Webber inlet flow study

Adam C.

GT40s Sponsor
Since I am basically snowed in, and recently got acces to a CFD package, I thought I would do a little investigation into the inlet side of everyones favorite form of fuell addition, the webber carb.

CFD stands for Computational Fluid Dynamics, or as one of my proffessors likes to call it, Colorful Fluid Dynamics, because of the pretty pictures it generates. Basically CFD is a way to simulate fluid flow. This can be internal flow, like in pipes, junctions, etc., or can be external flow, like around a car body (hmmm, possible future topic?).

What we are going to investigate is the effect of runner inlet geometry on flow. I have chosen this because it is a very simple case that doesn't require much computer power.

For our baseline case, we are going to look at a simple pipe pulling air from ambient. This could be an inlet pipe sticking way out of a wall into an empty room, or a webber carb with a really crappy velocity stack on it.

In our model, we are flowing air at standard temperature and pressure. We are applying a static pressure difference if 28 inH2O across our test piece. The test piece is 2" wide and 40" long. I know from experience that it takes a length of about 20 tube diameters for flow to become fully developed in a straight duct. This is important because we want to measure the total losses realized due to the inlet geometry. If we were to use a tube only 10" long, we would not be accurately measuring the losses due to the inlet. Of course no engine in the world has runners that are perfectly straight and 20 diameters long. We just want to accurately quantify the effects of the inlet. Anyway, on we go.

Below is the mesh used to simulate the pipe sticking into ambient. All the green lines show the cells that the area is broken up into. Generaly speaking, the more cells, the more accurate the result. If you squint, you can see the pipe walls in white. Don't wory, it will be more clear in the next picture.
 

Attachments

  • 29735-model.jpg
    29735-model.jpg
    95.3 KB · Views: 742

Adam C.

GT40s Sponsor
So now we are going to look at the velocity and direction of air flow at the inlet of this pipe. We have zoomed in to look only at the first few inches of the pipe to better see what is going on.

The arrows show the direction on flow. The color indicates the velocity. In this case, blue means zero velocity, while red is 110 m/s, or roughly Mach M=0.3.

The vlocity is practically zero only a few inches from the pipe inlet. As it gets closer to the inlet, it is rapidly accelerated into the pipe. Notice that once it is in the pipe the velocity is not uniform across the diameter. The velocity is higher in the center than at the sides. This is the primary source of the flow losses at the inlet. The little patch of red in the center indicates that there is effectively an area reduction as all the air is trying to squeze into the inlet of the pipe. After the air gets in, it then wants to redistribute itself so that the velocity is relatively uniform again. This is primarily accomplished through turbulent mixing, which is not good.

Also notice that some of the air is actually making an almost 180 degree trun to get into the pipe (the air that is below the inlet and outside of the tube). This is important to notice for later comparison.
 

Attachments

  • 29736-rad0.jpg
    29736-rad0.jpg
    91.2 KB · Views: 811

Adam C.

GT40s Sponsor
So how are we going to get this thing to flow better? Let's try putting a raduis on the inlet.

Now our inlet looks like some of the better velocity stacks that I have seen.

Although I have simulated several sizes I will only display the largest radius investigated. Here the radius at the inlet is equal to 50% of the pipe diameter. So if the pipe is 2", the radius is 1".

So what do we notice? Well more of the air outside of the inlet is moving toward the pipe. More importantly, there is no red spot in the center of the inlet. Instead the flow velocity is very uniform across the diameter of the pipe. Some may point out that although it is uniform, the velocity is pretty high. Good! That is what we want, the higher the average velocity, the more airflow into the engine.

Something is still bothering me however. Some of the air outside and below the inlet is making a 180 degree turn to get into the tube. Think about it. If you are driving along at 60MPH and make a turn, which is going to cost you more energy, making a 90 degree turn, or 180?

Lets see what we can do to limit the angle the air turns to get into the tube.
 

Attachments

  • 29737-rad5.jpg
    29737-rad5.jpg
    93.2 KB · Views: 769

Adam C.

GT40s Sponsor
To limit the angle in which the air is alowed to turn, we now mount our pipe inlet flush with the wall. It is no longer sticking out into the room. Now the air can only make a 90 degree or smaller turn into the inlet. In a way we are guiding the air in. Of course after I have compressed the bejesus out of the picture to fit under the size limit you can no lnger see the wall, but trust me, it is there.

Well that is a pretty picture. Look at all that air trying to get in. Compare this to the our baseline nasty sharp pipe in the first post.

Well it looks nice, but does that really mean that we are getting any more air in? You can't quantify air flow with these silly color pictures. We need to look at the mass flux numbers to tell for sure.
 

Attachments

  • 29739-rad05.jpg
    29739-rad05.jpg
    92.9 KB · Views: 1,538

Adam C.

GT40s Sponsor
So now we look at a plot of the mass flow rates for the inlets tested.

On the x-axis is the size of the inlet radius (r) divided by the tube diameter (D). For our baseline case it was zero, while the maximum of all tested was 1.

On the y-axis is the nondimensionalized mass flow rate. This is the mass flow rate of each configuration divided by the flow rate of our baseline case. So a nondimensionalized flow rate of 1.1 means it is 10% better than our baseline sharp inlet pipe.

There are two curves, one for the pipe that protrudes out into the room, and the other that is flush with the wall.

The flow rate is increased with increasing inlet radius. There are diminishing returns however. The maximum increase in flow tested was about 18%. That is a huge increase for such a simple modification.

The flush configuration beats the protruding configuration for all radii tested. Why does everybody have thier stacks just sticking up into the air? Even if you just used a piece of sheetmetal with holes cut in it and laid it over the stacks, you would get more airflow!

So does this mean that if you use flush velocity stacks with an inlet radius equal to the diameter of the primary runner that you are going to see an 18% increase in flow through your engine? Unfortunately not. The inlet to the primary is but one resistor which is in series with many other resistors. The largest of which is the intake valve. Still, every little bit helps!

Because our simulation is generic, the results can be applied to any case where flow is accelerated into a duct, be it the primaries, the throttle body inlet on a fuel injected engine, brake duct inlets, you name it.
 

Attachments

  • 29740-plot.jpg
    29740-plot.jpg
    94.2 KB · Views: 726
Very nice work Adam! Now I would be interested in a case similar to case 2 but where the radius continues for ~180 degrees instead of just 90 (Like the Air Horns on the TWM Throttle bodies). Is that still worse then the flush at 90 degree case?
 

Adam C.

GT40s Sponsor
Gary,

I believe that it would be almost identical to case2. Give me a day or two and I'll run it.
 
Nice stuff, as usual Adam. I think the answer to your question about why everybody runs individual stacks is that they look better! /ubbthreads/images/graemlins/smile.gif

Your suggestion of using a plate across the intake horns is a good one and one I would seriously consider testing if I wanted to eke out those last few horsepower. At this point I really have no idea how restrictive my 8-stack EFI setup is relative to other engine components.

The other comment I have about your analysis is that I don't think the maximum 18% increase in flow will scale up in a real application because I expect that there would be three stagnation points along each bank located midway between the intake bells. There would also be four stagnation points across the cylinder banks but since these intake bells are further apart I think they'd have less effect on overall flow.

It looks like your analysis was set up using a 2-D FE mesh. I suspect it's much more difficult to model this in 3-D to measure the effects of individual intake bells competing for airflow?

Again, very nice work. I'd love to see some Mk-1 aero analysis at some point in the future.
 

Adam C.

GT40s Sponsor
[ QUOTE ]
I expect that there would be three stagnation points along each bank located midway between the intake bells.

[/ QUOTE ]

Not quite, but this does bring up a good point. I have also seen bellmouths that are so large that they intersect eachother. This is also a no no. The resulting sharp edge at their intersection is not good.

[ QUOTE ]
I suspect it's much more difficult to model this in 3-D to measure the effects of individual intake bells competing for airflow?


[/ QUOTE ]

Doing one stack in 3-D is no problem, doing multiple is somewhat of a pain. Fortunately, the cylinders pretty much breath one at a time, and so don't compete for air. While on a V-8 there is some time when two cylinders have intake valves open at the same time, most of them (except 7 and 8 on 5.0L HO) are not next to eachother due to the firing order. We showed in the simulation that once you are a few inches from the inlet, the air is almost stagnant, and so they won't be competeing much. Now once you are talking about a single plane or EFI manifold, in which the air upstream of the primaries has some velocity, there is some interaction. Because the flow is unsteady, the interaction can be benificial or destructive, it just depends on the timing of the events.

[ QUOTE ]
I'd love to see some Mk-1 aero analysis at some point in the future.

[/ QUOTE ]

As soon as I get access to a CRAY or Baewolf cluster! We're talking way more power than the W.O.P.E.R. in wargames. /ubbthreads/images/graemlins/grin.gif Who wants to chip in on the 3+GB of ram it would require?

I may look at doing some type of simplified model in the future. I know it is a popular topic. Getting the ground effect due to the moving ground and tires is pretty difficult.
 

Jim Rosenthal

Supporter
Adam, I really enjoy reading this stuff and thank you for posting it. I do have a question: with an engine turning at several thousand revs/min, is it accurate to talk about pulsatile intake flow or does the rapidity of intake strokes combine to form continuous suction of air into the induction system? I would think it does. Also, how do you decide what is adequate flow size for air filters? My Webers were shipped from Pierce with K&N filters but I have no idea whether they are big enough.
 

Adam C.

GT40s Sponsor
Jim,

You question realy has two parts. The first is, "is it valid to assume steady flow when looking at flow losses in engines?" This has been studied extensively and the answer is yes. In fact, all good engine simulation codes use loss coefficients (think of them as resistance values) measured during steady flow tests on flow-benches. Even though things happen very quickly in engine flows, the length scale of the loss components is short enough that at any given instant of engine operation we can consider the flow to be "steady".

The second part of the question is "is the flow steady in the induction system" This depends on two things. Firstly if there is a point in the induction system in which all of the cylinder runners come together, like in a single plane manifold for instance, then there is a possibility that the flow could be considered steady in all of the ducting upstream of the common point, like in the carb and air cleaner. The second stipulation is that more than 4 cylinders need to come together in order to dampen eachothers oscilations out. This is because for more than 4 cylinders in a 4-stroke engine there is always more than one cylinder with the intake valve open. The more cylinders there are, the more valid the assumption of steady flow upstream of the plenum is. This is one of the reasons that a V12 sounds so smooth, and a flat 4 subaru just sounds butt nasty.

So what does this mean? It means that in todays fuel injected engines in which all of the primary runners come together in the plenum, the secondary runner, or the runner from the plenum to the air filter, actually has oscilating flow in it in engines with less than 5 cylinders, and therefore contributes to the resonance tuning of the induction system. The secondary manifolds on engines with more than 4 cylinders see nearly steady flow, and so do not contribute to tuning. The honda guys need to think about this when they are putting those silly "cold air" packages on their cars. Any change in the diameter or length of the ducting from the plenum to the air cleaner will change the tuning of the engine.

So if we have a V-8 why do we care? Because if you divide the intake manifold of a V-8 into two separate manifolds with 4 cylinders each you suddenly have changed the resonance of the system! If you connect both halves of the manifold with a valve, you can cange the resonance of the manifold with RPM. The effect is something like having a dual plane and a singel plane manifold at the flip of a switch.

This is used extesively in production engines. Most 5.4L F150's have it, as well as Jag V-6. Probably the best use of it is in my own car (96 probe GT) where Mazda did an awsome job of flattening the torque curve of the engine from 2K up to 7K.

A long winded answer to a short question, but I thinke this was the coolest development in engines during the 90's.

As for the air cleaners, I have not seen any rule of thumb for how to size air cleaners. You basically need to flow test them to see how restrictive they are compared to the rest of the system.

If you want you can send me a Webber, velocity stack, and air cleaner and I can flow test them for another thread. /ubbthreads/images/graemlins/grin.gif
 

Adam C.

GT40s Sponsor
Jim, I'm in Columbus Ohio, send it on down.

Gary,

I finally got around to your 180 deg. design. As I predicted, it flowed pretty close to case 2. Actually, it was slightly worse by 3%.

What was supprising to me was the velocity distribution. It took some time for the model to converge. As it was running I wondered if this could indicate that the flow pattern was somewhat unstable for some reason. Sure was. Look at how much air is coming around the right lip, while almost none is coming around the left. The result is that the velocity is not as uniform in the pipe. Pretty interesting. Now some may wonder "if the model is symetric, how could it of came to an asymetric velocity field". Well the pipe is symetric, but the mesh is not. I bet that I could remesh it and get it to preferentially pull from the left side as well.

Anyway, in the end we have illustrated that the greater angle you force or allow air to turn, the greater the flow losses. Any of you guys remember your high school physics? Work=Force X Distance A force is being applied to turn that flow around 180 deg to get it in the bellmouth. This work is not for free, it results in a greater flow loss.
 

Attachments

  • 29902-gary.jpg
    29902-gary.jpg
    90.2 KB · Views: 2,024
Back
Top