For a Type 1 Gage R&R I want to use 1 unit and measure it 30 times

[michalgajewskix]

Hi,

In my company (wind sensors manufactur) we try to introduce a set of tools to control our production and measurement system, problematic issue is the fact that no one knows how to do this correctly… my manager thinks about:

1. Cp, cpk indices and I and MR charts to control company production, how far from spec limits we are, how close to mean, etc.

2. Gage R&R Type 1 to verify a measurement system

I have read a few articles about this topic so far, but still this is not clear for me.Please correct me if I am wrong.

Type 1, because to start a calibration of our product (wind sensor) operator needs to press only a start button (so there are no variation between operators).

For a Gage R&R type 1 I want to use 1 unit and measure it 30 times, in this way I will have information about repeatibility. Can I add also 30 measurements of 30 different units (1 measurement per 1 unit) to get information about reproducibility?

I have problem with my outputs (in the attachments you can see graphs from minitab)

In TunnelDAta.xlsx you could find measurement data in two columns:
First – measurements of 30 units,
Second – 30 measurements of the same unit.

in Gage R&R type 1 I dont see any condition for NDC, can I use data which I used?

Thank you for any help.
Regards

 

[Miner]

Type 1, because to start a calibration of our product (wind sensor) operator needs to press only a start button (so there are no variation between operators).

Provided that you have clearly explained the system and have left nothing out, this is correct.

For a Gage R&R type 1 I want to use 1 unit and measure it 30 times, in this way I will have information about repeatibility.

Minitab recommends taking 50 measurements for a better estimate of the variation, but this is correct.

Can I add also 30 measurements of 30 different units (1 measurement per 1 unit) to get information about reproducibility?

No. This gives you an estimate of the process variation, not reproducibility.

I have problem with my outputs (in the attachments you can see graphs from minitab)

In TunnelDAta.xlsx you could find measurement data in two columns:
First – measurements of 30 units,
Second – 30 measurements of the same unit.

in Gage R&R type 1 I don't see any condition for NDC, can I use data which I used?

A Type 1 study does not calculate the ndc. ndc is a metric used only with a Type 2 R&R study. You can manually calculate ndc by dividing the standard deviation for Parts by the standard deviation for Gage, then multiply by 1.41 (which is equal to SQRT(2)) then truncate this value to an integer.

You have another problem. A Type 1 study requires a standard for the part that you measured. This should be the mean of repeated (>/= 10) measurements of the part by a measurement device of greater accuracy that the device used in the study. You appear to have used the nominal/target value for the standard. This will yield incorrect results for the Bias and any metrics that include the bias.

 

[Statistical Steven]

Couple of issues you need to resolve:

1. Your measurement system does not have enough discrimination as there seems to be only a few unique value (not a truly continuous value).

2. I would take 10 parts and measure each 3 times to get a better understanding of reproducibility (and repeatability)

3. Though an operator only pushes a button, if there is setup of the part, there can be operator induced variability

4. Cp, Cpk or SPC charts (I and MR) are process parameters. GRR is a measurement system analysis. They evaluate separate things. Always do GRR before any process indices.

Hi,

In my company (wind sensors manufactur) we try to introduce a set of tools to control our production and measurement system, problematic issue is the fact that no one knows how to do this correctly… my manager thinks about:

1. Cp, cpk indices and I and MR charts to control company production, how far from spec limits we are, how close to mean, etc.

2. Gage R&R Type 1 to verify a measurement system

I have read a few articles about this topic so far, but still this is not clear for me.Please correct me if I am wrong.

Type 1, because to start a calibration of our product (wind sensor) operator needs to press only a start button (so there are no variation between operators).

For a Gage R&R type 1 I want to use 1 unit and measure it 30 times, in this way I will have information about repeatibility. Can I add also 30 measurements of 30 different units (1 measurement per 1 unit) to get information about reproducibility?

I have problem with my outputs (in the attachments you can see graphs from minitab)

In TunnelDAta.xlsx you could find measurement data in two columns:
First – measurements of 30 units,
Second – 30 measurements of the same unit.

in Gage R&R type 1 I dont see any condition for NDC, can I use data which I used?

Thank you for any help.
Regards

 

[michalgajewskix]

Thank you very much for your answers

I have small problem with proper understanding them:

First aspect: I asked:

Can I add also 30 measurements of 30 different units (1 measurement per 1 unit) to get information about reproducibility?

Miner answer:

No. This gives you an estimate of the process variation, not reproducibility.

Statistical Steven answer:

2. I would take 10 parts and measure each 3 times to get a better understanding of reproducibility (and repeatability)

So how it is with reproducibility? I found info that to do reproducibility test a few operators (at least 2, better 3) need to measure one unit few times... Statistical Steven answer suggests something else...

Second aspect:

1. Your measurement system does not have enough discrimination as there seems to be only a few unique value (not a truly continuous value).

I changed discrimination from 0.1 to 0.01 m/s, new data and graphs are in the attachments (there are real measurement data). But still my Cg, Cgk are below 1, and %Var is above 30%, which means that process is not under control...

Correct me if I am wrong:
For this product my spec limits are +-0.3 m/s, so tolerance is 0.6 m/s, target is 10.00 m/s, so to do gage R&R I need measurement data which cover only 20% of tolerance (in this case data with max. spread of 0.12 m/s ??).

And If my real measurement data looks in this way, so what can I do with them to do Gage R&R in correct way?

Last question in the end:

in TunnelData2.xlsx there are three columns with data, first column is a reference wind speed, second is my product's measurement data, when I try to do Gage type 1 in minitab I could only select target value (in this case 10.00 m/s), so how can I include also variation of my reference tool ?
If I shouldn't included it, then which data use ( error = ref. - my product ) ?

Thank you for any hints,
Mike

 

[Statistical Steven]

See my answers in RED

Thank you very much for your answers

So how it is with reproducibility? I found info that to do reproducibility test a few operators (at least 2, better 3) need to measure one unit few times... Statistical Steven answer suggests something else...

Measuring multiple parts, multiple times is different than measuring multiple parts once. If you only measure a part once, your process/measurement variability is confounded. My point is that for the same 30 measurements, I would use less parts and more repeats. You can add in multiple operators to get the reproducibility too.

Second aspect:



I changed discrimination from 0.1 to 0.01 m/s, new data and graphs are in the attachments (there are real measurement data). But still my Cg, Cgk are below 1, and %Var is above 30%, which means that process is not under control...

Correct me if I am wrong:
For this product my spec limits are +-0.3 m/s, so tolerance is 0.6 m/s, target is 10.00 m/s, so to do gage R&R I need measurement data which cover only 20% of tolerance (in this case data with max. spread of 0.12 m/s ??).

And If my real measurement data looks in this way, so what can I do with them to do Gage R&R in correct way?
If the 20 data points are a single part measured 20 times,
then the measurement system is too variable for your specifications.
Last question in the end:

in TunnelData2.xlsx there are three columns with data, first column is a reference wind speed, second is my product's measurement data, when I try to do Gage type 1 in minitab I could only select target value (in this case 10.00 m/s), so how can I include also variation of my reference tool ?
If I shouldn't included it, then which data use ( error = ref. - my product ) ?

Thank you for any hints,
Mike

 

[Miner]

It sounds like you might be new to MSA. If so, you might want to start by reading my MSA Blog. They are sorted with the most recent first, so I would start with the oldest and work your way to the more recent ones.

Your specific question dealt with a Type 1 study which looks only at repeatability. I scoped my answer to a Type 1 study. Steven cautioned you not to assume that there was no operator impact if the results might be affected by how an operator loads a part into the test fixture (i.e., he widened the scope to a Type 2 R&R study).

Regarding the resolution, Steven was correct in this comment. I had noticed the same issue, but forgot to mention it. Poor resolution has differing impact on R&R results. It can hurt, help or not impact. It is a separate issue, just like bias. The priority of focus on ensuring a good measurement system is:

1. Ensure good resolution
2. Ensure zero bias (calibrate)
3. Ensure minimal variation (R&R)

I changed discrimination from 0.1 to 0.01 m/s, new data and graphs are in the attachments (there are real measurement data). But still my Cg, Cgk are below 1, and %Var is above 30%, which means that process is not under control...

It is not really a control issue. It is an issue of high gage repeatability variation.

Correct me if I am wrong:
For this product my spec limits are +-0.3 m/s, so tolerance is 0.6 m/s, target is 10.00 m/s, so to do gage R&R I need measurement data which cover only 20% of tolerance (in this case data with max. spread of 0.12 m/s ??).

And If my real measurement data looks in this way, so what can I do with them to do Gage R&R in correct way?

in TunnelData2.xlsx there are three columns with data, first column is a reference wind speed, second is my product's measurement data, when I try to do Gage type 1 in minitab I could only select target value (in this case 10.00 m/s), so how can I include also variation of my reference tool ?
If I shouldn't included it, then which data use ( error = ref. - my product ) ?

You are correct. The Type 1 R study assumes that you are making multiple measurements against an unchanging standard. It appears that you are taking parallel measurements of a changing characteristic where one measurement is your gage and the other measurement is a "Master" or reference gage. Is that correct?

Assuming that is the situation, I ran the analysis using the wind sensor speed error against a reference value of zero. That analysis is attached. It shows a statistically significant bias of 0.050, plus a repeatability issue. Regarding how to reduce the repeatability variation, consider: the stated repeatability of the device, variation in the wind speed itself, variation in wind speed between the location of the instruments. Another consideration. Not having worked with a wind tunnel, I visualize a giant fan on one end to provide the wind speed. If that is correct, I would expect the wind speed to pulsate as each fan blade passes. How does the measurement device handle these pulsations? Do you get instantaneous readings that fluctuate with the pulsation, or average them? How does the master device measure? Are they the same?