Freesteel Blog » Ducking a poke in the eye with a pencil

Ducking a poke in the eye with a pencil

Monday, May 1st, 2006 at 6:08 pm Written by:

People aren’t keeping us busy with bugs and feature requests on the Adaptive Clearing algorithm, probably because they’re not using it. For those who have been emailed copies of the executable, you’re supposed to tell us if it doesn’t work. I spent two years maintaining the post processors in NC Graphics; post processors are never right, especially the one that comes with this.

So, in absence of activity on that front, I have been compelled to write a pencil milling algorithm for Cimco. Since I’m using the same underlying structure as for the const scallop, the subdividing is quite free-form. The one I wrote in Depocam/Machining Strategist is based on quad-trees and is very complex. I remember there being a lot of memory issues, which made the structure even more compressed and hard to program.

No one can tell how this algorithm will perform by comparison to that one in the long term. Like any construct or machine, until it’s been tested, tuned, adjusted, and optimized for circumstances in the real environment, it’s hard to know how far it will go. With a more free-form structure than a quad-tree, there is a lot more room for change. A quad-tree, however, will always be a quad-tree; all that can happen is it gets implemented more efficiently.

No doubt the Danes will start finding problems with it during the next week, before I get stuck into the next algorithm. Did someone mention Rest Area Detection? Oooooh Noooooooooooo!

3 Comments

  • 1. Robert replies at 1st May 2006, 11:53 pm :

    How do you go about finding the CL points for the pencil toolpath? I’ve implemented one version but it’s based on a ZMap so it’s probably very different from yours.

    By the way, I’ve had thousands of people download MeshCAM and I’ve only heard from maybe a hundred or so users. When one user posts a bug I often have others add that they’ve had that problem too- they just never told me. No matter how many times I’ve asked, I’ve only gotten photos of the work produced with MeshCAM from 10 or 20 people.

  • 2. Neel replies at 2nd May 2006, 4:54 am :

    For Quad-tree , Octtree are you subdividing the triangle model based on triangle count or node count?. In the above picture the (2nd) the pencil trace seems to have found all fillet corners but some of the sharp corners are missing.
    The brochure are looking impressive and informative.

  • 3. Julian replies at 2nd May 2006, 11:42 am :

    Robert: My experience entirely. When writing Machining Strategist we had one user in particular who gave us such good feed-back he should have been credited (if there was any such thing as creditation for writing software). Users have no idea how important it is to be speaking back during development.

    There is no subdivision of triangles. The quadtree is a subdivision of the XY clearance plane.

    pencil tracing: Have you accounted for the side of the cutter in the picture? If it’s smaller than the fillet, then it won’t detect the fillet. The paths are solid, but only look dotted because of the 3D rendering puts them coincident with the surfaces.

    See next blog entry.

Leave a comment

XHTML: You can use these tags: <a href="" title=""> <abbr title=""> <acronym title=""> <blockquote cite=""> <code> <em> <strong>