efficiency. Modern tires are a given and we're (thus far) not particularly hampered with emissions equipment inspections so modern cats can be fitted. The cost of exchanging the OE Getrag manual for a newer ZF or Getrag (or whatever late model thing one finds) might be worth something. And of course things like VVEL are off the table barring an engine swap. It's probable that I'm just a pinhead tilting at windmills.

[Canuck]

Circling back to the original topic of approachable CFD, is there an available tool that will show the impact of the port's geometry on flow? I'm specifically interested in whether the port needs to be raised, or just the floor filled or just he roof raised or some of each or different things at different positions. Now maybe I'm out to lunch but I get the impression that the sub $1000 software packages use calculations based on very broad parameters like port volume and areas at specific points along the port but they know nothing of the port's position relative to the valve. Perhaps the only answer is scratching away at openFOAM or maybe SimScale. It won't tell me anything about flame front and the like but it's a step beyond where I'm at.

[gruntguru]

Why are a cam, a $3500 cylinder head overhaul/port and a set of headers worth a pathetic 15%-20% gain on an engine barely squeaking out 60 hp/litre? Or I suppose I could ignore all of it, toss the lumps on the scrapheap of history and be YALSS - yet another LS swap (to be fair to the LS platform, there's a reason they're so ubiquitous I would suppose).

72 hp/litre is not shabby for a 3.5 litre, 6 cylinder. There are only two ways to increase power for the DIY tuner - increase cylinder fill and increase rpm.

 

It is always going to be difficult (expensive) to increase rpm on an engine with big cylinders and a long crankshaft.

 

It is also difficult to increase cylinder fill (NA) on a well designed engine like the M30. If you already have the $3,500 cylinder head, cam and headers you are not going to find any significant gains with CFD. Forced induction or LS swap are your options. 

[Canuck]

No, I don't have a $3500 cylinder head. Or headers. I have an eBay turbo manifold and turbo and it does what it does.  My complaint doesn't lay there but with the stock unit waiting for a refresh build. Perhaps 72 hp/litre is "not shabby" but it's a long way from the mark. To suggest there's only cylinder fill and rpm seems short-sighted but I'm not here to argue.The point is I'd like better mileage if it's only going to make this much power in stock form and I'm not bound by production constraints or considerations or perhaps even packaging constraints. I can fab one-off metal parts, lay up composite bits, hell even print very-fancy-plastic bits, run narrower bearings, lighter rings, higher squish, fancy coatings... But it's all throwing darts with my eyes closed if I don't have a solid grip on the root cause(s).

[Greg Locock]

hp/litre isn't a very handy measure for engines, since it just measures how gutsy the durability guys are and how peaky the camshaft designers are feeling. I think torque/litre is probably a slightly better comparator.

 

However neither really gets to grip with bsfc. Typically optimum bsfc is at around rpm@max VE (and rpm@peak torque), but I can't remember if that is just one of those things or a demonstrable result.

 

Worse than that, typical driving sees very little WOT and bugger all rpms, in an auto at least. We monitored customer usage across hundreds of cars and 20000 km each and found zero counts at 4000 rpm or more (obviously they didn't do this on engineer's cars, the Norlane grand prix was a thing).

 

So if your goal is efficiency you have to throw away the boy racer box of bits, but I wouldn't put any great reliance on being able to calculate a decent bsfc map . There may be some pointers in Heywood.