Overclock.net banner

Industrial Robot CNC Mill Retrofit

13K views 41 replies 9 participants last post by  driftingforlife 
#1 ·
Edit: Changed the thread title seeing as how this has very little to do with a computer case...

Hey everyone, sorry I haven't been very active recently. Work took over and next thing I know months have gone by.

Anyways, I have been very busy. I managed to get my hands on some newly retired equipment from a fully automated assembly line that was scrapped to make way for "Worker Oriented - Job Creating" assembly work cells (Politics once again stifling the technology industry but what can you do...). These are some of my new toys:

Two of the work cells I was tasked with stripping before being hauled off to the dump:

http://s324.photobucket.com/user/InitialDriveGTR/media/20130720_132710.jpg.html

One of the cells with its guarding removed, leaving the bare robot:

http://s324.photobucket.com/user/InitialDriveGTR/media/20130718_121611.jpg.html

These robots are by Adept Robotics, they are called Cartesian robots, implementing X, Y, and Z axis in addition to Theta (Rotary table) and this one specifically had an industrial screw driver used for final stages of product assembly. We typically do not recycle old machines as it's cheaper to just throw them away. So I snagged three of these robots, and over the last month I have been working on converting them into CNC mills. I am developing two at the same time, one for my own personal use at home, and the other for our department to use for fixing messed up parts the machine shop gives us quite frequently. I'm using 600mm modules for the X and Y axis, and the Z axis is a 250mm module. Each axis has a 400W servo with disk brakes, and encoders with 65,536 points of accuracy per revolution. That means each axis can move incredibly fast, with a significantly heavy load, with repeatable accuracy of less than a thousandth of a millimeter.

I have been developing software for the robot so that it can interpret raw G-Code used traditionally in CNC milling applications. Using this method, the robot can now essentially be used as a CNC mill. I purchased a Bosch 1617EVS router last night, and will be making some brackets to mount it to the Z-axis later today. Yes I have been sharing my bedroom with an industrial robot. Turns out they don't make very good pets, as our electric bill has almost tripled since I brought it home. lol.

http://s324.photobucket.com/user/InitialDriveGTR/media/20130819_202057.jpg.html

The robot installed on the second floor of my parents shed; finishing the install this weekend:

http://s324.photobucket.com/user/InitialDriveGTR/media/20130827_185531.jpg.html

Some of the control used to interface and control the robot:

http://s324.photobucket.com/user/InitialDriveGTR/media/20130822_174526.jpg.html

Testing out th G-Code interpretation code with an expo marker (Many markers were brutally murdered in the process sadly)

http://s324.photobucket.com/user/InitialDriveGTR/media/20130820_202246.jpg.html

The first part! (It's not actually going to be used, but just to see how the calibration was working)

http://s324.photobucket.com/user/InitialDriveGTR/media/20130827_155736.jpg.html

The whole point of me showing that stuff is that before I start making performance automotive parts, I want to get some practice using the machine and to further develop the software and user interface. So what better than build a new computer case?!?!?
biggrin.gif


And for those who know my last build Arctic Rain, it's doing very well. I have been running it closed loop as I don't use it that often for gaming these days. Most of my time is now spent on my Sony Vaio Duo ultrabook (Which is fantastic btw)

I've been out of the desktop computer world for a few months now, so I don't know whats been going on in terms of performance specs etc. and this haswell stuff I've skimmed over a while back. So I need ideas. GO.
 
See less See more
8
#2 ·
:bow: You are the King of Thrift now. 3 servo based Cartesian robots that can be converted to CNC mills... Just thought you ought to know.
 
#5 ·
Very Nice!
 
#7 ·
#8 ·
Dude, I know I already said this, but you are my hero. How are you planning on cooling your bit with the CNC into the solid aluminum?
 
#10 ·
Oh yeah, I'm not saying that you can't, but the heat will degrade your milling bits more quickly. Well, I assume it will just because of high heat of metal to metal milling.

VVV Listen to TLHarrell over me ANYDAY.
thumb.gif
And I learned something new about CNC routers bits and the wear.
 
#11 ·
As long as you're hitting the necessary chip load, heat should be carried away from the cut by the chips pretty well and coolant mist can usually be optional. I just got my CNC machine, router for it comes tomorrow. I'll be milling aluminum with it as well. It's certainly not like the giant KOMO machine I used to program on, so I won't be hogging with a 1/2" bit 1/4" deep in the material. Probably much lighter passes, maybe 1/32 or 1/16, with a 1/4" or 1/8" bit. If I need coolant, I'll apply with a spray bottle.

It will, of course, depend on the aluminum type. Some is harder and will cut nicely. Some is more gummy and will tend toward sticking to the tooling.
 
#12 ·
Good to know. I'm going to be using mostly scrap aluminum from work, so I'm sure it will all be the high grade stuff. What kind of mill did you get?

Some more current pics:

DSC_7480.jpg

DSC_7478.jpg

I'm working on the code now to use laser sensors to automatically find the dimensions of the stock and the 0,0,0 location. Should make it easier when I have to flip the part over to machine the other side
 
#13 ·
I got the Probotix Comet, 25"x25"x5" working area, pretty solid little machine. I'm running a Porter Cable 892, 2.25HP router for the spindle, with a nice set of low TIR collets from Precise Bits. I also have a router speed controller on order from superpid.com. It's coming from Australia, so it'll probably be a couple weeks till it arrives.

Here's my first startup and homing from Monday night after uncrating it. Router arrived yesterday so isn't in place in the video.
 
#14 ·
Awesome! That is exactly the CNC I was looking at before I came across this robot.

This is the modifications I made to the spindle mount for the laser homing feature I'm adding into the robot.

http://s324.photobucket.com/user/InitialDriveGTR/media/20130912_164333.jpg.html

And a demo of how fast this thing is (It would never actually operate anywhere near this speed in any milling application):

 
#15 ·
drool.gif
Holy crap that's fast! Mine's running 300IPM rapids in X and Y, 60IPM in Z. I think it's probably more of a limit set in software, to be conservative and not end up with any dropped steps. I think the machine could handle double that easily. I don't really see it as necessary though. What's your speed on rapids? Looks like somewhere in the 1,000-1,200IPM range.

You'll probably find out that speed isn't really anywhere near necessary as you'll only do that on rapids, and your programming should eliminate as much of that as possible. Still, pretty darn cool. I'd definitely love to build a machine based on those linear servo systems. Maybe when I have a bigger shop I'll build a 4x8. I much prefer servos with encoders.

Are you using cutter diameter compensation?
 
#16 ·
Quote:
Originally Posted by TLHarrell View Post

drool.gif
Holy crap that's fast! Mine's running 300IPM rapids in X and Y, 60IPM in Z. I think it's probably more of a limit set in software, to be conservative and not end up with any dropped steps. I think the machine could handle double that easily. I don't really see it as necessary though. What's your speed on rapids? Looks like somewhere in the 1,000-1,200IPM range.

You'll probably find out that speed isn't really anywhere near necessary as you'll only do that on rapids, and your programming should eliminate as much of that as possible. Still, pretty darn cool. I'd definitely love to build a machine based on those linear servo systems. Maybe when I have a bigger shop I'll build a 4x8. I much prefer servos with encoders.

Are you using cutter diameter compensation?
I'm not sure what the max IPM is, nor what my rapids are. The robot's controller doesn't work in that way. The servos are 400W yaskawa servos with brakes, and Adept SmartAmps. The amplifiers are daisy chained together through the wire carriers via IEEE 1394 aka firewire. All I have to do in order to set the speed for all the axis is "SPEED 10 MMPS ALWAYS" for 600 MMPM travel at all times, or "SPEED 10 MMPS" for the next movement to be 600MMPM. I'm using software called "MeshCAM 5" to create my G-Code, so all the tool paths are generated beforehand including cutter compensations, circular paths etc.which means all the robot has to do is follow XYZ coordinates on a line-by-line basis and follow the feedrates given by the G-Codes. The Adept CX controller takes care of all the joint interpolated movements etc. and makes something like a CNC mill as simple as following a list of points. I think when we originally bought the robot it was about $40,000, so the features are pretty nice.
 
#17 ·
From my experience with CNC machines (I used to program for a huge 5'x12' KOMO router, all servo drive) I know all machines have a limit as far as acceleration. This is combination of the strength of the motors, frequency limitation of the drivers and physical weight and limits of the machine. Usually when setting something like this up, there's some testing you should do to determine the machine's absolute acceleration performance. This will affect how fast it accelerates/decelerates at corners, and how it follows programmed geometry. It's likely the controller itself already has some hard coded figures in it's firmware for these axes that it won't go beyond. You could probably just make some short codes in the manual mode to make it rapid from one end of the table to the other at different feedrates until it complains.

I'm running LinuxCNC (EMC2), and the DRO screen shows current velocity information. I also have MeshCAM 5 for doing the 3D stuff. 2.5D it doesn't seem to do quite as easily, especially straight from flat 2D shapes which 90% of what I'm making will be from. I got a free script which may do well for me that runs in Autocad, called ACAD2GCode. Looks promising. I still need to sort out my preamble and post for programs. Just getting back into the coding thing. It's been about 12 years since I've written this stuff.
 
  • Rep+
Reactions: InitialDriveGTR
#18 ·
Quote:
Originally Posted by TLHarrell View Post

From my experience with CNC machines (I used to program for a huge 5'x12' KOMO router, all servo drive) I know all machines have a limit as far as acceleration. This is combination of the strength of the motors, frequency limitation of the drivers and physical weight and limits of the machine. Usually when setting something like this up, there's some testing you should do to determine the machine's absolute acceleration performance. This will affect how fast it accelerates/decelerates at corners, and how it follows programmed geometry. It's likely the controller itself already has some hard coded figures in it's firmware for these axes that it won't go beyond. You could probably just make some short codes in the manual mode to make it rapid from one end of the table to the other at different feedrates until it complains.

I'm running LinuxCNC (EMC2), and the DRO screen shows current velocity information. I also have MeshCAM 5 for doing the 3D stuff. 2.5D it doesn't seem to do quite as easily, especially straight from flat 2D shapes which 90% of what I'm making will be from. I got a free script which may do well for me that runs in Autocad, called ACAD2GCode. Looks promising. I still need to sort out my preamble and post for programs. Just getting back into the coding thing. It's been about 12 years since I've written this stuff.
Found out the robot is capable of 3000IPM rapids. Acceleration is 16000mm/s^2.

Been having some issues with the X-axis stuttering, so instead of tuning the sero like I should have, I decided to upgrade!

DSC_7498.jpg

Problem now is I need a kinematics license for the controller to run the enhanced features of the robot..... Which is probably going to be a royal pain to get.
 
#19 ·
That's going to be a lot more rigid, and more accurate than having a single drive on the one side of the axis. That's exactly how my machine is set up as well, but there's some tweaking involved in getting it to home the double slaved axis. I got the machine turn-key though so I don't know how that part of it is dealt with.

I have found the current CAM setup I had to be insufficient. I use Autocad for generating my part geometry. MeshCAM takes .dxf or .stl files on import. STL won't work for what I want, as it turns circles into n-sided polygon meshes. The gcode produced doesn't have any circular interpolation. Not only do I see that as terribly inefficient code, it opens up inaccuracies that I don't want to have to deal with trying to work around. Exporting as .DXF didn't work out either. The software takes the linework rather literally and makes all of it extruded at one height. I fed it a square part with two circles in the center. One was to be punched through, the other to be a pocket cut which will produce a shoulder around the other hole. MeshCAM made a square part with a circular trench around a circular center part, and I couldn't find a way to fix it. ACAD2GCODE simply won't work for me because it won't create tabs. It's fine for everything else I threw at it, which is great as it's free software. The no tabs is a killer as I need to produce many parts from a screwed down sheet and have them not flying around after cutting. I'm also leaving them attached to the sheet for storage. So, I started running CAMBAM last night. Got by brain wrapped around it rather quickly, and have been digesting the manuals and finding other really cool stuff it can do. Think I'll end up purchasing that package. At $150 it's a very decent price for what it's capable of.

I'm still a bit jealous of your setup. Using servos with a closed loop feedback is awesome as it'll always hit the mark it's told to... no dropped steps like in a stepper only system. Can't wait to see where you go next with this beast.
 
#21 ·
I do not currently do service work through OCN. Just getting my CNC up and running and taking small steps toward a lot more insanely cool stuff. It's been 16 years since I last programmed g-code, so it'll take me a little time to get everything totally sorted out and making chips. Currently also buried in work through other sources.
 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top