Freesteel Blog » Really not enjoying this scallop

Really not enjoying this scallop

Friday, November 28th, 2008 at 12:38 pm Written by:

So much to get distracted from, what with codewikis to invent, people teaching me to use Pylons instead of Django, yet more templating languages I didn’t know about with so many layers of components that nothing makes clear exactly what each one of them does, as well as getting make-over done to undemocracy.com — that when you face bugs like this:

it’s easy not to want to do any more. But I’ve been going through things like this for about two weeks now.

What’s going on is that the scallop machining algorithm for 3-axis toolpaths somewhat fudges the issue when it comes to toolpaths on the vertical walls. It’s a special case. Until recently I couldn’t have more than one point of a toolpath on the same vertical wall at a time. This needed to be upgraded, because the Danes came up with a nasty little trick.

What you do is start with some sensible contours made around the shallow areas of the part. If you scalloped inwards from these, everything would be fine, because the scallop algorithm works well on these predominantly flat areas.

However, they’re sending the contour areas outwards a certain distance to smooth them. And what this does is place the contours all on top of one another on the vertical sides of the part where the scallop algorithm has a hard time getting it right, especially where it’s twisted simultaneously up and down a saddle shape.

So I spend a week debugging it, upgrading its capabilities (I can have up to 12 contours passing through the same XY area — half of them are invisible), get it all working lovely, send back the improvements.

And then all they do is double the offset outwards and it breaks all over again.

Although it is difficult to concentrate, so far I have been very lucky, because the fundamentals of the algorithm are sound. It’s just turning out that the some of the details, special cases, and submodules are misfiring.

But this, however, is the worst example I’ve seen. I haven’t started work on it, so I don’t yet know if my luck has run out.

Onwards to Euromold shortly. I’ve printed out a 5-axis algorithm on a sheet of paper to wave at people, to see what they think of it. The conventional response to the invention of an apparently new algorithm is to detail it incomprehensibly into an academic paper, include 5 pages of introduction reviewing all the other established 5-axis algorithms out there that don’t work (although the academics who think they know everything think they do and are going to argue with everything you say), include shed-loads of irrelevant citations, print it out in triplicate with the images on separate sheets of paper, sign it over to some evil Elsevier publication, wait three years, and get it thrown out by some distinguished South Korean professor who’s refereeing the paper anonymously (for no good reason, because anonymity was only ever justified to protect junior researchers from the wrath of senior professors who think they should get away with writing rubbish papers). With this drag of astonishingly mediocre past work which anything new has to answer to (never the other way round), you don’t wonder why there is so little progress. So I’m not bothering. I can afford to publish things independently of this broken academic accredited-publication system because my job doesn’t depend on it.

1 Comment

  • 1. Freesteel&hellip replies at 28th December 2008, 7:55 am :

    […] to finish the second round of elevation sketches in tunnel, then maybe work on the mother of all constant scallop bugs in hsmworks, and then perhaps a bit more work on the pylons conversion of undemocracy. At least […]

Leave a comment

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