Selecting Parts for an MSA (Measurement System Analysis)

[Black arrow]

Hi,

We will soon check if our measurement system is adequate for spc and cap.studies so % Study Var and ndc are important here. We have already selected parts and from my point of view it is important to select part from the normal process so those were picked during the whole day. So far so good i guess.

BUT My friend works at a company that also includes customer complaints parts in their MSA, complaints they have had during the year.
Is that correct approach?

 

[Miner]

Re: Select parts-MSA

If they are using %Tolerance, this will not affect their results. They should not do this if they are using %SV or ndc because it will bias the results and make them look better than reality.

 

[Bev D]

The intent is to have parts that are representative of the full range of variation that the process typically produces. Customer complaint parts are certainly part of this range but if you include too many of them you will skew your study (biased toward the worst parts)


I'm also not sure that a single day is representative of the full range of variation that your process will experience.

I do know that if you use the AIAG default of 10 parts you will have a poor estimate of the process variation regardless of the above concerns. 10 parts is just too small to reliably estimate the process variation. I typically use 30 parts (measured only twice).

 

[Black arrow]

Hi,

I selected 50 parts and measured them and calculated standard deviation. Then i selected 10 parts from those 50 so they will have same standard deviation (spread) as those 50 i selected first.

Hmm, Bev you said " process typically produces". How many customer complaints parts should he really select here? If he for example produce 500 parts and 2 of them are customer complaints, should he involved them? I could find it tricky here to define what is typically normal process.

 

[Bev D]

IHmm, Bev you said " process typically produces". How many customer complaints parts should he really select here? If he for example produce 500 parts and 2 of them are customer complaints, should he involved them? I could find it tricky here to define what is typically normal process.

it really depends. if the customer complaint parts have values that are outside the range of values captured by your internal parts then you shoul dadd some customer compaint parts. if they within the range of the internal part values there is no additional value that I can see at this point to adding customer complaint part.

 

[Black arrow]

it really depends. if the customer complaint parts have values that are outside the range of values captured by your internal parts then you shoul dadd some customer compaint parts.

Even though it is 1 customer complaint part of 1000 produced parts ?

 

[Bev D]

OK it's not about whether or not the parts are customer complaint parts. nor it about achieving some %TV or %PV goal. what is critical is to understand the variation of your measurement system.

so the question is: does the Customer complaint part(s) have a value that is larger or smaller than your in house sample set? if so, include it. if not don't include it.

these things are never cookbook unless you just want to 'check the box' and day you did it.

if you are truly trying to understand, you must think about your measuremetn system and how the process works to create variation in the characteristic of interest. you then design the study to answer the question(s) your are intending to understand.

 

[Black arrow]

Thanks Bev D for your support!!