I saw the posts on scanning, but my application is slightly different. I use the MPCNC at times to face slabs and cookies for making furniture.
I usually just create a model in fusion 360 in the rough shape of the stock, then set the job accordingly to face the wood, usually in multiple passes. Problem is that it can be doing a whole lot of nothing with the bit in the air.
I though of scanning the stock instead, bringing it into Fusion 360 and then setting the job to use that stock to have one job the more efficiently can plane the surface of the wood until it is completely flat, reducing the “air time” when it does no cutting.
Any suggestions on what scanner (like the xbox one, cheap) and what work flow would work for this kind of application?
Yes, I have thought of that as well, it would take care of the x and y coords. But it would not take care of high and low areas, which is why I was thinking of 3D. I don’t mean slightly low and high areas, but more like really odd formed shapes such as the part of a cookie that is near where the tree was cut, often you have a wedge shape there…
Could you take a picture of the side and use that to estimate a polygon?
I mean, you’re not needing exact heights. Just something close enough to knock down the material with a bit extra overlap.
Is it usually just wedge shape, or do you have random large bumps along the material? That would be harder to account for using just 2 pictures. Extrapolating heights from 2-D images can get more difficult, but you could use a mesh and push/pull points using images.
Autodesk for a while had a cellphone app you could take multiple pictures of an object and it’d create a 3d stl from them. Not sure if that still exists.
I feel like you need fairly precise measurements for this, am I wrong? Issue with the Xbox eye and photogametry (taking pictures to create a 3d model) is that raw accuracy is usually not that great (maybe 1/4") and will very with each scan. You can spend time cleaning up the model to get more accuracy but then all you did was shift your time from watching the cnc mill air to messing with 3d models.
I would think for this you would want to probe the surface. It’s time consuming but much safer than milling (I would have no problems letting my cnc probe for hours unattended) so you could set up the probing before you go to bed then in the morning you have the 3d surface to bring into f360 and mill faster while you have an eye on it.
You can get away with 1/4" accuracy or worse, as long as you always err on the side of cutting too much. Traveling through a part you thought was clear is going to be bad. But adding 1/4" to the first pass would avoid that risk. At the cost of time.
What kind of probe are you talking about @jeffeb3 ? Something like a BL Probe? Has anyone done a laser probe of some sort that doesn’t need to travel the Z up and down (lengthy) ? What software and such is needed to create the 3d model from the probes ?
There are also lot’s of Raspberry Pi sensors that are cost effective. Can the MPCNC / LR2 be controlled with say Octopi, and use a IR or Lidar or Time of Flight sensor to scan the bed?
Not that I know of. I was thinking using a touchplate and the bit as a probe. You don’t need the same resolution as the laser scanner video. I also suspect that would be a lot of work, or a lot of cost to get the laser scanner to work. I could be wrong.
How much resolution do you need? I would guess a probe grid of 3" between points would be plenty. You could even use a 3" diameter spatula as the touch plate and get pretty good. You are just trying to figure out where the board is already cut out, so you don’t waste a ton of time cutting air.
I know how I would do this. I would attach a raspberry pi and make a python script to control the CNC. It would calculate XY coordinates in a grid and move the tool, then run G30 to probe. The result would be stored for that point. Then you would need to export the results as a 3D format. (You could try to have it write out the gcode to flatten it right now, maybe). That 3D format would need to be pulled into a CAM program that create the gcode.
You would need a reliable way to set the XY and starting Z. If you took two hours between probing and cutting, the machine may have turned off and you would be off when you started again.
All of this requires software. Even my solution of a python script is a few hours, at least. If you want more polish, like integrating it into octoprint, or making a GUI or building an app, that all just takes more time.
The hardware is there. The firmware is also ready. The software needs to be written.
Instead of a probe, my first thought is a VL53L0X ToF laser range finder. They are cheap (around $10 USD). Their drawback is their ranging accuracy. It is about plus or minus 3% at best, but for this application, and being conservative on where to cut and not cut, that seems good enough.
Photogramerty can be accurate if you take enough of the right kind of photos and take the time to make sure the software stitches them together properly.
Unlike Jeff I wouldn’t use the touch plate. I would use a bltouch so that I could start a 10x10 point probe and just walk away while the thing was running. Maybe it’s just me but I generaly don’t care how long somthing takes if I’m not required to be there watching it.
I haven’t seen anyone but a laser sensor like that on a 3dp or cnc but it would be cool to do so!
Yeah I say bl but I would probably use a cheap $10 3DTouch. Rather than the name brand. And your right a microswitch on a stick should work just as well.