Wednesday, April 09, 2008
A few thoughts on warping
Recently, Chris (nophead) has been doing yoemanry duty in sorting out the problems associated with printing the various polymers that we hope to use while the rest of us worry about other issues. He recently published this little graph...
The chart displays the amount of upward curling that could be expected from various polymers printed at several fill percentages at each end of a 40 mm bar with, iirc, a 10x10 cross-section.
Chris, if I've understood correctly, has been doing such tests with the notion of identifying polymers and fill rates which give the smallest amount of warping after they've been pulled off of a printable surface.
I've been toying with the idea of trying to correct for warping by printing on a convex surface. The brute force approach to this would be to develop a mathematical model for warping and then design a support rick for a particular piece that we wanted accordingly. This morning, however, I am wondering if we could simply center any print on a standard convex support rick, the curvature of which would be determined by the polymer being printed.
It would seem that the first idea to test would be to try to print Chris' standard bars on top of a loose-fill support rick that corrected for the expected warping. If that worked a few other issues would have to be looked at.
The first would be whether the aspect ratio in the xy plane made much of a difference to the curvature of the rick. If it didn't it would seem that we should double the width of Chris' standard bar a few times to get the shape of the convex curve of the rick.
The second would be to see if the depth (z dimension) of the printed object affected the convex surface curvature as well.
It would really be nice if we could get by with a standard convex support rick for printing parts of a particular polymer. It would certainly make life a lot simpler.
(No additional content at 3DReplicators.com)
The chart displays the amount of upward curling that could be expected from various polymers printed at several fill percentages at each end of a 40 mm bar with, iirc, a 10x10 cross-section.
Chris, if I've understood correctly, has been doing such tests with the notion of identifying polymers and fill rates which give the smallest amount of warping after they've been pulled off of a printable surface.
I've been toying with the idea of trying to correct for warping by printing on a convex surface. The brute force approach to this would be to develop a mathematical model for warping and then design a support rick for a particular piece that we wanted accordingly. This morning, however, I am wondering if we could simply center any print on a standard convex support rick, the curvature of which would be determined by the polymer being printed.
It would seem that the first idea to test would be to try to print Chris' standard bars on top of a loose-fill support rick that corrected for the expected warping. If that worked a few other issues would have to be looked at.
The first would be whether the aspect ratio in the xy plane made much of a difference to the curvature of the rick. If it didn't it would seem that we should double the width of Chris' standard bar a few times to get the shape of the convex curve of the rick.
The second would be to see if the depth (z dimension) of the printed object affected the convex surface curvature as well.
It would really be nice if we could get by with a standard convex support rick for printing parts of a particular polymer. It would certainly make life a lot simpler.
(No additional content at 3DReplicators.com)
Comments:
<< Home
Hi Forrest,
My test bar is 40x10x20.
I can say for sure that there is no one curve fits all. A 100mm block warps more at the corners than a 40mm block but at the points +-20mm from its center the warping is less than the 40mm case.
Making the block wider will increase the corner lift because it gets tension from both dimensions.
Making it taller increases the warping but I think it has diminishing effect so gets pretty negligible above 20mm height as the top is nearly flat. That is why I chose my test height.
The bottom curves more than the top. That means that if you correct it by extruding onto a curved bed than you need extrude layers who's thickness varies as a function of x and y. That requires the filament flow rate to vary as well as 3D filament paths. Currently we don't having anything like the required extruder frequency response to do that.
The figures I am getting now are actually pretty good. I think I could make most of the Darwin bits with HDPE and they would function.
I am confident I can improve by at least 50% simply be putting the object in an oven and heating it to just below its melting point before cooling it again and removing it from the base. That is much easier than building it in an oven and I think it will yield similar benefits.
A 6" cube made out of PCBs soldered together orthogonally with spiral tracks for heating, wrapped in fiberglass, with one of Zach's thermocouple boards as the controller is what I have in mind.
It only needs to go up to about 120C so I think about 100W might do it.
My test bar is 40x10x20.
I can say for sure that there is no one curve fits all. A 100mm block warps more at the corners than a 40mm block but at the points +-20mm from its center the warping is less than the 40mm case.
Making the block wider will increase the corner lift because it gets tension from both dimensions.
Making it taller increases the warping but I think it has diminishing effect so gets pretty negligible above 20mm height as the top is nearly flat. That is why I chose my test height.
The bottom curves more than the top. That means that if you correct it by extruding onto a curved bed than you need extrude layers who's thickness varies as a function of x and y. That requires the filament flow rate to vary as well as 3D filament paths. Currently we don't having anything like the required extruder frequency response to do that.
The figures I am getting now are actually pretty good. I think I could make most of the Darwin bits with HDPE and they would function.
I am confident I can improve by at least 50% simply be putting the object in an oven and heating it to just below its melting point before cooling it again and removing it from the base. That is much easier than building it in an oven and I think it will yield similar benefits.
A 6" cube made out of PCBs soldered together orthogonally with spiral tracks for heating, wrapped in fiberglass, with one of Zach's thermocouple boards as the controller is what I have in mind.
It only needs to go up to about 120C so I think about 100W might do it.
Hi Nophead,
For the heated box, you could try insulating a styrofoam cooler with fiberglass insulation, so the styrofoam can help insulate but not go above its melting point.
If you want to make a hot box from scratch, you can make a solar box cooker:
http://solarcooking.wikia.com/wiki/
Minimum_Solar_Box_Cooker
with a PCB heater instead of the transparent top.
In either case, with all that insulation you might get by with just one heater board on the bottom.
For the heated box, you could try insulating a styrofoam cooler with fiberglass insulation, so the styrofoam can help insulate but not go above its melting point.
If you want to make a hot box from scratch, you can make a solar box cooker:
http://solarcooking.wikia.com/wiki/
Minimum_Solar_Box_Cooker
with a PCB heater instead of the transparent top.
In either case, with all that insulation you might get by with just one heater board on the bottom.
"Making the block wider will increase the corner lift because it gets tension from both dimensions."
Indeed, it should. If you center the convex surface on the xy centroid of the object you print the corners of the xy cross-section would be further from the centroid than would the corners of a print with a narrower aspect. Because of that you could certainly expect more curling and more correction from the convex rick.
Indeed, it should. If you center the convex surface on the xy centroid of the object you print the corners of the xy cross-section would be further from the centroid than would the corners of a print with a narrower aspect. Because of that you could certainly expect more curling and more correction from the convex rick.
Who needs a thermocouple?
If you want to get near the melting point of HDPE, why not use one of those resettable fuses?
They're made of cross-linked HDPE, filled with graphite to the point that it loses conductivity at Tg. It's similar to the ceramic heating elements in some space heaters, but tailored to your temperature range, and available from auto parts suppliers.
Bonus: it works on any voltage, including 240V and <1V, as long as the power supply can provide enough current. It might be worthwhile to put an incandescent bulb or some such in series with it.
If you want to get near the melting point of HDPE, why not use one of those resettable fuses?
They're made of cross-linked HDPE, filled with graphite to the point that it loses conductivity at Tg. It's similar to the ceramic heating elements in some space heaters, but tailored to your temperature range, and available from auto parts suppliers.
Bonus: it works on any voltage, including 240V and <1V, as long as the power supply can provide enough current. It might be worthwhile to put an incandescent bulb or some such in series with it.
The idea of a counter-curved bed was doomed from the get-go because the chart showing curvature versus material shows dramatically different results for 100% fill versus 50% for every material tested that way. Since a practical part will have some regions that are 100% filled and some which are either only partially filled - or perhaps just much thinner (the result would be the same) - there can be no single base shape that'll counteract the curvature for all kinds of part.
I suppose that in theory you could calculate the expected warpage using some heavy-duty FEM math and lay down a bed of filler material that's custom-sculpted to the shape of the object you were going to make.
I'd be quite surprised if that worked in practice.
I suppose that in theory you could calculate the expected warpage using some heavy-duty FEM math and lay down a bed of filler material that's custom-sculpted to the shape of the object you were going to make.
I'd be quite surprised if that worked in practice.
"It would really be nice if we could get by with a standard convex support rick for printing parts of a particular polymer. "
I should have said "a particular polymer and fill ratio".
I should have said "a particular polymer and fill ratio".
I don't think it'll really be worthwhile going into detailed analyses of warping before the machines start using support/filler material - it's quite possible that warping effects when constrained by support material will be radically different to those seen with current polymer-only runs.
forrest higgs said:
> I should have said "a particular polymer and fill ratio".
But that doesn't help - the "fill ratio" is only meaningful when the part is a big, chunky homogeneous thing. But if a part of it is big and chunky and needs 100% fill (for strength maybe) - but the other end needs a bunch of holes in it for fixing bolts or whatever - then that end may (in effect) only be 20% filled. Then the two ends of that single part will warp at different rates - and even if you have different ricks for different fill percentages, you'd be unlikely to have one that compensates for 100% fill at one end of the part and 20% fill at the other end (and if you do - you'll need more ricks for triangular parts and so on). In the end - for generalised part building - you'd need a custom rick for almost every part because the distribution of 'fill' isn't likely to be uniform.
> I should have said "a particular polymer and fill ratio".
But that doesn't help - the "fill ratio" is only meaningful when the part is a big, chunky homogeneous thing. But if a part of it is big and chunky and needs 100% fill (for strength maybe) - but the other end needs a bunch of holes in it for fixing bolts or whatever - then that end may (in effect) only be 20% filled. Then the two ends of that single part will warp at different rates - and even if you have different ricks for different fill percentages, you'd be unlikely to have one that compensates for 100% fill at one end of the part and 20% fill at the other end (and if you do - you'll need more ricks for triangular parts and so on). In the end - for generalised part building - you'd need a custom rick for almost every part because the distribution of 'fill' isn't likely to be uniform.
Well, I'm going to try it anyway. Even if I can cut down the amount of warping from what Chris (nophead) is getting now I'm ahead of the game.
If that doesn't work most of the maths that I'll need to do a more rigorous modeling of warping are in...
http://www-rpl.stanford.edu/user/files/papers/thesis_anickel.pdf
Post a Comment
If that doesn't work most of the maths that I'll need to do a more rigorous modeling of warping are in...
http://www-rpl.stanford.edu/user/files/papers/thesis_anickel.pdf
<< Home
