# 10 parts, 3 operators, 3 measurements GR&R study tolerance question

M

#### mkjohnson

Hello,

I am new to measurement systems analysis and I am doing a 10 part, 3 operator, 3 measurement GR&R study for a client. I have read through numerous posts on this forum about tolerance and GR&R, but haven't found a discussion that matches the scenario I am looking at.

Here is a simplified version of the problem:
- the client has a process that produces parts with a key parameter that varies between 0 and 2 (arbitrary units).
- parts measuring less than 1 will be considered 'A' grade and parts greater than 1 will be considered 'B' grade

Let's suppose the GR&R study shows Gage R&R at 5% of total variation. Do I think of the tolerance parameter as being related to actual process width and set it to 2, or do I set it to 1 because of the clients preference of considering parts from 0 - 1 to be within spec?

The situations I have read about with respect to tolerance are when the sample set in the GR&R study doesn't span the process range. In these situations, it makes sense to me to divide by a tolerance known through other means. But the situation I'm wondering is when the client's desired range for accepting a part is just a fraction of the process range.

Any help is appreciated.

- mkj

#### Stijloor

Super Moderator
Re: 10 part, 3 operator, 3 measurement GR&R study tolerance question

A Quick Bump!

Can someone help?

Thank you very much!

#### Miner

##### Forum Moderator
Re: 10 part, 3 operator, 3 measurement GR&R study tolerance question

Tolerances are independent of the process variation. Tolerances are from the Voice of the Customer, while the process variation is Voice of the Process.

There are different metrics for each. Use % Tolerance using the requirements as stated by the customer. This would be the Grade A range.

If necessary, you can assess the gage for process control using % SV or % Process and ndc.

These metrics assess the suitability of a gage for entirely different uses. % Tol is used to assess suitability as an inspection device. The other assess suitability as a process control gage.

M

#### mkjohnson

Re: 10 part, 3 operator, 3 measurement GR&R study tolerance question

Thank you for the answer and clarification.

I will read up on these concepts to be sure I understand the different metrics that can be used to assess the gage.

- mkj

#### Bev D

##### Heretical Statistician
Super Moderator
Re: 10 part, 3 operator, 3 measurement GR&R study tolerance question

Miner makes good points. I will suggest an alternative - treat this like a categorical data MSA. The traditional %tolerance math is incorrect and can either overstate or understate the actual contribution of measurement error to the observed variation.

treating this like a categorical MSA will provide a more direct measure of the ability of the system to properly grade parts as A or B...you can research how to do this type of study here.

M

#### mkjohnson

Re: 10 part, 3 operator, 3 measurement GR&R study tolerance question

Thanks for the advice. This is quite relevant to my study because in the simplified example, I only mentioned parts of grade A or B.

In reality, the parts can have four grades (A through D). Using a tolerance that is only 1/4 of the measurement range (to select A) ends up multiplying the GRR to tolerance ratio by 4. As a result, the statistical analysis didn't seem consistent with my intuition for how well the system can measure the parts.

I'll review the approach to see if it can be applied to the data I have already captured.

- mkj