MSA - take it ease.

Once, a very long time ago, when I was at school I learned one sentence from a poem by Adam Mickiewicz. by heart: "Feeling and faith speaks stronger to me than the glass and the eye of a wise man." Our poet in the poem "Romantic" criticizes scientist Jan Śniadecki (by the way, it was the man who introduced the Polish versions of terms differential and integral) for too much focus on the sciences. What can I know there? The only thing I can ask for is moderation and balance.

Okay, what does this have to do with MSA (Measurement System Analysis)? While improving processes and products, one needs to be able to use all our skills: hard statistical, engineering and process knowledge (glass and eye), but also intuition and ingenuity (feeling and faith).

Enough talking, let us get to work. Let us suppose we have a problem, e.g. too many defects. So what are we doing? Let us count how many defects there are, what kind, under what conditions they appear, etc. And at this point I must say stop! I do not know which company you work for, I do not know at what stage of technological development your company is, but I know one thing for sure - you need to check the Measurement System. Why do you blindly assume you do not need to? I understand, someone did this before you and you just believe it - I am ready to accept this argument (with some uncertainty). On the other hand, if you have heard that someone probably did it a few years ago, etc., it means that the time has come for you to do it before you start anything else.

I understand, I might not have convinced you. Well, let us imagine that you are driving a car at 50 km/h, a policeman stops you and says: the radar has recorded 75 km/h. Who among us would not question the measuring accuracy of the radar? We have to do exactly the same when measuring our processes. We ask whether the Measuring System can measure our processes and our products sufficiently accurately? If the answer is positive, we start collecting data and then improving processes. The Measurement System is our "glass" that is between the observer and the process or product. We must ensure that the "glass" gives a sharp image, does not have scratches and dirt. There is still the technical aspect - how to do MSA? At this point, I will not bore you with the technical details. I encourage you to read the next article about MSA details.

We can divide MSA into 2 categories:

measuring physical units, i.e. velocity, mass, volume, length, etc. - MSA for continuous variables,
scales, ratings, frequencies - MSA for attributive variables.

Regardless of variable type, an MSA should be performed. I will not go into any polemics with metrologists, as they only deal with MSA and probably know better, but I will present the approach used in the Six Sigma methodology. Very often, Six Sigma requires more than what is being used by companies. Even if we consider Six Sigma approach to be selective, it will be enough to complete a Six Sigma project successfully.

MSA for Continuous Variables

Consider an example - perhaps we can establish that we produce painted elements. The thickness of the varnish coat must be between 0.3 - 0.4 mm. All products with a coating thickness less than the lower limit are defects. In the case of thickness above the upper limit, we lose money on too thick coating.

What do we do if we want to improve the process. Let us record and count the defects. Let us make a Pareto diagram or an SPC diagram etc ... Stop. After all I said you have to do MSA. We are going to do MSA. We ask how the measurement is carried out, how often, under what conditions, by whom, if people are trained. Here you can hear:

Why do you need to know it?
The Measurement System is correct.
We do not calibrate measuring devices by ourselves, we send them to accredited laboratories.

In simple words, you should back off - at least according to the interviewees. Finally, you may hear the sentence: "It is not your business, everything is fine and you do not need to check it." This is where the true story comes to my mind. Suppose a woman is pregnant and is Rh- (which, as is well known, poses a risk of a serological conflict between mother and child). The doctor will want to test the pregnant woman's blood for anti-Rh antibodies. But doctor - says the woman - my husband also is Rh-, so the baby also is Rh-, which means the complications in this regard are not possible. As for you – said doctor - I am sure that you are the mother, but as for the husband, I do not know if he is the father. I have to check the presence of antibodies. And you have to act like that doctor. Listen to everyone and do your job. The point is not to question someone else's competence, to calibrate a measuring instrument, etc., but to check the Measurement System as a whole.

What does measurement and reading consist of?

Measuring instrument.
Operator.
Level of operator training.
Physical conditions: lighting, viewing angle, humidity, temperature, air movement, etc.
Procedures.
Software.
Typing error.

To cite such a long list, I may have scared you, but a well-prepared MSA should take less than 2-5 hours. During the training, we do the full MSA, including the analysis of the results in about 2 hours, so it is not that scary. The most time should be spent selecting the right parts. The measurement is probably less than 2 hours.

How to prepare MSA in the company?

Select a person responsible for the MSA, ie preparation, carrying out and analysis of the results. In the best case you should appoint yourself.
Select 2-3 operators - people who really do the measurements (not some people from nearby or metrology experts).
Determine which measuring instrument you want to check. All measurements must be done with the help of the same instrument.
Do not do any operator training, calibration of measuring instruments - you want to check how this system works now.
Select at least 10 parts (items you will measure). The whole spectrum of manufactured products should be taken into account, both from the tolerance range and outside the tolerance range. Good practice says: tolerance range +/- 20% of the tolerance range.
Each operator should carry out at least 2 series of measurements as follows:
- 1st operator - 1st measurement of all elements,
- 2nd operator - 1st measurement of all elements,
- 1st operator - 2nd measurement of all elements,
- 2nd operator - 2nd measurement of all elements etc.
When measuring, the operator cannot know which element is being measured and should measure the elements in a random order each time. The item number may only be known to the person responsible for the MSA.

What does the MSA look like?

Based on the collected data, we mainly calculate:

Repeatability - a factor determining whether a given operator is able to measure in a repeatable manner, i.e. whether she/he can measure the same elements in the same way (first R).
Reproducibility – a factor determining whether different operators can measure the same elements in the same way (second R).
Total error - so called Gage R&R.
Measurement error in relation to the tolerance range.

In the next article, I show you how to do an MSA analysis.

All the above-mentioned factors must remain within the established limits. What if they do not fit? Well - then you know that you need to do first. The project cannot be continued without first improving the Measurement System. Based on the analysis, you will know if the problem is with the operators, procedures, training, etc.

Author: Adam Cetera (LeanSigma.pl)
Creation date: 2018-08-30
Modification date: 2018-08-30

Leave a comment below if you would like to add something?