R chart from ANOVA gage R&R test

[abichou]

Hi,

So i have this R chart by the operator from an MSA, I don't get it , I want to understand how to analyze it get an interpretation from it.

I would be grateful for your help.

Of course any analyses or opinion about the other graphs would be appreciated.

 

[Miner]

Refer to my MSA blog on interpreting the charts. Then if you have additional questions, you can ask them here.

 

[abichou]

i had a look on the blog and i have question in the R chart graph by operator I can see that all the operators usually don't measure the same value on the same sample and also there is some sample on the LCL and above the UCL what does all of this means.

 

[Miner]

Repeatability variation is typically random, so I would not expect the same individual measurements for each operator. I would expect very similar average measurements though. As I look at the Range chart, I see that the gage has adequate resolution. There are many possible measurement values between the control limits, and there are not an excessive number of zero ranges. I also see that Uwe has the best repeatability. Uwe's method should be studied and used as a standard method. Joerg and Azuz have more repeatability variation, and Azuz appears to have had some difficulty measuring part # 6. I would investigate the reason for this and train Joerg and Azuz in Uwe's method.

 

[abichou]

Thank you very much for your answer, Also I got a high P to T value what is the reason for that, can you give me your opinion on these values from the gage evaluation table how can you analyze them .

 

[Miner]

This implies that your process variation (ndc=50) is much greater than your tolerance (%Tol=36.5). Is your tolerance input correctly? The parts do not appear to represent actual process variation (increasing trend) unless you have significant tool wear.

The output shows that you have problems with both repeatability and reproducibility. Breaking down reproducibility, you have problems with the operator/part interaction, which means that some operators have problems measuring specific parts.

 

[abichou]

This implies that your process variation (ndc=50) is much greater than your tolerance (%Tol=36.5). Is your tolerance input correctly? The parts do not appear to represent actual process variation (increasing trend) unless you have significant tool wear.

The output shows that you have problems with both repeatability and reproducibility. Breaking down reproducibility, you have problems with the operator/part interaction, which means that some operators have problems measuring specific parts.

Actually, I changed the tolerance maybe that's why so the table I gave is not the same from the charts I showed at the first time
I will attach the new values here maybe it helps

 

[Miner]

I don't see the % Tolerance column. Can you attach the data worksheet?

 

[abichou]

I don't see the % Tolerance column. Can you attach the data worksheet?

It's a Mini tab file

 

[Welshwizard]

Hi abichou,

Your points from each of your posts:

You want to be able to analyse, interpret and learn about the R Chart in the context of a Gauge R and R (GRR) Analysis.
- The R (Range) chart plots the maximum minus the minimum differences for the repeated measurements of each part by each operator, in this way it depicts the repeatability and consistency of the measurement process across the studied operators. The Range Chart has additional lines plotted to show the average range of the studied operators and also an Upper Control Limit (UCL)
The UCL filters the everyday random variation from the special cause variation thus if it is breached by a measurement it implies that there is a lack of consistency in the measurement process across your studied operators. The software goes on to characterise the usual numbers but they all rely on the process demonstrating random variation, which in your case it doesnt.
When you find the reason for the breach of the control limits you must put something in place to try to minimise it happening again because the implication is that this measurement process is important to you and you need to aspire to consistent measurements. You will find that if you do seek out and eradicate the special cause that the performance of the measurement process will improve for usually very little effort.

You want to understand why all the operators usually don't measure the same value on the same sample and also there is some sample on the LCL and above the UCL what does all of this means.

- The laws of natural variation and physics will dictate that each operator will on balance record different values for parts, this is to be expected, in fact if this doesn't happen you need to be cautious. The UCL helps you to filter and therefore not worry about the extent of the random variation but rather alert you when anything other than random variation is present, this is the essence of process behaviour charts such as this. Of course, if you only display random variation and the performance of the measurement process is still not what you want then you need to think about how you could improve things.

You have a relatively high P to T value, what does this mean?

- You are referring to the ratio of measurement variation to tolerance expressed as a percentage, this is known as the P/T ratio. You may have pre set ideas in your mind about good or bad values here based upon your learning or mandated by a customer.
The P/T ratio simply uses multiples of the standard deviation and divides it by the width of the tolerance, it implies that the higher the percentage from the ratio the higher the risk is for using the measurement process to classify parts.

What to do next?

- Ensure that the measurement study reflects what would happen in a production situation whilst using this measurement system, otherwise the outcomes mean little, for example:
Do you have a method that should be followed?
Was the study conducted in controlled or real life conditions?
Are the operators trained and familiar?
- Find and formally correct the reasons for the special causes for the operator, think about the impact on other operators who could be used on this task
- Re run the study, witness it and take notes
- If all is ok with consistency publish and ensure the disciplines are cascaded to the organisation, if not correct and re run
- Consider that if the operators are consistent this is the best that you can achieve without either time consuming repetitive measurements or another measurement device and process.

Final Observations

Looking at your reports without access to the raw data from experience I would say that the apparent differences between the operators are in the noise of the measurements therefore they are not there. This means that if you could find and eradicate the reason for the inconsistency for operator 3 measurements you could claim at the minimum the standard deviation for Repeatability if not better.

If this still does not match up to your requirements you best look for a different method.

Hope this helps