Complex Air Duct Sweep Help

[keaner]

Hi Modelers,

Using Creo 4.

I am designing a complex air duct. The duct has two different inlet/out profiles, sweeps downward and moves right to left. There is an mid wall inside the duct which at the inlet end sits about a third to the left hand side and then finishes in the middle at the outlet.

My design approach has been to complete a wire frame of reference sketches, datum points and splines to achieve the shape. I then use boundary blends to make up the different duct wall sections. I then merge the walls, thicken and remove the excess.

Wire frame

Boundary blends

Thickened - Outlet End

Rear Iso View - Inlet End

The challenge I face is to maintain a constant (or close to) smooth inner duct chamber profile on both sides. As air flows through each duct chamber the goal is for a stable airflow speed (little increase) and minimal pressure drop at the outlet. This is achieved by transitioning between the start/end profiles in a smooth fashion without collapsing the profile and keeping the cross sectional area as constant as possible.

The green lines shown the internal chamber profiles. The main problem is here as they are not smooth.

Please note I am aware of the mid duct wall not looking right. This is an easy fix (tuning of datum points). Outside profile solution is the main issue.

The below views show the areas of the duct geometry which aren't controlled and have blown out.

Side Elevation

Front Elevation

Plan View 

My questions:
1. This this the right modeling approach to tackle this part?
2. Or do I just need more reference sketches to control the sweep? 

Appreciate anyone who can help.

File attached.

Cheers,

Ray
 

[BHOoi]

I just blend your sketched sections and this is the outcome:

 The new features are grouped in "new" and new layers are with prefix "1-".

[keaner]

Cheers BHOoi,

Thanks for spending your time looking at this, appreciate it.

I've gone through your model tree and like your approach. Way more simple.

My plan is to modify the reference sketches that make up the middle body of the duct and use your approach to blend the inlet and outlet portions.

I will share my results in a day or so.

One question. What was the relevance of making the two reference sketches 4 and 5? I didn't see them being used in your features.

 

Cheers,

Ray

[BHOoi]

ooos, I should have deleted those 2 secs. I wanted to use sweep and added those 2 sketches as its trajectories  but realised later simple blend is  easier in this case. Sorry for the confusion.

[keaner]

No worries,

Could you advise how best to control the sketched sections that make up the middle duct blend?

They are not controlling the shape how I need.

 

I would like the middle duct to stay withing these reference sketches

Plan View

Side Elevation 

This is the current outcome

Not smooth transmission

More visible here in Iso. 

Appreciate it if you could take another look,

Cheers.

[BHOoi · Accepted solution]

I would say you should sketch out the path, "S" sections both on top and side. The intersection of these will  give you the 3D curve you need on the middle span. You can then sketch out the X-sections with reference to these 3D curves.

I attached an example  to illustrate the above approach using you existing end sections.

Hope this helps.

[keaner]

Thank you so much,

This helps me out immensely.

Fairly new to Creo, but have learned heaps from your modeling.

Just the mid wall now to add.

Have a great day,

Ray

[Patriot_1776]

Be aware that with Boundary Blends (or anything really), it's way too easy to OVER control it, and get ripples.  The rule of thumb is always use the least amount of sections and/or control curves.  I didn't add any intermediate sections and it's symmetric about the centerline, but one of the main tools of a Boundary Blend you didn't mention:  End Conditions.  When you use tangency, that's exactly ONLY what you get, but when you specify "Curvature" you can "Show Drag Handles" and push and pull on resultant surface to get what you want.  I've attached my quick "Duct" part, you can see from the "zebra" analysis that it's a very smooth transition.

Smoooooottttthhhhh....

[keaner]

Hi Frank,

Thanks for the tip and sharing your model.

Always good to see another approach and reasoning behind the method.

Very valuable.

Cheers,

Ray

[Patriot_1776]

Did you get a chance to play with the drag handles?  If so what did you think?  That's a VERY powerful tool that you can use to get exactly the shape you want, smoothly, without over-constraining things and getting ripples.

[keaner]

Hi Frank,

Didn't get a chance yet.

My main battle is controlling the reference sketch splines.

I use a side and top sketch and intersect these to create blends for the top, bottom and sides. The results at the moment isn't too pretty.

 

Have you experience controlling splines?

Cheers,

Ray 

[Patriot_1776]

Try this:  Don't use them for the actual surface, even as "guide curves".  I'd put the surfaces in the way I did just constraining them at the ends, then use the drag handles and get as close as you can.  Looks like this is a duct for an automotive HVAC (or similar).  As long as you clear whatever is causing the duct to curve like that, you're golden.  It doesn't have to be exact unless I'm missing something.

If you must use a spline, remove a LOT of those points if you want a smooth result.

[keaner]

Thanks for sticking with me on this Frank. Correct this is an air duct where I am trying to keep air velocity constant with min. pressure drop at the outlet.

Below is where i got to with your suggestions using drag handle. Great results. Just battling the red circled area as the surface reference is an angled one.

 

 

 

 

 

 

[keaner]

Sorry Frank. Another query. Having problems now thickening the merged shape. Could you advise how to solve this?

Cheers,

Ray

[Patriot_1776]

If I remember correctly from my time working at a NASA wind tunnel, it should really only matter the cross sectional area of the at the start vs the end to maintain the same velocity.  In other words, if you have a box, and on the inlet side you have a 4"X8" (height x width) opening, and on the outlet side you have an 8"x4" (different long axis) opening, the velocity/volume of air out of the outlet should remain the same, correct?  I mean, you might lose a little to turbulence, but...

So, I'd say as long as your duct misses any outside obstructions, and is reasonably smooth, you should be good.  That's why I suggested  the LEAST amount of control points, so you didn't have ripples and thus non-laminar flow.  Smoooootttthhhh is best.  Now, you could get all tricky and make sure the cross-sectional area was either constant or transitioned smoothly, but that's a whole 'nuther can of worms.  I did something like that in a large cast stainless manifold where the horizontal inlet was 6" diameter, and transitioned to a 5" vertical diameter in an "S" shape that turned 90deg down.  The end locations were defined, as was the top "S" curve, and to make the transition as smooth as possible I actually flattened the circular cross-section to an ellipse slightly then back to a circle, all the while using a graph to transition the cross-sectional area smoothly even in the elliptical section.  You could section the plenum anywhere along the spine, and the area would be whatever the graph said it should be at that point in the spine.  I.E. halfway down the trajectory the area would be exactly that of a 5.5" diameter circle, even if it was elliptical in shape.

I have Creo 4, so, if you're above that I can't look at the file.  I don't have time to open it today.  But, if there's just that one spot, you might have to cut out a small section of that side and make a small patch, with curvature constraints on both ends, then merge and try and offset.

Best of luck!

[keaner]

Thanks for taking the time to share your NASA story. Must of been a cool experience.

Yeah using less control points has definitely helped keeping things smooth. I will definitely focus on keeping the cross sections consistent 

I'm using Creo 4. If you get a chance tomorrow I would appreciate if you could take a look at the attached model. Having difficulty thickening the main duct after merging all the boundary blends.

Appreciate it Frank,

Ray

[Patriot_1776]

Didn't get the chance to upload your file yet, but I played with mine a little.  I didn't like the fact that the B/B surfs are split into multiple surfaces within the same quilt if you extend the references within the surface itself, makes no friggin sense, and removing them is a total PITA.  So, what I did was use the "Piece to Piece" option, then extended the surfaces manually afterwards.  This worked out MUCH better as now each surface is ONLY one surface.  It adds 2 extra features per surface, but, I like the results a lot better.  See what you think.

[keaner]

Hi Frank,

Tried completing the shape with 'one surface' on each side, but ran into difficulties on two sides (top and bottom). These sides wouldn't extend far enough with the edge looping and not staying straight. Thanks for the tip though.

My main issue is still the thickening of the main body as I can only achieve 0.3mm. Require 2mm. I think this limit is due to the sweeping tight shape.

[Patriot_1776]

Yeah, there IS a limit when extending surfaces.  What you might have to do is make oversized surface on all 4 sides (8 total surfaces to create) and using the piece-to-piece option, then trim everything.  I tried it both ways but mine worked with the extend so that's what I posted.

What you might have to do is if the thicken doesn't work after that, offset all 4 surfaces individually, and trim them and solidify after you have a quilt that will "hold water".  Or, make 4 total BB outer surfaces, and 4 inner ones to make an inner and outer quilt and solidify.  Something WILL work, it just might be more work.

[Pettersson]

Just skimmed this thread. I didn't see Swept Blend being mentioned. It seems a good tool for this kind of project, since you can control the cross section area along the sweep. Basically, you add a bunch of datum points along the curve (easily done with a pattern) and then tell the Swept Blend that at each of these points, the cross section area should be the same. More points = more control. I haven't actually worked with this kind of application, but I made a test part once to learn the technique. I'll upload it here. Maybe it can be of some help/interest? Check the Swept Blend feature and the Options tab.

[Patriot_1776]

Typically the more control points the more ripples in any curve or surface.  It's best for smooth surfaces to let the software do it's thing because you as a human never get it as accurate by trying to add points as the software will just using end conditions.  With the drag handles mentioned you can get almost any curved surface you want and it'll be far smoother than using a lot of sections or control points.  I typically don't use swept blends, I'd rather use sweeps with multiple trajectories (the old "Variable Section Sweep"), they're smoother.

[Patriot_1776]

Smooooth.

Ok, I played with it a little using your file as a starting point.  First, I took your Boundary Blends and made them with the "Piece To Piece" option, which allowed me to immediately get the 2mm thickness with me changing the Absolute Accuracy to .01.  That cleaned it up, but, I wanted to do something better.  I still don't like the way Creo made it 3 different surfaces (down from the "Natural" option's 5) for the quilt.  So, I created all 4 surfaces so large that they always extended the extents of any other quilt.  This made the quilts all one single surface.  Then it was just a matter of merging them.  If you do a "zebra" surface analysis you'll see you're not going to get smoother surfaces.  Sorry, I did delete some of your features further down the model tree before I thought you might need them later for something else.  The surface came out a little wonky on the inlet end because of that weird trapezoid surface (no surprise) so I extended then trimmed it after the thicken.  My features are all labeled and my surfaces are purple.

Is there going to be a CFD (flow) analysis done?  MY advice would be so make sure the surfaces do not double back on themselves ("S") if you can, and make your most abrupt turns on the inlet side so you can make the turns on the outlet side as smooth as possible.  You need time and distance to smooth out airflow, and the outlet end is the place to do it.  The perfect SUB-sonic shape is a raindrop because basically only surface tension, gravity, and aerodynamics are working on the shape to allow it to fall to earth as fast as possible.  Aerodynamically, it's more important to "close" the airflow over an object than it is to part it.  That's why raindrops are pretty blunt from the from to the maximum diameter, but have a long tail (like a LeMans race car), so there's time to close the airflow down.  Same thing when air is flowing INSIDE and object rather than OVER it.  Smoothing it at the end is most important.

Good luck, let us know how it works out (I'm curious), and I don't know if you can mark more than one post as the solution, but if not, you could "Kudo" it for me if you please.  It seems there's a ton of interest in this thread, but not a lot of comments.  Interesting...

Frank

[Patriot_1776]

On second thought, perhaps an exaggerated "S" MIGHT work better (if space/clearance permits) for smooth flow because it would give the airflow a longer distance/time to straighten itself out.  Hmmmm...  A CFD analysis would be able to tell.

Also, one thing we did at the exit of the plenum and before the start of the area where the models were placed in the 'tunnel, was there was a square grid that the airflow had to go through in order to straighten it out.  A separate plastic piece with a square or even hex pattern places in the outlet would be a simple solution.  You could even make all the bars a raindrop cross-section with the pointed end facing out and that would help.  Yes, there would be a little restriction (not too much for a raindrop shape), but if the ultimate goal is turbulence-free airflow, it might be worth it.  Looks like this is for a car HVAC system.

[keaner]

You're a smart cookie Frank 🙂

Correct an exaggerated "S" is where the next CFD cut got us. This shape (below) allows the air increased time to straighten and deliver more uniformly upon existing.

 

Challenge now is trying to model this in Creo. Other duct sizes are required in future. So a controllable 'family table' part would be the desired outcome. The other option is to import from the CFD. I have imported the CFD STEP model into Creo. It needed a little repair before I could solidify, but it worked. 

On adding a diffuser section part. I like this idea, but fear a pressure drop. It's all about the trade offs and what desired outcome you require.

Cheers,

Ray

 

[keaner]

Hi Frank,

Again a massive thanks for working on this. Love the new approach and the model looks great. Working through the model tree now unpicking your workings. Achieving that 2mm thickness is perfect. Just need to add the mid wall to split the duct internally.

To answer some of your questions:

Yes preliminary CFD work has been done by a colleague. I can't release too many details due to confidentiality, but I will share a side elevation first cut CFD below. The application for this duct is in domestic refrigeration area.

Interesting insight with SUB-sonic raindrop shape. Great to understand these fundamentals when trying to apply them to a solution. 

Cheers,

Ray

[Patriot_1776]

LOL  Thanks, but if I was THAT smart I'd be charging for all this!  🙂  ME, the work I've been doing lately is kinda blah, clean-up, so, every once in a while it's nice to be able to actually have a challenge, and this one looked fun.  I feel like an idiot chasing the "boolean" type surfaces, and something in the back of my mind just didn't like the approach.  So, about 3AM this morning, the solution hit me (like it typically does out of the blue when I'm not thinking of it).  It's like when I step away from things, BAM!, the solutions comes to me.  I'll post the STEP file and 3D PDF here, and I'll PM you the .prt file.

I will warn you, things are a mite touchy, as they typically are in these cases, I had to tweak, and tweak, to get that 2mm.  You make any major changes and it WILL fail.  So, to do any family table....well, you'd better be prepared to go through many tweak/fail cycles.  The splines at the sketch level are the key to tweaking things.  If you "modify" them like I did it's similar to using the boundary blend drag handles, you can determine the curvature from the fixed ends.  By tweaking and tweaking, I got the 2mm desired.  Whew!  I had to dig into my bag of tricks for this one!  I'm considering this done, gotta get back on other stuff.

Yeah, I find aerodynamics interesting, I learned a bunch, and what's better, for the short time I worked on the tunnel, it made me do more research to get an idea of how things work.  Not that this would affect what you're doing, but the physics change drastically from SUB-sonic to sonic.  It's crazy what happens.

Anyways, best of luck, and let me know how it all works out, just please don't post the part file.  Gotta keep some trick for myself, right?

[keaner]

Cheers Frank.

Noted on the necessary spline tweaking which will be inevitable as the other duct designs come together.

I have emailed you there for the part file.

Will keep you posted on the results.

Hopefully you can sleep easy now 😉

Thanks,

Ray

[Patriot_1776]

Cool, I'll send it to you tomorrow.  There are some sharp edges that, because of the thicken, waste material (very acute internal angle), but I figure that can be massaged out with a round or something.  Perhaps use that technique to do the internal surfaces first, round them, then thicken outward.  I just did it that way because the outside was defined.  The basics are there.

Did the 3D PDF and the STEP file here work?  I'm able to view it from my desktop but those 3D PDF's are weird.

Talk to ya tomorrow Ray!

F

[Patriot_1776]

Seems to be a popular thread for the lurkers...  🙂

[Patriot_1776]

Well, this has been fun.  Learned a few new tricks, thanks Ray!