hi, I have to do an msa in manufactoring. The controll is go/no go one on 100% of product. How many parts I have to give to my operator to controll? thank you very much best regards mattia

hi, I have to do an msa in manufactoring. The control is go/no go one on 100% of product. How many parts I have to give to my operator to control? thank you very much best regards mattia

Hello Mattia,

I assume from your post that you have to conduct an attribute gage type study? Have you looked at the various threads (http://elsmar.com/Forums/search.php?searchid=985250), posts and attachments (http://elsmar.com/Forums/fileslist.php?mode=allfiles&sortby=filename&pageamt=2&criteria=attribute+gage+study) here at The Cove? I am sure that you will find very useful information. Do you have access to a guidebook or the AIAG MSA manual?

Stijloor.

The AIAG standard reflected 50 parts, 3 Operators and 3 trials and some Customer specifics such as Ford also required this number. If you do not have a customer that dictates the this number or any other number, # of trials and # op Operators, then it would be up to your company.

I have had participants at a MSA seminar that stated that they used 20 parts and some use 30 parts but make sure that some of the parts a "bad". You have a choice of 2 or 3 Operators and 2 or 3 trials.

Again, it is up to you.

The AIAG standard I think that it's a good chooise or my study. But I want to know if the number of parts is influenced from yield of process and if the chooise of part is done random way from population or I must insered parts in percentile way... For example in 50 parts, 10 are from the first 20 percentile, 10 from the second and go on...

Thank you very much
best regards
Mattia

The AIAG standard I think that it's a good chooise or my study. But I want to know if the number of parts is influenced from yield of process and if the chooise of part is done random way from population or I must insered parts in percentile way... For example in 50 parts, 10 are from the first 20 percentile, 10 from the second and go on...

Thank you very much
best regards
Mattia
The AIAG standard I think that it's a good chooise or my study. But I want to know if the number of parts is influenced from yield of process and if the chooise of part is done random way from population or I must insered parts in percentile way... For example in 50 parts, 10 are from the first 20 percentile, 10 from the second and go on...

Thank you very much
best regards
Mattia

One would never be able to maintain a number of parts in each segment but you MUST have a number of BAD parts in the sample. If I were using a sample size of 40, then I should have around 8 BAD ones.

Hope this helps.

One would never be able to maintain a number of parts in each segment

Why not? How else can you have parts that represent the amount of tolerance being consumed by the process?

but you MUST have a number of BAD parts in the sample. If I were using a sample size of 40, then I should have around 8 BAD ones.


While it's true in general practice that "bad" parts should be included so as to verify that the measurement system detects them, sometimes it's just not practical, so the emphasis on MUST is not entirely accurate.

My idea is 50 parts, 2 operators (they are only 2), and 3 trials.
I want to put a number of bad parts proportionally at (1-yield of process for the specific difect in object), observed in measurements done by dea machine. And to maintain the same population characteristics I want to put any parts with other difects not involve in my analysis.
For you, Is it a good msa?
thank you very much,
best regards,

Mattia

Jim and David,

What if the process understudy is good that we cannot find bad parts? The Analytic Method in AIAG MSA Manual requires the use of out of tolerance parts besides conforming parts in order to perform the calculation, and its says must find bad parts. What if the process is really good and cannot find bad parts? Then how can we use this Analytic Method?

Jim and David,

What if the process understudy is good that we cannot find bad parts? The Analytic Method in AIAG MSA Manual requires the use of out of tolerance parts besides conforming parts in order to perform the calculation, and its says must find bad parts. What if the process is really good and cannot find bad parts? Then how can we use this Analytic Method?

In the past I have had Customer SQA representatives tell us to "create" a defective part if none can be found from process monitoring. This is also done as a practice for error proofing in line testing systems and sometimes is referred to as running a "rabbit" through the system. One thing you will need to do is ensure that the defective part is well controlled at all times in order to prevent someone from packing it for shipment. When operators, inspectors, et al are not accustomed to finding a defective part here and there there is a chance they wouldn't segregate it from the normal production parts.

For go no go gage we are checking on its accuracy and precision as per below,

Accuracy: What is the design intent of the control to detect the failure?
Do we use the worst case senario to design the control for the detection?
We should use the block gage or other standard to verify the control of the gage. Note: the control for the detection will wera out over time, fix the time to revisit the gage after releasing to line.

Precision: Are our operators have the same agreement when use the gage to check on the part which is good or bad?
Do their agreement match to the standard that we used in the study?
We should use the attribute agreement analysis to verify the agreement within the appraisers, between appraisers and against the standard that we known rating of each part.

:thanks:Thank u Big Foot!

Yew Jin,

Which method u referred to in AIAG MSA Manual? Is it the Hypothesis Test Method or Signal Detection Method? Or still use the Analytic Method but just to create some out of spec parts?:thanx:

There is a method available in Minitab - attribute agreement analysis which use kappa value calculation