I have done anealing of PLA, I tried a 3D benchy. The layers fused so well, there were no layer lines to the naked eye. The Benchy was now a perfect PLA puddle.
You just melted the surface. Cut it open and you'll see the layers are still there. There's no way FDM can compete with injection molding unless you're ready to shed $800K for an industrial printer from say Siemens or HP.
Hi Sthephan! I was able to get a +300% increase in impact resistance while annealing natural PLA. I tested according to ASTM D4508, for which the sample is smaller. As printed, it broke easily, but after 20min at 95°C they did not even break, using the same pendulum. I will be publishing the results next year at the Brazilian Congress of Manufacturing Engineering (COBEF2023).
I love what you are doing, taking 3D printer to the next level and breaking new ground and dispelling urban myths on 3D printed parts. Thanks and respect.
Keeping the print warm definitely helps with layer adhesion, same with slower printing speed (really slow) and slower cooling fan or just turned off. Which has me investigating how to turn off the cooling fan for internal fill, so that the fan only turns on for the shell. Furthermore, so the fan only turns on at a certain level of degree and ceiling layers. So that when you print vertical walls there's no cooling fan, but once you reach let's say 75 degrees it starts to kick in or ramp up for the outer shell if its printing outwards or the inner layer if its printing inwards. Might have to write my own code, haven't done that in a while lol.
5 лет назад
@@MrHeHim you don't need the fan for pla but it does help with bridging and you should be able to enable it just for bridging easy enough.
This is an absolutely valid remark. Although the part looks different it may actually be much more functional strength wise. He should definitely test it.
@@nekononiaow generally though being a puddle loses a crystalline structure. I imagine it'd actually be significantly weaker.......... at least weaker than the flat-printed hooks. Not sure about the upright ones though. But, I could be wrong too: after all that part is solid now
I'd like to see you test the hook that completely melted and see how it fails compared to the normal vertical hook. Great video always enjoy them thanks.
Hi! I looked a bit into that subject previously to improve my quadcopter strength. I found that if I disabled cooling and tweeked a bit the temperatures the parts were almost not deformed after the heat treatment. Of course, disabling cooling and the temperatures affect the look and it is not suitable for all parts but if you are only looking for strength it is good and you do not need to compensate for any deformation. Thanks for the work you do. It is quite interesting and helping. :-)
You tried sand but I was wondering how well something that sets like plaster might work? You cast a part, heat treat then wash away the plaster. It sounds like a ton of work but might be worth investigating
I was wondering also with plaster... when it has hardened, and assuming that the part is completely filled, could the part then me heated to much higher temperature to completely melt the PLA and sort of have a 'cast PLA' part ?
Using plaster is an interesting idea, but if you're going to all that trouble, then just do an investment casting. Once the part is covered in plaster, you can bake the filament out and cast metal inside. Takes some planning to cast properly, but you get much better parts.
The amount of effort required to make sure the plaster gets into every single nook and cranny, and then taken out when done, would not be worth the 10% gains while maintaining the same dimensions, if it even works. Instead of going to all that hassle, just use something 10% better from the start, it'll be cheaper in the long run too
Encasing in plaster and reflowing makes the part way stronger. Print 100%, mix plaster and immerse part, wait until set, then bake at 200C until temperature will be even throughout. Let cool and wash away plaster. Keeps a lot of detail, though any air trapped will create a void at the top. I suspect with further research I could print an integral reservoir to fill the void and let the air out of the piece. Fridge clips I was printing kept breaking along lamination lines, but reflowing them removes all lamination.
I've annealed PLA (plain and modified) by dunking the prints into boiling water for 5 minutes, then quenching with cold water (not my original idea, saw it in a youtube vid somewhere). I mainly do it for the improved temperature resistance, but that is almost always with functional, mechanical parts that need to fit with other parts, so warping is not acceptable. I've found the boiling water method produces almost no warping at all, other than the predictable shrinkage in x and y, and growth in z. The shrinkage is predictable enough to compensate for in the slicer. I've printed a fan duct that works fine on the MK2S, a few mm from the heaterblock, printing high temperature materials, so the heat resistance is good with this approach. It's also way quicker and easier than annealing in an oven. My process is to put the part into a mug or similar and pour boiling water over it, tumble the part to get rid of trapped air bubbles, then pour in cold water after 5 minutes. Once the part dries it's ready to use.
Ian Dawkins The more expensive HobbyKing stuff is made by MCPP: www.mcpp-3dp.com The engineering PLA - ‘PLA-X3’ - is very good but a little pricey. It has very good annealing behaviour.
Nylon will appear to be quite stiff and strong fresh from the oven. But take care, as the print adsorbs moisture from the air, the mechanical properties of the nylon change and it becomes more flexible and less stiff. Some people take the initial stiffness to indicate that Nylon can be stiffened from heat treatment, but they fail to acknowledge the temporary nature of it. Annealing Polymaker's PC-Max blend will be fine, but to anneal real, pure PC prints you need a precision temperature controlled oven. We're talking being able to control temps to a 10c rise over the course of an hour sort of thing. Without such control, you might as well not bother annealing pure PC as it won't be beneficial. Of course, it's better to just print PC in a 120c chamber to begin with so there's not any printed in stresses resulting from the part cooling too fast, but most people don't have setups capable of that.
There are different results changing the infill density. 10%, 50% and 99% behave differently, especially regarding the type of deformation (uniform in X and Y). Z shrinks less, meaning that there is not a big layer fusion. Also, the deformation depends of the infill pattern used. A ventilated oven usually distributes the temperature more evenly. I'm not sure of the benefits of the slowdown temperature because in any case the skin remane cooler than the interior, and the process of loosing energy depends mostly from the geometry of the part, creating deformations. I'm testing the results using the microwave oven...
I have tried boiling some PLA+ parts (kind of ring shaped) and letting them cool down to 40c in the water. They were thin walled parts, 2mm thick, so 10 minutes boiling was more than enough to change the temperature resistance. As a result the part was still hard at 80c while normal PLA gets rubbery at 50/55. But the same results with deformation and warping. Uneven depending on the axis. I will keep trying with other shapes and failed prints just for fun. I still think that this can be useful in parts that can be printed, annealed and then post proccesed to get the desired dimensions (sanding, drilling, etc. ). Great channel!! I really apreciate your work. Congrats!!
I recently tried drilling small holes on opposite side of my print. With about 20% infill I then used a syringe to push epoxy and silicone glue through two different parts. I used a transparent filament with about 2 or 3 walls. It drastically changed the feel of the parts. You should try testing some hooks that have epoxy or silicone in them. My only problem is that the epoxy got pretty hot.
Use epoxy which cures slowly (as opposed to regular epoxy). This way you generate the same amount of heat during chemical reaction but you have much longer time to dissipate it, therefore the max. temerature stays way lover.
Oh that's so awesome I just got into 3D printing, had some parts that needed extra rigidity, and was thinking of the exact same thing! cylindrical voids running the length of the print injected with something, or geometry resembling lightening cuts in the appropriate orientation filled with JB weld.
reacting to your question on 2:20: yes, used volcano pla (formfutura, crimps minimal, reasonnably cheap as I live in Belgium, for usa the shipping must be crazy) like you used on that coffee maker like a year ago (?) printed clotheshangers from that and they are holding up after one year of serious (ab)use : one time I noticed visitors put like 3 wet winter jackets (heavy!) on one hook and it didn't fail. Used your 'making parts stronger info' vid and did print them with 5 perimeters instead of high infill. clearly worked, thanx for all the research Stephan!!
As always, this video was incredibly informative and all the testing was very detailed. You really are an asset to the 3D printing community and I've seen no other channel go into detail in the way you do. Great job man
Great video. Another property that would be interesting to test is the creep resistance. I have had several PLA part fail over time under constant load (preloaded assembly for example). I guess that increase in crystalinity could reduce the creep rate. For the load, you could compare several heat treatments at 70% to 80% of ultimate strength for example.
I've been testing how annealing Inland PLA+ test hooks affects material creep for the past couple months. (I think Inland PLA+ is made by eSun, but I haven't been able to get confirmation yet.) I annealed half of the C creep test loops at 85°C for two hours on the print bed with some a paper foam insulation on top. I've had the two kinds of hooks loaded continuously with 3 lbs. (about 1.36 kg) for more than a month now. The gap on the unannealed part has grown from 2.0 mm to 18.1 mm, whereas the annealed part is steady at just 7.3 mm! I think it might be an alloy of PLA and PBT, both of which are semicrystalline.
I tried annealing myself, but the shrinkage causes buckling on walls, which is visibly clearly on the fan shroud. Annealing is only worth it for functional parts, but functional parts need to be accurate, so you can't really do anything with this. The only thing I still want to try is bolting down a part to a metal plate, and then annealing a part while it's constrained. I tried some filaments that were specifically advertised for annealing, 3dktop Berlin, volcano PLA and something else, but they still warp. The process is interesting from a scientific standpoint, but engineering wise, for now, seems entirely useless.
Weird thing I've seen in doing some annealing tests of my own is that two different brands of PLA shrunk and expanded in different directions. Both test parts were printed in same orientation and annealed in same orientation as well. Still one of them expanded in X direction and shrunk in Z and the other one did the usual shrinking in X and Y and expanded in Z.
Great video as always 👍. I have tried annealing high temperature PLA and found it to have a positive effect. I don't have your test setup but did a simple bend test by hand and found the annealed piece it to be a bit stronger. I did get some warping when I annealed at 100C but none when I used 85C. I had the best results (no warping) when I let the part warm up slowly with the oven rather than putting it into the oven after it was warm. I also let it cool down slowly after an hour of annealing. Supposedly the HTPLA also gets better impact resistens after annealing. It would be great if you could test HTPLA to get some real data on it. Keep up the good work 😀
@@japonicaren I have used the Proto Pasta HTPLA V2 and the HTPLA Carbon fiber. I really like the CF for functional parts that need to be strong and it prints great.
Hi. I swear by the. Using 3D fillies pla plus. 100 c for 30 min. Raises its temperature stability to over 100c. For parts that may get hot in a car it’s a must. Shrinkage is only 1.5 pct xy and expands 1 pct z. I don’t use a cool down step. Whip out of oven while hot and then clamp them down on a cold metal surface. Also have some wooden jigs to hold the critical dimensions in place during ht. All this adds up to quality stable parts that look great and take the very hot Australian sun. Strength is also much better. I have tried this on several pla brands and not all respond the same. Stephan. You should do these tests on ht pla or pla plus. People will be put off based on this video. It’s never a blobby disaster for me with ht. Just get the right material.
I have used PLA in my car. Initially PLA was softened by sun heated ambient the first time it was exposed. The next day and other days that were hotter it did not soften again.
I just tried annealing Proto-pasta PLA in sand and had great results. I put the parts in a glass baking dish, covered in 70 mesh high purity silica sand and left a probe in the center of the sand/glass plate setup. I used a convection oven. It took about an hour to reach 100C in the middle; I held it for 10 and then turned the oven off and left everything in there to cool. I was pleased with dimensional stability of my parts.
I would have been interested in finding out what the tests said against the hook that really melted. That one would really have merged the layers. So perhaps you should consider testing it anyway.
@_ David _ Yes. I appreciate that. It's more out of pure intrigue than anything else. If results are noticeably stronger. Then maybe an experiment with 100% infil with the piece placed in a mould in the oven to keep its overall shape and dimensions. I appreciate that this then more or less turns the part into an injection moulded part. But as I said. Just intrigued, that's all.
Thanks for keeping your channel "real". I can't stand other fake printer channels where the host talks funny and acts fake on camera for subs. Not to mention they don't even know how to properly tune a printer and post prints covered in cobweb strings
@@S41t4r4 Only one in particular off the top of my head. but everyone has seen the channel. its a great channel for new 3d printer users but anybody with some printing experience can see their lack of knowledge displayed by the quality of their prints.
I don't want to point fingers.... just keep your eyes open. Im not saying im a expert printer or anything but I shouldn't have less strings on my TPU prints than their PLA ones LOL
There are some factors in the heating and cooling. Since heat rises, the top of the part is the last part to cool, thus why you see different expansions in z versus x and y. If you could somehow put the parts in a 3D rotisserie as they go through the process, you should see a more uniform change in size. In large turbines, this is why they have to keep turning during cool down or they bow “upwards”.
Wouldnt it be cool to have a duble extrusion 3d printer that would have really high temp plastic in one nozzle and like pla in other. So it would print the shell of the part and than crank up the bed temp to 220°C and fill the inside with pla. It would make really strong parts
King Masterlord fair enough, however, it’s almost free for Germans to get an undergraduate or graduate degree. Getting a degree teaches you a lot at any well respected school.
@@Niloc1922 yeah that would be nice, but as an American in this day and age I'm in a unique position to value targeted self education and as-needed research and study preferentially.
I must say that the way you flipped the "untreated" and "annealed" text should be a crime against humanity. You should use vertical text in situations like that.
I have tried to print some parts at 100% infill, then put them into a plaster and heated them to nearly melting point of pla. The idea was that all layers will fuse together. Then the plaster is just "disolved" in water. But my test were unsuccesfull probably due to a low temperature. And i never tried again, but i thing that i could make the part really strong. Would be nice to see you test that.
What if, instead of whole-part-in-the-oven annealing, you heat treated specific areas with a small flame torch to fuse the layers, let each section cool before moving on to reduce warpage? You could do that to the stress-point of a vertically printed hook and see if it strengthens the strain point... The problem I guess is not overheating it...
I use a soldering iron with more filament to reinforce the layer this way. Heatguns can do but the warping is way worse without a lot of practice. Also the heat dont reach very far in, so to just dip the whole thing in epoxy make it just as strong if not stronger
You're more likely to add as many or more stress points as you remove. To remove stress you need to normalize temperatures between regions, and the only way to do that effectively is to heat the entire part as evenly as possible.
Please try annealing parts immersed in common household oils, castor, canola/rapeseed or vegetable oils in particular. Its would be interesting to see the different effects that each oil had (if any) due to chemical compatibility and the duration and temperature of the annealing process. If you use the gyroid infill pattern and had drain/fill holes in the part where it wouldn't affect strength then, as its a "porous", the oil could evenly contact all of the material evenly, reducing warping, and as it would allow the parts to be brought up to temperature evenly it should more throughly and evenly annealing the part. I'd suggest a process of annealing at 45c, 60c and ~75c for 20, 40 and 60 minutes. That would be a 3x3 matrix of results. Ideally one such test for each oil type. I think this is the most complete process that could be done in the home environment.
What if you try annealing in silicone mold to avoid warping during annealing process? Silicone form if something not difficult to make and it will keep form while heated.
I would be very interested to know if PLA deforms over time. I heard that it slowly deforms under pressure, but I don’t know if this is because the pressure is high enough to cause plastic deformation or if PLA deforms over time even if it is not in the plastic deformation pressure zone.
would it be possible to remelt the plastic and retain the parts shape. you could encase the part in plaster or something else that will retain its shape at high temps for support.
Awesome as always,!!! I do hope you try HT PLA’ s generally made from a raw 850 or 870 resin. These PLA’s are higher temp and generally take to annealing much better than the standard cheaper PLA resins most filaments are made from!
when you were placing them in tightly packed sand did you enclose the sand? I have found warping almost went away when tightly packed and closed in a container, that being said itw as way more effort than its worth and i'd never bother doing it again
I hope you cover HTPLA and PLA+ That was a great idea to use sand. I've used hot sand to bend PVC pipes without kinking for costume projects. It makes the heating more even and also prevents the PVC pipe from collapsing in on itself. I hope you continue to experiment using the sand technique and try the larger container to be thorough to see if that helps any. You also might want to try heating the sand first before pouring it in to see what affect it might have.
You can also anneal them using boiling water! ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-Rc85APSK9vo.html It's a much faster way compare to oven, which should help you get through more testing sooner. Maybe some interesting findings will come out using this method too, keep up the great work!
Thanks This was very insightfull and learned alot. I was wondering if this annealing process if it followed, a reflow soldering temerature curve where you heat the parts to a specific temp for a certain time that every object can reach that temp then ramp up the temp for short period then decrease the temp slowly. Probally achieve the same results..... Or will it?
Thanks for the video! Very interesting. Two things you may want to investigate in the future for annealing of PLA: (1) Some manufacturers make PLA that is specifically designed to be annealed. My understanding is that these types of PLA have solid particles included in the PLA that are supposed to help "seed" the microcrystalization process. (2) I think there was speculation that PLA with pigments (colored solid particles) would more easily crystalize than PLA with dyes (colored chemical dissolved in the PLA as a solution; no particles). If you have time, I would be really interested to see you perform similar tests on, say, white-opaque PLA (which very likely has pigments like titanium dioxide, zinc oxide, etc.) as well as PLA that is specifically designed to be annealed. I have no idea if there would be any difference in results. There are numerous academic papers on annealing PLA and PLA crystallinity. It would be interesting to see if your results are roughly the same as theirs. They studied things like annealing times and temperatures (crystallinity vs time at a given temperature), mechanical properties, and microstructure.
Interesting investigation as usual... I really adore the way you apply scientific method to your investigations. Keep up the good work. I wonder if the layer adhesion quality is related to the hot plastic being reactive to humidity or oxygen in the air, or perhaps some air is trapped at the interface between layers reducing the effective connection between them. Is it possible for you to make an investigation about that idea? [3D printing in a vacuum, different pressure levels or in an inert atmosphere]
Have you considered using an oil bath (maybe with a sous vide) to maintain more consistent annealing temps? Might help with warping. Looking forward to the stress relief video.
I've annealed HTPLA flywheels so that they can withstand running on hot motors. While I was successful in annealing it, I had a yield rate of about 1% of "acceptable" flywheels where they hadn't warped and secured snugly to the motors. The other 99% deformed enough to not be useful.. I couldn't figure out a pattern in the failures.. To put this into context, the flywheels are spinning at 35-45k RPM with clearances of about 1mm. I found it better to just print in regular PLA and redesign the wheels to include an impeller to pump air through the motors. With this, I can get fairly well balanced PLA flywheels spinning at around 75k RPM.
This sounds something like what we do when casting bullets our of lead alloys. Yes our temps are much higher (720 degrees F). We drop the hot bullet as fast as we can out of the mold as soon as it changes from a liquid to a solid into a bucket of water, the colder the better (Does not stay cold very long). We do this to increase hardness. It has something to do with crystals in the alloy. I wonder if cooling fast is possible and if it would make a difference? Thank you for your videos, I have ordered a 3D printer and have been watching your videos. I want to use my 3D Printer to make Antenna Parts for Amateur Radio.
IMO sous vide is nice, if you only have a crappy oven but otherwise it causes too much problems with submerging and water pressure why I wouldn't use it. I don't think the temperature needs to be accurate to 0.1°C.
I did see some tests from another RU-vidr. He tested typical annealing methods, and then used boiling water -- get water up to a rapid boil, dip the part for less than a second or two, then immediately dump in cold water to prevent deformation. From his tests it seemed stronger, but he didn't test with multiple samples or any controls. @CNC Kitchen, maybe you should try that method.
The video on annealing with boiling water is done by alex kenis ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-Rc85APSK9vo.html Great video by another great 3d printing RU-vidr
You know, if i want strongest parts possible, i will use 100% infill with 105% flow. I print pla at 215° way slower. Anealing these parts wont warp as much and the cristal sizes wont be limited by the latice. Also, as it warps less you could heat them a bit more. I would suggest using finer sand and with 100% infill, heathing them more wouldnt warp them at all. Using something like dry clay powder might make a temporary cast mold to contain the melting pla
Hi Stefan, thanks for taking the time to educate the larger 3D printing community. It's been bugging me a bit but I think the phenomenon of the annealed prints becoming less amorphous may not be correct. I suspect it's a bit more complex than that. I can't say I know the real answer but a typical thermoplastic should become more amorphous when heated above it's glass transition as the stresses that the printing has applied are relieved somewhat so that the polymers individual strings can go back to a relaxed state which is why we see some XY decreases and Z increases in size. Could the print lines be lensing the light? Maybe there's a polymer chemist here that can put me in my place :)
Oh, you're already on track to investigate inter-layer adhesion. Not looking like there is any particular advantage to heating for adhesion. What is really interesting is the smaller warppage when you stress-relieve first... that could prove useful. You might try burying the part in sand, then vibrating to settle the particulates, then compressing them with a lid/clamp. Locking the grains together might help minimize the possibility of the warping part displacing sand only held in place by gravity. It might also crush the part a little.... would have to test it.
5 лет назад
If you're worried about the bend modulus changing, could you use your 3d scanner to scan the deformed test piece then 3d print the control with that mesh and compare those? Also agree with what others have mentioned about trapped air potentially causing further layer adhesion. The fact that your XY shrinks and Z pillows in that cube corner calibration piece shows that air is trapped (albiet the infill space is more to blame there). Could you do zero perimeter pieces or cut/sand external faces to expose solid linear infills and test if giving air between extrusions a place to escape can improve your results further?
Thank you very much for this information! I was looking everywhere for a plastic that can stand both about 100ºC and some steam pressure. I´ll try now with PLA, PET and HIPS and annealing them. You helped me a lot!
The degree of annealing depends on time and temperature. 100°C is a "usual" temperature. I don't think lower temperatures would change alot besides needing more time for the process.
I just brought some Proto Pasta HTPLA V3 that is supposed to be designed for this very thing. so might be worth a test. I plan to try this using my reflow oven.
I would like to see more about a filament intended for heat treatment, like Volcano. Does it really stay within .3% dimensions when annealing, and does it really resist heat to 95c+?
I aneal PLA at 100C for Fluid Fittings to ensure they are watertight, after which I dip the parts in hot wax at around 100C to protect them from the water.
My experience. Feel free to apply the scientific method and increase your sample size. PLA SPECIFIC. I use a heat gun, which even on low reaches at least 250c almost instantly. It's very hard not to deform the parts using it. Layer adhesion did show a minor difference, but it was purely aesthetic, and did not provide integrity. Was unable to remove the layer lines or even correct simple mistakes without going well beyond the glass stage and ended up with a gob of plastic goo. I also tried Tamiya model cement, to no avail. Next up... Bondo.
I've only heat treated PEEK because the filament manufacturer said amorphous PEEK doesn't have the desirable properties. I suspect the idea that annealing parts improves layer bonding is on parts where there's residual internal stresses such as poorly printed ABS parts. Print ABS well and you probably wouldn't benefit. I suspect nothing will change with PETG because of its lack of crystallization. I'm not seeing where it's worth annealing PLA to avoid needing to print some other material.
Great video, thanks for the upload. In case you try it with ABS, I just finished a test myself. I think Thomas did something similar but he chose around 105°C for the annealing temperature for both PLA&ABS. I just did the annealing of ABS at 140°C. The reason being, that PLA has a glas transition temp (TG) of 60-65°C and ABS at 105°C. I've seen that the annealing of PLA works at around 100°C, so 40°C above its TG. Therefore ABS has to be annealed at around 145°C. My Part shrank but the geometry remained ok. All the best with your experiments! Edit: sorry it wasn't Thomas video with the ABS, it was your video as well. It's name was temperature resistance after annealing and you inserted the parts at 110°C (PLA,PETG,ABS). I would suggest setting the temperature 40°C above the TG of the material being annealed.
I have a high flow two stage water pump that I need to be able to withstand temperatures up to 80ºC. How do I go about this? I am afraid that annealing will warp the parts too much, thus they won't mechanically interface properly anymore, as a result I am guessing my best bet is to go with something other than PLA at this point. Problem is I live in a country which has... how to put it lightly... with certain recent actions decided to slightly separate from the rest of the world, as a result I am not much spoiled for choice.
Thanks for the great videos. Have you tried 'tempering' in hot water? The maker lab at a local college places PLA prints in water and then in a microwave till the water is hot but not boiling (guessing about 70-80C) then slow cool to room temperature followed by a second cycle where the water is brought to a boil and slow cooled. I have tried this at home on small parts and it does not seem to do much damage but not sure if the parts are stronger or more heat resistant.
I was thinking. you really want to melt the outer shell.. Because its the outer shell that takes most of the force. Then inner part is just there to stiffen it up. So in stead of trying to bake it for a long period of time. Bake it so fast that the inner structure don´t have time to heat up. Exactly how to do this i don´t really know. But it need to be heated up really fast and then cooled down fairly quickly as well. Something like submerging them in boiling oil. Or just dipping one side at the time. This should totally melt the outer shell. refuzing it, while the inner structure is still sollid, making so it don´t fall apart.
I wrote a small python program that generates gcode the ramps the heat bed temperature up and down at a controlled rate for annealing printed parts. Per all the documentation for annealing plastics, the ramp rate is important! Using an oven does NOT control the rate. Are you interested in testing this? I like the test equipment you have made and am curious if controlled temperature ramping changes the results.
You did not try to do these parts with solid infill? I suspect that the warping / deforming might have something to do with the air on the hollow infill. It has moisture in it and it can expand by different amounts in each pocket thus deforming the part. Also with solid infill the part should not shrink since there is still the same amount of material in the part
Using metal beads used during bead blasting should yield good results they come in sizes smaller than single grains of sand while weighing much more than sand. I have a bag of these for other experimental uses.
