Wet Pellet Plunger Extruder

[Ryan Carlyle]

So, pellet extruders kind of suck. Auger screws don't seem to have the necessary flow precision or retraction capabilities to be particularly effective 3D printer extruder drives. The filament-pusher extruder seems to be the only viscosity pump mechanism available that can control volume effectively enough for good 3DP results. 

Crazy idea time. 

I think a positive-displacement extruder mechanism is probably the best solution for pellet extruders.  Load a bunch of pellets into a syringe-type piston, heat and compress into a single melt, and use piston motion to push and retract the filament through the nozzle. People use paste-pusher extruders all the time; the only two big differences here are interstitial air removal and heating/sealing mechanisms. In fact, the sealing part is a pretty big issue -- very few elastomer sealing materials are good for 230C. O-rings that can survive that temp are idiotically expensive. (I should know, I work with them in my day job.) So, the question is, how do you pump molten plastic as a liquid, without needing high-temp pump seals?

Clearance seals (as are typically used in hydraulics) are probably out. They require precision machining and will seize up when the plastic is too cool. I do think a gerotor pump could move molten filament reasonably well, but it's still a very challenging service condition for any type of hydraulic pump. (EG, what happens when the shaft seals leak?)

So we need a pump system that handles cool material but lets us extrude hot material. I think you can do it using a pressure transfer fluid. What I'm imaging is a longish (12" x 1"?) steel tube that is actively heated through the bottom ~2-3" and heat-sink-cooled at the top. Load the piston full of pellets. Then pour in a low-viscosity, low-volatility, heat-resistant oil (less dense than filament) to fill the voids. Then insert a simple o-ring seal plunger on a jack-screw type pusher mechanism to act like a syringe plunger into the tube. (Alternatively, pump the oil into a closed chamber.) What should happen:

• The incompressible oil fills the voids between the pellets and transmits pressure from the plunger piston to the top of the melt pool.
  
• As the pellets melt, gravity separates the plastic from the oil and makes a stable melt pool at the bottom of the cylinder.
• Hydrostatic pressure applied by the piston generates the necessary nozzle pressure for extrusion.
• Active cooling keeps the piston heat cool enough for typical seal materials. Or, simply make the pump remote to the chamber through a hose/tube to restrict heat conductivity.

Complicated and messy? Sure. But I can imagine a big pressurized pellet reservoir like this getting much better volume control than a screw auger. 

[whosawhatsis]

Sorry, meant to send this on-list.

Would the viscosity of the molten plastic get low enough to reliably allow the oil to float to the top (and for that matter, are you sure that none of the oil would be absorbed by the molten plastic)? How would the thermal conductivity of the oil affect the thermal transition? I'm also not certain a priori that you can get enough range of speed on the piston to both extrude at an acceptable speed and resolution, and also get the rapid pressure changes you need for effective retraction. The molten plastic itself must be somewhat compressible, otherwise we wouldn't need to retract to relieve pressure.

There's also the problem with syringe systems that you have to have enough material loaded into it before you start to finish your print, because you won't be able to add any later.

--
You received this message because you are subscribed to the Google Groups "3DP Ideas" group.
To unsubscribe from this group and stop receiving emails from it, send an email to 3dp-ideas+...@googlegroups.com.
To post to this group, send email to 3dp-...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/3dp-ideas/f5c55c1d-4ec2-47f1-9eb2-e2611d430b75%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Petr Ptacek]

Just another crazy idea. Do you really need oil? If you tube is long enough, you can seal the cold end, then pressurize it with air. With cold pellets, air just escapes through the nozzle. As pellets melt, they will eventually clog the nozzle and extrude. You need to only regulate constant temperature of the hot end and constant pressure. Yes, it would be hard (impossible) to refill, but you just need air compressor with regulator, pressurized tank and heated tube with nozzle.

[Ryan Carlyle]

On Wednesday, April 22, 2015 at 10:38:41 PM UTC-5, Whosa whatsis wrote:

Sorry, meant to send this on-list.

Would the viscosity of the molten plastic get low enough to reliably allow the oil to float to the top (and for that matter, are you sure that none of the oil would be absorbed by the molten plastic)? How would the thermal conductivity of the oil affect the thermal transition? I'm also not certain a priori that you can get enough range of speed on the piston to both extrude at an acceptable speed and resolution, and also get the rapid pressure changes you need for effective retraction. The molten plastic itself must be somewhat compressible, otherwise we wouldn't need to retract to relieve pressure.

There's also the problem with syringe systems that you have to have enough material loaded into it before you start to finish your print, because you won't be able to add any later.

Reposting my off-list response:

We know oil floats on top of a filament melt pool pretty reliably and doesn't cause print issues because people oil their extruders all the time to reduce cold end jamming, and you can go hundreds or thousands of hours without needing to do it again. There's good evidence that a drop of oil in a normal extruder will primarily stay in the cold zone and is depleted at a pretty negligible rate. Only mixing due to retraction seems to cause any significant loss of oil out the nozzle. I take all this as reasonable evidence that an oil bath would probably work ok. 

It's a theory though. Might end up squirting oil pockets out all the time. I'm considering baffles or something to add shearing to break out stringers better. Typically though, high heat and low viscosity is all you need to break emulsions. I don't know if you can get the plastic viscosity low enough for that. 

On the pressure response, that's a fair concern, which is why I think a small, fast hydraulic pump might be preferable to a big slow piston. Lots of details to work out. 

Yes, the pellet capacity limit is an issue. But you can pause and refill, or use as many piston "reservoirs" as you want and then switch mid-print. In fact, assuming total system compressibility is low, you can even rig up multiple syringe feeds to the same nozzle for full color blending or material switching. Freeze plugs or a simple oil purge can fill unused cylinders. Because it's a positive-displacement system with no need for viscous seals around moving filament, most of the control and reliability issues with filament-mixing systems may go away. 

If you really wanted, everything could be automated with flowmeters for extrusion volume feedback, refill mechanisms, air purge systems, etc. It can get monstrously complicated, but that's why it's interesting. 

[Ryan Carlyle]

On Thursday, April 23, 2015 at 5:28:48 PM UTC-5, Petr Ptacek wrote:

Just another crazy idea. Do you really need oil? If you tube is long enough, you can seal the cold end, then pressurize it with air. With cold pellets, air just escapes through the nozzle. As pellets melt, they will eventually clog the nozzle and extrude. You need to only regulate constant temperature of the hot end and constant pressure. Yes, it would be hard (impossible) to refill, but you just need air compressor with regulator, pressurized tank and heated tube with nozzle.

First downside to air is the compressibility. You can't change air pressure quickly. That would basically force you to use a constant-nozzle-velocity system and some kind of molten plastic shutoff valve, which is an engineering challenge again. Although now that I think about it, you could run a needle valve stem all the way up the pellet hopper to the cold zone so there's no need for high-temp elastomer seals. Just one servo-controlled needle valve with a metal/metal cone tip seal. That could give you your flow rate control, too.

Other downside to air is Dieseling. We need at least 500 psi for good flow rate with a typical nozzle size. High pressure air contains enough oxygen to potentially auto-ignite flammable plastics. (Don't forget, auto-ignition temp is depressed by oxygen partial pressure.) So I think you'd need to use an inert gas like nitrogen. Which wouldn't be hard to pull from a nitrogen tank, but that's another consumable to worry about.

[whosawhatsis]

Some people believe that the oil works for a long time because, combined with the heat of the hot end, it seasons the inside of the steel tube, making it non-stick the same way as with a cast iron pan. Some people have even started cooking their thermal break with oil before assembly for this reason.

Also, when oiling filament, the oil is applied only to the outside. What I'm worried about here is that the pellets will deform as they are melting enough to seal up small pockets of oil before they deform enough to press all of the oil out of those spaces. Imagine a compressing a bunch water balloons in a tube. They would resist forming the corners that they would need to form to press all of the air out from between them before they formed a seal that would prevent the last of that air from escaping. The plastic should be solid enough at this point that the oil won't be able to float up through it, so I would expect pockets of oil to be pulled down into the melt zone. Assuming the oil and molten plastic aren't miscible, if you extrude slowly enough, they might float up faster than they are pushed down, but the plastic would remain viscous enough that it would have to be pretty damn slow, and even then, there would be a seal above it that would eventually result in one large pool of hot oil reaching the nozzle and squirting out all at once. Ouch.

BTW, I ran this by my business partner who has a background in fluid dynamics (used to work for JPL), and she pretty much agreed with my assessment.

It's an interesting idea, but if you decide to try it, remind me to be somewhere else at the time.

--

You received this message because you are subscribed to the Google Groups "3DP Ideas" group.
To unsubscribe from this group and stop receiving emails from it, send an email to 3dp-ideas+...@googlegroups.com.
To post to this group, send email to 3dp-...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/3dp-ideas/0d76f0ab-d4c5-4b01-ba7f-8948ef7e662e%40googlegroups.com.

[Ryan Carlyle]

On Thursday, April 23, 2015 at 7:39:26 PM UTC-5, Whosa whatsis wrote:

Some people believe that the oil works for a long time because, combined with the heat of the hot end, it seasons the inside of the steel tube, making it non-stick the same way as with a cast iron pan. Some people have even started cooking their thermal break with oil before assembly for this reason.

Also, when oiling filament, the oil is applied only to the outside. What I'm worried about here is that the pellets will deform as they are melting enough to seal up small pockets of oil before they deform enough to press all of the oil out of those spaces. Imagine a compressing a bunch water balloons in a tube. They would resist forming the corners that they would need to form to press all of the air out from between them before they formed a seal that would prevent the last of that air from escaping. The plastic should be solid enough at this point that the oil won't be able to float up through it, so I would expect pockets of oil to be pulled down into the melt zone. Assuming the oil and molten plastic aren't miscible, if you extrude slowly enough, they might float up faster than they are pushed down, but the plastic would remain viscous enough that it would have to be pretty damn slow, and even then, there would be a seal above it that would eventually result in one large pool of hot oil reaching the nozzle and squirting out all at once. Ouch.

BTW, I ran this by my business partner who has a background in fluid dynamics (used to work for JPL), and she pretty much agreed with my assessment.

It's an interesting idea, but if you decide to try it, remind me to be somewhere else at the time.

That is a very possible outcome. Depends greatly on polymer rheology effects that defy both intuition and typical fluid dynamics analysis. (I have some fluid dynamics background too.) Let me ask you, though -- why don't injection molding screw augers (or pellet extruding printers) embed significant amounts of air into the plastic as a foam? It's a fairly similar case of semi-molten pellets being squished into a solid mass. That should incorporate air pockets into the melt, but I don't believe it actually does. Auger shearing action would be more likely to convert large bubbles to many small bubbles (foam), not break them out of suspension.

Two-phase fluid mixing is highly driven by relative viscosity differences, more so than most other effects, and the viscosity of semi-molten polymer is so absurdly large that air and a light oil are both in the "near zero viscosity" ballpark relative to melting plastic. Factor in imperfect wettability and surface tension / capillary effects, and I think it's most probable that the melting plastic will maintain flow channels for oil expulsion and then coalesce into a solid pool simply due to surface-energy minimization. The polymer wants to be in a big blob because it's more attracted to itself than it is to the oil. The question is, how long does this take? What kind of residence time is required for full segregation? That drives your pressure chamber sizing. It might be too long to be practical. Or maybe you need to add energy to the system via some kind of vibration or mix-shearing. That would be complex, but not out of the question.

I don't buy the "seasoning" explanation for hot-end oiling. Many reasons:

• People commonly use oils with smoke points below the printing temp, and don't see any smoke/smell. 
  
• People commonly use oils with smoke points above the printing temp, which would mean they are not appreciably cooked onto the walls.
• People have tried water/glycol mix and it works similarly to oil. This strongly suggests that the fluid is not entering the hot end, and is acting more as a thermal transfer fluid than simply a lubricant. 

Retraction does introduce significant shear-mixing to the cap zone. The common view of filament motion as a smooth piston plug is only accurate for low surface energy hot ends with minimal plastic adhesion (PTFE lined or similar coatings) during fairly steady extrusion (such as load sequences or sustained extrusion). Repeated retraction breaks up the piston plug, and low-lubricity thermal barriers like a Rep2 or E3D never develop the piston plug behavior in the first place. They maintain internal pressure via a dynamic-shear cap zone effect, where the shearing action between the moving filament and stationary tube wall creates sufficient shear stress to contain melt pool pressure. This is a true dynamic viscosity seal -- the cap zone simply moves up and down in the thermal barrier's transition zone as needed for viscosity to vary until it reforms a stable equilibrium melt pool pressure. 

Dynamic viscous-seal cap zones look like this if you cool an all-metal hot end fast enough during extrusion:

Low surface energy extruders seem to spend more time containing pressure with a piston plug (where semi-melt deforms into a full-bore slug) that looks like this:

The viscous cap zone mechanism you see in all-metal hot ends only works if there is significant shearing action between the solid incoming filament and the semi-molten annular ring sticking to the wall. This shearing action SHOULD mix oil into the plastic, but it doesn't seem to do so. 

[whosawhatsis]

I've seen filament with air in it, forming a long bubble down the center that essentially makes it a tube. Other times, I've found small bubbles dispersed within the filament. As you mentioned, air is compressible, which probably changes the situation significantly. My understanding is that the augers are generally mounted horizontally, which means that if bubbles can float up through the molten plastic, they will still move toward the die rather than pooling under a hard plastic plug as I suggested. It will also move them to the side of the tube, so that it might pool against one side of the melt chamber, causing it to escape to the side of the filament rather than mixed into it. I also understand that proper mixing requires an auger with several "chambers" where the open space in the screw increases or decreases, which I suppose would effectively knead the plastic as it goes through, and would probably also help release the trapped air before the plastic is fully melted.

Off topic, on http://www.flashforge-usa.com/shop/flashforge-aluminum-tube.html, it says that that thermal break is made of aluminum (instead of steel, as one would expect). Is that true?

--

You received this message because you are subscribed to the Google Groups "3DP Ideas" group.
To unsubscribe from this group and stop receiving emails from it, send an email to 3dp-ideas+...@googlegroups.com.
To post to this group, send email to 3dp-...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/3dp-ideas/cc3706a4-3246-4da1-810a-171f12e3f9d6%40googlegroups.com.

[Ryan Carlyle]

Once air (or oil) is entrained as a bubble, I don't think it's likely to come out. Surface tension acts to keep the plastic intact and prevent bubble escape, at least while the melt temperature is low. To my mind, you're largely relying in the air (or oil) escaping via interstitial channels as the pellets merge into a single molten mass. I don't see gravity segregation alone as allowing bubbles to escape at a fast enough rate, particularly with small bubbles that have been sheared into a foam. (Foam-breaking with high viscosity fluids is notoriously difficult.) Who knows, though. Maybe bubbles can shear out and I'm over-estimating the entrainment.

Injection molding screws use different profiles along their length primarily as a way of tuning the plastic shearing action to ensure even melting. In particular, they use mechanical work-heating for around half of the total heat input to the material. (Ideal ratio depends on the material.) Plastic deformation hysteresis losses and viscous friction both generate much of the heat required to melt the pellets. Applying proper work-heating to the semi-solid chunks (rather than the thinner liquid bulk phase) requires reducing the clearance gap within the screw channel so that residual chunks continue to be sheared against the wall along a sizable portion of the screw length.

http://www.pitfallsinmolding.com/plasticatingpage.html

Injection molding machines also use check valves and a giant ram to apply the actual injection force. All the screw does is work the pellets, mix the melt to an even consistency, and provide enough pressure to retract the screw-ram assembly / fill the shot chamber. The screw is not generating all that much fluid pressure. (Just enough to reset the ram.) This is a big difference between injection-molding machines and 3DP pellet extruders to date.

How does all that relate to the wet-pellet extruder? Ehh, probably not much at all. Very different physical systems. 

Aaaand FF's mk10 (new style) thermal barrier is definitely steel. I have a few. Not the first time they've had goofy parts descriptions. 

To unsubscribe from this group and stop receiving emails from it, send an email to 3dp-ideas+unsubscribe@googlegroups.com.

[dan...@puptv.com]

There is another way that may be a better solution...  Daisy chain filament production machnie to a filament storage device and finally, to a normal 3d printer.

Augers work great extruding plastic if they not need to stop or retract repeatedly.

The solution that rarely seems to be discussed is to daisy chain both system...

First have a continuously running auger system that creates filament.

Next have a spooling system that takes up and releases significant quantities of filament as necessary. A filament bank. It stores filament when it is being made faster than it is being consumed, and loans out extra filament in the opposite case. The speed of the filament manufacturing can be gradually adjusted as necessary.

Finally, a normal printer.

Daniel - http://www.TriDPrinting.com/

[Ryan Carlyle]

Daniel, Jetguy is looking at doing that for his 4' cube bot (using a recent Kickstarter 'struder that claims to be able to run precisely enough to keep up with live printing). The big issue is diameter control -- it's hard enough to get consistent diameter filament production without also trying to slave it to the take-off speed from a printer. What happens if you get a jam or otherwise have to pause the print? You're basically going to have to stop the 'struder and cut out an off-spec section and reload the printer. It's doable, but a giant hassle and overall a lower reliability system.

The other challenge is the temporary storage medium. You have to handle a bight or loop of filament, not a free end. Spools are out because they can't effectively pay out from one end while paying in from the other end. I'm imagining a pair of pulleys arranged in a block and tackle arrangement, with the slack filament acting as the "wire" so that raising and lowering a weight provides the necessary length buffer. Again, doable, but more hassle.

[dan...@puptv.com]

A forward only pellet extrusion system seems like a good way to do it...  Getting a whole pellet based auger system to retract sounds like it would introduce some very difficult  issues.

Rather than adjusting the speed that the pellet extruder works, you could more easily adjust the speed of the printing.

There are a lot of ways to configure the filament storage system. Yes, a block and tackle system that can contract/expanded would work well.  A tube where the filament could coil as needed would probably  work rather well, but it might take some work to get the sizing and shape right, etc. Shorter storage systems could be as simple as a loop hanging off the side of the table.

Alternately, it may be that you can design the storage system to require very small deviations, as little as 1 cm for example. It would need to grow during the traverses to absorb the newly created filament. For example it could (expand/retract) to 5mm at the beginning of a traverse, then slowly to 6 mm during the traverse. It would then contract back to 1 mm in order to start printing again, and then over time contract to 0 mm or the starting position as printing continued.

Unfortunately this design might add some extra planning/slicing limitations, so the printer is never instructed to do repeated long duration retractions.,

Daniel - http://www.TriDPrinting.com/

[Ryan Carlyle]

I think this concept is more practical than my "fun and complicated" wet pellet extruder, and deserves a thread fork.