Hand processes - How can you error proof human systems with out automation?

[qualitytrec]

Hey all,
I do not know if this is the right place to ask this or not but here goes anyway.
I have a little problem over here that is going to bite us hard if it does not get fixed.
We recently recieved an 8-D for a missed process. It is legit. Our process was pulling unprocessed parts from a table processing them then restacking the parts back onto the same table. Bad practice if one is not careful I know. We did not get dinged until we started allowing everyone to run the operation.
Anyway we "corrected" the issue by having the processed parts stacked on a separate table. I have found no parts missing the op now but a new issue has popped up. For some reason the operators can not get the full number into the packaging consistantly now.
How do you error proof human systems with out automating or mechanizing the system. This program is very short and will be done late November so no automation will be done for cost benefit reasons, but I suspect that we have the same issues in some of our other part processes.
Mark

 

[howste]

Your question is very broad, so my answer is equally broad: Mistake Proofing. Try doing a search of the forums for "mistake proofing" and "poka yoke"

 

[qualitytrec]

OK i have spent a large amount of time just now reading through some stuff here. It does not give me any ideas. Just it is the system no it is OE no it is always the system. Oh come on and just puke me why don't you.
I want a practical way to reasonably protect my customer from getting unprocessed product and to keep them from getting short shipped when dealing with HUMAN based processes because in the situation automation and mechanization is toooo expensive. And sorting is not good enough. I accept that it might be part of a solution but as discussed elswhere it is only about 80% effective. and the longer you sort the less effective it becomes in my opinion. So how about a little help on the examples given please.
Mark

 

[Charmed]

Can you paint?

OK i have spent a large amount of time just now reading through some stuff here. It does not give me any ideas. Just it is the system no it is OE no it is always the system. Oh come on and just puke me why don't you.
I want a practical way to reasonably protect my customer from getting unprocessed product and to keep them from getting short shipped when dealing with HUMAN based processes because in the situation automation and mechanization is toooo expensive. And sorting is not good enough. I accept that it might be part of a solution but as discussed elswhere it is only about 80% effective. and the longer you sort the less effective it becomes in my opinion. So how about a little help on the examples given please.
Mark

Dear Markasmith:

Since we are dealing with distinguishing a processed and an unprocessed part, my be you try something simple like applying paint (in a corner) or sticking a small piece of duct tape, etc. to tell the processed part. Would this work for your situation? Just a thought.

Charmed

 

[Laura M]

Can you put a scale under the packaging and weigh count? I don't understand how the new process negatively impacted packaging.

I was involved with a weigh count identifying a missing sheet of paper, so scales can be usefull even for light weights.

There are alot of errorproofing techniques. Processes and unprocessed parts on the same table does lead to trouble. Without knowing more about the process, I can't help further. Most packaging issues are errorproofed with weigh count, or a light curtain (counting number of time a part has passed through a designated barrier). Sometimes the container color - black parts on a black tray are hard to see - but maybe a white tray. An accurate incoming part count could lead to a processed part count. Mmmm...I have leftovers, I wonder what went wrong....

Does this prompt any ideas for you?

 

[Bill Ryan - 2007]

Is "one piece flow" an option? In other words, is it possible for one operator to take one part from start (of whatever process is causing the issue(s)) to pack?

Have you asked the operators for their ideas?

 

[Claes Gefvenberg]

I have found no parts missing the op now but a new issue has popped up. For some reason the operators can not get the full number into the packaging consistantly now.

Strange. That is interesting... I have to ask: Is there any possibility that the problem or the potential for it existed before, and has been detected now? If they could do that before... Why not now?

I suspect that we have the same issues in some of our other part processes.

That is a good lead. If you have similar processes, have a look at them asap. You may find similar or identical problems. If you do: Good. They need fixing, and you may be on to a systematic error. If you don't: Even better - find out what makes the difference.

And hey: Good luck.

/Claes

 

[qualitytrec]

Since we are dealing with distinguishing a processed and an unprocessed part, my be you try something simple like applying paint (in a corner) or sticking a small piece of duct tape, etc. to tell the processed part. Would this work for your situation? Just a thought.

Actually the processed part is already distinguishable. The process it was missing was intallation of a cone head rivet. We can not mark the parts because of operations the customer has to have done to it would be negatively impacted.

Can you put a scale under the packaging and weigh count? I don't understand how the new process negatively impacted packaging.

I was involved with a weigh count identifying a missing sheet of paper, so scales can be usefull even for light weights.

This is a posibility but it would have to be a fairly large scale and we would need several of them if the issue proves to be spread through out the other products we run. The boxes we pack are 45"x48", the parts are 1.75 pounds each, in the box there are four layers, each layer has eight cells, each cell gets ten parts.
Here is what interest me, We never found an issue (does not mean it did not exist) when we were stacking the ten parts of each cell all together but I got over ruled in packaging and now they are packing them in two stacks of five. alot of the underpacked parts are missing five parts. I think that the operators are getting a visual cue that the layer is done when in fact they are missing five parts. That is what I liked about the ten part stacking method.

Is "one piece flow" an option? In other words, is it possible for one operator to take one part from start (of whatever process is causing the issue(s)) to pack?

Have you asked the operators for their ideas?

Bill while one peice flow will not work I am trying to keep the line balanced which is another point of frustration. The Prog (has a mechanical counter) and produces the 320 parts into a container. The three offline dies are in the same press and run the parts straight through (they also have a counter). The riveter takes and runs them and always seems off count (we have not found a way to connect a counter yet). I have often wondered if the operators are not using the counters. Oh wait, of course they are.

Strange. That is interesting... I have to ask: Is there any possibility that the problem or the potential for it existed before, and has been detected now? If they could do that before... Why not now?

It is a possibility, though with this product there was never a customer reject when I was running the station. Now that we seem to have the resources better balanced and I do not have to run it we right away have issues. I have seen many things that they were doing different up front and we have addressed all the obvious stuff.
I am still trying to figure out why we can not get 320 from prog to offline to rivet. My boss says it is because we ship a half container each week also. He does not know why we can not balance that half container in there with the rest, but we don't. It still does not make sense though. we are of 5 or 10 parts when we are off.
gotta go sort so talk later.
Mark