Feature request: Voronoi pattern and random infill cell surfaces
Posted by realthor
|
Feature request: Voronoi pattern and random infill cell surfaces November 04, 2015 03:16PM |
Registered: 9 years ago Posts: 1,035 |
Hello,
I was wondering about the feasibility to include as an infill pattern a Voronoi diagram that would make the infill choice denser around holes and corners and more sparse throughout the object. Also it might be helpful for the resulting tubes (for honeycomb for example) or walls (for grid or line) to follow the 3D shape of the object, a sort of 3D infill.
Regarding the second request I would like to have the possibility to print a skin even inside of the part we are printing, just as it would at the final layers. This would increase the rigidity of the part and the resulted closed cells would act as a thermal insulator, a sought for property in many aplications. Moreover I would like to have a slider as to set the fragmentation of the caps on individual tubes or group of tubes. The following sketch is to illustrate this:
I assume this kind of request can be done on github or via the slic3r website but wanted to put it here first maybe it is unreasonable or not useful. What do you think?
RepRap Lander concept on Concept Forge
RepRap Lander concept on RepRap Forums
My Things, mostly experimental stuff
I was wondering about the feasibility to include as an infill pattern a Voronoi diagram that would make the infill choice denser around holes and corners and more sparse throughout the object. Also it might be helpful for the resulting tubes (for honeycomb for example) or walls (for grid or line) to follow the 3D shape of the object, a sort of 3D infill.
Regarding the second request I would like to have the possibility to print a skin even inside of the part we are printing, just as it would at the final layers. This would increase the rigidity of the part and the resulted closed cells would act as a thermal insulator, a sought for property in many aplications. Moreover I would like to have a slider as to set the fragmentation of the caps on individual tubes or group of tubes. The following sketch is to illustrate this:
I assume this kind of request can be done on github or via the slic3r website but wanted to put it here first maybe it is unreasonable or not useful. What do you think?
RepRap Lander concept on Concept Forge
RepRap Lander concept on RepRap Forums
My Things, mostly experimental stuff
|
Re: Feature request: Voronoi pattern and random infill cell surfaces November 05, 2015 07:25AM |
Registered: 10 years ago Posts: 2,472 |
|
Re: Feature request: Voronoi pattern and random infill cell surfaces November 05, 2015 07:42AM |
Registered: 9 years ago Posts: 1,035 |
Wow, never saw that ... thanks for pointing it out.
RepRap Lander concept on Concept Forge
RepRap Lander concept on RepRap Forums
My Things, mostly experimental stuff
RepRap Lander concept on Concept Forge
RepRap Lander concept on RepRap Forums
My Things, mostly experimental stuff
|
Re: Feature request: Voronoi pattern and random infill cell surfaces November 06, 2015 08:45PM |
Registered: 10 years ago Posts: 444 |
|
Re: Feature request: Voronoi pattern and random infill cell surfaces November 07, 2015 04:55AM |
Registered: 9 years ago Posts: 1,035 |
I am after the closed cell insulation properties of such panels.
Other than that Strenght increases with every such layer but not sure how this helps.
RepRap Lander concept on Concept Forge
RepRap Lander concept on RepRap Forums
My Things, mostly experimental stuff
Other than that Strenght increases with every such layer but not sure how this helps.
RepRap Lander concept on Concept Forge
RepRap Lander concept on RepRap Forums
My Things, mostly experimental stuff
|
Re: Feature request: Voronoi pattern and random infill cell surfaces November 08, 2015 09:23AM |
Registered: 10 years ago Posts: 444 |
|
Re: Feature request: Voronoi pattern and random infill cell surfaces November 08, 2015 09:56AM |
Registered: 9 years ago Posts: 1,035 |
I would be curious too what they had in mind. I think we might have to ask the developers, there is zero info on the net or in the slic3r manuals about why would anybody want to use this.
RepRap Lander concept on Concept Forge
RepRap Lander concept on RepRap Forums
My Things, mostly experimental stuff
RepRap Lander concept on Concept Forge
RepRap Lander concept on RepRap Forums
My Things, mostly experimental stuff
|
Re: Feature request: Voronoi pattern and random infill cell surfaces November 08, 2015 10:25AM |
Registered: 11 years ago Posts: 5,780 |
As we have all seen from the top and bottom soild layers, a single solid layer usually doesn't fill all the gaps between the the infill and the perimeters- it usually takes two or three solid layers to get a really solid fill. I suspect that the "solid fill every X layers" option is intended to strengthen the part by providing some lateral support for the infill, especially if it is sparse infill, and especially if it is the "lines" pattern instead of the hex or rectilinear pattern. When you use "lines" infill, as the infill gets taller it starts to warp.
Ultra MegaMax Dominator 3D printer: [drmrehorst.blogspot.com]
Ultra MegaMax Dominator 3D printer: [drmrehorst.blogspot.com]
|
Re: Feature request: Voronoi pattern and random infill cell surfaces November 08, 2015 05:58PM |
Registered: 12 years ago Posts: 187 |
Why "Solid infill every n layers"?
A Structural Engineer's perspective:
For non-mechanical parts which are largely "unstressed", the infill pattern and density should have very limited impact on the functionality of the printed part. You can experiment with all of Slic3r's infill patterns, and you will probably find that any / all of them work fine.
However, if you are printing something which carries significant loads, the infill needs to be sufficiently stiff and strong to provide 3D support to the bottom, side and top faces, as well as possibly withstanding crushing loads through the thickness of the part in one or more directions. In most cases, you get 80% or more of the overall strength from the outermost shell of just a couple of mm thickness, so there are good reasons to use low density infill (speed of printing, efficient use of materials, etc), providing "just enough" infill material to give the required robustness for the application. For these applications, the density and profile of the infill can make a big difference.
Suppose you are using the "Concentric" infill pattern [manual.slic3r.org] for a 150 mm cube, say - you can imagine that some of the infill walls could be very slender and unsupported. They will be fine to provide printing support for the top surface, but if you then apply a vertical compression load onto the top face of the cube, it is likely that some of these infill walls will buckle, and the top face will lose its structural support. Now imagine squeezing the cube horizontally - the Concentric infill shells don't contact the side faces, so from a strength perspective, you are crushing a hollow shell, not a "solid" cube, and the crushing strength is limited by the number of perimeter shells in your model.
If you print the same part with the same infill pattern, but use "solid infill every 10 layers" say, those infill planes will provide lateral support to the thin vertical walls (and the perimeter walls), and will allow them to carry a lot more vertical load, for a very small additional amount of plastic. Even the horizontal crushing load will be significantly improved.
Better yet, if you use "Honeycomb" infill, the vertical walls are kept to a short chord-length, and self-intersect in a 3-way joint every few mm. Every layer is rotated by 120 degrees, so there are no weak lines extending through the model - the second layer prints across the joints in the previous layer, tying the corners together. This infill design is inherently capable of carrying very high loads in all 3 dimensions. (There is a reason that aerospace composites use honeycomb cores with structural skins!)
For my prints, 20% honeycomb infill (without using "solid infill every n layers") works fine, My parts have perhaps 75% - 80% of the full strength and stiffness as if the part was injection moulded as a solid part, but use a fraction of the amount of material, and that is plenty good enough for me. (In fact, when I print a replacement for a failed injection-moulded part, my replacements are generally much stronger, because I print a honeycomb-filled "solid" replacement, instead of a hollow open shell with a few thin diaphragm stiffeners.) Yes, honeycomb infill can be slower to print than some of the other infill patterns, especially for large parts, but the difference is minimal for the components I am working with. I haven't bothered exploring any of the other infill options because "if it ain't broke, don't fix it". There might be some parts for which a different infill pattern (with or without solid infill layers) might be better, but I haven't encountered any fundamental problems with 20% honeycomb infill.
If you want to experiment with the structural properties of various infill patterns on a macro scale, try making a cardboard deck using pieces of cardboard standing on edge and taped together so as to support a sheet of cardboard that can support your weight 100 mm or so off the ground. Chances are, you won't have much success until you try some sort of cellular / honeycomb support - or just use a few toilet roll tubes:
[www.youtube.com]
(Yes, I know the audio is in Cambodian, but the video works fine in any language!)
Follow my Mendel Prusa build here: [julianh72.blogspot.com]
A Structural Engineer's perspective:
For non-mechanical parts which are largely "unstressed", the infill pattern and density should have very limited impact on the functionality of the printed part. You can experiment with all of Slic3r's infill patterns, and you will probably find that any / all of them work fine.
However, if you are printing something which carries significant loads, the infill needs to be sufficiently stiff and strong to provide 3D support to the bottom, side and top faces, as well as possibly withstanding crushing loads through the thickness of the part in one or more directions. In most cases, you get 80% or more of the overall strength from the outermost shell of just a couple of mm thickness, so there are good reasons to use low density infill (speed of printing, efficient use of materials, etc), providing "just enough" infill material to give the required robustness for the application. For these applications, the density and profile of the infill can make a big difference.
Suppose you are using the "Concentric" infill pattern [manual.slic3r.org] for a 150 mm cube, say - you can imagine that some of the infill walls could be very slender and unsupported. They will be fine to provide printing support for the top surface, but if you then apply a vertical compression load onto the top face of the cube, it is likely that some of these infill walls will buckle, and the top face will lose its structural support. Now imagine squeezing the cube horizontally - the Concentric infill shells don't contact the side faces, so from a strength perspective, you are crushing a hollow shell, not a "solid" cube, and the crushing strength is limited by the number of perimeter shells in your model.
If you print the same part with the same infill pattern, but use "solid infill every 10 layers" say, those infill planes will provide lateral support to the thin vertical walls (and the perimeter walls), and will allow them to carry a lot more vertical load, for a very small additional amount of plastic. Even the horizontal crushing load will be significantly improved.
Better yet, if you use "Honeycomb" infill, the vertical walls are kept to a short chord-length, and self-intersect in a 3-way joint every few mm. Every layer is rotated by 120 degrees, so there are no weak lines extending through the model - the second layer prints across the joints in the previous layer, tying the corners together. This infill design is inherently capable of carrying very high loads in all 3 dimensions. (There is a reason that aerospace composites use honeycomb cores with structural skins!)
For my prints, 20% honeycomb infill (without using "solid infill every n layers") works fine, My parts have perhaps 75% - 80% of the full strength and stiffness as if the part was injection moulded as a solid part, but use a fraction of the amount of material, and that is plenty good enough for me. (In fact, when I print a replacement for a failed injection-moulded part, my replacements are generally much stronger, because I print a honeycomb-filled "solid" replacement, instead of a hollow open shell with a few thin diaphragm stiffeners.) Yes, honeycomb infill can be slower to print than some of the other infill patterns, especially for large parts, but the difference is minimal for the components I am working with. I haven't bothered exploring any of the other infill options because "if it ain't broke, don't fix it". There might be some parts for which a different infill pattern (with or without solid infill layers) might be better, but I haven't encountered any fundamental problems with 20% honeycomb infill.
If you want to experiment with the structural properties of various infill patterns on a macro scale, try making a cardboard deck using pieces of cardboard standing on edge and taped together so as to support a sheet of cardboard that can support your weight 100 mm or so off the ground. Chances are, you won't have much success until you try some sort of cellular / honeycomb support - or just use a few toilet roll tubes:
[www.youtube.com]
(Yes, I know the audio is in Cambodian, but the video works fine in any language!)
Follow my Mendel Prusa build here: [julianh72.blogspot.com]
|
Re: Feature request: Voronoi pattern and random infill cell surfaces November 08, 2015 06:19PM |
Registered: 9 years ago Posts: 1,035 |
Great explanation @julianh72, thank you it is just as I expected but I couldn't explain so well.
What about the Voronoi adaptive stress pattern density do you think we'll ever have those in the major open source slicers? I've read many articles citing the adaptive pattern density as an invaluable trick to compete with injection moulded parts strength for strength.
RepRap Lander concept on Concept Forge
RepRap Lander concept on RepRap Forums
My Things, mostly experimental stuff
What about the Voronoi adaptive stress pattern density do you think we'll ever have those in the major open source slicers? I've read many articles citing the adaptive pattern density as an invaluable trick to compete with injection moulded parts strength for strength.
RepRap Lander concept on Concept Forge
RepRap Lander concept on RepRap Forums
My Things, mostly experimental stuff
|
Re: Feature request: Voronoi pattern and random infill cell surfaces November 09, 2015 12:17AM |
Registered: 12 years ago Posts: 187 |
Quote
realthor
What about the Voronoi adaptive stress pattern density do you think we'll ever have those in the major open source slicers?
I learnt a long time ago to "never say never", but it will be a while before such a concept becomes part of the mainstream RepRap workflow in my opinion.
In order to do adaptive stress design, you need to do a stress analysis, and that means your modelling system needs to have some means of defining the loading, doing the stress analysis, and modifying the geometry to remove material where it isn't needed and adding it where the stresses go too high, all while maintaining the "design intent" (governing dimensions, etc). For the design of "trivial" parts (simple wall brackets, etc), it is possible to set up a parametric rule-based approach using just a few key parameters (spacing of screw holes, bracket out-stand, bracket load, etc), and optimise the design accordingly, but this becomes a hugely complicated task when you try to expand into generalised 3D part design. In the absence of a rational stress analysis, I'm not sure what rules you would apply to vary the geometry and / or the infill density according to the stress intensity?
There are plenty of professional engineers who can't perform meaningful finite element analysis even when professional-grade FEA software is embedded into their professional CAD systems, so I suggest it isn't quite ready for the mainstream RepRapper just yet!
If this area is of interest to you, there are numerous on-going research programs into the issues of adaptive / optimised design, but most people seem to be working on a particular sub-set or class of designs, rather than a general all-purpose optimiser. AutoDesk Within technology might also be worth looking into: [www.autodesk.com] (But don't expect AutoDesk to release it as Open Source software any time soon!)
Follow my Mendel Prusa build here: [julianh72.blogspot.com]
|
Re: Feature request: Voronoi pattern and random infill cell surfaces November 09, 2015 01:04AM |
Registered: 12 years ago Posts: 187 |
In many cases, I would argue that we're already there - at least as far as making RepRappable replacements for original parts.Quote
realthor
I've read many articles citing the adaptive pattern density as an invaluable trick to compete with injection moulded parts strength for strength.
Injection moulding imposes its own design constraints - the part has to be able to be released from a 2-part mould, so internal voids and overhangs are not practical. Injection moulded parts therefore tend to be hollow "shells" with a missing face, and internal diaphragm walls to provide the strength and stiffness. 3D printing has almost no limitation on where material can be placed. While it is possible to replicate the injection moulded shape shape using a 3D printer, there is no NEED to do so - you can easily print the load-bearing part using solid or honeycombed fill.
As an example, just yesterday I printed a replacement for a plastic crank handle for an adjustable saw bench which had failed (see attached "Injection Moulded Crank") - the original plastic part fractured where the three ribs joined the main boss. I could have just replicated this part completely, but instead of using a stiffened shell for the crank arm, I printed it as a solid using 20% honeycomb fill (see "Reprapped Crank"). The printed part weighs only fractionally more than the original, but it is MUCH stronger and stiffer, as I have eliminated the weakest points - the ends of the ribs. Sure, I could have done some fancy finite element analysis to optimise the design even further - and if my replacement design was going to be mass produced, that would probably be appropriate - but for a one-off build, there really was no point.
Follow my Mendel Prusa build here: [julianh72.blogspot.com]
|
Re: Feature request: Voronoi pattern and random infill cell surfaces November 09, 2015 03:58AM |
Registered: 9 years ago Posts: 1,035 |
Quote
julianh72
In many cases, I would argue that we're already there - at least as far as making RepRappable replacements for original parts.
Sure, in many cases we are there for sure. What about the few cases we aren't there? The software side of things can help tremendously without even knowing it. Voronoi patterns can make a part stronger at the exact spots that it is needed, reinforcing the corners and the holes and my personal opinion is that it's going to change a lot the stiffness of the parts as well as the weight. Maybe it's not noticeable with ABS or PLA but buy some nanocellulose reinforced PP or something fancy for your parts and you will feel the extra weight in your pocket.
Quote
julianh72
and if my replacement design was going to be mass produced, that would probably be appropriate - but for a one-off build, there really was no point.
Some time ago we didn't even thought of 3D printer parts as one-off buids but rather mere prototypes. Now we're at the point where one-off build are ok-ish. Let's conform to the trend and think of what is possible so we can do optimization in what we have and use the extra money on better filaments for example. That's why I believe no step is too small or unnecessary. I really hope OnShape will do FEA in the cloud and they will someday link slicers with it to run optimization algorythms in the slicing process.
RepRap Lander concept on Concept Forge
RepRap Lander concept on RepRap Forums
My Things, mostly experimental stuff
|
Re: Feature request: Voronoi pattern and random infill cell surfaces November 09, 2015 07:49AM |
Registered: 10 years ago Posts: 2,472 |
Quote
realthor
Some time ago we didn't even thought of 3D printer parts as one-off buids but rather mere prototypes. Now we're at the point where one-off build are ok-ish. Let's conform to the trend and think of what is possible so we can do optimization in what we have and use the extra money on better filaments for example. That's why I believe no step is too small or unnecessary. I really hope OnShape will do FEA in the cloud and they will someday link slicers with it to run optimization algorythms in the slicing process.
I agree. After all it is "only" software - by which I mean that there is a one-off cost in man-hours, but thereafter the "manufacturing cost" of software is equal to the support cost - and if the software is good this can be kept down to close to zero or made a chargeable addition. This means that popular software is extremely profitable. I am quite certain that one day CAD design tools will have built-in slicers so as to produce 3D printer output as an option. Having the slicer work directly on the CAD design instead of via an STL makes a lot of sense, and would allow better optimisation for the individual printer. Undersize holes etc. would be a thing of the past.
Dave
|
Re: Feature request: Voronoi pattern and random infill cell surfaces November 09, 2015 08:33AM |
Registered: 9 years ago Posts: 1,035 |
Thanks @dmould for the input.
I really hope that soon CAD software will be able to tap into the slicer's API and send it real geometry rather than STLs. But until that time we can start making the slicers do what we need. If enough people believe that this particular example, the Voronoi adaptive pattern is necessary we can raise a github request and chime all in to push the change. I can't help more than study the industry and see what can be done better to have usable 3D printed parts. Maybe that counts as an effort too
RepRap Lander concept on Concept Forge
RepRap Lander concept on RepRap Forums
My Things, mostly experimental stuff
I really hope that soon CAD software will be able to tap into the slicer's API and send it real geometry rather than STLs. But until that time we can start making the slicers do what we need. If enough people believe that this particular example, the Voronoi adaptive pattern is necessary we can raise a github request and chime all in to push the change. I can't help more than study the industry and see what can be done better to have usable 3D printed parts. Maybe that counts as an effort too
RepRap Lander concept on Concept Forge
RepRap Lander concept on RepRap Forums
My Things, mostly experimental stuff
Sorry, only registered users may post in this forum.