Convert Body into a T Spline

[rwillardphil]

I would like to generate the basic shape with normal modeling tools, and THEN convert into a T-spline to manipulate and sculpt as needed.  Is that possible?

[Anonymous]

Hi everyone,

It's July 2020 today, so the topic is quite old now .. 😛
But my problem looks basically similar to the the topic's creator one.

As far as what I read of the discussion, the problem doesn't look solved, or please tell why I'm stupid !

Here's what I want to do: 
1) Create a SOLID shape
2) Edit the top extrusion with T-Splines controls

Disabling history recording, and converting the selected face into a T-Spline element doesn't help.
The problem is exactly what was reported by the first guy: the created T-SPLINE face is squared !
Whereas what I want is that it has the outline of the face I want to edit.

Let me explain with images:

The goal I want to reach. (Photoshop image I created).

I start with a solid shape (because if I start with a T-SPLINE shape first, by extruding my outline sketch, it DOESN'T RESPECT the inital outline I just leaned on for the extrusion)

Then I would like to do is to EDIT the top face with T-SPLINES controls.
So I turned off history recording, and converted the top face in MODIFY>EDIT FACE.
Here is what I got:

Whereas here is what I wanted to get (photoshop I did to explain my sayings):

----> I need the red outline to be unchanged, and play with the top face, to give it some roundish shape.
----> I also tried to trim the created T-SPLINE to the selected sketch, but I don't find how to ...

Until now, I've been working with Autodesk ALIAS AUTOMOTIVE, which is awesome ! I could have done it with it, but I need a software that allows me to do SOLID/SURFACE together.
Fusion 360 looks really great, but I still need tips from experienced guys to achieve my goal with it 😉

THANKS FOR ANY HELP ! 😄

Clément
Industrial Designer

 

[TrippyLighting]

You can work with surfaces and solids, but surfaces are not to be mistaken for T-Splines.

You can create surfaces from curves, just as you can in alias with the tools in the surface tab and I would believe that is the better approach for this object.

T-Splines are more similar to how you would model with Sub-D surfaces. Alias since the 2020 release now provides the ability to Sub-D model, but again that is a very different animal.

A T-Spline is an intermediary between a quad mesh and a NURBS surface with natural boundaries, so in essence an untrimmed NURBS surface. Curved surfaces with a quad layout usually convert fine into a T-spline as the topologies involved are similar. However, trimmed NURBS surfaces usually convert into rectangular T-Splines patches because of the apparent topology mismatch. 

If you are familiar with Sub-D modeling I am sure you'll agree that no one experienced in Sub-D modeling would attempt to even approximate the top face of that button with a rectangular grid topology.

[Anonymous]

Thanks for your reply !

Actually, I'm not familiar with Sub-D surfaces, that's why I moved to Fusion 360: to have a try at this "in-between" method.

Also, please excuse me for the terms I use, I may used "T-Spline" or "Sub-D" wrong.
Here is a pic of another design of mine, which I assume is SUB-D modeling, right ? That's what I called T-SPLINE in my previous message.

 
In fine, what you recommand is to use the surface method, is that correct ?
Isn't there any way to create a surface (with a determined outline by a sketch) that I could edit with controls like those on the picture above ?

I guess the only way to do it with Sub-D modeling is to create faces manually, trying to fit the outline as precisely as I can.

[Anonymous]

I was also trying to figure out this simple workflow that seems to work with most modeling software. If Fusion 360 allows to blend solids, surfaces and Tsplines, and that is probably why it is called "fusion" it should offer a simplified workflow that goes from polygonal solids to Tspline without having the needs of converting to a mesh (exporting to Recap because you need quad faces) and then convert in tspline. Also a re-topology simple tool would help in that sense. with these 2 little buttons you saved a lot of time for someone that want to model a poligonal shape and then want to smooth it using Tspline.

[TrippyLighting]

---

@Anonymous wrote:

I was also trying to figure out this simple workflow that seems to work with most modeling software. 

---

Can you elaborate on what you mean with "simple workflow"?

To my knowledge, there is not a single 3D software out there that converts a polygonal solid consisting mostly of trimmed NURBS surfaces into vertex-aligned quad-mesh.

Definitely not on the push of a single button.

Re-topology is used to convert triangulated meshes into quad meshes. Manual re-topology that is! Re-meshing is NOT retopology. Manual re-topology requires considerable skill and also is not possible on the push of a simple button.

Then if you do use a triangulated mesh as an intermediary to convert a CAD model (BRep, trimmed NURBS) into a quad-mesh, you lose vital data. CAD objects are comprised of mathematically precise, resolution free surfaces, stitched into a BRep in the case of solid models.

A triangulated mesh e.g. in form of a .stl has a finite resolution, and the topology information is lost. Neither the topology nor the mathematical precision can be fully regained.

[therealsamchaney]

The short answer is no, surprisingly and disappointingly, Fusion 360 does not have a reliable way to convert from a solid body to a T-Spline. This is really disappointing to me since the whole point of Fusion 360 is that it's a fusion of different modeling environments like parametric Solid (BRep) and Form (T-Splines). Yet, there is no good way to convert a body from one to the other. 

The long answer is that there is a workaround that sort of works, though I would say the quality of what you end up with isn't worth the effort. To do this, first save the body as an STL (which is a mesh). I would set the refinement to low so you don't end up with too many triangles.

Here's the annoying part. T-splines require quad (rectangle) based meshes, but Fusion 360 can only convert BReps to tri (triangle) meshes and can only create STL files with tri meshes. So now you have to convert this STL we made into a quad mesh which requires another piece of software. I would recommend Blender because its free. Here is a good tutorial of using Blender to convert to a good quality quad mesh.
https://www.youtube.com/watch?v=l88hQj16IhA

You could also use Autodesk's ReCap Photo but I honestly can't even figure out how to download that unless you purchase ReCap Pro which is $40 a month. Not worth it for just that functionality in my opinion.

Lastly, you need to import this mesh into Fusion by clicking on Insert > Insert Mesh. Then, go to the Solid environment tab, and click Modify > Convert, then select Quad Mesh to T-Splines from the drop down, and you're finally done! 


The process is described towards the end of this video.

Again, I really think all of this should just be natively handled within Fusion. Autodesk certainly has the codebase to convert a BRep to a quad mesh, and then convert that into T-Splines. 

Anyway, hope this was helpful.
-Sam

[davebYYPCU]

Normal tools can get it done...

Might help....

[therealsamchaney]

Which tools? The only built-in tool that I know of that is close is Convert > BRep Face To T-Spline. This doesn't convert an entire body though, it converts a single face. If you select all of the faces that make up a solid body, you will end up with a bunch of separate T-Spline surfaces and some very wonky results.

If you have a good workflow for converting a somewhat complex solid body into a single T-Spline form (not a bunch of separate T-Spline surfaces), I would love to see a screencast or a step by step description.

I opened your f3d file and I don't see how that is supposed to answer the question. There are no T-Splines in this design, and the timeline looks like this is just a solid BRep body made using BRep surfaces, all solid modeling environment. T-Splines requre turning off the timeline so even if you used a T-Spline at some point, there would be no way to know, and no way for others to see your process because the timeline before it would be deleted.

[TrippyLighting]

The reason that there is  no good way is because the two types of geometry simply not backward-forward compatible.

Once you are fluent in Sub-D / T-Spline modeling and CAD(BRep/NURBS) modeling you’ll understand why at least intuitively.

Trimmed NURBS surfaces, which is what the majority of CAD models consist of, cannot simply be converted into T-Splines. When a T-Spline is converted into a NURBS surface, these are NURBS patches with un-trimmed boundaries.

Prismatic CAD geometry does not lend itself well to reverse Engnieering  by using tools such as Instant Meshes. At least not when precision is required. If you’ve done that a few times you’ll know why.

[therealsamchaney]

I know that it's never going to be a perfect 1 to 1 conversion back and forth, but it should be more than possible to get close enough to make a usable T-Spline form from a BRep. I know enough about boundary representations, B-splines and the like to understand the limitations with programmatically converting them to a mesh, but it's possible to get close enough for many uses.

Fusion 360 can already convert it's BReps to meshes, and it does a passable job. The only issue is that it always uses tri meshes. But, we know that Autodesk already makes products like ReCap Photo which can convert tri meshes into quad meshes, and we know that Fusion can convert quad meshes to T-Splines. So, it follows that Autodesk should be able to add a simple BRep to T-Spline conversion tool. It might not be perfect but with a few adjustment sliders, it should be passable.

Look at the Export As STL function. That is converting a BRep to a mesh, and they are very close to the original body, good enough for renders and 3D printing on SLA machines with no visible defects in the mesh. If it just converted to a quad-based mesh instead, it could then convert that to a T-Spline. I don't see where there issue lies. Are you saying it's that much harder to conver to a quad mesh than a tri mesh?

It's somewhat like an analog to digital conversion. Of course you're going to get some losses in accuracy but with the right algorithms and settings you can still get a pretty good end result.

[davebYYPCU]

The example knob request was looking for / discussing a top section as a one step t-spline, similar to replace face tutorials, and it's noted that the function is not there.

I made the knob supplied  - with std Fusion tools, not professing that it is a one step method, nor a T-spline.

Was done to allow / show some thinking outside the square.  

You may find that Fusion Generative Design outputs meshes editable in the T-Spline environment, but have no experience with GD.

Might help....

[TrippyLighting]

I've looked through some of your other posts and it seems you have a lot of suggestions about what AD should implement. Can you show what you work on and where the lack of functionality you describe in this thread hinders you in your work?

[therealsamchaney]

Yes I do have a lot of suggestions for improvements to Fusion 360 because I use it almost every day, usually for hours at a time and have found areas where I think it could be improved. I make these suggestions because I believe in the product and want to see it function to its highest potential, and because I'm very invested in it and the community surrounding it. Since I am a programmer and a mechanical engineer, I believe that I have a good foundation to make reasonable and valuable suggestions.

I also post a lot of questions, bug reports and answers. I act as an official "Fusion 360 Mentor" in the Facebook group and try to help out there whenever I can. I also post many of my designs on my Thingiverse, and contrary to many others, I include my f3d files because most of my designs are fully parametric, and have intricate relationships between the parameters which allow for fast and easy customization.

In terms of how converting a solid body to a T-spline would be beneficial to me, I am currently working on modeling a shakuhachi (a type of Japanese bamboo flute). These flutes are very difficult to find in the US and are very expensive so I would like to make highly playable and beautiful 3D printable versions.

I would like it to be a flexible design to be able to generate various flutes of varying lengths, tunings, and thicknesses. So, I am using parametric design to generate a single bamboo joint component which can then be copied and scaled to make a length of bamboo for the final flute. Parametric design is great for this as I can simply modify a few parameters and generate different bamboo sections with different properties.

However, the design is representing a natural material which has subtle but important variations. T-Splines would be great for making these minor adjustments, pulling a bit here, and twisting a bit there.

I could make this completely in the solid modeling environment, but then it would be very difficult to get the natural variations. I could use the random number function (and I plan on trying that too) but still, it won't give the fine control that T-Splines do to give it the natural bamboo look and feel. 

Conversely, I could model it completely in T-Splines, but then I would lose out on the power of the parametric modeling and would have to make each different flute basically from scratch.

So here I hope you can see that a great workflow would be to design the brunt of the flute in the solid modeling environment using parameters, then convert to T-Splines at the end for the final finishing touches.

Here is a current file showing the parametrically designed bamboo section. This is what I would currently like to convert into a T-Spline. This is a work in progress but you can see how I can use the parameters to easily change the geometry to get close to the natural bamboo shape, but that I would need to tweak it a bit further to get it to be fully accurate.

[TrippyLighting]

I am a Geman mechatronics engineer with 3 decades of professional experience in the field. 

I got interested in computer graphics during university and have programmed computer graphics algorithms down to  x86 assembler level.  Today I do some embedded programming.

I've used CAD and 3D modeling for over 30 years professionally also since college. I've used 3D CAD since 1998 (Solid Works). Today I use mostly Fusion 360 and ZW3D but infrequently also use Inventor, Solid Works and Autodesk Alias. I have used the 3D modeling Suite Blender for 18 years and played around with a number of other applications e.g.Houdini, which is why I am very familiar with Sub-D modeling and T-Splines. Advanced surfacing techniques are still developing (Alias is quite a beast to learn)

As it pertains to community involvement I've got 16,000+ posts on this forum over 1900 posted solutions 800+ screencasts and who knows how many bugs I've reported. I've been part of the now-defunct Fusion 360 customer advisory board and have had the opportunity to meet a number of the Fusion 360 team members in person during the meetings and during Autodesk University where I thought a Fusion 360 class in conjunction with @jeff_strater. I have used Fusion 360 since November 2014. 

I have no problem with you pointing out areas of improvement, I do that frequently myself.

What I have a problem with is that your partial knowledge leads you to believe that some of these things are easy to implement and that could not be further from the truth.

I would also suggest that the Fusion 360 Facebook group is hardly the clientele I would judge by whether or not a particular functionality is of benefit to professional  Fusion 360 users. I can assure you it isn't.

[therealsamchaney]

I would like to say that the only reason I listed my credentials was because it seemed you were calling into question the validity of my recommendations so I wanted to back them up.

As for your credentials, that's all well and good; nobody was calling your credentials into question.

You say that you only take issue with me stating that this would be easy, but I never said that and I'm not insinuating it. Of course the conversion between a solid BRep body and a T-Spline would be a difficult algorithm to crack, but my point is that Autodesk already has the codebase and tooling required to make it possible:

1. Fusion can convert BRep bodies to STL which are triangular meshes
2. ReCap Photo can convert triangular meshes to quad meshes
3. Fusion can convert quad meshes to T-Splines

Of course, the isn't the most efficient or effective workflow for doing this conversion, because both BReps and T-splines are mathematical boundary representations of a solid body, not meshes (so there's not really a need to convert to a mesh first), but it would still work. I know that the T-spline you'd end up with using these steps would be sub-par and probably contain mixed polygons, n-gons and the like.

The point isn't that I think Autodesk should actually implement the conversion this way, but to show that they already have the code tools needed to make a conversion toll that would work well.

There are also already free open source auto re-topology tools like Instant Meshes which you can see in this video. Instant meshes uses an approach of a mostly automated algorithm-based remeshing along with a little bit of intuitive manual manipulation which I think would also be a great approach for converting from a BRep solid body to a T-Spline. 

Long story short, I wasn't saying this would be easy, I'm saying Autodesk has already put the time and energy into developing the codebase and tools that would be required to make it work, and I personally think it would be worth it.

[TrippyLighting]

---

@therealsamchaney wrote:

Long story short, I wasn't saying this would be easy, I'm saying Autodesk has already put the time and energy into developing the codebase and tools that would be required to make it work, and I personally think it would be worth it.

---

That is a fair statement that I would support!

Not only is the codebase, in areas of Fuoisn 360 that conversion is already done!

The Generative design tools use relatively simple BRep input geometry and convert those it o volumetric meshes for the purpose of FEM. The results of the generative design algorithms are then surface meshes.  Those get automatically re-meshed and converted into T-Splines and then again into. B-Rep.

However, the geometry created by the generative design algorithms lends itself very well to re-meshing because of its very organic/flowing nature.

As I understand it there is work being done to make these re-meshing tools available as tools in Fusion 360. 

I am indeed very familiar with Instant Meshes and have conducted many experiments with it.  I created a YouTube tutorial 2 years ago that shows workflows and geometry where the conversion from a mesh into a BRep makes sense.

[Intuos5]

If I may butt in on this, 3ds Max' remeshing modifier has been in development for several years and publicly launched this year. Considering that these developers work on a mesh based program and the amount of effort it takes to create an automatic remesher, this isn't your average small feature enhancement. There are various remeshing algorithms out there (instant meshes, Zremesher, Exoside's quad remesher plugin for Blender/ 3ds Max from the folks who made Zremesher) dynamesh, to be fair 3ds Max remesher contains 3 different algorithms for specific use cases). Some offer topological control by allowing the user to draw the flow of (mesh) edges. That said, remeshing changes a surface just like tracing a pixelated image changes the image's outline. At the moment your best workaround is probably to export a mesh and use a mesh modeler to either manually retopologise the tesselated base mesh, the export, or use any of the aforementioned remeshing solutions available to you. To then import back into Fusion.

Anyways, the Fusion product team may be able to leverage the work done by the 3ds Max devs, but don't expect this to be integrated anytime soon.

P. S. Autodesk's Vred seems to have an amazing Nurbs surface tesselation engine, however that's only available for a modest price almost non can afford.

[christian.munn]

Hi everyone

   Just checking in to see if Fusion has made a simple(r) solution to the OP's original question in the past 2 years?  I've found myself in the same situation as user clement.cham, where I have made my model, but now I want to add some contour to the model to bring it in to final form using the sculpt funcion. 

    I read the previous debate about all the conversion between Brep, T-splines, tri-mesh, quad-mesh, NURBS, SUB-D etc, etc, etc doesn't leave you with a super accurate model. And I'm on board with user therealsamchaney when he says that finite accuracy isn't always that important. All I'm wanting to do is contour a fish shape that I can 3d print to use as a mold for pouring soft plastic fishing lures. I don't need 1 micron accuracy for that. 

    My dilemma is that I can easily import a canvas of a fish, trace the image for a nice fish shape, then extrude the sketch for dimension. That part's as easy as breathing! You know what's not so easy? Using the form environment to get the same fish shape while sculpting, at least for me...  I don't know how many hours of life I've currently wasted, or how many times I've jumped up from my desk while cussing in anger while trying to get this fish form, but it's way too much! 

   I've used the solutions that were shown in previous posts in this thread, and was met with the same problems as the other users. Yes I can achieve weird T-spline rectangle faces,  but manipulating those are even harder than trying to form the fish shape from scratch. I get that developing the code to make the BRep to T-spline and quad-mesh work isn't easy, but intuitively I feel like I should be able to to take a solid model and sculpt it, somehow. I don't know how, I'm not a software engineer, but in my mind it should be a thing. 

    In any case, I suppose I'll try that Blender work-around to see if I can achieve what I'm trying to do here. Or, if anything's changed in the past few years since the OP, I'd love to know. Thanks!

[christian.munn]

Sorry!

    It wasn't Blender, it was was InstantMesh. I was able to export the solid model from Fusion as an .OBJ, then open it in IM. From there I followed the steps in the video provided by @therealsamchaney to turn it into a quad-mesh. I then brought that new post-conversion  .OBJ file back into Fusion, and in the solid workspace>Modify>Convert>Quad-Mesh to T-Spline>OK. 

    The conversion really broke the model up into an unmanageable amount of boxes, but I'll play with IM some more to see if I can get it to produce something with less resolution. But, for now, it was a valid work-around to get something that I can now sculpt in the Form workspace. Thanks @therealsamchaney for the tip!

[TrippyLighting]

---

@christian.munn wrote:

...

    I read the previous debate about all the conversion between Brep, T-splines, tri-mesh, quad-mesh, NURBS, SUB-D etc, etc, etc doesn't leave you with a super accurate model. And I'm on board with user therealsamchaney when he says that finite accuracy isn't always that important. All I'm wanting to do is contour a fish shape that I can 3d print to use as a mold for pouring soft plastic fishing lures. I don't need 1

---

Agreed!

---

@christian.munn wrote:

...

    My dilemma is that I can easily import a canvas of a fish, trace the image for a nice fish shape, then extrude the sketch for dimension. That part's as easy as breathing! You know what's not so easy? Using the form environment to get the same fish shape while sculpting, at least for me...  I

---

If you post a fish canvas you wan to model, I'll try to show you how. With T-Splines!

For technical/mathematical reasons a solid body cannot be simply converted into a T-Spline