How to test operator to see if they can catch the defective parts?

[Berger]

Hi, Top management would like to see the plan for testing new operator to see if they can catch quality defect part during their production or not. The idea is they want us to create bad parts during the normal production run and let operator do the regular job and expecting they find these bad parts. Basically, we will not let operator know of this event and these bad parts cannot be marks/identify which will be notice by operator. I am scratching my head on how to do this? In order to create bad parts, someone has to stop the machine and change tool to create hole size, angle etc.,. The only defect we can create is raw material defect where we can make raw material bad and this will be a visual defect only. Now, This is a risk to ship this parts to customer as this will be tested during normal production run un less we need to hold all product from this lot and do a 200% inspection. Anyone have experience or any suggestion on how we can test operator for the quality defect during the production run? Please and thank you ..

Ps. I know we can create various defective parts and mix with other good parts and ask them to inspecting them during their quality check, which we can create more defects to fail during CMM, gauges and visual, but management want to be on their production process during a regular check.

 

[John Predmore]

I think you asked 2 questions, a) how to make bad quality part that isn't obvious and b) how to find the bad part in case the new operator missed it, in the batch of finished parts. You know better than I your process and quality requirements. You mentions a tool to make a hole, so I will assume some type of drilling operation in a block of raw material. You didn't say if there is any gaging involved.

a) you could slip in a block of different raw material which machines differently, a different grade of steel or heat treated differently. If the raw piece has to fit in a fixture, you could make a workpiece that doesn't fit the nest properly. The height of the block could be slightly lower or higher, maybe.

b) Let's assume the pieces do not have serial numbers. If the parts are placed sequentially in stacks or trays after drilling, the location in the tray would correspond to the position in the sequence. One way to make a hidden "red rabbit" is to color the part with ultraviolet dye, then use a black light to find it in the finished parts tray.

 

[Guest]

Hi, Top management would like to see the plan for testing new operator to see if they can catch quality defect part during their production or not.

Your top management need a lesson in process control. Deming would be spinning in his grave if he could read this. Why waste peoples' time on this futility? Fix the reasons for defects being caused. Do they know about PFMEA?

 

[Berger]

I think you asked 2 questions, a) how to make bad quality part that isn't obvious and b) how to find the bad part in case the new operator missed it, in the batch of finished parts. You know better than I your process and quality requirements. You mentions a tool to make a hole, so I will assume some type of drilling operation in a block of raw material. You didn't say if there is any gaging involved.

Yes, operators requires to check part with gauges on Finish good. It will be easy if I can throw the bad part on the finish good pallet before they checking it, but management need engineering/quality to make them create bad part (not us) and they need to be able to capture those bad parts. Engineering cannot just stop the machine and changing tool eg. different drills to make oversize/undersize hole, surface or angle without operator knowing it. M/C cannot do that itself. But, your suggestion below is really help me to get more on this.

a) you could slip in a block of different raw material which machines differently, a different grade of steel or heat treated differently. If the raw piece has to fit in a fixture, you could make a workpiece that doesn't fit the nest properly. The height of the block could be slightly lower or higher, maybe.

Yes this is a good one to add on, we can make part to not fitting to the nest, so operator should be quickly identify the defective raw material and not just place it anyway.

b) Let's assume the pieces do not have serial numbers. If the parts are placed sequentially in stacks or trays after drilling, the location in the tray would correspond to the position in the sequence. One way to make a hidden " is to color the part with ultraviolet dye, then use a black light to find it in the finished parts tray.

Again, yes we should use the ultraviolet dye to identify this bad part later on.

Thank you:

 

[Ron Rompen]

What you appear to be describing is (in actuality) a GRR study for operator inspection. Treat it as such.......take a population of KNOWN good parts, and add in your KNOWN bad parts. Number each of the parts (you want a fairly large sample size for this) and keep a record of which parts are known bad and good. Review the parts which were rejected by the operators, to see if they detected all the known bad parts, and if they rejected any of the known good parts.

 

[Miner]

I agree. This is called an attribute agreement analysis. This is a much safer way to get the information that you need without risking the NCs escaping.

 

[Sidney Vianna]

The end goal should always be to develop production processes so capable/robust that you could eliminate inspections as they would become non-value added. That's the essence of Deming's 3rd point of "Cease dependence upon inspection as a way to achieve quality."

It does NOT mean STOP ALL INSPECTION.

For multiple reasons, many organizations will not be able to develop processes that are reliably defect-free; so, while still aiming at better process capabilities via improvement initiatives, they have to rely on some form of inspection and/or testing (verification) for certain attributes to minimize escapes moving downstream. When that is the case, the inspection process itself needs to be effective and there is nothing wrong in making sure the inspectors are capable of detecting the out of spec characteristics.

 

[Semoi]

During my 6 sigma training I ask whether or not I am allowed to perform gage RR studies without telling the operators. Just like your management I would like to know the "real life performance" and not only the "when-I-am-extremely-careful performance". My trainer advised me to always involve the shift leaders or the group leaders and insure that the operators are properly "anonymised" before the data is evaluated. He told me that I could get into serious trouble with the labor union. Although I doubt that this really happens, I can see how it creates an environment of mistrust. Personally, I would not like to be evaluated without somebody telling me. Thus, the same is probably to for the operators working in production.

So, my advice is that you make sure to tell everybody in advance what you are going to do and what will be the outcome of the study. Do not underestimate how important it is that colleagues in operation trust you. Just remember, if your study shows that the process should be optimised, you will need their help/respect for doing so.

 

[Tidge]

I've been part of, and evaluated others, attribute agreement analysis (with measurements). I've never had it be anonymous. This is different than trying to "catch" operators allowing escapes. In my own experiences, we've discovered that some operators allow systematic bias into their measurements... and in one extreme case, we discovered an operator was measuring the wrong dimension (yet still passing the parts). If the union guys want to check each other's work all the time, that isn't necessarily a bad thing.

I wouldn't try to inject bad parts into a batch to see if people can catch bad parts. Others above have explained this very well. I would simply say that I do not want to be in the business of having a factory whose expertise is making non-conforming parts.