Contraption: Filament winder build

rocket_troy

Well-Known Member
Joined
Sep 9, 2013
Messages
524
Reaction score
274
Location
Melbourne, Australia
Actually, a wind angle of ~55° (as shown in the video) IS optimal for my primary intended application of motor cases. You primarily need hoop strength so you want a wind angle not too far from a typical hoop wind, but you also have to have some axial strength to keep the closures in. I can dig up and link the paper I found on the subject later.
Yes, for a pressure vessel the optimum winding angle is completely different to something representing a load bearing column or something along those lines ie. something experiencing compressive bending loads as opposed to pure tensile in all directions.
The challenge for a straight (internally pressurized) tube of composite construction is attaching the ends. You can't really optimally utilise many of the standard methods available to you for a plain alloy material; although frankenbolt retention has been utilised before with success with some composite tubing although your winding angles might need to reflect the stress distributions of typical alloy tubing than that optimised for pressure vessels as you'd be removing chunks out from the longitudinally stressed structure.
You can glue the ends in to avoid that, but then you need to be ultra careful about the design of those bonded joints with very close attention paid to the hoop *strain* of the tube.

anyway... Great Job! I'll be following with interest.

TP
 
Last edited:

G_T

Well-Known Member
Joined
Dec 29, 2011
Messages
2,772
Reaction score
728
Actually, a wind angle of ~55° (as shown in the video) IS optimal for my primary intended application of motor cases. You primarily need hoop strength so you want a wind angle not too far from a typical hoop wind, but you also have to have some axial strength to keep the closures in. I can dig up and link the paper I found on the subject later.
It sounds like the paper might have tried to do it with a single wind angle, which is simple but sub-optimal. Hoop strength for instance is fiber perpendicular to the long axis of the case. Being 35 degrees off that means your fiber is contributing only about a quarter of its strength in that direction. Being 55 degrees off the long axis means the fibers contribute even less in that direction. Net result is you need a lot more fiber to do the job. That's if you stick with a single winding angle.

A better approach:


You can find other videos showing the winding process, and these can give you a good idea of the winding angles. In manufacturing it is reasonably well optimized.

Gerald
 
Joined
Aug 8, 2018
Messages
14
Reaction score
5
That's a beautiful wind sequence! Prepare for some more self-taught lessons when you get resin in that, though I don't think you're gonna too much trouble.

I'm all moved into my new place and the old place is closing on Friday. I'm starting get back into my winder with some control theory, I have some issues to resolve/ go around with using an induction motor for the mandrel drive... Should be an interesting update, but not as pretty as yours.
Wow, darn nice design on that, I am impressed. Are you sharing any of the design details? If so, I would love to look at them.

I used the XWinder model for a while before I redesigned it.

From the video I think I saw something you may want to redesign. Was the last point the fiber was contacting on the delivery system a roller? If so, you will want to change that to a sliding surface instead of a rolling surface.

Rollers are a bad idea in any filament system like this. A stray fiber or two can get pulled around a roller, and suddenly you have the filament getting sucked into a knot. Carbon fiber is so strong that those stray fibers don't break, and once they wind around a roller, it is game over for that run.
The more advanced systems I have seen all use Stainless fixed rods or Delrin surfaces for sliding over. I have generally heard these called Doctor Blades but I am not sure that's an accurate use of the term.

If that part isn't rolling you are all good. This is in no way trying to be critical. I just want you to avoid the painful lesson I learned when a roller in my system effectively became a highly leveraged winch. It literally bent aluminum extrusions on the system before I could stop it.

Best of luck, and I will be following your progress.
 

eggplant

L3 | NAR 93664, TRA 17791
Joined
Jan 5, 2011
Messages
1,788
Reaction score
459
Wow, darn nice design on that, I am impressed. Are you sharing any of the design details? If so, I would love to look at them.

I used the XWinder model for a while before I redesigned it.

From the video I think I saw something you may want to redesign. Was the last point the fiber was contacting on the delivery system a roller? If so, you will want to change that to a sliding surface instead of a rolling surface.

Rollers are a bad idea in any filament system like this. A stray fiber or two can get pulled around a roller, and suddenly you have the filament getting sucked into a knot. Carbon fiber is so strong that those stray fibers don't break, and once they wind around a roller, it is game over for that run.
The more advanced systems I have seen all use Stainless fixed rods or Delrin surfaces for sliding over. I have generally heard these called Doctor Blades but I am not sure that's an accurate use of the term.

If that part isn't rolling you are all good. This is in no way trying to be critical. I just want you to avoid the painful lesson I learned when a roller in my system effectively became a highly leveraged winch. It literally bent aluminum extrusions on the system before I could stop it.

Best of luck, and I will be following your progress.

Thanks! Yup, you can get the CAD here and the toolpath generation script here.

Interesting, I do have a few rollers in the fiber path. I haven't seen this so far myself, but that does make intuitive sense. I wonder if it will start happening once resin is involved? Either way, I went with rollers because then I could cheaply print them rather than buying aluminum rod for each surface. That said, I already used some smooth aluminum rod for the tow centering mechanism and the resin bath. Maybe I should swap more over. I will certainly post an update if I have this issue.
 

G_T

Well-Known Member
Joined
Dec 29, 2011
Messages
2,772
Reaction score
728
Aluminum might not be the best choice. Aluminum + carbon fiber = battery. You get galvanic corrosion. Stainless would be a better choice - or possibly Delrin as mentioned, though I'd expect the Delrin to need replacing periodically. Friction will be lower with Delrin.

You do have the option of winding the cylinder dry, then using a resin infusion technique. I understand some places are experimenting with that, though I have no details. Sorry!

You might be able to use a hybrid technique where you roll a thin layer of epoxy on the mandrel, then wind a layer, then pause. Use a fairly dry foam roller to roll out the resin through the fibers, then apply a little more. Wind the next layer. Repeat until done... Then do the shrink tape thing. This way your fiber handling stays dry, avoiding the fun of getting the proper wetout. Foam roller should be the thinnest foam you can find. Cut into short sections and use the narrow roller handle. Warning though - if you are not used to working with carbon fiber, wetout isn't as easy as for Silane coated fiberglass.
 

kramer714

Well-Known Member
Joined
Jan 20, 2009
Messages
802
Reaction score
121
Rollers are a bad idea in any filament system like this. A stray fiber or two can get pulled around a roller, and suddenly you have the filament getting sucked into a knot. Carbon fiber is so strong that those stray fibers don't break, and once they wind around a roller, it is game over for that run.
The more advanced systems I have seen all use Stainless fixed rods or Delrin surfaces for sliding over. I have generally heard these called Doctor Blades but I am not sure that's an accurate use of the term.
I disagree......

When I was doing filament winding (day job - composites engineer) we used both rollers, combs and polished eyelets (all at the same time). Rollers work well for spreading the tows into a 'flat ribbon' with the combs aligning them. We used rollers to either submerge the fibers or to put them over a transfer wheel. Again rollers were used to remove excess resin (letting it fall back into the resin bath). A polished eyelet was used at the point of fiber placement on the mandrel. The diameter of the eyelet is important, it needs to be large enough so when the fibers lay down they are in a uniform ribbon. Ther are companies that make textile eyelets.

Tube Design

For References 90 degrees is a hoop wrap, 0 degrees runs straight end to end (this will be useful for my next comments)

Angles, the tubes really want a combination of angles.
For pressure (think motor case) a combination of +/- 15 degree (or 20 if you cant get down that far) and 90 degree (hoop wrap) it is best to start with the hoop wrap and end with the hoop wrap. You want 2x the thickness of hoop wrap to the +/-15. If you have a wide ribbon you may want to drop some of the tows for the hoop wrap. to get a better wrap and pull out excess resin.

For bending stiffness, think tube failing from compression loading (tube failed in buckling) a combination of +/- 15 degree (or 20 if you cant get down that far) and 90 degree (hoop wrap) it is best to start with the hoop wrap and end with the hoop wrap. You want 2x the thickness of the +/-15 to the hoop.

Quasi Isotropic (the tube acts like a metal, same strength and stiffness in all directions) - a combination of +/-22.5 and +/- 67.5 degree in equal thicknesses. best with the +/- 67.5 as the inner and outer layers. If you read about QI Lamaintes made from cloth at 0/90 , +/-45, filament wound tubes at 22.5 x 67.5 is the same thing just shifted 22.5 degrees so you can wind it. Easy but very few tube applications have the same stress in all directions, Filament winding allows you to optimize the material properties

Resin - Curing, It is best to use a resin the vitrifies at room temperature, then SLOWLY heat it for a 'post cure' Room temp epoxy will get tougher and stronger with a post cure. Temperature depends on the epoxy chemistry you are using. Poor mans oven - Parked car in the sun....

Mike (previously the composites guy) K.
 

Scrapmaster87

Well-Known Member
Joined
May 7, 2019
Messages
205
Reaction score
272
Location
Southeast PA
Aluminum might not be the best choice. Aluminum + carbon fiber = battery. You get galvanic corrosion. Stainless would be a better choice - or possibly Delrin as mentioned, though I'd expect the Delrin to need replacing periodically. Friction will be lower with Delrin.

You do have the option of winding the cylinder dry, then using a resin infusion technique. I understand some places are experimenting with that, though I have no details. Sorry!

You might be able to use a hybrid technique where you roll a thin layer of epoxy on the mandrel, then wind a layer, then pause. Use a fairly dry foam roller to roll out the resin through the fibers, then apply a little more. Wind the next layer. Repeat until done... Then do the shrink tape thing. This way your fiber handling stays dry, avoiding the fun of getting the proper wetout. Foam roller should be the thinnest foam you can find. Cut into short sections and use the narrow roller handle. Warning though - if you are not used to working with carbon fiber, wetout isn't as easy as for Silane coated fiberglass.
My winder(s) are dry winders and I start on a dry mandrel. Once I get a few passes in with the filament locking itself in place, I turn the speed down and start blotting resin onto the filament with a chip brush. Once I have full coverage with resin I'll start turning the speed back up, blotting the dry spots as they show. I've fount that once I get past the initial coverage, the wind is very low maintenance.
 

Scrapmaster87

Well-Known Member
Joined
May 7, 2019
Messages
205
Reaction score
272
Location
Southeast PA
View attachment 531401
I'll post a longer update soon with many more details, but it made a carbon fiber tube! And for a first attempt, I'm pretty happy with the results.
That looks exactly like the last glass tube I made. BEAUTIFUL!

For finishing I chucked that up in my layup lathe (before I took it apart) to sand it smooth. I had to stop several times to skim-coat it in resin because I was getting into the fibers.

20 years ago when my dad and I were making CF and Kevlar tubes and fins we always had a sacrificial outermost e-glass layer. This way we're protecting the good fibers when finishing.
 

eggplant

L3 | NAR 93664, TRA 17791
Joined
Jan 5, 2011
Messages
1,788
Reaction score
459
Seeing as I'm now up to having made three tubes, I figure it is time for another update.

Part of why I waited until now was that I wanted a tube I was truly satisfied with before typing up my notes. The first tube (my last post) was better than I had expected for a first try, but had some issues that I'll write up below. The next one was a two-steps-forward, one-step-back experience because it fixed some issues from the first tube, but introduced new ones. Finally, the tube I made today incorporated all of the best parts of both tubes into something that would just take a tiny bit of sanding to make perfect. I'll document them in separate posts.

Starting with the original tube, I made a video to summarize the process:

There were a lot of things that I just guessed on, either because I couldn't find a single answer on them or because there were enough other unknowns that it didn't cross my mind while planning things out. For example:
  • I didn't have my mold release spray with me because it was with my propellant making supplies, so the mylar on the mandrel ended up adhered to the inside of the tube.
  • I didn't realize that only one side of the heat shrink tape is release-coated (and managed to pick the wrong side), so getting it off of the tube was a huge pain.
  • I guessed on the overlap for the heat shrink tape, and I guessed wrong. I did a full there-and-back hoop wind to apply it which means you have two opposing spirals to sand out, and I also overlapped it by about a third of the width of the tape, which just means that you have different amounts of compression in different areas and as a result, really deep/wide spirals in the tube. If my life depended on it I could get it smooth, but that would suck.
Those missteps aside, I was pleased with the first tube turning out as... well... a tube! Nothing wind-endingly catastrophic took place and the tow behaved almost better than it does dry. The "squeegee" mechanism I came up with (just a piece of rubber with a slit cut in it) did a good job of keeping the resin in check and also created the cool lava-lamp effect you see in the video. I was feeling pretty good about it after that early success, so I reprinted the single-use resin bath and rollers and got ready to try again, hoping to achieve a smoother surface finish.

20220807_224259.jpg
 

eggplant

L3 | NAR 93664, TRA 17791
Joined
Jan 5, 2011
Messages
1,788
Reaction score
459
For the second tube:
  • I kept the wind schedule almost the same, but added an additional hoop layer to the start to make the tube more appropriate for a motor case. I plan to glue in closures and hydrostatic test these tubes in the future to see the impact of wind angle and the number of layers, but for now it seemed like a good change.
  • I bought a second can of mold release spray so I don't have to remember to move it from the other shop where I make propellant.
  • I faced a dilemma on the heat shrink tape. Discussion with some rocket friends made it clear that I should only wind it in one direction and be careful to make sure every part of the tube gets covered the same amount, but it still wasn't clear what that amount should be. Hoping to save tape, I decided to wrap it so everywhere gets exactly one layer of tape (no overlap). This seemed right because I have the more aggressive, 20% shrink tape while the friends overlap many times but use 5% tape.
The setup was smooth compared to last time and tow was rolling quickly. Unfortunately, the squeegee seemed to not work anywhere near as well as during the first wind. The first wind was on the resin-rich side, but totally workable especially with heat shrink tape. This time, huge amounts of resin were coming along with the tow. I had to refill the bath much earlier than the first wind, and by about half way though it was nearing having used as much resin as the entire first wind did. For reference, compare this to the video in the last post:
20220827_170343.jpg
Last time the tube was just shiny during the wind, now it had huge globs of epoxy on it. Not great! I decided to just not put any more resin in the bath once I hit the same mass I had put in last time. This seemed to just about work out, because though the late-layer tow went on to the tube effectively dry, there was more than enough excess resin to wet it out once in place. Still not ideal. It have two theories about why this happened, but that's for the next post.

Unsurprisingly, the tape squeezed out a ton of resin even with it only being wrapped on once. Last wind, I had applied the heat gun to the tube for a long time as I wasn't sure of when the tape would be "done", and noticed that the epoxy had hardened up well before the expected pot life. This is typically fine, but for high temperature epoxy I wanted to follow the instructions closer so I can assume the datasheet properties. The data sheet just says "3 hours at 25 C, 2 hours at 155 C". As such, I only applied heat with the gun for a couple of minutes and then did the rest of the cure at room temperature, then put the tube into a preheated oven. Almost immediately, I saw a bunch of tiny bubbles form. After waiting for the cure cycle the data sheet calls for, I pulled the tube out, (easily) unrolled the tape, and saw this:
20220827_223101.jpg

The bubbles totally obscured the cool filament-wound pattern! That's like the whole reason we do this, right? :p Either way, it suggested that something was wrong with the cure, which means I definitely wouldn't trust this tube at high temperatures. It seemed like the tape shifted a bit because it doesn't like being applied at this (relatively) steep angle, producing a couple of ridges without any compression and some valleys with double compression. At least the mylar came out this time, producing a nice, smooth interior:
20220902_192138.jpg

But clearly I still had some thinking to do to resolve the issues on the outside of the tube.
 
Last edited:

eggplant

L3 | NAR 93664, TRA 17791
Joined
Jan 5, 2011
Messages
1,788
Reaction score
459
That brings us to the third tube, which I wound today. Compared to the last one, I:
  • Replaced the "squeegee" with a freshly-cut one instead of just cleaning it like last time, thinking that maybe epoxy cured in the slit and made it not work properly anymore. I realized I also wasn't careful to center the bath under the roller above it last time, which could have led to the squeegee only really scraping one side of the tow. I made sure it was centered this time.
  • Applied the tape with a 50% overlap, meaning that everywhere got exactly two layers of tape. I was worried that this would lead to small regions of 1 or 3 layers because of the tape slipping like it did last time, but it seems that the shallower angle made the tape wind on much better.
  • Ran the heat gun for a bit longer than last time, and then put the tube in the oven at 30 C for the rest of the room temperature part of the cure (close enough). Once it was time to increase the temperature, I did so with the tube in the oven to gradually heat it.
Not much else to say in this post! The changes above produced a much nicer tube, without any regressions compared to the last tubes:20220904_184232.jpg
If this were a body tube, all I'd need to do is a light skim coat of epoxy to fill in the pinholes and then sand it to get rid of the very minor tape spiral. For a motor case I think it is basically already perfect!
 

Maxwelljets

Well-Known Member
Joined
Sep 19, 2015
Messages
326
Reaction score
199
It sounds like the paper might have tried to do it with a single wind angle, which is simple but sub-optimal. Hoop strength for instance is fiber perpendicular to the long axis of the case. Being 35 degrees off that means your fiber is contributing only about a quarter of its strength in that direction. Being 55 degrees off the long axis means the fibers contribute even less in that direction. Net result is you need a lot more fiber to do the job. That's if you stick with a single winding angle.

A better approach:


You can find other videos showing the winding process, and these can give you a good idea of the winding angles. In manufacturing it is reasonably well optimized.

Gerald

While that is true if you wanted only hoop strength, or only axial strength, 55° wind angle is still optimal because a pressure vessel sees both at the same time. Analyzing composites gets weird because of the anisotropy.

I ran two different scenarios through some composite analysis software I got from a professor when I took a composites course. It is unfortunately unable to simulate tubes, only plates, but by setting up the y axis load to be twice the x axis load, the analysis still works. The first scenario involved wind angles at 0 degrees (axial) and 90 degrees (hoop), with twice as many fibers in the hoop direction as the axial direction. This is what you might expect would be ideal if you were used to analyzing loads in different directions independently. The second scenario I ran involved an equal number of fibers running in the ±55° directions.

Scenario 1:
1662344160483.png

Scenario 2:
1662344086338.png

The "load ratio" (LR) is essentially the safety factor for the composite. It's the ratio of the load it would take in order to cause the composite panel to fail to the current load on the panel. As you can see, the ±55° panel has a higher load ratio, and thus a higher burst pressure. The reason 55° is optimal is because that is the angle where the fibers are contributing twice as much hoop strength as they are axial. i.e. Fibers at that angle are being loaded purely in tension when pressure is applied to the pressure vessel. If I remove the hoop stress on the tube, and leave only the axial stress, it turns out the composite would actually fail *below* the test load I was applying:
1662344773080.png
This is due to the weird anisotropies inherent to composites. Without the hoop load pulling the fibers in that direction, they end up being loaded in shear instead of pure tension.


TL;DR: intuition from isotropic materials isn't that useful with composites, and the ±55° wind angle really is optimal for pressure vessels.
 
Last edited:

REK

Sponsor
TRF Sponsor
TRF Supporter
Joined
Feb 21, 2012
Messages
3,076
Reaction score
106
Location
El Paso, Texas
You want to overlap your tape in 1/8” increments. It compacts way more and will leave a better surface finish. Most especially if your using dunstone heat shrink tape.

4DC6686F-B930-46D1-AF08-D69E0F4DEAEB.jpeg

When it comes to filament winding however, Dunstone’s heat shrink tape is overkill and I prefer you get the one from Rockwest Composites. I’ve used it with great success, so long as you apply enough tension to it. Again space it in 1/8” increments.

Also it is cheaper and you get a giant roll for the price.
 

eggplant

L3 | NAR 93664, TRA 17791
Joined
Jan 5, 2011
Messages
1,788
Reaction score
459
As mentioned in previous posts, my main goal with this filament winder project is to manufacture motor cases. I still have some improvements to make to the winder itself before I can wind the 3"/4", L-O cases I'm picturing, such as lengthening the main rails and increasing the speed of winding by tuning it better. Still, I wanted a check to make sure I was on the right track for winding pressure vessels, so I kicked off a (hopefully) quick project to prove that concept. First, a hydrostatic test article.

I started by winding my longest tube yet, using a 54mm liner as the mandrel. 1 hoop layer, 3 layers of 55 degree helical winds. It went about as smoothly as the last 3" wind and the mylar meant that the resulting tube was a perfect fit over the liner, snug without being too tight. The primary reason I'm excited about using custom CF tubes for motor cases is the weight savings, but I've also been burned a couple of times by buying liners/motor cases that didn't fit well with their counterpart and either required an incredible amount of sanding or just buying a replacement. These tubes are made to the ground exterior of the liner, so they fit perfectly. You can also fix the liner-fit problem with case bonding, but that's for another time...
20220910_213107.jpg

I then turned closures from circles cut out of a sheet of 1" LE phenolic, one solid and one with a hole for a tapered aluminum NPT fitting that I also made. The closures have steps on them for interfacing with the liner and taper on the other end to give more space to pour glue into.
20220918_135314.jpg

I cut about 6" from the tube to test, hoping to use the rest for a motor. I chose to include a liner even in the hydrostatic test article because it will provide some amount of hoop strength in the motors themselves, so I want it present in the test too. If I switch over to casebonding these motors I will have to redo the test without the liner, but that's not a problem because I don't think I'll try that until 3" or 4" motors and I'd want to test those anyway. I RTV'd the aluminum fitting into one closure, and glued the endcaps into the liner with Scotchweld DP420NS. I chose this epoxy because I had a fresh tube of it around for another project.

1664332278423.png

While that part cured I used a flap wheel and acetone to clean the inside of the case. The entire outside of the liner/closure assembly was coated with the same epoxy and slid into the tube, then the tapered regions were filled in with more epoxy. The DP420NS was not the right choice here because the "NS" stands for "no sag", and you want it to flow into the narrow space. In the future I'll use something much less viscous, maybe the Stycast 3050 that AT/RCS includes for their DMS motors.

Once cured, I filled the case up with water and rigged it up to the pump. In the process, I encountered a setback. I had intended to mill a feature into the aluminum part to allow me to grab it with a wrench when torqueing the fitting in, but I didn't get around to it and figured that the RTV would be enough. Nope, right as I was done screwing it in, it started spinning in the closure, breaking the seal. This made it a struggle to get the case started pressurizing and there was a constant slow leak out of the fitting. I decided to try anyway, because the alternative would just be throwing all of this away and making a new one. This single-use hardware is going to require me to not cut any corners.
1664333049303.png
Still, with enough pumping to account for the leak, the case held pressure, and actually reached ~1500 PSI before the I heard a pop and pressure dropped. It seems the epoxy cracked and the closure shifted. Seems like the closure attachment was the weak link, and the tube itself was unfazed by 1500 psi!
20220926_202731.jpg
Post-mortem cross section. I was pleased by just how thin this liner+case assembly ended up. Of course, there's room for mass savings on the closures.
20220926_212201.jpg

That's it for now. My next step is another hydrostatic test with a few tweaks (and no leaks), then a K motor with a silly mass fraction (that might work). I'll probably start a thread over in the research section for that to keep this one focused on the winder.
 

jim_kc

Active Member
TRF Supporter
Joined
May 31, 2018
Messages
34
Reaction score
14
Nice progress, that tube looks great. It's been fun watching this project come along all the way up to a completed part.

Good luck with the motors.
 

eggplant

L3 | NAR 93664, TRA 17791
Joined
Jan 5, 2011
Messages
1,788
Reaction score
459
Lots of small process updates on the winder itself that I'll type up at some point in the future. Summary is that I learned a lot about mold release. The update for today is that I finally accomplished my goal of firing a motor in a custom filament wound tube! It is been about 11 months since I started really dreaming about it, so it feels great to have done it. Hit a mass fraction of ~0.7, with easy enough avenues to ~0.73. Longer motors should be even better, maybe 0.75 or a bit more. I hope a rocket built around a motor like this could exceed 0.6 mass fraction. Next step is scaling up to a 3" motor, GEM3.

 

RGClark

Mathematician
Joined
Aug 25, 2017
Messages
275
Reaction score
16
Location
Philadelphia, PA USA
Lots of small process updates on the winder itself that I'll type up at some point in the future. Summary is that I learned a lot about mold release. The update for today is that I finally accomplished my goal of firing a motor in a custom filament wound tube! It is been about 11 months since I started really dreaming about it, so it feels great to have done it. Hit a mass fraction of ~0.7, with easy enough avenues to ~0.73. Longer motors should be even better, maybe 0.75 or a bit more. I hope a rocket built around a motor like this could exceed 0.6 mass fraction. Next step is scaling up to a 3" motor, GEM3.



Thanks for the very informative post!

Edit: for the 0.7 propellant fraction was that comparing just propellant to motor casing weight or were other weights such as nozzle or retaining plug also included in the dry weight?

Bob Clark
 
Last edited:

Latest posts

Top