Intake flows

[woollymoof]

I've looked for other posts on this but can't find any.

What sort of numbers are people getting on the flowbench for their intake, for one runner and all four?

Our bench can't go above 17"H2O, if people are willing to post what they're getting for this drop, that would be great.

[woollymoof]

I've looked for other posts on this but can't find any.

What sort of numbers are people getting on the flowbench for their intake, for one runner and all four?

Our bench can't go above 17"H2O, if people are willing to post what they're getting for this drop, that would be great.

[Frank]

woollymoof, if that is your real name http://fsae.com/groupee_common/emoticons/icon_smile.gif

we measured plenum pressure whilst the engine was making maximum power

it was about 20"

so testing restrictors at 15"-20" is a good idea

ours choke at about 40" and get 135CFM, but of coarse this is irrelivent information, because the engine wont pull 40".. in fact any restrictor with a good throat design will reach about 135 CFM, regardless of exit diameter

to be honest, i cant remember the 20" figure right now.. something tells me it was 112CFM

i will check

never tested single runners, as i never thought it a wise use of time

[BeaverGuy]

I would say that flow through individual runners is probably more important than through all four at once. You never have all four cylinders on the intake stroke at one time so I don't see how that data is of any use. However, you do ususally have two cylinders with a small overlap of intake strokes and that can probably be approximated with flow through just one runner. Also the flow through a single runner, if tested on each runner should give you an idea of the flow distribution between the runners and show you how good of a plenum you have.

[woollymoof]

True, you wouldn't have all four cylinders on the intake stroke at the same time, but even if you did the flow bench would no way near represent the flow in the plenum. Pulling on all four runners at a time is sort of like an average of what the restrictor and diffuser would see.

Our results are 105.5 CFM at 17" on all four and we average 76 CFM on one.

[BeaverGuy]

I don't care if the volume of flow on the bench matches that on the engine. It isn't the flow rate that interests me. It is the variation in flow rate from runner to runner.

Even if I was interested in raw flow numbers I still don't see how flow through all four runners at once is any better an approximation of what the engine flows would be. The flow in the engine is a highly unsteady pulsed flow. That can't be approximated on a bench unless you are opening and closing valves.

Now if someone out there can show me why it at all makes more sense to draw through all the runners at once please do. http://fsae.com/groupee_common/emoticons/icon_smile.gif

I hope to be doing some of this testing in the next week or so. I will post some numbers then. Until then, I will try and find some data from a couple years ago when they did flow through all four runners and post that.

[woollymoof]

<BLOCKQUOTE class="ip-ubbcode-quote"><font size="-1">quote:</font><HR> Now if someone out there can show me why it at all makes more sense to draw through all the runners at once please do. <HR></BLOCKQUOTE>

<BLOCKQUOTE class="ip-ubbcode-quote"><font size="-1">quote:</font><HR> Pulling on all four runners at a time is sort of like an average of what the restrictor and diffuser would see. <HR></BLOCKQUOTE>

About drawing through one runner at a time. I'd argue that the flow distribution you get across the runners doesn't represent reality, as suggested in a post about CFD a little while back for the reasons you mentioned above - highly unsteady pulsed flow.

[Frank]

ever swapped plenums on a dyno?

do it, see if you can measure the difference..

dont get me wrong.. i think it is great seeing all you guys tear engines apart and doing really interesting stuff, but I believe the main reason most people usually do this stuff is because the "restriction" is in the port

personally, id concetrate on the restrictor, and individual runner lengths, then make the car lighter

gees, im gonna challenge for a bronze medal in "most posts"

[Chris Boyden]

Is there a standard pressure drop that flow numbers are evaluated at on a flow bench?

When we went to a local guy here with a flow bench, he used 14" and 28" of water to run the testing.

It's cool to throw out numbers, but unless we can compare apples to apples then it's just confusing.

[Mark100]

I would have to agree with BeaverGuy and woollymoof, that flow testing an intake and using the flow data to try to determine the degree of cylinder-to-cylinder variability is next to worthless. As has already been stated before, the flow is highly pulsed flow, and I will add that the assumption of steady state flow after the throat is very poor. It is probably an ok assumption before the throat depending on what you are trying to accomplish.

How many people have taken their flow data and compared that with some form of dyno data that indicated cylinder-to-cylinder balance?

I would add that I think steady state CFD applied to cylinder-to-cylinder balance is also pretty worthless. Basically you have a computerized flow bench (without getting into mentioing CFD accuracy and methodology). Even unsteady 1-D methods such as using WAVE, GT Power, etc. can have problems getting cylinder-to-cylinder balance correct.

Frank, you mention that the plenum pressure is 20" (" H20, I am assuming). Where are you taking the pressure measurement? Also, this is an averaged pressure, and is not necesarily what the restrictor is seeing across it in actual pressure ratio. Due to pressure dynamics you will get more than 40" of pressure difference across the restrictor, at an instant in time. Averaged pressure ratio is another matter.

Don't get me wrong, I think flow testing is very useful, but has it's limits when applied to the pulsed flow nature of an intake manifold.

[BeaverGuy]

I did not say that flow testing the manifold on a bench to determine cylinder to cylinder variations was useless. I said it was more relevant than flowing through all of the runners to see what the max flow rate through the system is. Deteriming maximum flows for comparison sake is fine but just to get a number is pointless. Do I think it is a perfect method of measuring cylinder to cylinder variation? No I don't. Do I think there is a better method? Yes, but most of us can't afford the tools to measure the flows while on the dyno in an unobtrusive manner.

Now to answer some other things, yes we have swapped plenums on the dyno, and we did notice a difference,nearly 5 ft-lbs at 5500 RPM. The change wasn't what I would expect but there may be other reasonings behind it.

Steady state CFD, yes I would say it is worthless because it is just as easy to set up an unsteady analysis. You only need to simplify when you have to, and in the case of the flow bench that is exactly what is goin on. You can't do an unsteady test real easily and you can't test on the engine that easily or cheaply.

and I direct you to Intake Testing (http://fsae.com/eve/forums?a=tpc&s=763607348&f=125607348&m=99810976611) for someone elses opinion on intake testing.

I think that is everything I wanted to say.

[John Bucknell]

Believe it or not, we OEMs (DCX and GM) use flow benches almost exclusively to understand and develop momentum effects on distribution - usually confirmed by individual fuel-air measurements on dyno within a percentage or two. Acoustics impacts distribution hardly at all in my experience (that is to say plenums can change the overall vol eff, but not individual distribution unless the change impacts the momentum effects).

I'm always willing to listen to your explanation of your development process and how it was correlated to the dyno.

[GTmule]

I tend to agree, john. I think the "pusling" of the flow affects the OVERALL flow capacity (VE) of the manifold more than the distribution. in the simplest example, if one runner is a smaller diameter than the others or has a more "pinched off" opening, this will show up on the flowbench as a runner with less flow.

The harmonics would seem to come into effect more when you're looking at the big picture (i.e. how much will it flow ON the engine?), since the runners, of course, are drawing at different times.

I'd not put ANY stock in flow bench numbers, as far as deciding how much the ENGINE is flowing, or even the relative efficiency of the manifold when it's ON an engine. All that being said, liek all things with knobs, motors, fans, dials, gages, and the like...flow benches are FUN to play with. http://fsae.com/groupee_common/emoticons/icon_smile.gif

But what do I know?

-Chuck

[Frank]

mark,

thats 20" water
taken from a static pressure tap in the plenum

[Chris Boyden]

I found this from a google search.

http://www.highperformancepontiac.com/tech/0210hpp_flow/

There is also an interesting conversion chart to convert flows at a given pressure to flows at another pressure.

"A flow bench test is done at a fixed test pressure. Due to this, the test value will need to be identified. Higher test pressures yield higher flow numbers. SuperFlow Corporation, the industry standard for flow benches, suggests 25 inches of water. The test pressure is read on a manometer attached to the bench while a rotary valve is controlled by the operator to arrive at this value. The aftermarket performance industry commonly flows at 28 inches of water, yielding higher flow numbers from the same port. A port that flows 200 cfm at 25 inches of water will now flow 212 cfm at 28 inches. Shops that have smaller or older flow equipment may not have the capacity in cfm to move a sufficient air volume to create higher test pressures. This style of equipment is usually operated at 10 inches of water. The results can be converted to any value with the chart provided in this text."

[Ben Inkster]

Hey Guys

I have done alot of flow bench testing, steady-state CFD, cyclic CFD and dyno testing to answer these questions for my own research.

So to cut a long 100 page document short, I found that steady state flow through the manifold was highly inappropriate and often contradictory to actual mass flow distribution and volumetric efficiency between cylinders. Particularly at around the 4000-7000 rpm region, manifold resonance was the dominant factor with a capitol D. But at around 12000 rpm it became far more like the steady state predictions.

Obviously, these kinds of studies are dependant upon manifold geometries and engine types, however the only reason I have investigated this stuff is because all previous UWA manifolds have been severely affected by manifold resonance (nearly 20Nm difference) and not steady state flow efficiencies.

Anyway, thats my 2cents (AUD), so its probably only worth 1cent (US).

Cheers

-Ben Inkster

intake guy
University of Western Austrlia

[BeaverGuy]

Ben,

are you still using a side entry plenum? I haven't actaully seen any pictures of your car but I have a copy of a UWA paper on the intake system design from a couple years ago and it seemed to hint at issues like that with the side entry plenum.

It would seem that a side entry plenum would be affected by the resonance more than a top entry design. I say this because the air has to pass by all of the intake runners to to reach the back where as on a top entry it doesn't. Thus it would seem to make steady state comparisons useless for a side entry design.

[Mark100]

Beaverguy - Sorry, didn't mean to put words into your mouth.

John - I understand how you get from AFR, back to mass flow through the port, and that if all the AFR's are the same, then distribution is equal (as long as injectors are injecting the same amount of fuel, or injector imbalance is taken into account), but how do you break down the contribution of momentum and pressure wave effects, and the relative contribution of each to cylinder-to-cylinder air distribution?

Frank - Thanks for the info.

Chris - An increase in pressure may not necessarily increase the flow to the same value as the chart given by the website. In Gordon Blair's book, "Desing and Simulation of Four-Stroke Engines", he goes into some detail on the pressure sensitivity of flow coefficients for different cylinder heads. Basically he shows some graphs which show contour plots of flow coefficient, plotted against pressure ratio and unitless lift (L/D). Some heads show more sensitivity than others. Worth a look if your interested.

[Mark100]

Ben - what kind of cyclic simulation where you doing? Coupled simulation?

[John Bucknell]

Mark -

As others have said earlier, the absolute numbers on the flow bench are not important - but rather the percentage flow of each runner relative to the ports (which are usually identical). Worst case for distribution is almost always peak flow, but once the test is set up it is easy to do a sweep. And yes, fuel-air ratio tests are done with flow-matched nominal injectors to isolate the variation due to fuelling. So if the dyno tests match the relative maldistribution of the bench, I call that correlated. It is important to remember to use a pressure ratio high enough that you get fully developed flow for decent on-engine correlation, Ricardo has a set of recommendations that are based on the orifice size (ie valve size) that INCREASES the necessary pressure ratio with DECREASING orifice size. Superflow's recommendations are based on this basic fact, and GM and DCX use 25" H20 (which are approximate in-cylinder pressures during an intake stroke on most unrestricted engines at peak power) and race teams tend to use 28" H20 - so if you are using lower pressure ratios than that, you might get misleading info.

Ben -

My professional experience has yet to explore restricted engine performance - I'm sceptical about your distribution claims, but data always wins. Having said that, correlation on high overlap engines can be difficult due to very misleading fuel-air ratio signals due to short-circuiting (edit - this is a issue on unrestricted engines, probably less so on higher exhaust to intake pressure ratio engines. This suggestion probably wouldn't reveal anything useful to Ben).

[DPH]

John,

You have me a bit worried. From what I gather you would agree that if through flow bench testing one develops an intake manifold/port combination that gives equatible percent flow distribution one should expect correlation with numbers given by AFR within a few percent. Conversly, I am assuming that you would expect that if given equal AFR numbers first you would then predict the flow bench data would show equal flow distribution. And thus this makes flow bench testing the good old engineering indicator and process for designing intake manifolds with good air distribution.

Now take a step back and be a little more skeptical and look at it from a whole systems approach. Could you devise an experiment that would show that the strong correlation you suggest is partly attributable to for a lack of a better word luck? How about if we take a fanstatic flow bench designed intake manifold and hook it up to a lousy unequal "extraction" exhaust manifold. We would expect lousy agreement between flow bench and AFR numbers (with flow equal on the bench, but AFR unqual on the dyno), right? Now what if we took a lousy intake manifold with poor air distribution, could we design an exhaust manifold to even up the distribution? Sure.

I would avoid over-hyping the flow bench, it is a good tool, but it only models a subset of the reality an IC engine operates in. In this case it is a bit like a pyschic (who here has money problems or lost a loved one recently or knows someone with a first name that starts with a "J", I must have a connection to the great beyond), we want to believe the results more then we should.



On the momentumn or "acoustic" tuning debate you believe a flow bench shows that the engine is largely or completely tuned by momentumn. Using solely flow bench data how do you show momentumn is significant over "acoustics" in a real IC engine? How do you analytically decouple the two?

<BLOCKQUOTE class="ip-ubbcode-quote"><font size="-1">quote:</font><HR>Originally posted by John Bucknell:
Mark -

As others have said earlier, the absolute numbers on the flow bench are not important - but rather the percentage flow of each runner relative to the ports (which are usually identical). Worst case for distribution is almost always peak flow, but once the test is set up it is easy to do a sweep. And yes, fuel-air ratio tests are done with flow-matched nominal injectors to isolate the variation due to fuelling. So if the dyno tests match the relative maldistribution of the bench, I call that correlated. It is important to remember to use a pressure ratio high enough that you get fully developed flow for decent on-engine correlation, Ricardo has a set of recommendations that are based on the orifice size (ie valve size) that INCREASES the necessary pressure ratio with DECREASING orifice size. Superflow's recommendations are based on this basic fact, and GM and DCX use 25" H20 (which are approximate in-cylinder pressures during an intake stroke on most unrestricted engines at peak power) and race teams tend to use 28" H20 - so if you are using lower pressure ratios than that, you might get misleading info.

Ben -

My professional experience has yet to explore restricted engine performance - I'm sceptical about your distribution claims, but data always wins. Having said that, correlation on high overlap engines can be difficult due to very misleading fuel-air ratio signals due to short-circuiting.... <HR></BLOCKQUOTE>

[John Bucknell]

<BLOCKQUOTE class="ip-ubbcode-quote"><font size="-1">quote:</font><HR>Originally posted by DPH:
On the momentumn or "acoustic" tuning debate you believe a flow bench shows that the engine is largely or completely tuned by momentumn. Using solely flow bench data how do you show momentumn is significant over "acoustics" in a real IC engine? How do you analytically decouple the two?
<HR></BLOCKQUOTE>

Actually, all I said was momentum dominates the distribution portion of intake manifold design. 1-D simulation code cannot adequately describe 3D flows (yet, they're almost there), so we must find other tools to correlate the momentum-impacted portions of the model. Flow benches are merely one kind of tool, which at today's level of geometry generation and CFD outperform dramatically for productivity. You can iterate through complex geometries, isolate problem areas, modify and retest at a rate that makes comptutational methods seem like a waste of time. That isn't to say I don't use CFD, just that it is only one tool - one I use for time-based analysis mostly.

As a judge, my job is to make sure you understand the problem you have attempted to solve. I'm an engineering professional in powertrain development - and I know the subject matter (powertrain development process) is extremely poorly documented, so I am not here to tell you what to do but rather provide insight into how it is done in industry so you don't have to invent the whole field on your own thus perhaps saving you the most precious resource of all, time. If you bring data and stand your ground well when I grill you in the design tent - you score points, doesn't matter whether I would have done it that way.

Finally, you have to really try hard with an exhaust design to mess up distribution of a good steady-state designed intake manifold. As long as the distance to the first change in cross-section is within 10-20% from cylinder to cylinder, it'll work just fine in a high overlap engine - low overlap are even less sensitive to exhaust runner length equality. It might be horrible aerodynamically, and cost you a ton of backpressure work but backpressure doesn't impact distribution enough to matter (it's a 4-stroke, remember). Conversely, a single entry intake is extremely easy to mess up - I've never seen an aerodynamically-ignorant design even come close to being satisfactory, and believe me I see poor designs every day.

[Mark100]

John, first let me state that I am not in the competition anymore, and I am not being judged. Also, I realize you are in industry, and you probably are experienced, but that by itself doesn't preclude me to believe anything at face value. However, I am also always open to explanation and discussion. Further, I have friends in the automotive industry and they are never without a story of how people are getting things wrong. Some of them even lurk on this forum. Being in industry by itself doesn't convince me of anything, without some discussion.

I already agree with you that the absolute numbers from a flow bench are not that important, however I disagree that a flow bench will necessarily give the proper indication of cylinder-to-cylinder airflow / VE distribution.

From my experience (with high-rpm, high-overlap engines), a flow bench test will show the momentum effects of pressure loss and this will manifest itself to a small degree in cylinder-to-cylinder airflow imbalance on the actual engine. However, most of the imbalance will come from the wave effects. I've never seen an airflow test (with runners individually flowed) that shows the airflow ranking change as a function of pressure ratio. By airflow ranking, I mean which cylinder is getting the highest airflow (or volumetric efficiency, etc). With an intake on an actual high-rpm engine, the cylinder airflow ranking changes drastically over the rpm range. So for example, at one rpm, #2 might be getting the most air, and at another rpm #4 might be getting the most air. If I looked at one rpm point only, I might get the airflow data to match engine data as far as ranking goes. However, looking across the rpm range, I have never seen an airflow bench match the engine data in regards to airflow/VE imbalance/ranking. This is due to wave effects. So to reiterate, if the cyl-to-cyl relative imbalance (how far each cyl is off from another; standard deviation of VE if you like) and the cylinder airflow/VE ranking is changing across the rpm band, how is an airflow test ever going to reflect this?


You stated that your experience hasn't been with restricted engines as much. Perhaps lower rev engines are more predominately effected by pressure losses in the intake. Also, I have been told by a Chrysler engineer who has worked on restricted engines, that in general a restricted engine will have higher cylinder-to-cylinder airflow/VE imbalances. Perhaps this would explain some of the difference between our experiences.

As far as 1D simulations go, I have found that with a decent model they get the ranking/order and general trend of cylinder-to-cylinder airflow imbalance fairly well. They can't do separation and 3D flow, because, well, they are 1D. On the flow side of things I don't see 1D elements getting any better as they are already explicity solving the 1D N-S equations. However, using 1D elements in a quasi-3D setup has been shown to work very well in predicting cylinder-to-cylinder imbalance, where a simpler 1D model didn't get it right. Going to a 1D-3D coupled model will yield much better results, but at a very large computation penalty, as you probably already
know. I think a coupled model will allow you to see the airflow in a a way that an airflow bench never could.

Agreed though that everything has it's uses and limitations. And agreed that you could run a couple flow bench tests before you get a coupled simulation done.

So, what I am saying, is that at least with high-rpm engines, everthing I've seen/done shows that wave effects dominate cylinder airflow imbalance. I would have to generally agree with Ben, as what he is stating agrees with
everything I have seen/done. I would also agree that at higher rpms, the intake flow tends to be somewhat steady-state (especially before the throat), but anything a bit after the throat, even at high revs; say 14,000rpm), could not really be called steady state.

[John Bucknell]

These are the kinds of discussions I always hope to have when I read a technically-oriented forum. Sharing info and experience is why I'm here.

First, let me reiterate that I am (and continue to try to be) as objective as possible in my role as a judge. Thus my statements regarding bringing data - I have no problems with the fact that my experience is finite. But I do have sufficient theoretical background that if something doesn't sound right, I'll try to sound out how what I am hearing fits in my understanding of the way things work. So for the benefit of everyone here, my direct background in industry is PFI automotive 4-strokes from 4 to 8 cylinders (boosted especially) with a single throttle and speeds never exceeding 8500 rpm.

Ultimately, the readers of this thread will form their own opinions about how to develop an intake manifold - so I want to address Mark's points and illustrate what I understand to be true about them and where my experience disagrees and hope we can identify where manifolds are sensitive to design to get you the reader pointed down the best-possible direction (for you).

First and foremost, distribution is THE hardest thing about designing a single-inlet manifold if you are trying to maximize performance. On a production engine, if we can maintain a best-to-worst distribution on the order of 5% at WOT across the speed range we call it good. Why? Well for one AFR is difficult to measure on an individual cylinder basis even with the best tools (EGT is notoriously bad indicator), as wide-range O2 need local pressure and temperature correction (which leads to many sensors crowding the exhaust port impacting the engine performance) and carbon-balance methods via emissions trains have similar problems (which can be several percent unless accounted for). On top of that, production injectors can have +/- 3% variation (assuming your fuel rail is perfectly distributing). Worst of all with short runners (~250mm and less valve-to-plenum, again in my experience) the pressure waves force the wall film of fuel up the runner into the plenum where it will redistribute to other cylinders at low to medium engine speeds. In order to isolate the above affects, we will run gaseous fuels injected upstream of the throttle just to understand what is going on. Almost always the AFR maldistribution increases with airflow, if it doesn't we can usually isolate a measurement error. It is this last test that leads me to believe my prior statement regarding momentum effects - which I guess needs to be qualified in that I am merely talking about airflow distribution. So my point is that it is hard to understand distribution completely on the dyno, especially for high speed engines which necessarily have short intake runner lengths, high overlap, and exhaust manifold tuning.

The 1-D modelling world has a variety of quasi-3D geometry tools, again making for very promising trends in future modelling (assuming a decent library of boundary conditions, tough for students). Mark and I both agree on this, full computational solutions are coming.

Controlling the fuel distribution and accurately measuring that airflow on engine lead to most of my correlation issues with my 1-D model. So I will run my 1-D model as a virtual flow bench and tweak it (with great difficulty I might add) until it matches my flow bench (or CFD) data before continuing with dynamic analysis. This has resulted in best results for me, even though I find 1-D analysis is best prior to geometry generation (ie drawing parts) to provide direction on what to do with the geometry as opposed to analyzing what the geometry is doing. That is to say, it is one of the bigger knobs - but only gets you within 6-10% of the final answer, and is intensely time consuming to build and then collect enough data to correlate.

Okay, next to last comment (for those who haven't had their eyes glaze over yet) - the explanation of how acoustics can impact distribution to a greater degree than momentum. As far as I can tell, it can't (with regards to manifold design at least). That is to say, there isn't anything you would do different with your manifold design to improve your distribution than what your steady-state tools tell you. And that I think is the distinction where we're getting our shorts in a knot. That has been my experience, and again we're talking about a few percentage points.

Ultimately distribution is a fuelling issue, so a statement about the importance of all the discussion so far. The penalty for running between ~15% lean of LBT to ~25% rich of LBT is only about 4% worst case for torque (assuming your spark is MBT, which does move around a little bit with AFR). So if your injectors are pretty close, and your malidistribution from best to worst is 10% on high octane fuel (ie the leanest cylinder isn't knocking) you are penalizing yourself a small amount in the scheme of things - unless you really want that 4% on the cylinder that's misbehaving http://fsae.com/groupee_common/emoticons/icon_smile.gif.

If on the other hand, you are interested in emissions - you had better get all the cylinders running right on the money if you want your catalysts to work. So I spend a lot of time on it, relatively speaking.

To summarize, don't use flow bench data independent of anything else. Use your head. And don't regurgitate any of what I've said in the design tent as a suck up method http://fsae.com/groupee_common/emoticons/icon_smile.gif

[Ben Inkster]

<pre class="ip-ubbcode-code-pre"> Ben,

are you still using a side entry plenum? I haven't actaully seen any pictures of your car but I have a copy of a UWA paper on the intake system design from a couple years ago and it seemed to hint at issues like that with the side entry plenum.</pre>

BeaverGuy

yes i am still using a side entry plenum, but the main reason for this is because I am using a variable geometry intake manifold. I did notice big differences between a top and side entry plenum in flow near the restrictor but this was using some crude simulations I have done, so I would rather not comment.


<pre class="ip-ubbcode-code-pre">Ben - what kind of cyclic simulation where you doing? Coupled simulation? </pre>

Mark100,

I am using Ricardo Wave coupled with Ricardo Vectis. Seems to work really well, and I have validated the results using pressure transducers on the intake ports of an F4i on the dyno. The coupled techniques increases accuracy from 1D sims but not as much as you might think. I found the best outcome from coupled simulations were the flow animations.

Cheers

-Ben Inkster

intake guy
University of Western Australia

[John Bucknell]

Sorry for killing this thread dead. I thought it was interesting... anybody else have any other questions about intake manifolds I can sound overbearing about?

[BeaverGuy]

I did some CFD analysis of our current manifold and noticed an interesting phenomenon. Our taller plenum had a reasonably even distribution. Then I did an analysis of a shorter plenum. I expected the number 2 and 3 runners to have a larger share of the flow than the 1 and 4 runners. However, the outer runners were getting considerably more air than the inner runners.

I didn't really understand it at first then I looked at at the firing order and the flow with respect to crank position. a 1-2-4-3 firing order and the outer runners were at their maximum flow just as the adjacent cylinders were starting to open. The outer runners were pulling the air out of the adjacent inner runners. I figure that this wasn't a problem with the taller plenum because there was enough volume that the flow trajectory to the outer runners didn't have to pass right by the inner runners.

I think a 1-3-4-2 as opposed to the 1-2-4-3 firing order would alleviate this problem. But obviously that isn't really feasible. And was wondering if this is possibly the reason for the 1-3-4-2 order on some automotive applications and how isolating the runenrs to deal with this type of problem is handled.


I appologize for the large picture but my image editing program didn't want to open it. This is the larger of the two plenums, approximate height 4.1" the shorter is 2" both with an elliptical cross section.
http://oregonstate.edu/~gilletjo/images/intake.jpg

[John Bucknell]

Good observation. I'm surprised there are any 4-cylinders still with an outer-to-inner firing order. The short answer is the 2" high plenum is too small and unsteady effects are biting you. In rough terms, you need some volume ready to fill the cylinder every cycle. If that volume extends up into the plenum, the runner entrance acts like a vacuum cleaner and pulls indiscriminately at any available air. With a small plenum, usually the available air is laterally in one direction for the end cylinders. One way to isolate this is to design your plenum such that each cylinder has relatively the same amount of air in every direction, ie stick some 'ears' outboard of the outer cylinders. The other way is to do just what you did, move the opposite wall further away. Some development is usually necessary to figure out what the smallest throttled volume is (best response) without getting into distribution issues.

Mark -

Answering this, I realized maybe there is another bit where we aren't talking about the same thing. Unsteady effects like the above example I don't bin with acoustics - I segregate these by time scale. Unsteady flow impacts are a stroke or more in duration, acoustics in the 1/4 to 1/8th stroke (or less) range. I can see in several circumstances unsteady flow impacting distribution heavily - usually outside of my design space, but I can see where the effects you have experienced might result.

Damn, that sounds like back-pedalling....

[DPH]

Ben,

Could you tell us more about the experimental methods you used to gather the pressure data?
Did you provide a report to your team that you would be willling to share? It is excellent to see individuals working all the tools at their disposal. Experimental and numerical methods work well toegether when they are not done in isolation.

<BLOCKQUOTE class="ip-ubbcode-quote"><font size="-1">quote:</font><HR> I am using Ricardo Wave coupled with Ricardo Vectis. Seems to work really well, and I have validated the results using pressure transducers on the intake ports of an F4i on the dyno. The coupled techniques increases accuracy from 1D sims but not as much as you might think. I found the best outcome from coupled simulations were the flow animations. <HR></BLOCKQUOTE>

John,

You mentioned: <BLOCKQUOTE class="ip-ubbcode-quote"><font size="-1">quote:</font><HR> Worst of all with short runners (~250mm and less valve-to-plenum, again in my experience) the pressure waves force the wall film of fuel up the runner into the plenum where it will redistribute to other cylinders at low to medium engine speeds. <HR></BLOCKQUOTE>

Could you describe the instrumentation you used to determine the wall film break up was caused by pressure wave forces and not momentum effects? Wall film experiments and models can be difficult to get right.

[John Bucknell]

<BLOCKQUOTE class="ip-ubbcode-quote"><font size="-1">quote:</font><HR>Originally posted by DPH:

Could you describe the instrumentation you used to determine the wall film break up was caused by pressure wave forces and not momentum effects? Wall film experiments and models can be difficult to get right. <HR></BLOCKQUOTE>

This engine where we discovered this baffled us for a while (almost too long as it wasn't passing emissions tests). The gaseous-fuelled version did not show the distribution issue (confusing, as this momentum method had worked everywhere else). It was a log manifold (injectors at the cylinder head flange as per typical port fuel injection practice) and we kept noticing the indolene dye (red in our case) opposite the runner openings on the inside of the plenum (cast aluminum manifold) when we had the engine apart, intriguing in that the stains were almost exactly the shape of the runner cross-section. Finally we instrumented the exhaust runners with emissions sample lines and adjusted fuelling on an individual cylinder basis until they were all even at the speeds we were having issues with. With this test we were able to demonstrate that fuelling was radically different, most fuel needed near the open end of the log and least fuel at the closed end. Since it was aluminum we hacked off the plenum and modified it with a larger volume and the problem got better at medium speeds but still was terrible at idle speeds and lower loads. We tried our other tricks (the extra volume at the ends) and nothing happened. So the bright mechanic that worked for me suggested since it didn't happen with other manifolds on the same engine (no fuel stains on that plenum) - we might try a runner length somewhere between the short one and the long one. This was tough because the vehicle didn't have the room for a longer runner (it was supercharged so didn't need the runner length for tuning), but we gave it a shot and surprised us with working the first try. Subsequently we iterated up and down in runner length to find the sensitivity range and came up with the 250 mm number (which subsequently got implemented at great expense by changing the sheet metal). Since then, that rule of thumb has been used on several other programs successfully. So the experiment yielded results in short order that could be duplicated (two days of working on it), and weeks of arguing back and forth among the engineering community as to the mechanism (because nobody wanted to spend the money on the chassis change). Not totally scientific because we couldn't devise a test where we could measure the issue directly until we stuck a video scope in the manifold and recorded the little rivers of fuel (flourescent dye).

[BeaverGuy]

John,

When you are measuring the runner distribution on the flow bench, are you flowing each runner individually or all at the same time? I didn't find it all that clear.

[Nick McNaughton]

<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by Mark100:
However, most of the imbalance will come from the wave effects. I've never seen an airflow test (with runners individually flowed) that shows the airflow ranking change as a function of pressure ratio. By airflow ranking, I mean which cylinder is getting the highest airflow (or volumetric efficiency, etc). With an intake on an actual high-rpm engine, the cylinder airflow ranking changes drastically over the rpm range. So for example, at one rpm, #2 might be getting the most air, and at another rpm #4 might be getting the most air. If I looked at one rpm point only, I might get the airflow data to match engine data as far as ranking goes. However, looking across the rpm range, I have never seen an airflow bench match the engine data in regards to airflow/VE imbalance/ranking. This is due to wave effects. </div></BLOCKQUOTE>

Mark,
Admittedly, I've only had third-hand experience with intake design, but I don't quite understand the mechanism you're describing here. We've seen effects similar to what you describe, however we've got the added complication of runner inlets that move around with engine speed. Are the wave effects you describe as causing the airflow ranking to change across the RPM range because of the runner geometry alone, are they linked to the plenum geometry surrounding the runner inlets, or is it something else entirely?

Any more detail on how to understand and deal with these issues would be appreciated.

Cheers,
Nick

[John Bucknell]

<BLOCKQUOTE class="ip-ubbcode-quote"><div class="ip-ubbcode-quote-title">quote:</div><div class="ip-ubbcode-quote-content">Originally posted by BeaverGuy:
John,

When you are measuring the runner distribution on the flow bench, are you flowing each runner individually or all at the same time? I didn't find it all that clear. </div></BLOCKQUOTE>

Typical setup on a bench is a complete cylinder head and manifold, with the ports not being tested masked off at the gasket interface (head to manifold). The head is moved such that each cylinder is tested individually, masking and unmasking the ports so that only one runner is flowing at a time. Sorry if this wasn't clearer earlier...

[Andycostin]

John,
With regards to the flow bench testing, do you mount the cylinder head directly onto the flowbench, or use a tube to simulate the cylinder that you are flowing into? If so, I expect that the length of this cylinder would have an effect on the flow?

[John Bucknell]

Andrew,

We use something called a bore adapter, which is typically exactly the bore of the cylinder on the ID and a reference OD on the outside (that mates to the bench). The height of said bore adapter is about 250mm, as that is usually at least a production engine stroke and also sufficient to smooth out any flow irregularities induced by the bench (which are typically v. small). We also make a plate that slips over an outside shoulder of said bore adapter that provides a place to locate with head bolts (or similar fastener) - often the adapter has as many holes as the entire head has cylinder bores and can cause the head to be easily lifted up and transferred from bore to bore without requiring time-consuming realignment.

[Nishant Jain]

I'm going to touch this old topic once again. Hopefully get some inputs which I'm able to make sense of.

So in what applications do FSAE teams use flowbenches? We don't have a flowbench system in college yet. But we're planning on setting one up, so we can run some restrictor designs on it. I also want to check flow through cylinder head.

[@John: It's only through the one line about 25" of water in your post that I finally understood why teams go for that particular number. Thanks!]

Now, elsewhere on the net, engine-heads use flowbenches for "porting" - changing the design of the intake port for better flow.

Absorbing what I've been reading on the forum these past few days on the whole subject, my question is: Understanding that the steady state results of a flow bench can only appear to be like an instantaneous statistic from a transient analysis which would be the order of the day, in what ways can a flow bench analysis really benefit intake design?

With all the rapid cut and thrust between impact on momentum flows or not, I'm beginning to wonder if a flow bench is even such a good idea.