[SCzEngrgGroup]

After building a PDB/ATC for my knee mill a few months ago, I have now designed a second, greatly improved PDB/ATC, primarily for my new bed mill, but it will be almost trivially adaptable to almost any mill. Error handling has turned into a very interesting problem as there are soooooo many ways things can go wonky.

I am curious how commercial ATCs deal with, and recover from jams/errors. Suppose that for whatever reason, a tool does not load correctly. What happens? What does the user have to do to resume normal operation? I've implemented one method of error handling in my current ATC, but I'm really curious how commercial systems deal with this. For that matter, what does the Tormach ATC do if, for example, the tool jams going into the spindle, or if the PDB doesn't release the tool?

Regards,
Ray L.

[Al_The_Man]

In Commercial systems the logic varies according to the complexity of the T.Changer and also depends on who writes the logic, but typically when the machine stops due to a malfunction, the operator goes to MDI and enters an assigned M code that exercises each function independently to get the system back in the correct sequence.
Al.

[SCzEngrgGroup]

In Commercial systems the logic varies according to the complexity of the T.Changer and also depends on who writes the logic, but typically when the machine stops due to a malfunction, the operator goes to MDI and enters an assigned M code that exercises each function independently to get the system back in the correct sequence.
Al.

Al,

Well, whaddya know! That's almost exactly what I've implemented in mine. When an error occurs, it pops up a dialog telling the operator *exactly* what failed, and gives him the option to either abort G-code execution (if any), or correct the problem. If he chooses to correct the problem, I pop up a dialog with buttons allowing the operator to manually exercise all aspects - raising/lowering the "lift", swinging the carousel between the park position and the spindle, rotating the carousel (either direction), homing the carousel, viewing all state variables (the tool the interpreter expects to be loaded, the tool the ATC thinks is loaded, the current carousel position, all sensor data, etc.), sensor data, etc. Still need to add the PDB controls.

[ame=http://www.youtube.com/watch?v=zjGi1WBTMBE]My Bridgeport Mill Toolchanger - YouTube[/ame]

Regards,
Ray L.

[wwendorf]

Very nice Tool Changer Ray. What software are you running your mill with? That doesn't look like Mach3 and it doesn't look like you're running Linux.

Wade

[SCzEngrgGroup]

Wade,

It's my own CNC controller app, running on KFlop. Mach3 was far too unstable for my taste. I needed a hardware motion controller (high-count servo drives), but wanted more functionality in the controller app than either LinuxCNC, Mach3 or the standard KFlop app provided. So, I wrote my own, in c#, running on top of the KFlop DLLS. It has been really wonderful. Where Mach3 gave me endless random problems, this has proven dead reliable since the first day I put it into service over a year ago, and I've customized the functionality to my machine, and my work flow. It's not the prettiest thing in the world (all built using standard windows controls), but it works extremely well. Someday, if I get some spare time, I'll make some pretty custom controls for it.

Regards,
Ray L.

[samco]

Very cool -

Just to mention - Everything that I have seen could have been implemented in linuxcnc...

(here is a non conventional tool chain setup entirely in linuxcnc.)
https://www.youtube.com/watch?v=39q6kvrSBSk

I didn't implemented any error handling as the old control had none. (estop - set the arms in correct location by hand - restart) But it would surely be possible.

sam

[Al_The_Man]

Incidentally the latest Mazak machines have a programmable chip in the base of each tool holder so the machine knows the location of any tool, allowing the placement of any tool in any convenient pocket and also placed at random, and other features.
Al.

[samco]

That is what I love about this setup.. Stick a tool anywhere in the chain - machine doesn't care. Got to give it to the engineers in the 60's when this machine was designed...

sam

Incidentally the latest Mazak machines have a programmable chip in the base of each tool holder so the machine knows the location of any tool, allowing the placement of any tool in any convenient pocket and also placed at random, and other features.
Al.

[SCzEngrgGroup]

Just to mention - Everything that I have seen could have been implemented in linuxcnc...

As good as LinuxCNC/EMC is, for many, probably most, of us, it's simply not a viable option, as it means giving up all of our other software - CAD, CAM, and countless utilities. Unlike with Mach3, KFlop allows me to actually USE my PC even while the machine is running, so I can be working on the CAD or CAM for the next job, while the current one is running. Using LinuxCNC, I'd need another PC to do that. I've been praying for 20 years for Windows to be supplanted by some *nix variant, but, sadly, it's still beyond the horizon.

Regards,
Ray L.

[samco]

If that is your process - then yes linuxcnc probably isn't a good fit.

(there are things like virtualbox that allows you to run other os's within linux)

Not to mention linux cad/cam options. There are a lot of things I do at the machine. (surf , edit/write programs) but I don't cad/cam there. (considering I would have to be standing up the whole time )

Anyway - nice work on the tool changer and app for kflop.

sam

[TXFred]

I am curious how commercial ATCs deal with, and recover from jams/errors. Suppose that for whatever reason, a tool does not load correctly.

Haas halts if any sensors indicate a problem. If the spindle dogs aren't aligned, the tool changer cannot move all the way in, and a sensor fails to trigger. If the tool changer cannot retract fully, another sensor fails to trigger. I assume there's also a drawbar sensor to prevent the head from lifting while the tool is still held in the spindle.

There's a Tool Recovery button that starts a dialogue. It's something like this. (From memory)

Is there a tool in the spindle? Y/N
____Y Is there a tool in the tool changer?
________Y Is tray in?
____________Y Tool in spindle must be same as tool in tray. Fire drawbar, lift head, retract tool tray.)
____________N Tool in spindle is not the same as tool in tray. Retract tool tray.)
________N (Retract tray)
____N (Retract tray)

It always recovered from errors, but I don't know how many logical paths existed to handle various failures. The Haas changer was fairly dumb. It used a Geneva mechanism to align itself, and it didn't know what tools were in the tray. In addition, tool numbers = tray position numbers, so a ten tool ATC could not use tool #11.

I think you'd be better off with a mach screenset that offers manual control of the changer. That's how the Tormach ATC does it.

[SCzEngrgGroup]

Haas halts if any sensors indicate a problem. If the spindle dogs aren't aligned, the tool changer cannot move all the way in, and a sensor fails to trigger. If the tool changer cannot retract fully, another sensor fails to trigger. I assume there's also a drawbar sensor to prevent the head from lifting while the tool is still held in the spindle.

There's a Tool Recovery button that starts a dialogue. It's something like this. (From memory)

Is there a tool in the spindle? Y/N
____Y Is there a tool in the tool changer?
________Y Is tray in?
____________Y Tool in spindle must be same as tool in tray. Fire drawbar, lift head, retract tool tray.)
____________N Tool in spindle is not the same as tool in tray. Retract tool tray.)
________N (Retract tray)
____N (Retract tray)

It always recovered from errors, but I don't know how many logical paths existed to handle various failures. The Haas changer was fairly dumb. It used a Geneva mechanism to align itself, and it didn't know what tools were in the tray. In addition, tool numbers = tray position numbers, so a ten tool ATC could not use tool #11.

I think you'd be better off with a mach screenset that offers manual control of the changer. That's how the Tormach ATC does it.

Pretty much where I ended up. I tested it today, and it works fine. When an error is detected, the ATC stops where it is, updates the internal state to indicate the toolchange succeeded, then throws up a dialog telling the user exactly what went wrong. When that dialog is dismissed, it pops up a dialog giving the user all the controls needed to manually operate the changer mechanisms, so he can clear the error, and complete the toolchange manually. When he exits that dialog, the program picks up execution right where it left off, as though nothing had gone wrong. I tested every failure I could think of, and it resumed properly in every case.

Regards,
Ray L.

[dbrija]

Very cool -

Just to mention - Everything that I have seen could have been implemented in linuxcnc...

(here is a non conventional tool chain setup entirely in linuxcnc.)
https://www.youtube.com/watch?v=39q6kvrSBSk

I didn't implemented any error handling as the old control had none. (estop - set the arms in correct location by hand - restart) But it would surely be possible.

sam

Absolutely! I am currently doing this for my 1100... Very much like what Ray has done, and Hoss did with his ATC. I'm using proximity sensors to sense the tool in the spindle and the pocket, using that information to stop if something goes wrong. Not trying to cover every possibility, just the main ones (like carousel didn't fully move into position, tool left in the spindle at last powerdown, air pressure sensor in case I forget to power on the compressor, etc...).

Learning Glade, and all the HAL stuff... but its been pretty painless so far. Almost easy.... I've done most of what I consider the hard part (the bulk of the software, wiring the BOB, etc...) and now just need to finish doing the mechanical design and get it built.... Shooting for 24 tools.

[MichaelHenry]

I tested every failure I could think of, and it resumed properly in every case.

It can be risky to test Murphy's law that way.