What's new
What's new

240,000 data points to mill

Joined
Aug 13, 2018
Yes I'm new to the CNC world but learning the hard way. I have had a unique cylinder head combustion chamber laser scanned and now have a quarter million data points that need to be massaged into a usable milling program. After welding up the seat pockets and all the remaining surfaces to be re-machined back to OEM profiles/ contours, the combustion chambers don't need mold surface finish but can live with a respectable surface finish. The problem I have is 240,000 points wont work due to time and memory. I'm drip feeding into an older Fanuc milling machine with 4800 baud max capacity. Not sure how or what program to post process the points into the best program. I have read about arc fitting, step over cusp calculations, look ahead drip feeding programs for feed rate adjustments and some other ideas. I'm not in the position to purchase a newer machine at this time and need to make my machine work for at least eighteen months, then I may move onto a newer controller/ machine.
Has anyone been confronted with many data point conversions into a workable program? I know throwing out some of the points can be done but which ones? Is there a best practice post processing method for assisting in doing what I need to do? And can an inexperienced programmer tackle such a humongous program?
 
Yes I'm new to the CNC world but learning the hard way. I have had a unique cylinder head combustion chamber laser scanned and now have a quarter million data points that need to be massaged into a usable milling program. After welding up the seat pockets and all the remaining surfaces to be re-machined back to OEM profiles/ contours, the combustion chambers don't need mold surface finish but can live with a respectable surface finish. The problem I have is 240,000 points wont work due to time and memory. I'm drip feeding into an older Fanuc milling machine with 4800 baud max capacity. Not sure how or what program to post process the points into the best program. I have read about arc fitting, step over cusp calculations, look ahead drip feeding programs for feed rate adjustments and some other ideas. I'm not in the position to purchase a newer machine at this time and need to make my machine work for at least eighteen months, then I may move onto a newer controller/ machine.
Has anyone been confronted with many data point conversions into a workable program? I know throwing out some of the points can be done but which ones? Is there a best practice post processing method for assisting in doing what I need to do? And can an inexperienced programmer tackle such a humongous program?

Have you tried knocking down the number of surfaces? I'm assuming that you're trying to machine from a .stl file. If this is the case, you can bring it into MeshLab, which is free to download, and use quadric edge decimation.
 
Seems like Laser Scanner guy is out to lunch to me. I've worked on laser scanned 3D parts for a customer. Customer sent hand-made parts off to laser, they scan and then clean up surfaces to some tolerance, then produce a .step file or whatever with nice, easy to machine surfaces.

I'd ask for that before you tear your hair out.
 
Usually a scanned surface will be a .stl file, a bunch of polygons, made from the point cloud. If the resolution of the scan is really high, you can end up with a huge file, that's difficult to work with.

The picture is from a scan of a fish, the original file had so many surfaces that it wasn't really possible for me to work with it in Solidworks. I think the OP has a similar problem, but I can't be sure without seeing the file.
 

Attachments

  • FEESH!.jpg
    FEESH!.jpg
    94.5 KB · Views: 266
A pro shop should be able to take a point cloud of typical engineering shapes (i.e. not a skull) and turn it into regular features or NURBS (3D spline surfaces) using software like Geomagic Design X.
 
Tony talked about doing this to replicate an issue with engine wear to reproduce sounds... Now in his case they actually needed to replicate the surface they scanned.
 
Laser Scanned Date

Ok fellas here is one of the laser scanned STL programs. Unfortunately I don't have access to the arc fitting process, may need to go back to the scanner. They sell high end post processing software of some kind but didn't offer any assistance. They knew what I was doing and what I would be doing it with. Maybe they were out to lunch on my money!
Unfortunately the file extension is a X_T format and the system won't let me upload. Tried an export but not being a computer genius, can't figure out what rout to take. What can I convert a X_T file into so it will come through?
 
Can you share the file on google drive or Dropbox? Do you have both the STL and an X_T? Both would be potentially worth looking at.
 
Good luck, we`ll keep watching the thread. My father's 70-something half sister has been collecting all sorts of information about my grandfather that seems to have been a total mystery to everyone. Anyway she's assembling these scanned documents in an open office doc and uploading to google drive but the files are like 80Megs so it's causing all kinds of problems I'm trying to help her with. I'm sure I can get them into PDFs which will be much smaller but I have to get them to Word first. Arg!!! In some ways it's a similar problem to dealing with your ancient Fanuc...
 
Slow down.
For starters a "quarter million data points" is not huge and sounds like your scan source did their job well.
Every point will be subject to measurement error so these are not absolute numbers
You need sort of mold finish or better in a head. (duh)

Unclear, you are are going to profile mill this .... maybe with .010 diameter ball-nose cutters?
Unless I missed the boat someone else made this. Get the how to go with your reverse engineering numbers.
You may not have that how but grasping it makes life easier.

It is interesting to me that you have a a "unique cylinder head combustion chamber"...... Sort of a lifelong quest in both four and twos.
Screw the making it, I'd love to see this design and it's merits.
Bob
 
My thoughts...
I am an old world Machinist, drug kicking and screaming into the new world.
So 3, maybe 4 years ago, I was still programming in hand-written G-code utilizing canned cycles and loops.
Then, I went Cad-Cam.
The very same part went from say... 500 lines of code, to 12,000 lines of point-to-point code.
My machines didn't have the capacity to even load it.
Today's Cad-Cam has absolutely no regard for memory.

Do a memory up-grade, and don't worry over it.
I can still write the same part in 500 lines... but I don't care.
The Bridgeport DX-32 control had megabytes of memory.
I just up-graded it to where it did not care anymore.
File size these days... don't worry over it.
What I do miss is being able to read the code through for a quick spot-check.
A quarter million points... can't do it.
 
Mastercam will also filter that path. If you make a parallel toolpath in the X or Y direction only it will turn those thousands of points into hundreds of G2/G3 paths.

You might be able to just use a 3d toolpath driven by and STL with a containment boundry. Then you can skirt around making surfaces from them.

You really want good short code to drip feed. You might even not have to drip feed if you can filter it enough.
 








 
Back
Top