really poor performance
If you are interested in either helping Fusion improve, or some suggestions as to how to make your particular workflow better, please share your design. Otherwise, there is little we can do to help.
If you want to help us analyze Fusion performance, the design + the particular actions on that design that are slow is the best way to do that. Performance is something we want to focus on in the near future, so the more sample designs and workflows we can collect, the better.
Jeff Strater
Engineering Director
If you're having performance problems, then you have a workflow problem😃 By the way Autodesk will sign an NDA so you don't have to worry about your proprietary info.
ETFrench
@jeff_strater I helped @RScott9399 offline to overcome some of the hurdles in this project, but I did warn that there are more stumbling blocks. I think I can shed some light on the problems he's facing.
Robert has a triangulated mesh model of a car. The model was initially conceived in a CAD, or NURBS surface modeling software and contains each of the external panels of the model. As is usual these are are individual meshes separated by a panel gap. Also as it was initially a surface and not a solid model the meshes don't enclose a volume watertight.
The aim is to create one or more solid models from these meshes. Those would then be sliced to create toolpaths for later machining out of modeling foam.
There are two ways to achieve that.
1. recreate a true CAD model by remodeling each individual panel using surfacing techniques. This would require substantial surfacing skills and at least a couple of hundred hours.
2. Remesh the model into a quad mesh. This is the path Robert has chosen. He used a Fusion 360 plugin for this and having seen the quad mesh, I can say that it works indeed very well, however ....
The problem with the 2nd approach is that in order to retain a certain level of detail a very dense mesh has to be created. Real CAD models (NURBS & BRep) are mathematically precise models that don't have a finite resolution and have topology (in a mathematical sense). When they are converted into triangular meshes a substantial amount of information is lost. The most substantial one is the topology. A triangulated mesh has no topology, so re-meshes are limited in what they can recognize as a feature. that is particularly true for more "organic" geometry as the flowing lines of a vehicle body.
As such, a dense mesh is needed to retain a certain amount of detail. If I am not mistaken each T-Spline quad face represents a 3-degree NURBS "patch" and has a 4x4 grid of NURBS control points. My guesstimate is that the density of NURBS control points in Robert's NURBS models resulting from T-Spline conversion is at least factor 200 larger than if this geometry was designed with normal surfacing tools.
Of course thickening, the infinitely thin surface in the T-PSLine environment is then going to double the amount of data needed. It is no surprise to me that the result is an incredibly sluggish model.
@RScott9399 modeling actions in any current CAD system including Solid Works, Autodesk Inventor are handled by the geometric modeling kernel and are mostly single core! So for the most part, most of the multi-core architecture of your Mac Pro is sitting idle when modeling. A $1200 computer with better single-core performance, meaning a high base frequency and higher boost frequency is probably going to run faster your Mac Pro machine in terms of CAD modeling performance.
Also, one reply to one of the comments you made in one of your other threads. You mentioned that in Alias the surface of the mesh model looks very smooth and beautiful. It also looks smooth and beautiful in Blender. All 3D applications nowadays use Phong shading if told so. It is a mathematical trick (averaging of surface normals) to fool your eye into believing there is a smooth surface.
All 3D software has to tesselate (triangulate) geometry in order to display it on a computer screen. So all sorts of shading tricks are used to make the model look smooth.
The Workflow in this case is to a good degree pre-determined by the data that is the input to the design process.
I forgot to mention a third way to get from a triangulated mesh to a CAD/NURBS model. That is manual re-topology. This also requires considerable skill and is not usually used to re-create CAD models with technical surfaces.
I disagree. You won't have a solid workflow until you have your first parts machined!
Just out of curiosity, if you export your design to your hard drive as a .f3d file, what's the file size?
The technologies for re-meshing and converting quad meshes into CAD surfaces have only appeared in CAD software over the last 10 years. Fusion 360 users can consider themselves lucky that this is even available at such a low price point.
I personally would not approach this project with the workflow or the software you have chosen. Reverse engineering technical surfaces from triangulated meshes or point clouds, if one wants to achieve a high-quality cad surface, requires specialized tools and in this case a good understanding of surface modeling, or Sub-D mesh modeling, or even both.
In either case, it requires a lot of time!
A specialized tool for this sort of work that can create CAD/NURBS surfaces directly from point clouds or triangulated meshes is Geomagic Design-X for 3D systems. The Artec website lists this for $19,000 a license!
A professional tool for specialized applications with a price tag to match. While I do own a license for another CAD tool (ZW3D, approximately $3000 for a perpetual license) Design-X is completely out of my price range.
Another approach where you would also end up with a quad mesh, albeit at a much lower mesh density, could be a manual retopology workflow (Topogun, or Blender's Retopo-Flow plugin). This requires a lot of mesh modeling practice and understanding to do this well and not lose important details. While I have used this on an A-Pillar I personally don't have enough experience to use this for an entire vehicle.
As such your project is limited by the quality of the data you are working with, your budget, and your knowledge. Perhaps the time you can spend is also limited?
This isn't anyone's fault.
P.S. I have already mentioned that the dense mesh created by remeshing is created a very heavy CAD model. That model will slow down any CAD model. However, it will definitely have an effect on FUoisn 360 when saving or loading the model Fuiosn 360 stores data in the cloud, and loading and saving are generally slower than doing that locally.
@Hfrossard wrote:
I do agree that Fusion 360 is not nearly as good as Inventor or SolidWorks.
It has a LOT of potential, but Autodesk needs to stop foccusing on niche or new features (like generative design or Cam simulation) and make basic features better and reliable.
I also agree that It is sad that they always ask for our designs. The most complex designs are made for companies that cannot share designs like this.
My recommendation would be that before making a post, to actually read the thread and then make a decision whether or not the post has anything to contribute to the thread.
I've worked with SolidWorks since 1998 (and still do) and I know enough about Inventor to be dangerous. As should be evident by now, I understand the difficulties of this project on a much deeper level than the OP and can assure you it would be almost as difficult to achieve in SolidWorks or Inventor.
I find it perfectly OK for Autodesk or for the folks here on the forum to ask a user for Data. I also find it perfectly OK for most users to say "No, sorry, I can not share because it is proprietary.
I find it incredibly unprofessional and outright disturbing when users completely lose their s**t because they've been asked to share their designs. A simple "no" is all that is needed.
If a user does work with proprietary data, and that is usually clear right from the start, then it is the user's responsibility, to make sure they have the skill set to deal with problems.
I know that paints a bit of a black-&-white picture but across the last 18,000+ post and 2000+ solutions on his forum, I think I've come across enough users who did not have the needed skills.
@Hfrossard wrote:
Why take things personally just because you disagree with the opinion of others?
The post started stating the difficulties with Fusion 360, and I just want to joint OP with his opinion, becaus I have the same experience.
You are mistaken when you assume I take this personally.
I prefer to have a nuanced discussion around issues. That requires a bit of research and homework. It isn't too difficult to go through another user's previous post to see what they are posting and what problems they may encounter.
I would encourage you to go through @RScott9399's posts and see what his project entails and then compare that with the areas in Fusion 360 you are encountering difficulties with and see if there is common ground. And then, again, come to a rational decision whether or not your post contributes something to the thread.
I think it is perfectly fine to report performance problems!
This did not surprise me, but what I was trying to convey was that it isn't anyone's fault.
It is simply the result of a process.
Your warning to other users not to use Fusion 360 is completely warranted, but it is worthless information if it isn't explained what "large" means.
I also warn people not to use Fusion 360 for "large" projects, but that is a completely different "large" then your project, and make no mistake, your project is "large".
Professionally I work for a midsized company that designs and builds turn-key high-end factory automation systems. A current concept model for a system that is still fluent to work on and easy to navigate has over 4000 components. This is either on my work laptop or on my 2017 MacBook Pro. And that is in Fusion 360.
Yet, I can not recommend Fusion 360 to our mechanical design department, for the actual production design. A parametric production design would easily triple the component count and Fusion 360 lacks a lot of functionality in other areas to be able to compete with SolidWorks in that area of machine design. I was never designed for that either!
"Large" in your project is not determined by component count. The large size of your model(s) is simply is a result of the high-density quad meshes that create very data heavy NURBS surfaces. If you watch this guy model cars in alias (he's a pro and works for well know name brand design studios) and compare the density of NURBS control points in his models with one of your models you'll see what I mean.
@Hfrossard - Thank you for the encouraging words about Fusion's approach. "The thing is that I think Autodesk is failing to see we are having this basic issues". I assure you, we do see this. Your points are all valid. There are certainly workflows where Fusion becomes unproductive to use, once a model reaches a certain size. We are very interested in those workflows, because we very much would like to fix those. However, we need to know more about your process and your model so that we can do that. If you are willing to share some more information about process/model, that would help us in our goal to improve Fusion performance.
Second, if we are able to find out more about your process and model, it is possible that we can help you even now. There are some workflows that are known to result in poor performance (for example, large sketch patterns, large feature patterns set to "Adjust", shelling complex geometry, etc). For larger designs, deciding what is an internal component vs an external one also can improve performance. If we know a bit more, it is possible that people on this forum can recommend alternate workflows that will make life more pleasant while we try to improve Fusion. But, without that information, all we can do is speculate.
Sometimes, asking to share designs can touch a sore spot. Point heard loud and clear. You obviously have the right to say no, but I think an all-caps response to what I think is a reasonable question is, itself, also somewhat unreasonable. Hopefully everyone can understand how crucial that information is to both helping us improve Fusion, but also in helping you optimize your processes. Saying "any medium-sized design is unusable in Fusion", without more context, is just not enough. I know I can create what I think is a "medium-sized" design and have it be very efficient. But, my definition of "medium" and yours might be different, and the performance with different kinds of geometry will also be different. So, all we are asking for is more information...
Jeff Strater
Engineering Director
@Hfrossard wrote:
I will open a new thread for the complaints I have in order not to
overpopulate this one.
Yes, please!
