Process Stability & Sampling - Can Sampling detect an Unstable Process?

[philip1985]

Hi all!

My first post on the forum. Nice to meet you!

Anyway, I am in the manufacturing industry for medical products, and would like to seek clarification on a few issues.

Firstly, regarding new production introduction, the management has tasked me to do 100% inspection on finished goods, the reason being that the process is unstable and they told me that only in a stable process can I resort to sampling. By the way, our company is currently using the ANSI/ASQ Z1.4-2008 Level I at AQL of 0.65%. I always thought that the inspection should be independent of the process stability. In fact, if the process is so unstable, I should be able to pick it out via sampling. Therefore, this criteria of having a stable process before I can use such sampling tools boggles me. Clarification is really welcome here.

Secondly, Level I samping seems to be abit lax here, I hear that most industries are currently using at least Level II sampling standard. Moreover, I am in the medical manufacturing industry. Any input will be great!

Lastly, on a specific note, let's say I want to find out the confidence level I have if I find no defects on a size 50 sample from a daily batch of 250 units. How do I go about it? AQL is at 0.65% as well.

Thanks alot for any help! Really appreciated!

 

[Stijloor]

A Quick Bump!

Can someone help?

Thank you very much!!

Stijloor.

 

[Steve Prevette]

Hi all!

Firstly, regarding new production introduction, the management has tasked me to do 100% inspection on finished goods, the reason being that the process is unstable and they told me that only in a stable process can I resort to sampling. By the way, our company is currently using the ANSI/ASQ Z1.4-2008 Level I at AQL of 0.65%. I always thought that the inspection should be independent of the process stability. In fact, if the process is so unstable, I should be able to pick it out via sampling. Therefore, this criteria of having a stable process before I can use such sampling tools boggles me. Clarification is really welcome here.

Thanks alot for any help! Really appreciated!

If a process is unstable, you increase the risk that a short term fluctuation won't be detected since you aren't measuring and plotting every result.

There are some criteria for increased sampling frequency and decreased sampling frequency in the standard. However, I'd suggest shifting temporarily to a 100% sampling for 25 to 50 sequential product, and plotting that on a control chart to see if it is in control or not. If in control, I'd suggest starting random sampling and also plot the results on a control chart to detect subtle shifts across several sequential results. If not in control, work on getting the product in control, using future updates of the control chart to help you.

 

[Jim Wynne]

Hi all!

My first post on the forum. Nice to meet you!

Anyway, I am in the manufacturing industry for medical products, and would like to seek clarification on a few issues.

Firstly, regarding new production introduction, the management has tasked me to do 100% inspection on finished goods, the reason being that the process is unstable and they told me that only in a stable process can I resort to sampling. By the way, our company is currently using the ANSI/ASQ Z1.4-2008 Level I at AQL of 0.65%. I always thought that the inspection should be independent of the process stability. In fact, if the process is so unstable, I should be able to pick it out via sampling. Therefore, this criteria of having a stable process before I can use such sampling tools boggles me. Clarification is really welcome here.

Secondly, Level I samping seems to be abit lax here, I hear that most industries are currently using at least Level II sampling standard. Moreover, I am in the medical manufacturing industry. Any input will be great!

Lastly, on a specific note, let's say I want to find out the confidence level I have if I find no defects on a size 50 sample from a daily batch of 250 units. How do I go about it? AQL is at 0.65% as well.

Thanks alot for any help! Really appreciated!

In addition to Steve's response, "unstable" doesn't necessarily mean that some parts are nonconforming. In this context, "unstable" refers to statistical stability. It's possible to have a process that's out of statistical control but still producing conforming product.

 

[Tim Folkerts]

Lastly, on a specific note, let's say I want to find out the confidence level I have if I find no defects on a size 50 sample from a daily batch of 250 units. How do I go about it? AQL is at 0.65% as well.

You might read through this thread: elsmar.com/Forums/showthread.php?t=33830
I think it will have the info you want about confidence levels.

 

[Steve Prevette]

The binomial function in Excel spreadsheet may be used to answer the question.

The confidence level basically is the 100% minus the probability of getting the result. But I need to state it about a defect rate. I'll use your AQL in this example.

What is the probability of getting 0 defects out of 50 if the chance of a defect were 0.65%?

=BINOMDIST(0,50,0.0065,TRUE)
which equals 0.72

So, I am only 28% confident that the defect rate is 0.65% or less.

 

[philip1985]

Thanks all for the prompt and detailed replies!

Something I forgot to mention, is that I am doing an attribute inspection, meaning that units are either pass or fail on a certain attribute. Does that change anything?

Thanks once again!

 

[Statistical Steven]

What is the probability of getting 0 defects out of 50 if the chance of a defect were 0.65%?

=BINOMDIST(0,50,0.0065,TRUE)
which equals 0.72

So, I am only 28% confident that the defect rate is 0.65% or less.

Just to add here....if you observe 0 defects in a sample of 50, you have only a 27% probability of sampling from a population whose percent defective of 0.65%. It would take a sample size of 458 to have 95% confidence.

 

[bobdoering]

What can be detected has a lot to do with nature of the instability. Intermittent versus systemic have two different probabilities of being caught. Also has a lot to do with the homogeneity of the material.

 

[philip1985]

The binomial function in Excel spreadsheet may be used to answer the question.

The confidence level basically is the 100% minus the probability of getting the result. But I need to state it about a defect rate. I'll use your AQL in this example.

What is the probability of getting 0 defects out of 50 if the chance of a defect were 0.65%?

=BINOMDIST(0,50,0.0065,TRUE)
which equals 0.72

So, I am only 28% confident that the defect rate is 0.65% or less.

Thanks!

Using the same formula, if I asked myself what the probability was of getting 0 defects out of 20 if the chance of a defect were 0.65%,

=BINOMDIST(0,20,0.0065,TRUE) = 0.88

which means that I am only 12% confident that the defect rate is 0.65% or less.

If that is the case, why are we using such low sampling numbers if the confidence levels are so low? Can someone enlighten me? I feel like I am missing something critical here...