Confirming correct change part

I have a machine that requires different sized change parts to run different products and having the wrong change part will cause a pretty serious crash. Right now, we simply require you to remove the part to home and each time you change jobs. I'd like to confirm the part is correct on future machines (and hopefully remove the annoying removal/re-installation of the part when it doesn't need to be changed). I just need to be able to differentiate between 4 or 5 parts.

I'm about 95% sold on RFID for this, but if there is a better option, I'm open to hearing it. I know we've done barcode reading in the past, but ran into a problem with the barcode getting scuffed up. I'm stuck with Sercos III and CANopen as my available protocols, Modbus TCP is doable as well, but it would be a hassle. Annoying since Pepperl-Fuchs has an RFID product I like but not for Sercos/CANopen and Banner/Turk's CANopen RFID stuff is a little bulky. Sick was bulky and not Sercos/CANopen.

Have you looked at a part recognition system?
We used Cognex with good success.

regards,
james

I think a vision system is a little over kill. I'm just checking to see if a hunk of aluminum that clamps in the machine is the right hunk of aluminum for the recipe/job or not.

Do they make light curtains/grids made for measuring? If so you can set that up to read the size of the object set down.

A laser sensor that will measure the distance from the block to it, if the block is set down at the same point that should be able to tell the size of the object.

Weight sensors that will take the pre-clamp weight and not let the clamps work unless it is right.

We've used a set of proxes that detect rest-buttons or drill spots on components to identify them.

I think right now we are either going with multiple retro-reflective sensors as a kind of bit code (cheapest option for small number of possible parts) or RFID.

This machine currently has 5 different possible parts, so I just need three sensors and the IO points to take them. $100 for a workable solution, but it is kind of a hack job.

RFID can check for a part number match, but it also leaves us open to simply store the parameters of the part on the tag, so new parts would program themselves into the machine.

I think at this point we are committed to RFID as a sales option and if they don't buy it, we may do what we do now (require you to remove the part to home, change jobs, etc) or do the retro-reflective bit code.

danatomega said:

We've used a set of proxes that detect rest-buttons or drill spots on components to identify them.

Click to expand...

Good idea, way cleaner than the retro-reflective. I could even drill the same pattern in each one and code by plugging holes with little steel bolts for better prox performance!

CapinWinky said:

Good idea, way cleaner than the retro-reflective. I could even drill the same pattern in each one and code by plugging holes with little steel bolts for better prox performance!

Click to expand...

I've seen this done before, it works well and is much simpler than the alternatives as long as you don't have too many parts to change out. You can create a binary code with (2^n - 1) combinations where n is the number of prox sensors. The -1 is because you probably don't want to use the combination of all sensors off, since it could also mean that no part is installed.

and having the wrong change part will cause a pretty serious crash

Click to expand...

how serious is this crash? I mean, what happens when using prox or opto sensors and there is a defective sensor or wire break?

Good point on a sensor that doesn't read when it should. The crash, if as spectacular as possible, would definitely make it where they couldn't run a particular job (by destroying the change part), and a medium risk of destroying some automatic adjustment mechanisms. It has low risk of taking the machine down 100%, and since I'm confirming the part instead of adjusting to it, the sensors would have to fail just so to detect the part for the job.

We've decided to plan on RFID with the prox array as backup if the RFID system doesn't work out or we don't have time to work with it.

Instead of using all possible combinations of prox switches represent a part, consider using more proxes and limit the valid patterns.

For example, get seven parts identified by using four proxes instead of three. Only use patterns with two or four proxes on as valid. That gives you seven possibilities, but any pattern where a single prox fails in either the on state or the off state will result in an invalid pattern.

Use the PLC to do a redundancy check. e.g. Between changes make sure all sensors have gone on/off-off/on.

and since I'm confirming the part instead of adjusting to it

Click to expand...

Ah I see, it's more like a double check

Steve Bailey said:

Instead of using all possible combinations of prox switches represent a part, consider using more proxes and limit the valid patterns.

For example, get seven parts identified by using four proxes instead of three. Only use patterns with two or four proxes on as valid. That gives you seven possibilities, but any pattern where a single prox fails in either the on state or the off state will result in an invalid pattern.

Click to expand...

We use this method on our machines that slide in on trolleys. There can be assembly, test, or print stations at any of the docking points. There are a set of proxes that identify which type of station is docked.