MSA For non replicable test + no operator influence

[RUBEN.MARTIN]

Hello
Please excuse my poor English and if this topic has already been discussed in another thread but I can't find clear information on it

I need to know how to correctly perform an MSA for a measuring device without operator influence + with non-replicable samples
The situation is the following: we have a press with a force reader so that the operator inserts the piece into a housing, always in the same position and when activating the equipment, he performs a press where the sensor offers the value of the applied force. In this case, there is only an upper limit that is 1,500 N maximum

I don't know if the nested Anova method is the right one, but this method offers variation between operators, but for the example described above, I consider that the influence of the operators is null


I have a similar question for measurements with CMM, where the parts are placed on a fixed support and the measurement is carried out automatically by a automatized program. In this case the measurement is replicable, but.... should I consider the null operator influence? And therefore use the option of 1 operator x 2 repetitions x 10 samples?


I use Minitab as main software tool

 

[Miner]

The first scenario is a difficult one. A Type 1 R study is not appropriate since you do not have a reference sample that can be replicated. A non-replicable Type 2 R&R study isn't appropriate either since there is no operator effect. I can think of two possible options:

• Are you able to use split specimens? This is where you split a larger sample into smaller samples that are assumed to be homogeneous in properties. If this is feasible, you can use these split specimens as the subgroups in an Xbar-R chart then use the within subgroup variation as the Repeatability.
• Are you able to substitute a load cell or spring for the piece? This method is used to evaluate load testers such as an Instron.

For the CMM scenario, are you sure that the operator has no influence or are you just assuming that? There are many cases where operators do have an impact from how they load the part. Even if there is no influence, would it hurt to prove that using the standard Type 2 study? You could use a Type 1 R study using one part as the reference. You would have to ignore the bias and Cgk output since you do not have a true reference, but the Cg would be valid.

 

[Bev D]

Miner is correct. For the first non-replicable test, the standard way is to cut each piece into two parts and use those as the 2 ‘repeats’. If you ‘fail’ the MSA there are 2 reasons: the measurement system isn’t repeatable OR there is substantial within piece variation. Either way you learn something.

The second scenario is very common. Having performed hundreds of CMM MSAs, my advice is to JUST DO IT. If there truly is no operator influence the MSA will show that. If there is, the MSA will show that.

 

[Johnnymo62]

I agree, you need to prove there is no Operator influence.

 

[Bev D]

Opinion is cheap.
I’ve seen way too many times when operator influence is important, even the largest component of measurement variation. It’s really cheap to use 2 operators and PROVE that your opinion is right, rather than stick with your opinion and get bit. A real Guage R&R (repeatability and reproducibility - aka operator) or a real MSA is intended to take opinion out of things and deal with facts and data.

 

[Scanton]

I have to agree with the opinion that operator having no influence needs to be proven before it can be discounted.

We have a non-contact CMM where a component is placed between two centres and even that showed operator influence on the MSA study, I bet a penny to a pound that an MSA study on your CMM where operators place a component on a fixed support also shows operator influence.