Another aspiring fishing lure designer :)

Hi All,

Another new user trialling Rhino. Sorry if you all attend to these regularly - but I’d greatly appreciate some guidance here.

1st up - long time cadd user for more boring civil design duties… I’ll possibly have some future questions related to my work and how Rhino can assist, but for now: Hobby / Passion project → Fishing Lure Design.

Ultimately, my goal is:
→ Design hollow carefully engineered lures with lovely visual exteriors and carefully engineered interiors
→ Print prototypes on 3D printer to refine and test and balance
→ Ultimately run production via 3rd party plastics company using ABS or PC plastic via moulds or any other high productivity processes(?)

My current design workflow has been in Blender, using very high SubD and mesh density and then displacement and UV maps to use dispalcement greyscale pngs (created in inkscape) to displace the shaping of a lure exterior. It looks stunning (in my opinion) and is such a nice non-destructive workflow for me (create blank; subD down to ridiculous density ~2M polys; want to update or tweak push / pull?.. just update vector map and it adjusts for me without tweaking base mesh…). Results look like this:


But now, with Google AI as my mildly competent but wildly scattered tutor, I’ve been told: "hey this looks great, just don’t forget you can’t actually use that to create production moulds from a dense mesh like this → it needs to be nurbs or another mathematic geometry for CNC to mill out the moulds…”… This is months after starting this journey lol… With Google AI steering me down this rabbit hole of displacement which I now think I love… Grr…

Anyway, I’ve been trying to integrate this high density mesh or displacement workflow into rhino as my “engineering platform” ffor creating the int ernals and actual shells / halves, but yes: the massive 2M poly mesh does not play well; and using a mesh in rhino is not it’s strength.

I’ve tried all blender options I can find to remesh or simplify and all are no good so far. I’ve tried rhino’s quad remesh (at e.g. 20,000 which is high but significantly less than the e.g. 2M poly original):

It’s honestly pretty good… but it’s falling apart (e.g. side fins lose detail), unpredictable (some geometry jumps around incorrectly) and not a good workflow in my opinion for those reasons…

So…

QUESTIONS:
a) Is this true… a 2M poly mesh is unmillable with CNC or not suited to high production moulding? (e.g. Must be nurbs or sudD or another math / curve system)? E.g. if I just stayed in blender with crazy poly counts, is the result likely to be unusable for production moulds and therefore this is a broken design pathway I need to abandon…?

b) Is my intended workflow of 2M poly Blender object (displace / shape / model) → Rhino (split, engineer internals, create final lure halves for moulds and printing) salvageable?

c) is there some kind of practical nurbs alternative for using a displacement mesh or vectors in a non destructive way where I can expect this level of detail (as per original blender images above)?

d)…is it really that: if I want to use rhino and trial it properly, I should probably just design lures from scratch in rhino and use its strengths → nurbs etc… and just abandon the vector displacement approach altogether for this trial, and approach design with the aesthetic capabilities of rhino as the goal, not with the approach of “how can I displace using grayscale maps with high detail in rhino like I want to”… because it’s not really the design pathway rhino is good at?

I honestly greatly appreciate any insight or assistance. I’m about to explore d) above, but want to make sure I’m not steering down a long new journey when I can actually salvage my blender displacement workflow…

Take a look at this topic. The idea that you need a NURBS model for CNC machining is a myth—or at least it used to be true, but not anymore. I design molds and machine models using only NURBS, only meshes, or even a combination of both in the same mold.

Thats a very promising start to my time on this forum lol :slight_smile:

Thanks very much for the reply… I’ll dig intop that referenced post!

Actually - because this has been a long day trying new things here and there and an even longer week doing the same (due to theis concept of a dense poly mesh not suiting moulds and cnc)… Can you assist in confirming:

My epic density mesh with 2Million polys is not actually a dead end after all, and my modelling technique is perfectly fine for CNC or other mould manufacturing some day (provided I keep up pratical reasonable design rules for the moulds so things actual mould / cure and release correctly).

That would be an incredible relief for me :slight_smile:

I work in ZBrush with polygon meshes that exceed 40 million polygons. Is that enough for you? I can even go up to 100–120 million polygons. (Of course, for mold making that’s usually unnecessary, so I normally reduce the polygon count.)
Take a look at this example I posted. It was around 10 million polygons.


This model has more than 60 million faces.

Unbelievably good news. Thanks so much for chiming in - totally made my day :)…

I might be back with questions, but for now im just going to enjoy the fact that my design process isn’t broken!

That’s not entirely true, it depends on the capabilities and preference of the service and the end use. Many toolmakers refuse to CNC mill from a mesh. There are multiple reasons for that. That being said, one could search for a service that will CNC mill from a mesh. However, there are compromises. The best surface finish straight from the CNC will be from a NURBS model.

@JonnyBam Keep in mind that any pattern I work on is created on a flat plane first. Then I use FlowAlongSrf, ShapeMap in Grasshopper, or any other Grasshopper method to apply the pattern to the target surface.

Regardless of the pattern, I often start with NURBS. However, when I need to apply complex details, such as floral or other organic designs, I either create them directly as meshes or convert them to meshes. That’s because fitting them onto the target surface requires much less computation than using NURBS, which can become unnecessarily heavy in these cases.

So, wherever there is patterned detail I use meshes, while the smooth surfaces remain NURBS.
mesh and nurbs




The gray part is a NURBS surface, while the yellow part is a mesh.


nc


Ok - thanks @cdordoni and @brvdln …

More reading for me… I wonder if I should explore more remesh options still. In blender I’ve tried shirnkwraps and remeshs. In rhino I’m tried quadremesh as mentioned… I’ve assume shrinkwrap will be a poor outcome like blender was, but who knows… maybe it’s significantly better as a toNurbs solution?

Maybe I’m missing a whole tool set for this stuff… not sure. There was some kind of voxel thing I glossed over. I mean, as a hypothetical, does turning my 2M into a subD and even then toNurbs’ing it (EDIT: It does warn me that I’ll likely crash at this step… so yes, I guess this is pushing things lol)… The mesh is quad based anyway… does that completely solve the issue? Or is the possble cnc issue more about the density of “any” object. I ask because: well, sure it takes a few minutes to calculate, but I CAN toSubD my mesh… once calced it actually pans and moves kinda fine… I’m sure I can toNurbs it too… SubD below (which is an ideal conversion with full detail):

I mention this mainly because after seeing examples from @brvdln my “complex” meshes look pretty darn simple lol :slight_smile: … and the idea of rhino calcing away at a toSubD for 2-5 minutes seems light duty after some of what I read!!..

Not discounting the comment that my mesh might be able to be milled as is though… which is still probably the simplest path…

The workflow I usually use is:

Option 1: Mesh → SubD → ToNURBS, which generates quad-based NURBS patches.

Option 2: Mesh → SubD → ToNURBS using the Packed option, which generates larger patch-like NURBS surfaces.

In this example, I used the Packed option in ToNURBS, so the entire model was converted and kept as NURBS.

Option 3: Machine the mesh directly without converting it to NURBS.

It really depends on your workflow and your needs.
Take a look at this link
For example, I used ShrinkWrap on the mesh, and the offset shown as a quad SubD is a reference surface for generating the toolpath. The final part is the outer object.

I’m sure I’ll look like the simpleton here, but I’ve found that there are some good video tutorials on YouTube.
In a big general way, I would suggest that you try looking at what you want to do as jewelry design and check out some of the jewelry design vids and the often used commands to do so.

Use of 3d scan or a mesh with small detail relative to a large area (high freqeuncy detail) should be concerned that a workflow converting to NURBS may compromise the detail, resulting in a softened or melted effect. This is an inherent limitation regardless of the software used. Increasing the sampling / resolution to get better detail in NURBS results in more patches and can be problematic in processing in CAM software. Rhino’s workflow to generate a more detailed NURB model from a mesh may fail for a similar resolution mesh compared to other dedicated reverse engineering software like Geomagic.

Thanks all for the replies. It’s a great help.

It’s odd too… I started the rhino trial specifically with the intent of just seeing how to use it to engineer the internals… and I’ve now realised that there are some hiccups wit h my method.

On one hand - it IS great to hear that high density meshes can be cnc’d (t hanks again @brvdln !)… On the other hand, concerns from others are mirrored by my local Aus plastics manufacturers… All I’ve contacted have come back with a pretty clear “brep only please” statement after I asked if t hey can cnc a high density mesh…

I think my current battle plan is:

- → I have a great displacement workflow for aesthetics with high density mesh. I’ll put that on the shelf as a technique.

- → I’m going to dive into math modelling (nurbs etc) and see how far I can gget visually. I assume I’ll realise I can go far and beyond my needs. Whether it’s a process I like… not sure yet. But I like t he idea of having pure design and b-rep accessibility by default… so I’ll explore this for a bit.

- → Probably should use my valuable rhino time experimenting with functions I might use it for. I’ve jumped into Fusion too, and… well.. sorry all… but I find it much more approachable for raw design and shaping :)… e.g. I find lofting in fusion a joy. I find it overwhelmingly confusing in rhino (and error prone). So I wonder if I use fusion for the broad shaping and engineering; what might I wish I trialled rhino for? Grasshopper probably.

Rambling a bit - sorry. Anyway, thanks for all the help.

One last question: Is there a way to confirm whether a cnc / plastics company allows high density mesh?.. or is ther e a good global supplier of cnc moulds t hat you guys would use / rely on for that… and then I take the mould to whoever I like for injection / plastics…? A littel unclear on this side of it all sorry :slight_smile:

Probably the companies you’ve contacted simply don’t accept meshes because B-Reps are easier to work with and they don’t have a workflow for handling meshes. Or they may be capable of working with them but simply choose not to.

I think their answer is “no” for the same reason we give that answer to new clients. Building a mold from mesh-based geometry often isn’t worth the extra time and cost involved, especially if the final production run is only going to be a small number of parts.

One possible solution is to try converting the mesh using Mesh To SubD followed by ToNURBS, or to manually rebuild the surfaces through reverse engineering. Which approach is best depends on the quality of the mesh and the complexity of the model.

Yes - good advice again @brvdln…

I think in my head it’s worth exploring non mesh options. The simplicity and “all plastic shops can do b-rep” reality means it might be worth at least starting with solid performing low detail (i.e. not facny looking) but well engineered lure :slight_smile: .. I’m still of course hoping for the nice details I’ve discussed..

I did try Mesh → Remesh (chose 100k target) [intent was to try and reduce poly count] → ToSubD → ToNurbs with Packed on…

Result is still very high density… Says 134k… I kind of expected packed to tidy up and simplify a lot of the tail.. it’s just from a basic Blender subD quad orginal (with high density poly)… I assumed the math would be pretty simple back there…?

My next silly question is: This is stupidly high for B-rep, right?.. i.e. it doesn’t solve my “they won’t accept a mesh” issue right?

Assuming that’s the case, I’m going to jump back in to rhino and try some basic modelling… worst to worst I’ll target simple designs in either fusion or rhino. I must admit, Fusions glossy exterior is followed up by some really odd design limitations… I’m realising like anywhere, programs seem to have their strengths. My guess is that Rhino is more of a learning curve but ultimately more capable (in terms of modelling and raw abilities).

If you all don’t mind, I might keep bouncing along in this thread while things are roughly on topic :slight_smile: … any assistance is greatly appreciated!

Having a lot of surfaces after the conversion shouldn’t be a problem for the toolmaker (it mostly depends on how powerful the PC is).

That said, I recommend rebuilding the mesh using curves extracted with the Section tool. The sections you get are first-degree curves, so you should use Rebuild and convert them to at least second-degree curves; otherwise, you’ll end up with an infinite number of straight segments.

Once you’ve done that, you’ll have a series of profiles running across the entire mesh that you can use to reconstruct the surface, for example with a simple Loft. Think of rebuilding an aircraft fuselage or a boat hull—the process is very similar. Once the overall shape is defined, you can move on to refining the details.

I’m linking two videos I made for a modeler who asked me how I reconstruct a high heel. I recorded them so he could go back and review any part of the process he might have forgotten. There’s no audio because we were talking about unrelated things while I was recording, so what you see is just the video showing my reverse engineering workflow.

You’ll find the part where I use the Section tool to extract the curves I need. Every now and then you’ll notice me doing some very basic operations, but that’s simply because I was explaining the commands as I went along.