Full Version: Modelling questions

Show messages:  1-20  21-40  41-58

I have a couple of questions about moi3d modelling approach.
To not spread amongst different threads i guess it's better to ask them in one thread here.
Any advices or tricks are welcome.

Q_01: need advice from cad gurus.
I want reconstruct this kind of shape in moi3d from scratch.


I tried different approaches but the result shape is not quite right.
The first in mind is to use loft. But the result (if we use loft from this 4 curves) is kinda flat around the edges.
I know that i can use iso operator to create more loft profiles from source shape.
But i insist that we need to reconstruct source shape from scratch. So we shall not use source shape as maquette.


The second approach is to use rail revolve. But it doesn't have "scaling rail" option
and in result we have shape without thickness difference by X Axis.



I also can use any operator (sweep, loft, revolve etc) and then use flow deformer to get
closer to source shape but its kinda slow and not efficient way.

I dunno maybe blend operator may be helpful somehow. I saw some tricks with it.

Attachments:
SRCSHAPE.3dm

Image Attachments:
LOFT_001.jpg  RAILREVOLVE_001.jpg  RAILREVOLVE_002.jpg  SRCSHAPE.jpg 

---

Hello Bravlin,

it's always difficult to reconstruct things that results of strech operations or organic shapes from subd operations, but that's the right place to discuss such things.
My proposal:
get 4 profile curves and 2 rails
join the segments and rebuilt them with a tolerance of the treshold 0.005 units
make the sweep

get a curve for a railrevolve from the original also and rebuilt it
create a railrevolve axis perpendicular to the curves end/center
create a isocurve on the railrevolve and trim there
build a G2-Blend between the trimmed railrevolve and the sweep

Hope it helps!

Have a nice weekend
Karsten

Attachments:
SRCSHAPE.3dm

---

Hi Bravlin,
The biconcave script will give a 1/4 lobe profile in the +x direction, mirror the 1/4 lobe profile and join to make a 1/2 lobe profile of a bilobe, in +x direction.
This half lobe profile can be 1d scaled to stretch it out.

Using biconcave again for a thicker 1/4 lobe profile. Mirror it also, and join to itself, and rotate to -x direction.

The above two profiles are the two sweep profiles.

Draw a tiny circle at the origin for a sweep rail.
Draw a larger circle at the radius of the second lobe profile, for a second sweep rail.

Do the Sweep

- Brian

Image Attachments:
BiconcaveSweep.png 

---

Why not take a simple sphere, Show Points and make some tweaks with the 4 views ?

Add some control points if default are not sufficient...

---

And RailRevolve and flow - for exact reconstruction it is necessary to make base and target surfaces with a more complicated inner structure:-)

Image Attachments:
RR-Flow.png 

---

Network works
Cheers
Barry

Revised version to allow for differences of underside from top side.
Projected straight lines onto original solid to get cross sections & outer profile.
Divided cross sections on intersection point.
Select all and network.

Image Attachments:
Screenshot (245).png 

---

Hi Bravlin, it can be difficult to create sort of blobby melty all-over-smooth shapes like that in a CAD program like MoI where the focus is more on precision. Sub-d modeling is probably a better approach.

- Michael

---

Thanks for such a quick feedback guys.


I examine Karsten approach to this shape reconstruction.
At first glance everything looks great. But i found surface artifacts.
I expect that some of them may be solved if we try to make profile curves (or rail curves) of equal density.
But unfortunately it doesn't help much. I thinks about blend use but not quite sure. Since blend and main shape
have different surface density and wires, i guess it may cause problems later when we would try to add details on section between blend and main shape.

I would examine other approaches a bit later.

Image Attachments:
ALG_001.jpg 

---

Hi Bravlin,

re:
> I examine Karsten approach to this shape reconstruction.
> At first look everything looks great. But i found surface artifacts.

The first one looks like only a display artifact, not in the actual surface. If you export to a mesh format and make the mesh higher density it should go away.

It's a type of artifact that happens when the display mesh ends up with some sparse longer triangles across an area where the normals are changing.

The regular viewport display is more tuned to generate display meshes quickly so that you don't have wait extra time on every modeling operation. To get a better view of surface quality you can set up a keyboard shortcut that does a SaveAs to a mesh format like: SaveAs c:\bugs\test.obj then crank up the density and cancel when you're done without actually writing the file.

The other one though running through the centerline looks like a tangent discontinuity. That could be improved by trimming away some space and putting in a G2 blend surface like in the attached version.

- Michael

Attachments:
SRCSHAPE2.3dm

---

Ok. Moving forward.
I tried network approach by Barry-H. It's pretty straight forward. Network have useful options to tweak result shape via Custom and Uniform mode.
Network also doesn't rely much on a profile density so you may use dense and not evenly spaced curves.
Unfortunately result shape builded by network operator have one of artifacts i mentioned earlier.
I tried to export this shape to .obj and i have one additional question.

- " How should i export this type of geometry (botomm right shot) ? "
If we leave vertex with such high valency it definitely cause artifacts on render.

Attachments:
Q001NETWORK.3dm

Image Attachments:
Q001.jpg 

---

Hi Bravlin - high valence worries is something that applies to sub-d modeling, not to static mesh rendering as long as vertex normals are coming through and being used in the render.

But what can happen with network surfaces with poles are surface shaping defects with little ripples in the surface as different shapes shrink down and collapse to a point.

One technique that can help alleviate that is to include a circle before the pole in the network curves, that helps to even out shaping before things collapse to a point, see attached version.

- Michael

Attachments:
Q001NETWORK_2.3dm

---

Should there be a small circle on the bottom side as well?
- Brian

---

Hi Brian,

re:
> Should there be a small circle on the bottom side as well?

Yes if you wanted that same location on the bottom side to look very smooth under close examination you would want to do the same thing there as well.

But if it's not going to ever be visible that might not be needed.

- Michael

---

bemfarmer, today i tried biconcave scripts. I can't see the benefits of this approach.
I guess it may be useful in some kind of procedural modelling session or node graph.

---

Hi Bravlin,
The script is just a quick way to generate some 2d curves...
- Brian

---

Q_02:
Another one modelling question. It's about right patch creation.
Here is a test project. It has 2 layers SRC and WRK.
How should i construct patch on WRK layer so it look the same as on layer SRC (it has a bit curved profile).
I suppose that in such a case it's easier to create some kind of a curved plane and trim it, but each time
i tried it always have gaps and naked edges in result after join process.
I also tried "SeparateSelectOuterLoopsDeleteTrim" but still cant understand how to use it in rebuild(retrim) surface process.
I also noticed that when we convert edges to curves we should not rebuild result curve this way we have better chance to avoid gaps after join process.

Attachments:
Q_02.3dm

Image Attachments:
Q_02.jpg 

---

Hi Bravlin,

I gave it a shot, but I had to clean up to a point that I started wondering which parts of the model are important and should be retained. There were multiple overlapping edges and some double faces all around the model which can make it difficult to select the right lines for the commands to work. How did you model the MDK enclosure if I may ask? I will give it a shot again later when I have some time.

Good luck!

---

Hi Bravlin, so unless you had a really specific need to do it this way you would not want to approach your modeling steps so that you have to fill in an area like that.

Instead you generally want to have extended pieces that then all get booleaned or trimmed to common cutting surfaces, that's the best way for things to match up.

For filling it in though, one way would be to use the "Patch" command in Rhino which fits a bendy surface to a set of points sampled along curves. However, it is often quite finicky to get it to be accurate enough to join cleanly. Since this one isn't too bendy it could work though. In the future I also want to add in a similar surface fitting mechanism into MoI as well but there is not one like that currently.

A little later here I'll see about giving you some steps for reverse engineering the extrusion.

But again, you don't want to build things in such a way that you put yourself in this situation of needing to build a surface to a hole like that. It's better to work with solids and 2D curves from the beginning.

- Michael

---

Hi Bravlin, some of the naked edges could be coming from some new bottom surfaces that no longer align with the original extrusion.

The process for reverse engineering the extrusion is kind of tricky, the steps are something like extract the edges as curves, prepare some cutting lines to cut away portions that are not on the silhouette, trim the curves leaving no doubled areas in the Y direction, then compress it in the Front view using the edit frame corner grip with "Flat" snap.

That will then leave the flattened silhouette which can be extruded. It's also good to rebuild it to reduce segmentation and also there are a few spots where control points get very nearly stacked up at the end of some curves and those need a little adjustment to move the first interior control point.

When that is extruded it is then accurate enough to be trimmed with those edges, except at the bottom where some new surfaces stick out a little more than they did on the original. 3DM model file is attached.

Here are some screenshots of the process. At some point I want to make a curve silhouette tool to automate these steps.
















- Michael

Attachments:
Q_02_extrude_3dm.zip

Image Attachments:
Bravlin_extrude1.jpg  Bravlin_extrude2.jpg  Bravlin_extrude3.jpg  Bravlin_extrude4.jpg  Bravlin_extrude5.jpg 

---

Today i tried this algorithm of patch creation and have problems with naked edges again.
Guess i missed something important in trimming process or in surface creation.
I assume that one of a steps should guarantee the right contact of result surfaces.
I just cant figure which one. I inspect the project you provide Michael. And it works like a charm.
So i recorded short video of whole process.
https://www.dropbox.com/s/8nms9s7buihvk56/2019-02-03_12-16-25.mp4?dl=0

I also add project file to inspect if necessary.

Attachments:
Q_02_ALG01.3dm

---

Show messages:  1-20  21-40  41-58