Search the Elsmar Cove!
**Search ALL of** with DuckDuckGo Especially for content not in the forum
Such as files in the Cove "Members" Directory

Necessity of Expanded GR&R for Multiple Test Systems



Hi All,
I have searched some but was not able to find specifically what I was looking for so hopefully this doesn't repeat an existing topic.

I have a situation where I have two identically configured automated test systems. I have performed the basic GR&R on each system independently (Minitab) using the traditional 10x3x3 method for data. Each system tests to an acceptable level for the different test points. (each test point is analyzed independently.

When I combine the data and add the system ID as an additional variable to perform an Extended GR&R (again, in Minitab) I get, as I pretty much expect to, results that are not as good. In my case, I pretty much figure the majority of this is based on the tolerance ranges of the components of the test systems.

While I get this and understand that it indicates some work ahead to improve the test systems, my manager does not. He keeps lobbying to use each system anyway since each one passed their individual GR&R.
While I have a fundamental understanding of the GR&R process and can generally interpret the results, I am by no means a statistical expert and have difficulty explaining why the multiple gauge analysis is needed and the consequences of no doing it and deploying two systems that are mis-matched to an unacceptable level.

What I need is some type of documentation, white paper, manual, etc that I can use to make my manager (and others) understand the points I am trying to make. It needs to cover the use of the expanded GRR specifically being used with two independent measurement systems and the consequences of ignoring those results compared to two independent GRR?s.

I can upload an example Minitab file if needed as an example of what I am working with.

Thanks for any feedback on this



Involved In Discussions
It sounds like your people do not understand that automated test systems should be no different from a caliper or a micrometer. No one would accept using one caliper that consistently reads differently from another. Especially if you were measuring something like a gage block where you are fairly certain what the measured value should be close to.

Here is a suggestion to help your people understand the need for the two systems to be in line with one another. Get two calipers or some other "simple" measurement device and a calibrated standard to measure. Temporarily modify one of these items (in a reversible non-destructive manner of course) and then present these to your manager. Ask him which gage he would like the company to use to determine if product is acceptable.

You will likely get a comment along the lines of the modified gage being obviously not the one that they would use because they can clearly see that it will give them incorrect results. Then you can start to explain that your two test systems are very similar to these simple gages. The big difference is that it is immediately apparent that the simple gages are going to provide different results. You could even do a gage study with them and the data would indicate that there is something different. Determining if complicated test systems are "obviously" different is not so easy. Also there often is no equivalent to a gage block to put into your test system with known values that you can compare to. You only have test data to guide you. And the only way to allow the test data guide you is to use the various test systems as additional variables in your gage studies.

This is a difficult subject to get people that are not involved with gage studies to understand for sure. Especially when most people spend all their life thinking that measurement is an exact process.

Keep at it and good luck.


I appreciate that input but I am not confident that example would do much to sway the mindset here. I can certainly see the analogy but to those that don't want to be convinced, I think I am going to need something a bit stronger.

It is unfortunate that, right now and for the immediate future at least, there is no 'standard' available to calibrate the "system" as there would be for a device (ie. gauge block for a caliper).

The automated test system is comprised of multiple controlling or measuring components, each with it's own tolerance and calibration performed. As the product being tested is an active, dynamic product that generates the work being controlled and measured, it is challenging to develop a standard or set of standards to perform a system calibration. It is going to be a project within itslef.

Am trying to go multiple routes here and am probably going to have to solicit the quality group to intervene and spend money to have an authority that cannot (easily) be ignored provide a company wide eye opening session.

Will keep plodding away.......

Top Bottom