MEASUREMENT SYSTEMS ANALYSIS (MSA)-Based R&R Studies

Sometimes, when holding a new position in a company, we do not have the time to be properly trained due to the daily pressure, and have no other choice than using formulas and tools without knowing very well their origines or meanings. All we know is what we have been briefly told, that is, that if the result ranges between “X” and “Y”, then it is OK, and if it is out of this range, then it is NOK. Actually, such methods could make us take wrong decisions sometimes, because we do not have the sufficient knowledge to be able to do as good an analysis as possible.

I can take the example of the calculation of the traditional R&R study (repeatability and reproducibility). Usually, we use an Excel sheet already filled in with formulas – and we assume they have already been approved by someone else within our company – in which we add data (input) and get an “OK”, “OK but need of improvements”, or “NOK” result.

For today’s post, I have decided to look for more information about R&R, and the first thing to know is where it comes from. The concept was first mentioned in the “Measurement Systems Analysis Reference Manual” created by Chrysler, Ford and GM at the beginning of the 90s. Appart from the R&R study, the manual described other concepts to know whether or not our measuring system is right.

REMEMBER:
Repeatability: Measuring capacity of the device to provide correct results.
Reproducibility: Variability created by operators.

The method most frequently used to get the input and then calculate the R&R is the crossed-designs method. The crossed-designs method consists in having all the operators measuring all the parts several times.

DATA-BASED (INPUT) CALCULATIONS

• The average and range method (approximate method) . Method usually used. It is simple and can be made manually.
• ANOVA Method. Exact method. Based on the analysis variance. Means tedious calculations if made by hand. Analyses if there is any difference between parts, any difference between operators and any interaction part-operator. Can be made with software such as MINITAB (I have also read that Excel can make this calculation but I have never tried)

BUT WHICH CONSEQUENCES DOES IT HAVE ON THE MEASUREMENT SYSTEM?

• The measuring device: The checking fixture in itself, the design and concept, the alignment elements, materials and wear-and-tears, holders, fixations, measuring elements, etc. User-friendliness of the device.
• Operators, their specific training to use measurement systems of this type, people’s capacity or discipline to follow written or spoken instructions.
• Measurement guidelines (methods): How to fasten devices, how to mount parts, how to collect data, and so on.
• Specification: The measurements are compared with a specification, or reference value. The specification range does not affect the measurements but is an important factor that may affect the viability of the measurement system.
• Parts (what is being measured): Some parts are easier to measure than others. For example, a measurement system may be right to measure the length of a solid block, but not for a rubber part.

SHOULD WE REFUSE A MEASUREMENT SYSTEM IF AT THE END OF THE CALCULATIONS THE GRR (GAUGE R&R) EQUALS TO MORE THAN 10%?

According to the section about GRR errors of the AIAG normative (Automotive Industry Action Group), the final decision of accepting or refusing the measurement system is to be decided between the customer and the supplier; the decision will depend on the criticality of the measurement characteristic.

CONCLUSION

When I wrote this post, I did not mean to detail the complex mathematics involved in the calculation of the error of a measurement system but to define its frame, know where this matter comes from and explain that there is more than one data collection and calculation methodologies. There are other concepts that I have not mentionned and are also part of the MSA, such as linearity, stability or uncertainty value but of course… this is what the MANUAL is for.