Tool breakage detection system

More
01 Feb 2017 23:51 #87139 by dextaca
Hello
I'm wondering if anyone implemented or tried to implement a tool breakage algorithm based on VFD / Spindle tool load?
Thank you

Please Log in or Create an account to join the conversation.

More
02 Feb 2017 13:11 #87151 by andypugh
I haven't heard of anyone doing that.
The difficulty I see is in telling the difference between a broken tool and a non-engaged feed.

If you can see a way round that, then I would consider setting up a "lincurve" HAL component to map the base-line spindle load against spindle speed, and then use a "comp" component to detect when the load is below the curve. The motion type (rapid or feed) is available on a HAL pin.

I guess a 2 minute filter on the low-force signal might work.

Depending on the CAM, you could perhaps filter on program lines, ie if a line of code is a feed and never crosses the threshold then something might be wrong.

Have you tried monitoring spindle load? You might be surprised how little it varies between engaged and unengaged for light cuts and small cutters.

Please Log in or Create an account to join the conversation.

More
03 Feb 2017 02:16 #87178 by Sparky961
I've always imagined a system that senses vibration and load to keep tabs on whether things are ok. The idea would work best for production, or long running parts. You'd monitor it yourself the first time and tell it "this is the base line". From there, either you set trip points for each tool or cut path, or when something starts loading up due to dulling you flag it so the system knows what's considered "bad".

Extreme deviations would cause an E-stop because it would mean a broken tool or launched part.

This could probably be done with sensitive accelerometers, which can be had inexpensively on a chip.

Please Log in or Create an account to join the conversation.

More
03 Feb 2017 14:25 #87212 by Todd Zuercher
I've seen some advertisements for such systems. (I'm sure they are quite expensive.) They appear to work like this, you make a known good run and the system records readings from it's sensors (both load and vibration), then compares readings from that good run to subsequent runs watching for readings out of tolerance and stops the machine or takes appropriate action. (too high for a tool that is about to break or is too dull, or too low for a broken tool.)

Please Log in or Create an account to join the conversation.

More
03 Feb 2017 19:02 #87240 by tommylight
I have seen one of those systems on a VMC in Germany from a US company.
Unfortunately, after 2 days of testing and adjusting and calibration they where so confident that they would intentionaly ram the tool at fool speed on the stock, well it did not end well at all. It broke the 1 inch tool from the side and continued to ram the spindle down till it tilted the head by at least 20 degrees, al the while sparks flying all over and even the ground shaking. It could have been a glitch, but i was terified watching that happen ( so were some 20 other ), and waisting a perfectly new machine.
Later i went on to discuss the system with the technician, but after he refused to give any info.
Now back on topic.
Anyone have a shock sensors from car alarms??? There are so called dual zone ones that have 2 separayely adjustable sensitiities with 2 outputs, they would do the job quite nicely wired directly to Linuxcnc. Use the high sensitivity one to activate at the lightest cuts, and the low sensitivity one to activate at max load. In Linuxcnc it should be wired so with and feed it should have hisens output on and if the lowsens activates stop the machine and raise Z axis. If there is a feed and none of the outputs are active, the tool is gone ( or stock ) or cutting air so just seting an alarm is enough, the machine can safely continue to work.
I do have some of those sensors around my shop ( i did alarm and security systems instalations for over 15 years ), but i do not have a VMC.
I do not know how much they do cost but a complete cheap car alarm is from 30 euro upwards, so very cheap, and ver effective for that job. Having 4 or 5 of those attached to a machine would work like a charm.

Please Log in or Create an account to join the conversation.

More
03 Feb 2017 20:50 #87255 by Todd Zuercher
Here is a video.

Please Log in or Create an account to join the conversation.

More
03 Feb 2017 21:06 #87257 by Sparky961
Yup, pretty much what I had in mind. It confirms that most things I imagine have already been commercialized.

Please Log in or Create an account to join the conversation.

More
03 Feb 2017 21:42 #87258 by andypugh

Here is a video.


Demonstrated on a tiny lathe, and not even using a real cutting tool.

Monitoring axis force might be more telling than spindle load, probably depending on the machine.

Please Log in or Create an account to join the conversation.

More
03 Feb 2017 22:07 #87260 by Todd Zuercher
Ideally, I would think you would want to monitor several things, axis loads, spindle loads, and vibration or noise. But unfortunately you would still need a base line run.

Please Log in or Create an account to join the conversation.

More
04 Feb 2017 03:33 #87274 by tommylight
That was ......well disapointing, the video i mean. It is not doing the only thing it is made for, cutting.
Bare with me for a moment:
All of you have smartphones, install the Physics Toolbox, open it, swipe from left, chose accelerometers, place the phone on the machine somewher that it will not fall while cutting, turn on the machine and check the screen, move all axes and check the screen, now cut something doing light cuts.
The point is, the diference betwen a not cutting machine and a cutting machine is huge, even at light cuts.
The other point is, i am fairly certain that the system i described above with at least 2 or 3 sensors is much more likely to work much much better that the systems described and seen on the web, case in point the video above, at a 20 to 40 $ price tops.
I have never seen those on a production machine ever, i have seen plenty of machines. The only one i have seen was the one i mentioned before, costing 40.000, mounted on a mori costing at least 200.000 and it all failed miserably on the first test, after 2 days of tunning.
I am not blaming the technology or the experts or anything or anyone, it is just that some things do not work beter by complicating them, never ever. I could add this last line to another thread going on in this forum, on another subject that has the same bottom line, namely simple is better.
Sorry for barging in, i will retreat now.

Please Log in or Create an account to join the conversation.

Time to create page: 0.204 seconds
Powered by Kunena Forum