Drill Bit Breakage Sensor Ideas

Greetings!

I have been given the task of searching for methods of detecting broken tooling for our CNC Drilling machine.

Our drill machine is a CNC operated gang-drilling machine use to drill holes into paper rolls. This machine utilizes 180 drill heads driven in tandem by a spine shaft. We are looking for a good way to help us sense when drills have broken during a drilling cycle.

The main challenges we have are as follows:

1.) With up to 180 drills that could possibly be in use at a given time, a laser sensor on each drill head would likely prove to be too cost-prohibitive and cumbersome to integrate to the machine. It may also be difficult to use such technology during use with our coolant system, or when we are making heavy chips.

2.) Ideally, we want to be able to detect broken tools before, during or after a drill cycle is complete. If a drill breaks and sticks in the metal shell of a roll during entry, the broken end remaining in the collet may continue to feed into the roll, and can cause significant damage to the roll cover. We are typically drilling into a pre-existing hole, and when a bit breaks, but does not stick in the hole, we want to be able to detect the broken tool in the collet after it retracts, and before it goes to the next cycle.

3.) The cutting forces generated by this machine are created by dozens of tools cutting at the same time, so overload sensing would be very difficult to detect for a single broken bit. Our most commonly used drills are fairly small (.125"-.176"), and do not generate spikes in cutting forces when they break.

My questions:

1.) Are there any options out there that could visually sense that many tools at one time, and perhaps be integrated into our control to stop the machine if/when it detects a break?

2.) Are there any options that use audio frequency detection? When bits break outside the shell, they make a pretty distinct noise which we feel could potentially be used as a source of detection.

I have attached a few images of the machine to show the setup of the drills. Any advice is greatly appreciated!

Thanks!

Connor

Apologies for the sideways pictures..they should be rotated clockwise. Still getting used to the site

Can you monitor the current at the drill spindle? I suspect that at the moment of breakage there would be an increase in RPM and a drop in current. The controller would compensate nearly instantly but there should be a detectable change for a very short time, probably less than a millisecond or two.

Another possibility is the current to the servo driving the Z. A broken drill should cause a change in current.

Once you have determined what is possible to detect it should be possible to instrument it. Fancier would be LabView with appropriate input circuitry, simpler would be a data logger with an alarm window set.

Edit: Confession time - I responded without opening the pictures. Too many spindles to make what I suggested practical.

Interesting problem :-)

first thought was a pressurised collar through which the drill passes on its way into the job. It should be possible to detect pressure changes as it moves.

second thought, a mechanical lever/switch that traverses down the whole row of tools checking each drill tip as it passes after each drill cycle.

third thought, an automated visual inspection (camera) of the holes after drilling using pattern recognition software. (this might be getting a bit CSI :-) )

Bill

Question - why don't you punch the holes?

We produce custom assembly/production line stations. We use a green light source behind the holes and a camera with a filter (matched to the source) to detect the holes. The coolant can cause problems if you don't use an appropriate, matched light source/filter.

interesting indeed,

I´d say you are already having some good thoughts. Should be very possible to "hear" with a mic and suitable software. Two or Three mics could maybe even tell you which one. This is however most likely "build your own development and verify one off kind of thing". Essentially it could be similar concept as for Electronic target but the application and sound should be rather different.

Like Bill said Vision Cam could also be possible.
I think here you can also get very good help and on-site testing from system providers. Sick, Cognex are two names coming to my mind. Challenge will most likely be lighting, you need good constant light and contrast. It´s better today than some years back but still. Good light and contrast is necessary.
One set up could be enough cameras to see all the drills every time they come up and then you can get alarm if one is broken before next drill sequence.

Continous visual monitoring of drill length after retraction to detect broken ones is certainly possible. No idea what would be most economical with current technology but back in the 1980's I helped demo a simple system that used a 16 x 60 element linear photodiode array to to something vaguely similar. Idea was to use a distorting lens to compress the image field of view onto the detector and see which elements were shadowed by the object of interest. In your case the objects would be the drills showing up as a dark line in the 16 element direction when pulled back out of cut. I guess on thta system each drill image would be 2 or 3 elments wide. If the drill is whole then say 15 elements would be covered. If broken less than 15 would be covered. The difference being pretty easily detectable. Need decent lighting on the other side of the drills and some sort of retracting cover to keep coolant off the optics except when inspecting. Would need subsidary optics in the form of a lenslet array directy over the detctor as well as the primary lens to get the right balance of image size between drill image and intervening spaces. Basically the drill image needs to be made wider and the space image compressed. At the time systems using custom molded optics to do such things were seen as a coming technology and expected to be affordable. No idea whether it ever happend as a commercial product as it wasn't really my field.

Diffractive opical systems producing arrays of points or lines are said to be fairly affordable these days so maybe something similar could be done that way. Probably need a scanned laser diode source if the diffractive bit is on the imaging side. Plain laser diodes would do if a set of lines were projected onto the drills. Need maybe 10 or 20 sources. If you went for line array projection an off the shelf linear array camera would probably work fine.

Audio detection of breakage would work well if the sound is sufficiently distinctive. I've used similar with geart success to supress mechnically generated noise in the output of an IR detection system with inherent, and seriously unwanted, vibration sensitivity. Tuned SAW filter is the elegent and inherently fast way. But these days microchips have MIPS to burn so run the sound through a continous Fast Fourier Transform analysis and look for the characteristic combination of frequencies. I imagine suitable programmable FFT modules can be got off the shelf. Heck some of the more sophisticated PC sound cards might be up for it.

If you don't need continuous monitoring of drill length then use audio detection of breakage to trigger machine stop and force drill retraction. A fairly basic solid state camera could be mounted on some sort of sliding carriage and driven along a rail covering the whole length of the machine looking for short drills. There will be a trade off between using natural light and a more complex image analysis program or including a monochromatic light source with simpler alaysis. I'd use a modulated multi line source with a linear array camera so as to get multiple measurements off each drill. But thats me. I don't like single point detection and have an inordinate affection for systems slightly over engineered to be essentially bullet proof. That one comes from being exposed to SPRITE infra red detectors at an impressionable age!

Clive

How much money do you have to spend.?
Would buying longer life or changing drills more often be cheaper?
Certainly you can do this with enough cameras if you can turn the coolant off for the check but it won't be low buck.
Both acoustic and power sensing are going to be very iffy here and very, very hard to tune.
A change in drill vendors or a change in the vendors grinding machines will toss you out to no man's land unless you have someone real good inhouse on the software side on this type feedback stuff.
Bob

Acoustic won't be hard to tune. Especially if you accept a small false alarm rate. Around 5% or better should be doable. The snap sound spike will be quite distinctive. I see no reason why a divergence band can't be made wide enough to accommodate different makes or types of drills ca. Fairly trivial exercise actually for a competent person with some experience in the field. Especially if typical "snap sound data" exists for drills. If such data doesn't exist I'd darn well want to know why not. If I were an experienced person in this field I'd expect to have a robust demonstrator up and running in a few days. The mechanical noise suppression system mentioned in my previous post took two of us 4 days to build and de-bug. Instrumentation quality reel to reel tape recorder and discrete components in those days too. The wanted optical signal was buried, at least 10dB down on the noise too.

Visual inspection ought to be low buck but last time I was involved in that sort of thing the consultants and camera suppliers wanted big bucks for little work. So I rolled my own via LabVIEW in a couple of weeks. Wet finger estimate for a rail mount scanning system with linear array camera and low cost diffractive element laser diode line source would be £5,000 to £10,000 if done by an experienced, up to date person. Plastic diffractive optics are pretty cheap and you should be able to get a line pattern off the shelf. Adequate diode modules are almost commodity items. Best guess is £500 to £1,000 for the line source. For example Thor labs will sell you a decent linear array camera with pattern matching software under £1,000 (before tax). OK its not ideal being single row so it needs to be mounted at an angle to check the whole drill length but it does illustrate what's out there. For comparison a basic semi-scientific USB CMOS camera is £260 or so from Thor complete with PC software but whether you can easily set it up to do line length analysis I don't know. Skads of cheaper stuff out there but integration may cost you much more than the savings.

Bottom line is it should be pretty easy for a guy who knows.

Clive

Quote bill
[second thought, a mechanical lever/switch that traverses down the whole row of tools checking each drill tip as it passes after each drill cycle.]
Or a rider bar fixture and micro switch to each cutter (that might crash), a switch rail that would be mounted with the job that would have an arm to ride each tool. But yes if they break long the riding arms might not work well. Or if the arms are made long so made to bend up at feed might tangle long chips (with shot or few long chips it would workkwell to shut down or turn on light to aleart operator so shou down. . We had much the same problem in a manufacturing line where a broken something would take out a line of cutters. our remedy was a tool change frequency that ran most tools at about 60% of expected life.
You might try a high helix split point so giving less drag at a chisel point resistance and leaning more to lift than twist at break through at pulling in the burr. Might seek the heavy web drill and taper thin at point.
If the part is thin and wide you might benefit with having a heavy spring hold down at one or a few places to that would push down as the drilling was being done to avoid pull up as drills pass through. All those drill passing through at the same tome is a lot of pull-up. With finding pull -up you might alter length so perhaps only 25 or 33% make pass through at the very same time. Ye s increasing cycle time for longer feed.

Special drill would have a hump at flute face to retain chips closer to web so adding perhaps 30%+ more strength to breaking.. with having a split point and a gash chip breaker. Yes big bucks for special drills.

By the looks of it, it appears that a visual monitor of some sort is probably our best path to investigate. Though the process of drilling with this machine is fairly straight forward, there are so many variables to consider when it comes to monitoring these tools. It's a very unique machine and a very unique process indeed!

Thanks all for the advice, and please keep it coming. As we investigate further, I will post with any updates and share what we learn!

Connor U

Time might be a concern for checking but it is not mentioned. In such a typical process 1 second is way too long to scan the row.
If you could afford the time a laser probe on a rail would work and be cheap but it would not be fast.
Labview is sort of a joke in machine vision. No high speed work here.
The coolant really screws with most camera/video systems. Stopping and restarting that flow every check may or may not be a oh-shit.
That is one hell of a field for one line scan to cover not to mention the coolant.
180 drills, the snap is way more than 10db buried inside. Not like a 4 spindle head which would be like falling off a log.

Does a lot of the the drill break off and they are always near the same dia. and in line?
How about a simple mag prox on a slide rail that scans the row and you id the signature for different configurations.
Robust, collant proof and reasonably fast. Sort of a steel magnetic bar code id.
Would find missing drills, not gonna find .010 missing off the tip.
We do something like this to id fixture numbers with notches cut in them as they are placed into the grinders so we know which offsets to use.
Basically as they slide into place we count the notches and spacing with a prox so each has a unique signature for the controller to reference.
(and then there is the brute force method of a simple swing down bar with 180 proxes on it)
Bob

I should also speak a little to the drilling process as well. This may help explain some of the problems that create broken drill bits on our machine.

Typically, we are drilling into a newly covered paper roll, which consists of a metal shell (either brass or stainless steel) covered by layers of a hard composite and polyurethane (or rubber). This steel shell has a pattern of through-drilled holes along the length of the roll, and it is the job of this machine to drill holes into this cover which match up to the holes in the steel shell. Typically, we are drilling these cover holes about .016" to .032" undersize to the shell hole. This may seem like a lot of room for a CNC machine tool, but when we are dealing with old rolls that have distortion, or if we have any alignment issues during setup, the drills may end up contacting the metal shell, and thus breaking. The cover materials are also fairly rough on the drills, and contribute to the breakage too.

The main concern when a bit breaks has to do primarily with the risk associated in damaging the cover. As of now, the only option we have is to have the operator be at the switch during the entire process, and use the cycle stop button to pause the machine. With some drilling steps taking as much as 5-10 days, this is not a very reliable method to catch these broken bits. And even if the operator is at the switch, sometimes breakages occur before the operator can see that they happened.

With this post, we hope to find a viable option to aid the operator in catching broken bits before they cause damage, whether by alerting and prompting the operator to act, or by directly interfacing with the machine and stopping the cycle.

Connor U

Conner looks like you are a new guy to PM so be sure to tell us how you solve the problem

Buck

Spring contacts that are calibrated to make contact with the drill tips upon retraction would 'take inventory' at every withdrawl. Any missing drill ends would be noted, machine stopped, and corresponding indicator light illuminated to mark the breakage. The sensor could easily be adjusted to capture even a 1/8" or so tip break-off. Pretty immune to coolant and swarf build-up, and much simpler/cheaper than vision, audio, or laser systems. Sort of like michiganbuck's method, but without microswitches -- just low-voltage, low-current, thru LED, to ground.
Secondary sensing could be verifying thru-hole by light (low light level indicates incomplete hole) or physical probe with mechanical jamming protection.

Chip

Chip Chester said:

Spring contacts that are calibrated to make contact with the drill tips upon retraction would 'take inventory' at every withdrawl. ...

Click to expand...

I like this outside of spring contacts on running drills every cycle. Million plus cycles and wear on the sensor end? Carbide tips on the switches?
Not seeing stopping the the drill or the coolant for each check every cycle. Both of which would make things nice and much easier, both of which would cost much cycle time and $$$.
Maybe different points of view as my customers demand million cycle no charge warranties on such devices in the fine print on the purchase orders.
What works on the bench can be very different that what is robust in a production environment.
You want such a system to run for decades and never screw up with as little maintenance time as possible. Zero maintenance is best. No cleaning a lens or window.
You do not want any bad parts getting out over hundreds or even thousands of millions of cycles.
These numbers may seem out of the range of thinking to the point where they make no sense.

Try building and inspecting spark plugs.
We used to build a million a day here. Not a month, not a week, a friggen day. Try checking all those threads and gaps 100%. Where did they all go???
How can the world use that many spark plugs every single day? Literally train cars running though the kilns.
Then, think of the carbide form tools being eaten up on the shells in the Acmes.
Bob

Drill breakage

Connor_U. said:

Greetings!

I have been given the task of searching for methods of detecting broken tooling for our CNC Drilling machine.

Our drill machine is a CNC operated gang-drilling machine use to drill holes into paper rolls. This machine utilizes 180 drill heads driven in tandem by a spine shaft. We are looking for a good way to help us sense when drills have broken during a drilling cycle.

The main challenges we have are as follows:

1.) With up to 180 drills that could possibly be in use at a given time, a laser sensor on each drill head would likely prove to be too cost-prohibitive and cumbersome to integrate to the machine. It may also be difficult to use such technology during use with our coolant system, or when we are making heavy chips.

2.) Ideally, we want to be able to detect broken tools before, during or after a drill cycle is complete. If a drill breaks and sticks in the metal shell of a roll during entry, the broken end remaining in the collet may continue to feed into the roll, and can cause significant damage to the roll cover. We are typically drilling into a pre-existing hole, and when a bit breaks, but does not stick in the hole, we want to be able to detect the broken tool in the collet after it retracts, and before it goes to the next cycle.

3.) The cutting forces generated by this machine are created by dozens of tools cutting at the same time, so overload sensing would be very difficult to detect for a single broken bit. Our most commonly used drills are fairly small (.125"-.176"), and do not generate spikes in cutting forces when they break.

My questions:

1.) Are there any options out there that could visually sense that many tools at one time, and perhaps be integrated into our control to stop the machine if/when it detects a break?

2.) Are there any options that use audio frequency detection? When bits break outside the shell, they make a pretty distinct noise which we feel could potentially be used as a source of detection.

I have attached a few images of the machine to show the setup of the drills. Any advice is greatly appreciated!

Thanks!

Connor

View attachment 166303View attachment 166304View attachment 166305

Click to expand...

First, ask yourself what is causing the drill breakage? I would start by looking at more strokes to clear the chips so the flute don,t fill and bind the drill and cause the breakage. Start keeping records for the length of time before the drills become dull and for breakage. Do Satisical analysis to determine how long the machine can be run till drills need sharpening. Then start a maintenance program and stick to it. I don't see how trying to sense drill breakage with that many drills has a chance to work.
Good Luck
Rogertoolmaker

Not much way to do this "CHEAP". My suggestion is 8 mm proximity sensors on a mechanical pivot mounted on the top of the carriage. During index of the roll, they pop down and check the drills. This is probably the least expensive route. You are going to need quite a few sensors and cables and probably and a couple input cards in the PLC rack, but it will be reliable and the coolant won't bother it. It won't detect 0.02" missing but it will detect a drill broken off by 1/4" or so.

I would think that there would be a pretty unique sound transient when the drill breaks that would be transmitted through the frame of the machine. With a hard mounted piezoelectric sensor you might be able to pick this sound out of the background machine noise.

wheels17 said:

I would think that there would be a pretty unique sound transient when the drill breaks that would be transmitted through the frame of the machine. With a hard mounted piezoelectric sensor you might be able to pick this sound out of the background machine noise.

Click to expand...

That seems like a great idea but I just don't see it working 98% of the time even with all the cpu power we have now.
Have you done such as this? Any good software techniques for pulling this out of the noise in the signal from the machine itself cycling or someone dropping a hammer or a forklift dropping a load?
People are real good at this, computers not so much.
My dog will fool my grinding gap sensors by getting excited over a biscuit near a machine
Bob