MSA for Torque Dynamometer in Stepping Motor

Hi experts,
it's my first question here in forum. I hope you can help me, I am not a motor specialist, either MSA specialist and I am already fed up.

I am struggling to convience customer that:
1. Our equipment for torque measurement is capable
2. Our two different equipment for torque measurement can be used alternatively

My engineers keep saying:
The torque of the motor alone and its measurement contain various factors of variability. Metaphorically speaking, it is like the difference between LED and fluorescent light. LED is a semiconductor and infinitely constant. On the contrary, the fluorescent paint within the fluorescent light emits light responding to the signal input by the glow starter. If the power supply is unstable or the glow starter has a problem, it may bring discomfort (the light flickers and dims) that man can feel. Which means that it is not always constant. Stepping motors are just like fluorescent lights.
The variability of the object to be measured is too big to use the stepping motor torque for judgment with GR&R tool.
However customer keeps requesting GR&R and does not accept our explanations of motor's nature.

So what we did:
Situation to point #1
We performed MSA for pull in and pull out torque and presented results using total variations and, as we expected, results were very poor: pull in ~50%, pull out ~30%. So customer asked to calculate it using tolerances, results: pull out < 10%, pull in from 6% to 60%.
So we are stuck here and don't have idea what to do next.
I was thinking that maybe non repricable MSA would be better, as there are too many variations, but I am not sure.
These results are only for equipment #1, for the equipment #2 we didn't perform MSA yet.

Situation to point #2
we measured 50 different motors (once per piece) on equipment#1 and repeated measurement on these 50 motors on the equipment#2 to correlate results. As we expected we got correlation at about 0,55, what is not acceptable for customer. Moreover customer has compared the standard deviation from MSA of equipment#1 with standard deviation of 50 motors for equipment#2 result - S=0.069 versus 0.197. So customer claimed this big difference and forbade to use equipment #1 for design validations.

I have no idea what to do

Thanks for any idea, anything would be helpfull.

This is definitely a sticky issue. If I understand it correctly, the product itself varies significantly. This is similar to measuring a part with a lot of Within Part Variation. Only, in your case, the variation is temporal instead of positional.

In these situations, I try to separate the test variation from the product variation by substituting an equivalent with less variation, such as a standard. For example, consider a complicated electro-mechanical device with many moving components and a lot of friction. The activation force is of interest, but the device itself varies in force from activation cycle to activation cycle. Therefore, you can substitute a spring with a similar force. Unfortunately, I am not sure what you would used as a standard in your situation.

Once you have done this, you can compare the two tests using a Bland-Altman plot or a Youden Plot (both of which have been discussed extensively in the forum).

I believe that the customer is fundamentally wrong in their approach to MSA.
What you should be proving is that the system is capable of measuring dynamic torque to the tolerances specified.
This could be achieved by using a known source of reliable consistent torque. ( like a standard ).

A simple analogy is how do I know if a machine I am using to tensile test my product is accurate. Logic would suggest an MSA using a suitable range of standard test pieces. But our customers ignore logic and tell us to use our actual product, which then presents us with this problem.

I find it very frustrating that those with a very high grasp of statistical mathematics can be so stupid in application.

I know from experience with engine dynamometer's that car engine's produce varying amounts of torque and power , with time, running conditions, and build variation ( and its a joke when they quote to single decimal Kw or Nm ). I'd kill to see them demonstrate MSA of their own engine dynamometers based on their production engines. It simply wouldn't work.