Gyroid

[Lucas]

Nice to meet you guys, first post here. Hope it's in the right place.

I have been modeling some Gyroids recently, but I can’t improve it anymore, would appreciate some help. Below I described the steps to achieve the best result I got until now, so it might be easier to spot where the flaw/improvement is (Also am attaching some files here - SW2021 Version):

- Create a cube and convert it to 3DSketch (with origin in the center)
- Draw Arcs and constrain them on the planes of the cube (have also done it with Fit Spline and made them tangent, but the result was not neat)
- Delete Body (Original Cube used for the first 3DSketch)
- Convert Entities (Arcs) in different 3DSketches and Extrude Surfaces with them
- Create 6 Planes parallels to the originals and create Split Lines where they intersect the Surfaces (so it is possible to select de edges without the corners)
- Create another 3DSketch, convert the Split Lines and draw tangent Splines connecting the lines (reset their handles) [1]
- Fill Surface with edge tangency and the 3DSketch of Splines [2]
- 6 Fill Surfaces with edge tangency in the corners [3]
- Knit
- Thicken

[1] Guess a point of improvement would be not using splines at all, what can be done here? I tried using Curve Through Reference Points, but the result was worse.
[2] Also tried Curvature on settings and uncheck Optimize surface. But I have some problems trying to compare different results, how would you suggest doing it?
[3] Tried different distances for those parallels planes (changing the Split Line position), but the result was kind of the same independent of those Filled Surfaces sizes

Later is where the trouble comes:
- Create an Assembly and start connecting the surfaces (using planes as reference Mates).
Somehow the tangency on the Edges are not really tangent and it is notable the singularity on the corners (suppressing the Thicken helps to analyse better).
- After creating a “cell” assembling 8 surfaces, create a New Part inside the assembly and Join it all together in one part.
- Combine to merge all the bodies.
*Somehow the corners get all knitted and neat here and I have no idea why. (Did not use “Force surface contact” in Join Feature - when I used it the waiting time was so long that I gave up.)
- After the unexpected result, pattern it twice to create a large volume with it.
*Notice that the corner will not be properly merging, but repeating the previous steps (Create New Part -> Join -> Combine) will fix it again. ???? lol


-> Does anyone know why the Combine feature is knitting the corners? This is the result I wanted but it should be already merged when connecting the bodies in the Assembly. Also, how can I make the geometry perfectly tangent, like a gyroid should be? Guess the main problem is in the Filled Surfaces in the corners

-> Any suggestions on the steps are welcome.

I exported many STL files and checked them in Materialise Magics, did not require any fixing for the structure (which was great ), but it is noticeable the lack of smoothness in the joints.

Pictures in SW and Magics with/without Combine Feature

Since some files are more them 50MB I hosted it on Drive: https://drive.google.com/drive/folders/ ... sp=sharing

[zwei]

Just some quick question...
1. Why do you use spline? Maybe i am missing something here, but isnt arc already sufficient?
2. When dont you use Pattern + Move/Copy bodies instead of creating a new assembly and then creating a new part again


Seem like an interesting surfacing exercise

[Lucas]

Just some quick question...
1. Why do you use spline? Maybe i am missing something here, but isnt arc already sufficient?
2. When dont you use Pattern + Move/Copy bodies instead of creating a new assembly and then creating a new part again


Seem like an interesting surfacing exercise

In this file I only used splines on the "Tangent Spline (3DSketch)" . It wans't possible to make a 3D tangent arc in both Split Lines for the contour; Since the split lines are in 3D the connector should bend to fit in both. Maybe there's a way to fit tangent arcs, without using those split lines...

About the Assembly, I was doing this method before when working with lattice. Somehow the software handles better heavy files in assembly mode (can also lightweight them) and it is also better to rotate the part to see it. Doing everything in one part file was taking a lot of processing time and patience. lol

[matt]

I wasn't able to download the files.....ok, I got them, 3rd party cookies...

[zwei]

I had tried playing with it...

1. You can actually create a surface using all the sketch on cube face using Fill surface while using the internal sketch as constraint curve.

2. Seem like it is not that easy to get the surface to transition smoothly at the joint...

image.png (42.17 KiB) Viewed 9869 times

As i spent more time than i expected, in the end i sort of give up and brute force my way through
What i did is just generate the gyroid as isosurface in matlab and export as STL in 5min

image.png (89.98 KiB) Viewed 9869 times

[Lucas]

I had tried playing with it...

1. You can actually create a surface using all the sketch on cube face using Fill surface while using the internal sketch as constraint curve.


2. Seem like it is not that easy to get the surface to transition smoothly at the joint...


As i spent more time than i expected, in the end i sort of give up and brute force my way through
What i did is just generate the gyroid as isosurface in matlab and export as STL in 5min

Have been there. Using constraints in the Filled Surface will always make them nasty haha

Is it possible to do it on Octave or other free software? Guess I could find a way to work with imported STL files if the mesh quality is good

[matt]

A couple of things:

SW hung up 3-4 times and my machine blue screened once working on this.

You can simplify things with symmetry. The main shape of the part is essentially the same shape 12 times.

I tried a few things:

1) each surface is made of 12 symmetrical bits, so I mirrored the 1st one and then circular patterned the rest.

image.png (33.73 KiB) Viewed 9853 times

image.png (36.58 KiB) Viewed 9853 times

image.png (44.51 KiB) Viewed 9853 times

2) I made a single fill feature straight from your arcs

3) You can also do it with a Boundary if you draw lines from the center of the cube to the corners, and use those lines as Dir1 and use the arcs as a closed loop in Dir2. It may not have the tangency properties you're looking for, but it's an interesting way to make a shape.

[MJuric]

So I played with this a bit and the way I approached it was to break the first unit down even further. You end up with two pieces that you can then make the entire thing out of that are simple lofts with four edges.

I started playing with this trying to figure out the complex surface using planes and. As I kept making the surface smaller and smaller units I ran into a spot where the surface only comprised of the four edges so I lofted that surface. I then realized that I should be able to have a mirror of that to complete the arc....then realized that was the only two pieces I needed.

Part one

End essentially it's mirror
Part two

I then created the first step assembly by using the two pieces in every other position.

image.png (67.34 KiB) Viewed 9844 times

Then made a cell from the previous assy

Then a block of cells

I'm sure it's not a perfect Gyroid because I just faked in the arcs but looks to me like you could make the arcs in the initial pieces anything you wanted as long as the end points were in the right place and tangent to the mating part or cell.

[Lucas]

Thank you for your answers.

@matt, I'm sorry, I might have saved some files with high resolutions in image quality, maybe it wont crash if you lower it. Usually I drop to the minimum when assembling then increase it after, so I can evaluate better.

About your methods, very interesting the first and the third one. Never thought it would be possible to do this with boundary, I will check it. lol

@MJuric, this is a nice approach. but SolidWorks only keep contact constraint with sketches, so the surface will lack smoothness in the edges. This method works well with NURBS

******************************
Some zebra results:

This is just with contact constraints (just need the arc sketch to do, and lacks tangency when assembled)

It is not possible to make all the edges tangent or curvature continuous, but it is possible to create leaving one with contact constraint. The shape is not good though:

With curvature and one contact:

Here I have tangency constraints in all edges:

With curvature it is not neat :/

Curvature in every edge:

I think there might be another way to make this contour without splines, so curvature continuous will look better, but I can't figure another way… have any suggestions?

[Lucas]

Improving it a little bit. Reduced the Filled Surfaces sizes almost to the minimum and the Thicken to to the maximum.
More than this the Filled Surfaces would start to bend and the Combine in the assembly would result in zero thickness. Also sharing the files:

[matt]

Improving it a little bit. Reduced the Filled Surfaces sizes almost to the minimum and the Thicken to to the maximum.
More than this the Filled Surfaces would start to bend and the Combine in the assembly would result in zero thickness. Also sharing the files:
image.png
image.png

Very cool models, and a fun challenge too.

Thanks for posting this!

[Lucas]

Very cool models, and a fun challenge too.

Thanks for posting this!

Sure, my pleasure. I got obsessed with this lol

Trying with Boundary now, but can only constraint half of it properly

[MJuric]

@MJuric, this is a nice approach. but SolidWorks only keep contact constraint with sketches, so the surface will lack smoothness in the edges. This method works well with NURBS

Not a surface guy so how does this smoothness effect the model? IE if you had a sphere that was a single surface and then split that sphere into two halves the zebra stripes would not show a continual line on the split version but aren't both version the same "Smoothness"?

[Lucas]

Not a surface guy so how does this smoothness effect the model? IE if you had a sphere that was a single surface and then split that sphere into two halves the zebra stripes would not show a continual line on the split version but aren't both version the same "Smoothness"?

It affects mostly the appearance. A gyroid is just a surface bended in many directions, it doesn't have any notable joints. Here is a more detailed explanation about curvature continuity:

https://www.augi.com/articles/detail/an ... continuity





Unfortunately looks like SolidWorks can only make G1 Gyroids natively, maybe with add-in G2 /G3 might be possible…

[MJuric]

It affects mostly the appearance.

Ok, that makes more sense. I was thinking it was an actual "Flaw" as is the case some times when there is a gap between edges of a surface.

[matt]

Not a surface guy so how does this smoothness effect the model? IE if you had a sphere that was a single surface and then split that sphere into two halves the zebra stripes would not show a continual line on the split version but aren't both version the same "Smoothness"?

That type of curvature control can also be important in strength, dynamics, fluid dynamics, cams, and all sorts of motion. Accelerating with a sharp in the path will lead to a shock, and the smoother it is, the more you can distribute the shock, making parts last longer, or for less turbulence in flow, or removing a stress concentration in a structure.

[MJuric]

That type of curvature control can also be important in strength, dynamics, fluid dynamics, cams, and all sorts of motion. Accelerating with a sharp in the path will lead to a shock, and the smoother it is, the more you can distribute the shock, making parts last longer, or for less turbulence in flow, or removing a stress concentration in a structure.

I'm aware of the effect of an actual lack of smoothness in the real world.

What I was wondering is what is the difference in SW when you have two surfaces vs one surface both of which are essentially the same smoothness.

When I drew my Gyroid, unless SW models it incorrectly, the smoothness of what I drew should be the same as that of a single surface. However if I look at the Zebra stripes it shows a "Break" at the joint areas.

In other words from what I can tell my model would be as smooth if I manufactured it as a single surface because all the joint areas are tangent...again unless SW breaks that because it can't create the single surface properly.

[matt]

In other words from what I can tell my model would be as smooth if I manufactured it as a single surface because all the joint areas are tangent...again unless SW breaks that because it can't create the single surface properly.

The settings that you use in a feature do not always really control the outcome. That's why we have the evaluation tools, especially the Deviation Analysis so we can tell numerically how far off from tangent a connection between faces really is. There is (almost) always a little tolerance between complex surfaces.

It's great to model the smallest possible surface, and then mirror/pattern bodies to get the full shape you need, but it's the connections between the bodies that causes the problem. So you avoid the problem by making larger bodies with fewer connections, and trying to make fool-proof connections. It's one of the hair-pulling circular conundrums in SW surfacing. The easy methods give you bad models, and the methods that give you good models are (seemingly) impossible.

This is why we sometimes make bad connections and then patch them.

https://episodes.dezignstuff.com/blog/

[zwei]

Is it possible to do it on Octave or other free software? Guess I could find a way to work with imported STL files if the mesh quality is good

Yes i believe it is possible, as long as you could generate the "points" for the gyroid, you could convert them into STL using some script.

[MJuric]

The settings that you use in a feature do not always really control the outcome. That's why we have the evaluation tools, especially the Deviation Analysis so we can tell numerically how far off from tangent a connection between faces really is. There is (almost) always a little tolerance between complex surfaces.

It's great to model the smallest possible surface, and then mirror/pattern bodies to get the full shape you need, but it's the connections between the bodies that causes the problem. So you avoid the problem by making larger bodies with fewer connections, and trying to make fool-proof connections. It's one of the hair-pulling circular conundrums in SW surfacing. The easy methods give you bad models, and the methods that give you good models are (seemingly) impossible.

This is why we sometimes make bad connections and then patch them.

https://episodes.dezignstuff.com/blog/

Ok, so what you're telling me is that SW creates surfaces that aren't really what they are supposed to be

I don't work with surfaces often and 99.9% of the time it's to fix an imported part where I'm just trying to get it to be a solid so I really don't care much about how it "smooth" it is.

That being said, what is the purpose of making a software that doesn't make surfaces as one would expect? I can see having options to manipulate the surface to make it so it is not as one would expect, but I would think the default would be "Ok the surface should start and end perfectly normal", which would mean mating joints would be what you sketched not "Well something we decided it should be".

Probably going into the weeds here but how does SW "Decide", "Nahhh, I'm not making this one right/tangent....oh but that one I am.".

Edit to add: BTW this is information that I wish I never had....now things like this will bother me when in the past I would have blissfully ignorantly have gone on my way thinking "Welp, those two surfaces are tangent...well because that's the way I drew them

[matt]

Ok, so what you're telling me is that SW creates surfaces that aren't really what they are supposed to be

Well, when you're lucky, they're close. For example, I consider a model like this to be very close:

and one like this to be kind of marginal

Obviously, "by the sketches and features" I made, these edges should all be perfect, but in practice it doesn't work that way. So half the trick of "surfacing" is to get the results you intend from tools that may or may not work that way today.

I don't work with surfaces often and 99.9% of the time it's to fix an imported part where I'm just trying to get it to be a solid so I really don't care much about how it "smooth" it is.

That being said, what is the purpose of making a software that doesn't make surfaces as one would expect?

The purpose is that SW costs ~$4k, and NX costs ~$20k. Eventually that difference will pay for itself when you have to keep wrestling stuff like this.

Probably going into the weeds here but how does SW "Decide", "Nahhh, I'm not making this one right/tangent....oh but that one I am.".

Edit to add: BTW this is information that I wish I never had....now things like this will bother me when in the past I would have blissfully ignorantly have gone on my way thinking "Welp, those two surfaces are tangent...well because that's the way I drew them

I don't know anyone who can really put that in words, but after decades of working with the stuff, you get an intuition. Some things will work fine, and others will stray. The more you expect it to do automatically, the worse the results will be. The harder you force it into a corner, the worse the results will be. You try to set up a situation where the right thing just happens naturally, which is a lot harder than it sounds.

[MJuric]

The purpose is that SW costs ~$4k, and NX costs ~$20k. Eventually that difference will pay for itself when you have to keep wrestling stuff like this.

I haven't used a higher end system in years. So does something like NX et al actually create surfaces as you would expect them to? If so I just can't imagine someone who does that kind of work sticking with SW. Mold makers, Die makers, plastic guys etc etc. Why on earth would you ever stick with a tool that essentially doesn't do what you want it to and should do?

I don't know anyone who can really put that in words, but after decades of working with the stuff, you get an intuition. Some things will work fine, and others will stray. The more you expect it to do automatically, the worse the results will be. The harder you force it into a corner, the worse the results will be. You try to set up a situation where the right thing just happens naturally, which is a lot harder than it sounds.

That's a painful learning curve when you start out of the gate with something that won't work.

Thanks

[Lucas]

3) You can also do it with a Boundary if you draw lines from the center of the cube to the corners, and use those lines as Dir1 and use the arcs as a closed loop in Dir2. It may not have the tangency properties you're looking for, but it's an interesting way to make a shape.

I am trying to make this model, but no matter what I cannot select all the 6 open groups. How did you do it?

[MJuric]

I am trying to make this model, but no matter what I cannot select all the 6 open groups. How did you do it?

Not answering your question as I'm still on the smoothness issue. If you manage to make this piece aren't you still going to have to piece it with a bunch of other pieces in order to make the Gyroid? Isn't that going to result in the same smoothness issues although fewer of them? Or is this a case of if you do this area it decides to do it right and a pieced together version is smooth but if you do this piece with six pieces it does it wrong and makes it not smooth?

The more I know the more I realize I don't know anything

[zwei]

Not answering your question as I'm still on the smoothness issue. If you manage to make this piece aren't you still going to have to piece it with a bunch of other pieces in order to make the Gyroid? Isn't that going to result in the same smoothness issues although fewer of them? Or is this a case of if you do this area it decides to do it right and a pieced together version is smooth but if you do this piece with six pieces it does it wrong and makes it not smooth?

The more I know the more I realize I don't know anything

Just my 2 cents, but I dont think there is way to create a PERFECT model in SOLIDWORKS for this kind of complex lattice minimal surface structure. You will have a little bit of imperfection here and there, especially at the joint.

To get a theoretically perfect model, using a mathematical model to generate a mesh might end up being more accurate.

I think in the end it depends on how much "imperfection" you can accept.

[MJuric]

Just my 2 cents, but I dont think there is way to create a PERFECT model in SOLIDWORKS for this kind of complex lattice minimal surface structure. You will have a little bit of imperfection here and there, especially at the joint.

To get a theoretically perfect model, using a mathematical model to generate a mesh might end up being more accurate.

I think in the end it depends on how much "imperfection" you can accept.

I agree, I'm just curious as to why Lucas's approach is better than doing it the way I was. Again not a surface guy so why would doing the complexity of one surface instead of a more simple single surface six times end up in a "Smoother" end model?

I have no idea what the answer is here which is why I'm curious.

I messed with my model and evaluated it and it off considerably. I'm just wondering if the other approach, for some reason, ends up with less error and or why.

[matt]

I am trying to make this model, but no matter what I cannot select all the 6 open groups. How did you do it?

You have to use the open and closed groups with the selection manager.

[Lucas]

@MJuric I am not surface guy too, actually I might be the least experienced guy in here. lol
I started studying surface modeling recently and this gyroid problem is teaching me the limitations of each tool, since I am checking almost every approach.

I wanted to check boundary surface with splines as influence curves just because I never used boundary surface like that. (the result would be worse since it is not possible to constraint it with the tangent edges)

Did with regular splines with vertical tangency and fit splines with tangency at the top plane:

As @Zhen-Wei Tee said, it is not possible to get a perfect model in SW (in Catia/NX neither), but some methods will get closer than others. I guess the best one for this problem is doing it with Filled Surface.
Got the solution from this guy here (he actually is a surface guy lol), he also compared results between Catia and NX and explained them, if you want to check it:



NX has two tools that SW doesn't, the Bridge Curve (which I used splines instead), and Studio Surface. But the result is close enough

You have to use the open and closed groups with the selection manager.

Yeah, I am trying that, but when I select the fifth line the software doesn't allow me to get the sixth... Anyway, I just made it in separate sketches, maybe there was some constraint causing this problem...

[zxys001]

Nice to meet you guys, first post here. Hope it's in the right place.
..good work!

,... keep simplifying.
Although, it's fun to create,.. I agree with some comments,.. this could be created using a formula (many mesh versions of it).

[Lucas]

It's a different shape, but looks neat.
Mind sharing the files?

[MJuric]

@MJuric I am not surface guy too, actually I might be the least experienced guy in here. lol

I think like @matt said there's probably a steep learning curve in there. I'm sure that one way or the other will give you better results but apparently the only way to know which way, at least with SW, is experience.

Maybe not surprisingly that is not much different than anything else in SW and for the most part one of the more frustrating aspect of the software. Just because you CAN do something in SW does not mean you SHOULD. This is a bit different, and certainly more normal in SW, than other systems I have used where there seem to be far greater limitations on "Doing something you shouldn't"

It can take years of using the software before you can just look at a model and say "Well you could model that 10 different ways. Five of them will end up with an unstable model that will eventually explode and only one or two will actually be stable and fast."

In other systems I've used it would be more like "Well you could model that in four different ways, only one will explode, two are not the best but still good.

[zxys001]

It's a different shape, but looks neat.
Mind sharing the files?

Oh, details... yeah, it's different.
I lost the version you have.
And, yes, it was a challenge and I also did not get a good zebra from either.

[matt]

I am trying to make this model, but no matter what I cannot select all the 6 open groups. How did you do it?

I made a little movie to show it. I used the multiple arrow option on the selection manager.

[Lucas]

I made a little movie to show it. I used the multiple arrow option on the selection manager.

Wow, I see now! Was trying with Open Loop, and when I tried with Select Group wasn't clicking OK after selecting every line. Thanks for the video!

Oh, details... yeah, it's different.
I lost the version you have.
And, yes, it was a challenge and I also did not get a good zebra from either.

No problemo, thanks

I think like @matt said there's probably a steep learning curve in there. I'm sure that one way or the other will give you better results but apparently the only way to know which way, at least with SW, is experience.

Yeah, I agree too! Right now just found another trouble trying Rib Feature on cylindrical surfaces

[zwei]

Yeah, I agree too! Right now just found another trouble trying Rib Feature on cylindrical surfaces

Off topic
Well... this is because if you use rib on cylindrical surface it will cause ZTG

There are lots of workaround, here are some of them:
1. Change your rib profile to "protrude" in your cylinder
2. Extrude a square → rib → cut away the square to form cylinder
3. Use a normal extrude instead of rib, and apply a DxD chamfer

Personally i prefer to stick to "basic" features whenever possible, there are a few features in SOLIDWORKS that is useful in paper but often cause issue if not use properly.

[Lucas]

Off topic

Posted here, it is not ZTG :

viewtopic.php?f=16&t=740