Gage R&R MSA - Variation, Tolerance, Variance by different methods

[Blaest]

Can someone give me a hint please..

Conducting a MSA, evaluating the results:
• A&R, GRR Variation related to TV (Variation)
• A&R, GRR Variation related to Tolerance
• ANOVA, GRR related to TV (Variation)
• ANOVA, GRR related to TV (Variance)

The above methods give me way different GRR values, and I do not know how to handle this!

I do understand that ANOVA is preferred above A&R
I do understand ANOVA, TV related to Variation is not additive
I do understand ANOVA, TV related to Variance is additive
I do understand a good measurement system should be GRR<10%

Example:
1000mm +/-10
TV=1,085
ANOVA, Variation related to TV: GRR=27,9%
A&R, Variation related to Tolerance: GRR=9,2%

According to ANOVA the variation is too high, but compared to A&R it is less than 10%
Why should I not use this measurement system as the Variation related to Tolerance indicates?

 

[Blaest]

Re: GRR MSA: Variation, Tolerance, Variance by different methods

I partly found my answer,

If the Gage is used for Product Control -> %Tolerance
If the Gage is used for Process Control -> %Variation

So far so good,

If I use the Gage for Processs Control and use ANOVA I have the options:
• ANOVA, GRR related to TV (Variation) = 17,9%
• ANOVA, GRR related to TV (Variance) = 7,8%

Variance is ok - Variation is poor!
How should I evaluate this Gage?

 

[Miner]

Re: GRR MSA: Variation, Tolerance, Variance by different methods

If I use the Gage for Processs Control and use ANOVA I have the options:
• ANOVA, GRR related to TV (Variation) = 17,9%
• ANOVA, GRR related to TV (Variance) = 7,8%

Variance is ok - Variation is poor!
How should I evaluate this Gage?

• ANOVA, GRR related to TV (Variation) = 17,9% This is % Study Variation or better % Process Variation
• ANOVA, GRR related to TV (Variance) = 7,8% This is % Contribution

Use the % Study Variation (%SV) or % Process Variation (%PV) to determine suitability of the gauge.

% Contribution is not currently used to determine whether the gauge is suitable, but to determine the major source(s) of measurement variation in order to take action to improve the measurement system.

 

[Blaest]

Re: GRR MSA: Variation, Tolerance, Variance by different methods

Thanks Miner,

I would be able to have:
• A&R, GRR Variation related to Tolerance > 10%
• ANOVA, GRR related to TV (Variation) < 10%

According to:
If the Gage is used for Product Control -> %Tolerance
If the Gage is used for Process Control -> %Variation

I would have a gage that is:
NOT OK forProduct Control
OK for Process Control

That makes no sense to me,

Would I be wrong to assume that both would need to be OK to used for Process Control?

Or should I interpret this differently?

 

[Miner]

Re: GRR MSA: Variation, Tolerance, Variance by different methods

This scenario could happen when you have an incapable process, that is, where the process variation is much greater than the specifications. Under this scenario, you could have a gage that is acceptable for process control, but not acceptable for product control.

Fortunately, this scenario is not to common.

 

[Blaest]

Re: GRR MSA: Variation, Tolerance, Variance by different methods

Thanks Miner,

And you are right, that is my situation,

The reason is that I select my Parts to spread across the measuring range of the gage, for exampel a caliper 0-150mm

I am experimenting with uncertainty budget(GUM/VDA) and evaluating the gage over its range rather than at a specific dimension/tolerance (I migth very well be off ground here, but logically it makes sense to me..)

See Topic: 68299
(I am not allowed to post a hyperlink)

 

[Miner]

Re: GRR MSA: Variation, Tolerance, Variance by different methods

The reason is that I select my Parts to spread across the measuring range of the gage, for exampel a caliper 0-150mm

In that case, only % Tolerance has any meaning. The % Study Variation was rendered meaningless since the samples are not representative of the process.

 

[Blaest]

Re: GRR MSA: Variation, Tolerance, Variance by different methods

That makes sense..

What I am using is:
EV in unit (like 0,021mm)
AV in unit (like 0,19mm)

These values do not depend on the characteristics or the specifications.
These values would relate to the Equipment and the Appraisers alone.

I transfer EV & AV in units to GUM calculations.

 

[Statistical Steven]

Re: GRR MSA: Variation, Tolerance, Variance by different methods

Bottom line is, you should use it! The bottom with %GRR as it relates to total variation is that if the part to part variability is small then your measurement system looks poor and if the part to part variability is high, your measurement system might look better. When you have specification, that is the preferred method of GRR for MSA studies in my opinion.

Can someone give me a hint please..

Conducting a MSA, evaluating the results:
• A&R, GRR Variation related to TV (Variation)
• A&R, GRR Variation related to Tolerance
• ANOVA, GRR related to TV (Variation)
• ANOVA, GRR related to TV (Variance)

The above methods give me way different GRR values, and I do not know how to handle this!

I do understand that ANOVA is preferred above A&R
I do understand ANOVA, TV related to Variation is not additive
I do understand ANOVA, TV related to Variance is additive
I do understand a good measurement system should be GRR<10%

Example:
1000mm +/-10
TV=1,085
ANOVA, Variation related to TV: GRR=27,9%
A&R, Variation related to Tolerance: GRR=9,2%

According to ANOVA the variation is too high, but compared to A&R it is less than 10%
Why should I not use this measurement system as the Variation related to Tolerance indicates?

 

[Blaest]

Re: GRR MSA: Variation, Tolerance, Variance by different methods

Thanks for your input..

I do not have the statistical background to back me up - so I rely on logical thinking, this forum and intuition.. :

I have also been thinking about the fact that %GRR is a relative value related to the part variation - and I never really got my head around it to make logically sense, and has also been tempted to use specification instead.

Thinking about it:
-the higher part variation of the process, the more likely it is the Cpk will not be ok
-the higher part variation of the process, the more likely it is the %GRR is ok

I guess the last pin is why it is so important to have parts that represents the WHOLE process

So, I need to find a variation in between to satisfy both, either by:
-Less variation from Appraiser/Equipment (= lower %GRR score)
-Higher tolerences (= better Cpk-value)

I guess It would be a good idea to conduct a capability index of the measurement system prior its first use - like a Measurement System Capability Type-1 Study as described in VDA (Cg=1,33=10% of tolerance) to ensure EV is under control.

Any other input much appreciated as I am still trying to get my head around this..