Reducing Variation via Multiple Readings

[jelly1921]

In Section G of MSA 4th edition, it recommended a way to reduce variation by increasing the times of readings, if the current variation of MS is not acceptable ( over 30%) --- does it also apply to 10% < GRR < 30%?

In other words, if I did GRR analysis ( 10 parts, 3 appraisers and 3 times), and it turns out GRR = 26.68% as shown in the mannual, can I increase the reading times to let GRR < 10% ?

 

[Miner]

Yes, you can.

To determine the improvement made by a particular sample size, divide by the square root of that sample size. For example, you intend to take the average of four measurements as the measurement result. Divide the measurement variation for individual measurements by the square root of 4, which is 2. Therefore, averaging 4 individual measurements will reduce youre measurement variation by half. The GRR of ~ 26% would be reduced to ~ 13%.

However, there is always a price. You must take 4x as many measurements and calculate averages. This may be worthwhile in the short term such as for a study, but is rarely practical in the long term unless the measurements are easy and made infrequently.

 

[Bev D]

expanding on Miners' comments: if you use the average of multiple readings to improve the variation of your gage study, then you also have to use the average of mulitple readings when measuring production parts. are you prepared to do that? Remember the goal is not to pass the MSA, but to ensure that you have a gage system that will be effective in measuring your parts.

 

[Golfman25]

For us simpletons, this is the kind of "statistical trick" that makes MSA hard to swallow.

 

[Bev D]

ah but it's not a statistical trick unless you make it one.

I have used this approach often in order to have confidence in my measurements until I can improve the measurement system - and this is the point of an MSA: to assess your system awiththe intent of understanding and mitigating any weaknesses.

Of course if you only perform an MSA to check the box, then yes this is a statistical trick, but then so is the MSA itself...

 

[Miner]

To clarify Bev's intent. It is not trickery, if you continue using the average of multiple measurements after the MSA is complete. It is trickery if you do it only to pass the MSA, then discontinue the practice, but not the measurements.

I have used to technique for short term studies where it was not worthwhile upgrading a measurement system only for a short study. This was done to prove out a process change where ongoing measurements were not necessary.