MSA - Gage R&R Study Analysis - SAMPLE - Case Study

ZeeTX · Jun 29, 2005

Hi everyone:
I have come accorss a rather strange DATA SHEET in which all the measured values of a part(s) are exactly the same as the measured values of the other operators..!

Here's a Brief description of the DATA SHEET below.

PART = 2x2 Inch FOAM piece. LSL = 2.000 Inch ; USL = 2.250 Inch;
No. of parts = 5; No. of Trials = 3; No. of Operators = 3;
Measured Characteristc - Length; Gage Name - RULER; Gage Number - 1/16 Inch = 0.0625 Inch

PS. Images of the GRR Data sheets are attached for reference (Click on the Images for Bigger size)

From the above Data sheets:
- There is NO Equipment variation (Zero);
- NO Appraiser/Operator variation (Zero) [This is not possible unless Assuming all the operators measure the part together influencing each others readings..!];
- R&R is also Zero;

-- There is only Part Variation = 0.131 and %PV = 52.4

In this Case Study I'd like to know:

Q. Referring to the above GRR Study; What would one should change to Improve the Measurement System..?

Q. If GRR Study is to be repeated; What would one do differently (Assuming all operators measured the part together)..?

:thanx:

Jim Wynne · Jun 29, 2005

ZeeTX said:
From the above Data sheets:
- There is NO Equipment variation (Zero);

The device is a scale (ruler); why does this surprise you?

ZeeTX said:
- NO Appraiser/Operator variation (Zero) [This is not possible unless Assuming all the operators measure the part together influencing each others readings..!];

Of course it's possible, given the device and the object being measured, and assuming reasonably competent operators.

ZeeTX said:
Q. Referring to the above GRR Study; What would one should change to Improve the Measurement System..?

It appears that the device lacks the requisite sensitivity; the tolerance spread is .250 and the resolution of the device is (apparently) .0625. In order to be kosher, the device should be sensitive to 1/10 of the tolerance spread (.025 in this case). If you were to use calipers instead of a scale you would undoubtedly see more of the variation you're looking for.

ZeeTX said:
Q. If GRR Study is to be repeated; What would one do differently (Assuming all operators measured the part together)..?

See the answer to #1.

ZeeTX · Jun 29, 2005

JSW05 said:
The device is a scale (ruler); why does this surprise you?

It does bother me a little bit because a ruler whose precison is 0.0625 Inch (which is very high and the ruler cost is cheap); What is the necessity that ruler has the exact accuracy throughout the length of the ruler and the operators using it to measure the part might not use the same/exact position(s) of the ruler always..!

JSW05 said:
Of course it's possible, given the device and the object being measured, and assuming reasonably competent operators.

Resonable competent operators will have very very less variation but is it possible to have no variation..? Don't you think that the Operators measuring the part influenced each other's readings..?

JSW05 said:
It appears that the device lacks the requisite sensitivity; the tolerance spread is .250 and the resolution of the device is (apparently) .0625. In order to be kosher, the device should be sensitive to 1/10 of the tolerance spread (.025 in this case). If you were to use calipers instead of a scale you would undoubtedly see more of the variation you're looking for.

[/color]See the answer to #1.

Yes, I too came up with the same conclusion that:

- The Gage (Scale or Ruler) used to measure the part (Foam piece) is not right since unit of measurement should atleast be:
1/10 of the Tol (0.250) = 0.0250 but it actually is 0.0625

- But when I do XbarR Analysis: The Precision to Tol ratio = 0 (** P/Tol < 0.1 - Adequate Measurement system; P/Tol>0.3 - Unaccepatable Measurement system).

So this states that measurement system used is Adequate...!!..?? (I'm totally confused here.. :frust:

)

Jim Wynne · Jun 29, 2005

ZeeTX said:
Yes, I too came up with the same conclusion that:

- The Gage (Scale or Ruler) used to measure the part (Foam piece) is not right since unit of measurement should atleast be:
1/10 of the Tol (0.250) = 0.0250 but it actually is 0.0625

- But when I do XbarR Analysis: The Precision to Tol ratio = 0 (** P/Tol < 0.1 - Adequate Measurement system; P/Tol>0.3 - Unaccepatable Measurement system).

So this states that measurement system used is Adequate...!!..?? (I'm totally confused here.. )

But you've already determined that the device is not sensitive enough to reveal possible issues in the system. There's no point in going further, because at this point that are only a few possibilities:

The tolerance is too tight (while 1/4" might seem generous, sometimes a tolerance like that will be assigned arbitrarily).
There's actually danger of approving nonconforming material, (i.e., the tolerance is appropriate) so a more sensitive device is needed.

ZeeTX · Jun 29, 2005

JSW05 said:
But you've already determined that the device is not sensitive enough to reveal possible issues in the system. There's no point in going further, because at this point that are only a few possibilities:

The tolerance is too tight (while 1/4" might seem generous, sometimes a tolerance like that will be assigned arbitrarily).

There's actually danger of approving nonconforming material, (i.e., the tolerance is appropriate) so a more sensitive device is needed.

You mean Tolerance is not appropriate assuming that it is too tight and assigned arbitrarily..?

Below is the Sigma Product Vs Sigma Total Graph which shows the distribution a little bit skewed to left; Only some part of the distribution is going below LSL (2.000 Inch in this case).

This graph shows that there are/will be parts out of spec (2.000-2.250) but not in large quantity.

Q. So how does it has more danger of approving nonconforming material with respect to this graph..?

:thanx:

Jim Wynne · Jun 29, 2005

ZeeTX said:
You mean Tolerance is not appropriate assuming that it is too tight and assigned arbitrarily..?

I don't know whether the tolerance is appropriate or not. That's what you need to find out, if you don't already know.

ZeeTX said:
Below is the Sigma Product Vs Sigma Total Graph which shows the distribution a little bit skewed to left; Only some part of the distribution is going below LSL (2.000 Inch in this case).

Where did you get the data needed for the graph? Using the inadequately-sensitive device? Please go back and look at the two points in my last post.
BTW, the distribution in the image is shifted, not skewed.

Rob Nix · Jun 29, 2005

What is the risk if the parts are a tiny bit out?

What is the likelihood that they would be way out of spec?

It seems to me that you are putting far more effort into this project than you have to. Why not just make some attribute (go / no go) gage up, and chart nothing. Your scale (ruler) discriminates to 1/16", and your spec is 1/4" which is 1 to 4. So it is practically a go/no go check anyway.

This also seems a prime candidate for some error-proofing device.

ZeeTX · Jun 29, 2005

JSW05 said:
I don't know whether the tolerance is appropriate or not. That's what you need to find out, if you don't already know.

Where did you get the data needed for the graph? Using the inadequately-sensitive device? Please go back and look at the two points in my last post.
BTW, the distribution in the image is shifted, not skewed.

The Data sheets are attached as thumbnails in the first post itself. And by meaning "inadequately-sensitive device" (RULER), yes.

I do agree 100% with you that the device used for measuring the part is not sensitive enough to measure the part close to its true value.

Q. So how can I determine that the Tol is appropriate or not..? Sorry, I do not know how Tol can be verified since this is just a case study not a actual manufactured part whose Tol can be verified..

ZeeTX · Jun 29, 2005

Rob Nix said:
What is the risk if the parts are a tiny bit out?

What is the likelihood that they would be way out of spec?

It seems to me that you are putting far more effort into this project than you have to. Why not just make some attribute (go / no go) gage up, and chart nothing. Your scale (ruler) discriminates to 1/16", and your spec is 1/4" which is 1 to 4. So it is practically a go/no go check anyway.

Q. How do you determine that this part is suitable for a go/no go check..? Are there any specific ratios (1:4 in this case) to determine that..?

And also the Part in this case is a peice of "Foam" (2x2 Inch). If the part is held even in hand for measuring, the shape of the part is modified giving a different dimension value..!!

Q. So how can it be used for a Go/no go check when the shape of the part measured is altered..?

* The best way to measure the part (Foam) might be drwaing the shape profile around the the characteristc being measured and then measure the dimensions (or)

* Placing the object on a shawdow projector that projects the shape (Characteristic being measured) of the part to a scale and then measuring the shadow with respect to scale.

Rob Nix said:
This also seems a prime candidate for some error-proofing device.

Q. You mean to say that scale (Ruler) needs to be calibrated..?

:thanks:

Jim Wynne · Jun 29, 2005

ZeeTX said:
Q. So how can I determine that the Tol is appropriate or not..? Sorry, I do not know how Tol can be verified since this is just a case study not a actual manufactured part whose Tol can be verified..

You need to tell us what the assumptions are if the question is hypothetical. Where did the case study come from?

MSA - Gage R&R Study Analysis - SAMPLE - Case Study

ZeeTX

Jim Wynne

ZeeTX

Jim Wynne

ZeeTX

Jim Wynne

Rob Nix

ZeeTX

ZeeTX

Jim Wynne

Similar threads