|
|
|
CFD modelling of good and not so good ducts
Well, quicker than I had expected, here's some pics showing the differences between simple and correct ducting. The differences are subtle, but the effect on fan performance should be noticeable. BTW, don't ask why I was still at work at midnight so close to holidays!
For those technically minded, here's a description of the cfd model. BTW, we use CFX 5.6 at work, I wish I was rich enough to have it at home (and be able to afford the computer to run it!). To model the flow, I constructed a duct 150mm long, 77mm dia inlet with a 3mm radius lip. The fan section is 90mm dia with a 38mm dia spinner of close to elliptical section. Two options were modelled: one had a straight transition from 77mm to the 90mm fan, the other went from 77mm dia to 81.5mm dia inline with the tip of the spinner and was then "area ruled" around the spinner up to the 90mm fan. Rather than model a full duct, I modelled it as 1/8 of a full revolution, making the "sliced" faces periodic pairs. This is typical cfd stuff and cuts computation time down considerably without sacrificing accuray. The inlet was given atmospheric pressure. The fan was modelled at the outlet section as a momentum source of 0.25kg/s . The flow was steady state. The first pic shows the streamlines for the simple duct. The colour of the streamline represents it's velocity - from blue to green to yellow to red in order of increasing velocity. Notice at the lip the red region. This is most likely flow separation - the area inside the separated flow will be chaotic and high velocity. Second point of note is the thick white line drawn across the streamlines ahead of the spinner. Look at the variation of streamline colour across this line. It shows how the flow slows right down ahead of the fan. Hard to imagine, but this is most likely what's causing the inlet lip to separate. Next look at the pink line. This is where the fan blades would be. Again notice the large variation in colour. The flow at the hub is much slower than the tip. This causes a static pressure gradient across the face of the fan which is negative towards the tip of the fan ie. pressure is falling towards the tip. Given that most fans are designed for free vortex flow ie. equal pressure across the face of the fan (not sure about DS's fans, but Wemo's look like they are) there is no opposing pressure gradient designed into the blading so there will be more radial flow. Granted there is no fan present, and the upwash from it's blades will have an effect, but not as pronounced as not presenting it with an even flow. |
|
Last edited by stumax; Dec 15, 2003 at 06:37 PM.
|
|
|
|
|
The next pic is a streamline plot for the more correct duct. First point (and the pic doesn't show this as well as the full 3D representation I get) is that the area near the lip may or may not have separated - it's borderline, and I would normally fine down the mesh around there and run it again, but it's good enough. Next notice the thick white line. See how the variation in velocity is not as marked as the previous duct. The pink line shows only a slight variation in velocity across the duct. Much better thatn the previous one as the pressure across where the fan would be is more uniform. This means that the fan blades will have the angle of attack that they were designed for.
|
|
Last edited by stumax; Dec 15, 2003 at 09:21 AM.
|
|
|
|
The next pic is a velocity plot for the simple duct. Once again notice the flow at the lip, the variation in velocity across the thick white and pink lines and the overall non-uniformity of the flow velocity after the spinner. What is interesting is that you can see the thickening of the boundary layer along the duct wall, shown by a thickening dark blue area.
|
|
|
|
|
|
Finally, here's a velocity plot for the more correct duct. Again, look at the inlet lip - this time the difference is more noticeable (not sure why the previous streamline plot had so much red there - must be a result of shrinking the image?), and the flow seems to accelerate but not separate. Next look at the thick white and pink lines. Notice how much less the area in front of the spinner is decelerated, and how much more uniform the flow past the spinner is. To me, this means that the fan will receive better, more uniform air, and this should be independent of airspeed (provided we're substantially subsonic!).
I hope this helps a few people solve their ducting riddles! Stu Maxwell StumaxAircraft stumax@hotmail.com |
|
|
|
|
|
Very interesting Stu, you're doing a great service to the EDF community by running these simulations. This gives us a visual idea of what's going on. Imagine how much blue and red there would be inside the ducting of a Kyosho T-33...all the colors of the rainbow!
Another simulation I'd like to see is bifurcated ducting like that found on the F-18/F-22/F-15 etc, both with and without streamlined cones that fit up against the hub of the fan (the spinner is left off since the cone, attached to the inner ducting wall, takes it's place). Think of the Harrier inlets if you've ever looked into one. The inner wall cone could serve the same function as your near constant cross section modeled here does. A whole half duct might need to be modeled though so it would take more time. Another interesting sim would be the area behind the motor, with and without a streamlined cone. I imagine the inlet ducting is more important though, since motor cooling can be reduced with the cones, and having a complex shaped exhaust duct means more to construct and fit in the tight confines of a fuselage. Very interesting, thanks a bunch Professor Stu! |
|
|
|
|
|
Yeah amazing stuff Stu.
Barry |
|
|
|
|
|
STU.
Most excellent work! I guess that a vac. formed transition tube is in order for the aft end of the duct as well?? From your similation could you tell if a smooth transition was significantly better that a cone shaped transition at the start of the spinner. IF it is not a significant difference then perhaps a straight 77mm to the start of the spinner then a straight walled transition tube could be fabricated easily and attached to the straight tube and we can be done with it. If required ,should the transition simulate the curve of the spinner. If so ,we will need at least two different transition shapes. One for the Midi and one for the DS fan. This is very similar to what the BVM ICDF unit incorperates( Area ruling) to allow for constant pressure throughout the ducting. My intake should perform satisfactorily as designed with a 3mm radius to a 77mm inner diameter. Would a decrease in diameter WITHIN the ducting to less than 77mm be of any value? It would act like an augmentor tube. Thoughts??... comments?.. I am not a fluid dynamics engineer but am glad you are!! Bruce |
|
|
|
|
|
Very impressive chunk of data.
What effects would have on thrust in our real EDF world; 10%, 30%??? PM sent, also. |
|
|
|
|
|
here's cross section of my intended Huck intake part...the top lip is more accurate than the bottom as far as smooth radius is concerned. Should perform OK...
|
|
|
|
|
|
Great work, I too am interested in seeing what is more efficient in a bifuricated inlet. Take for example, the inlet on the Pyrahna. Very short lead duct and a cylinder leading from the splitter plate to the front of the fan. If anything, it should be accelerating through this area.
|
|
|
|
||
|
Quote:
|
|
|
||
|
|
|
Thanks guys! Ed, the bifurcated ducting without the spinner will show a similar thing, only worse as the flows from left and right sides are mixing. The area behind the motor is also important, you'll notice Daniel has a cone on his fan - he's no dummy. Once again, it works without, but it should work better with. The difference, as you point out is motor cooling, and it may be that the motor gets hotter, losing eficiency, and that's why in a minifan we don't see much difference. However, the flow inside the cone must also obey the rules. There's no point having space around the motor for air to flow and then suddenly expanding it inside a cone behind the motor. The air must be guided, the area expanded to allow for thermal expansion, the ideally accelerated to freestream velocity where it exits the cone.
Bruce, a vac formed transition would be the go - it only needs to be short and can have a flaired lip to go around the fan and around the rolled tube at the other end. You wouldn't want to close it down below 77mm because you then have to open it back up before the fan. Any change in velocity inside the duct is bad, we want ot minimise this for least losses. Larry, thanks for picking that up (in the PM). Everyone go back to post 1, I edited it, made a slipup about the sign of the pressure gradient across the fan - it was 2am and I was getting square eyes . As for how much it would affect things - I have absolutely no idea! It should affect it, maybe only a few %, however, the fan may compensate. Our fans are never running at their design point anyway, so everything is a compromise. James, you're correct, accelerating flow is good for an inlet as it is less prone to separation. With bifurcated ducts the most important thing to me is that they meet with a zero angle transition so that the flows are aligned separately before they meet. Barry, Steve, no worries! One thing, though guys. this stuff isn't gospel - don't live by it. CFD is a very useful tool only when backed up by real world testing. If real world testing shows that either of these ducts would work the same, then the CFD model needs refining. |
|
Last edited by stumax; Dec 15, 2003 at 05:53 PM.
|
|
||
|
Re: CFD modelling of good and not so good ductsQuote:
Perhaps you could quantify the difference in velocities we're talking about here. What velocity does blue represent and what velocity does red represent? TIA, Dan |
|
|
||
|
|
|
Thanks, Dan. I don't have access to the ball-buster computer today (selfish buggers are running F1 engine flow modelling - can't they get their priorities in order?) but I think the spectral green was around 50m/s, dark blue down to about 10m/s and red was a whopping 90m/s, don't hold me to that, though! I'll verify that when I can, and maybe post a pic with a scale on it. Both cases had the same ranges in the plots to allow for direct comparison.
Stu Maxwell StumaxAircraft stumax@hotmail;.com |
|
|