87% Confidence Level for the Product Quality Control for Various Products

[Higgy23]

Hello Covers.

First off, this is my first post but I have been a visitor for many years off and on.

I hate to do this, but I have run into an issue that I am not sure how to handle and I have not found what I am looking for on here for 5 days, due to the vagueness of my situation. So here it goes:

I have been working with a US company that works with a large number of international factories on a huge variety of product, mostly sporting goods. I am working as a consultant to help with QC protocols that both help them achieve their goals, but also satisfies a sports specific certification that is mandated for a lot of their product (NOCSAE). Anyway, the national meeting is in 2 weeks and I received a note saying that one of the NOCSAE documents (001-13M13) will have a revision that recommends a demonstrated 87% confidence level for the product quality control for various products, of which affects my product line. This statement is very vague and to me I can only think of 2 areas in which this could make sense:

1. Confidence and reliability
2. Control chart upper and lower limits.

Without any additional information, I really don't know what to expect. With that amount of information, does anybody out there have any ideas as to what I might expect to see at the national meeting? I do not have a way to get in touch with the association or the person that sent the note until next weekend. Is there something obvious that I am missing and looking at this to be much more complex than it needs to be?

Any and all help is greatly appreciated. I apologize my first post isn't more challenging for all the experts on here.

Thanks, Brian

 

[t.PoN]

87% confidence level suggest that the limits are drawn on 1.51 sigma rather the usual upper control chart 3 sigma.

if you are accepting based on samples, then only 13% are allowed to be outside the limits (for example only one for 10 samples).

If you are working on Process capability index, it is Cp= (USL - LSL)/3Sigma not 6 sigma.

 

[Higgy23]

Thank you TPON. I was leaning towards the 1.51 sigma for the Upper and lower control limits, but just wasn't sure if there was something else that I should have been considering with the information presented to me. However, since we are currently accepting lots following ANSI Z1.4 and AQL guidelines, it is possible that they are removing the AQL portion of the standard and going straight to an acceptance level based on 13%. This gives me a little more to dive into for some preparations for the meetings next week. Thanks again for your time.
Brian

 

[Bev D]

This is a very unusually worded request / requirement. once you find out what they mean can you come back and let us know?

 

[Higgy23]

I was able to get a little more information as to the 87% confidence level. Here is an excerpt from the new protocol that will be introduced next week at the NOCSAE meeting:

6.2. Firms that manufacture protective products shall conduct a reasonable testing program that includes ongoing QC/QA protocols that demonstrate at least a 99% confidence level that all of the products that bear the NOCSAE logo/compliance language are in compliance with the standard relative to all portions of the applicable standard except impact resultant pass/fail criteria of less than 1200 SI. The preceding sentence is not to allow any helmet with a tested score in excess of lower level thresholds to pass but is to limit the statistical analysis to the 1200 SI threshold only, where SI is the measure.

6.2.1. Specific exemptions to the 99% confidence level have been made only to those products that epidemiologically do not present a risk of catastrophic injury. Those exemptions are as follows:
NOCSAE DOC. 001
10
　
6.2.1.1. Impact attenuation on soccer shin guards demonstrated confidence level of greater than 87%.
6.2.1.2. All mechanical pass/fail criteria to football gloves demonstrated confidence level of greater than 87%.
6.2.1.3. Face guard test mechanical failure criteria demonstrated confidence level of greater than 87%.

The red in the main paragraph is describing what is meant by confidence level, whether it is 99% or 87%.

This causes a new confusion for me. The majority of the NOCSAE protocols call out the need to create a sampling program using ANSI Z1.4 and AQL. I cannot see where there is necessarily a correlation between AQL and confidence level. Just about everything that know and have researched calls out that Inspection Level II would approximate a 95% confidence level for lot acceptance or rejection, based on the results from the sample testing. So I am not really sure what they are trying to accomplish by calling out either a 99% or 87% confidence levels. It kind of makes sense to me that they are trying to specify the sample size in a different manner than following the tables in ANSI Z1.4, but I am not sure that I know how to tackle it. It is not as if the difference between 95% confidence and 99% confidence is by increasing the sample size letter by 1 row. And the same for 87% reducing the letters by 1 row.

Not sure where to go from here.

 

[ISAN16]

Hi,

we are practice this format.

 

[Steve Prevette]

I would lean towards this is referring to option 1. If the company states they want a 95% quality rate, then this would become "I want to be 87% confident that no more than 5% of product are defective".

Control chart limits are always set at three standard deviations, so the statement of 87% confidence should not change what you do with your control limits. It may change your interpretation of the control chart results.

 

[Higgy23]

Thank you all who took the time to respond to this thread. I spent a couple days at the conference and got a better understanding of what the are looking for.

Steve - your are spot on. The explanation I was given was that the confidence levels had to do with the SPC during production, not the QC inspection of finished goods. The example given was on a helmet requiring a 99% confidence level. Each of the various measurements made through production should be held to a tighter pass/fail or variable tolerance than a soccer shin guard which needs to be at 87% confidence level. So, one measurement that is taken during the production is the weight of the helmet shell after injection molding the part. This weight needs to have a tight enough tolerance, or maybe better stated, be in tight enough control that 99% of all the helmets pass this check point. Where this gets confusing again for me is how to interpret this. As of right now, the weight target as well as the tolerance has already been given to me. So should I be looking at changing the tolerances in such a way that my confidence levels are met, or should I be increasing the number of samples measured through the process to meet this? If my control chart UCL and LCL are driven by the data, then what should I be doing to show the governing body that we have a QC/QA protocol in place for the product that meets the desired confidence level? I am sorry if there is an obvious answer to this. I haven't found it yet. Maybe I am asking the wrong questions in my searches. A nudge in the right direction here would be greatly appreciated. Thanks, Brian

 

[Steve Prevette]

. Where this gets confusing again for me is how to interpret this. As of right now, the weight target as well as the tolerance has already been given to me. So should I be looking at changing the tolerances in such a way that my confidence levels are met, or should I be increasing the number of samples measured through the process to meet this? If my control chart UCL and LCL are driven by the data, then what should I be doing to show the governing body that we have a QC/QA protocol in place for the product that meets the desired confidence level? I am sorry if there is an obvious answer to this. I haven't found it yet. Maybe I am asking the wrong questions in my searches. A nudge in the right direction here would be greatly appreciated. Thanks, Brian

Dr. Wheeler explains this very well as the "Voice of the Customer" versus the "Voice of the Process". The "Voice of the Customer" has given you what the targets, promises, and confidence level requirements are. Short of renegotiating with the customer (who may be internal) and changing their requirements, these requirements are a given.

The "Voice of the Data" is what you get on the Control Chart, or in Acceptance Sampling. These show if the product (and its production process) is "capable" of meeting the requirements, the Voice of the Customer. IF the production is NOT capable of meeting the requirements, no amount of altering the measurement process will make the product capable (unless there was some underlying measurement error). You are then faced with - how do I change the production process in order to meet the requirements reliably.