AQL and Sampling Plan as they relate to Functional In-Process Inspections

[Pudge 72]

OK Cove.

Throughout my career I have always been weak in the area of Sampling Plans and AQL levels as they relate to functional In-Process inspections whether it is the operator checking the part or as a QC function.
So, here goes the question. Has anyone created or does anyone have a good plan in Excel or Word for determining the amount of pieces to be inspected on an hourly basis as would be required per a Control Plan for production parts? Or, a document that explains a good methodology for the selection of # of pieces and time that would be acceptable to a customer?

IE - I check a feature on a part that is a length let's say. This Length is listed as a Feature that is Inspected on the Control Plan, but, it is not a customer critical. I inspect 5 pcs. per hour. I do not do CPK on this feature as it is more of a "preventive" inspection so that I can see a drift during the run to assist with an adjustment if necessary.
On this part, I have let's say 5 or 6 of these types of features to inspect and I am measuring 5 pcs. per hour.

I do 5 pcs. per hour because that is what I have always done and it seems to be economical whether it is the operator or QC - seems to be effective. But, what is the real or better way to arrive at the amount of pieces or time frame in the beginning of the project? I can't change the IP based on rolling data because it is submitted with my Control Plan - it needs to be developed on the Front end.
Any Ideas guys?? I hate the norm and never thought that this method was quite right.

[Jim Wynne]

Originally Posted by Pudge 72

OK Cove.

Throughout my career I have always been weak in the area of Sampling Plans and AQL levels as they relate to functional In-Process inspections whether it is the operator checking the part or as a QC function.
So, here goes the question. Has anyone created or does anyone have a good plan in Excel or Word for determining the amount of pieces to be inspected on an hourly basis as would be required per a Control Plan for production parts? Or, a document that explains a good methodology for the selection of # of pieces and time that would be acceptable to a customer?

IE - I check a feature on a part that is a length let's say. This Length is listed as a Feature that is Inspected on the Control Plan, but, it is not a customer critical. I inspect 5 pcs. per hour. I do not do CPK on this feature as it is more of a "preventive" inspection so that I can see a drift during the run to assist with an adjustment if necessary.
On this part, I have let's say 5 or 6 of these types of features to inspect and I am measuring 5 pcs. per hour.

I do 5 pcs. per hour because that is what I have always done and it seems to be economical whether it is the operator or QC - seems to be effective. But, what is the real or better way to arrive at the amount of pieces or time frame in the beginning of the project? I can't change the IP based on rolling data because it is submitted with my Control Plan - it needs to be developed on the Front end.
Any Ideas guys?? I hate the norm and never thought that this method was quite right.

Use of a sampling plan with an associated AQL or OPerating Characteristic Curve is usually done with finished lots of material, as in the case of final or incoming inspection. Determining sampling quantities and intervals for in-process checks is a little like setting calibration intervals. You do it based on experience and history. You should monitor closely in the beginning (what "closely" means will vary with the processes involved) and then gradually expand the interval until you're reasonably sure nothing bad is happening between inspections. The goal of these checks should be process control--knowing how to control the variables that can affect product quality.

[Pudge 72]

Excellent.
This is exactly the type of feedback I am looking for.
I agree with you completely Jim. The philosophy you suggest is along my lines of thinking - but, based on my expierience, it will be hard to implement as it almost falls into QE territory whereas you base the inspection interval on historical data - the problem falls in where someone has to analyze all of that not to mention that it has to run for a while to gather this data.
So, while it could be said that in the end you will have a cost savings based on eliminating unecessary inspection, you will spend a lot to get to that point.
That is kind of my ultimate dilemma, in this day and age of listening to how much QC (by the operator or direct QC) is a non-value added activity, we don't seem to have good definitions in place for analyzing In-Process data to arrive at these points and eliminate the waste during production which is where the rubber meets the road - not Final inspection.

[Jim Wynne]

Originally Posted by Pudge 72

That is kind of my ultimate dilemma, in this day and age of listening to how much QC (by the operator or direct QC) is a non-value added activity, we don't seem to have good definitions in place for analyzing In-Process data to arrive at these points and eliminate the waste during production which is where the rubber meets the road - not Final inspection.

That's exactly what statistics is for.

[Pudge 72]

I know, but I am sure that you feel my pain when I try and explain and justify someone analyzing data - even if it is for improvement. Just wondering if you or anyone else had a better method right out of development - IE - CPK of 1.75 = 3 pcs. per hour for In-Process Inspection instead of 5 based on a certain lot size or something. I am not sure, I know how to apply the statistics on the historical data, but to change all of the paperwork for sampling sizes is a pain - probably trying to take a lazy way out that is really not there.