Dear Guys,

Kindly need your help on interpret the graphical result of GR&R. Base on the attached file, what I can observed is

1. The study was done base on part tolerance instead of total variation. Believe there is a reason behind such the customer would like to see whether the process control is able to meet the part tolerance and suitability of gage for product acceptance to the specification. As a end result the study show pretty fine
2. The GR&R value that calculate in total variation is seem to be over than 30%. By right the study should fail. Seems the interest is in different view, somehow it is accepted.
3. The impact will be seeing in NDC, the value will be lower than 5 due to GR&R value is over than 30% (calculate in TV) – Correct me if my understanding is wrong
4. R Chart by operator – I can’t really explain about the graph as no idea on it. As far as I know the point should not fall out side the control limit. Believe you guys can help me to understand the purpose of this graph.
5. X Bar Chart By operator – it seems most of the point is fall outside the control limits, this show the variation is due to difference between part instead of measuring system. However fixing operator misnan and zurin can show much more good result.
6. Data by part number – show the variation between parts, the white dot show the different value (range) is seeing for each trial. We also can conclude the part problem as question why a certain part shows different reading for most of the trial (range). Correct me if my understanding is wrong
7. Data By operator – Show the operator variation in term of reproducibility. A straight line a cross the appraisal shows the good reproducibility.
8. Operator part interaction – indicating the significant interaction between each Part and Operator. I can’t really understand as what the graphs try to deliver in stead of interaction. It is related to reproducibility?


Periannan :notme:

Dear Guys,

Kindly need your help on interpret the graphical result of GR&R. Base on the attached file, what I can observed is
You only included the graphical portion of the analysis, so some of this response is based on my experience versus your actual data. It also assumes that the parts selected adequately represent the actual process variation. If they do not, the study was not valid for assessing gage suitability for process control, and my comments are irrelevant.

The first chart to evaluate on MSA graphs is the RANGE chart. There should be a minimum of 5 possible range values below the upper control limit. Your graph shows 1. This means that this gage has inadequate resolution. When a gage has inadequate resolution, the rest of the MSA is invalid. You must address the resolution, then repeat the study.

Having said that, I will try to answer the other questions in light of the resolution issue to clarify these issues.

1. The study was done base on part tolerance instead of total variation. Believe there is a reason behind such the customer would like to see whether the process control is able to meet the part tolerance and suitability of gage for product acceptance to the specification. As a end result the study show pretty fine Decide in advance for what purpose the gage is to be used, part inspection to tolerance, or process control? A gage may be acceptable for use in inspection, but not acceptable for use in process control. P/T Ratio is used to assess part inspection. %GRR and ndc are used to assess process control.

2. The GR&R value that calculate in total variation is seem to be over than 30%. By right the study should fail. Seems the interest is in different view, somehow it is accepted. If P/T Ratio (%Tol) was less than 30%, the gage could be used for inspection, even though %TV is greater than 30%. See response to question 1.

3. The impact will be seeing in NDC, the value will be lower than 5 due to GR&R value is over than 30% (calculate in TV) – Correct me if my understanding is wrong If the gage is used for process control, you are correct. These measures have been negatively influenced by the poor resolution.

4. R Chart by operator – I can’t really explain about the graph as no idea on it. As far as I know the point should not fall out side the control limit. Believe you guys can help me to understand the purpose of this graph.The gage resolution is so poor that any range other than zero falls outside the limit.

5. X Bar Chart By operator – it seems most of the point is fall outside the control limits, this show the variation is due to difference between part instead of measuring system. However fixing operator misnan and zurin can show much more good result.The criteria for evaluating this chart is that more than 50% of the points should be out of control limits. On this chart, out is good.

6. Data by part number – show the variation between parts, the white dot show the different value (range) is seeing for each trial. We also can conclude the part problem as question why a certain part shows different reading for most of the trial (range). Correct me if my understanding is wrong You want to see a tight pattern of dots around each part. You want to be able to see the difference between parts.

7. Data By operator – Show the operator variation in term of reproducibility. A straight line a cross the appraisal shows the good reproducibility. Correct.

8. Operator part interaction – indicating the significant interaction between each Part and Operator. I can’t really understand as what the graphs try to deliver in stead of interaction. It is related to reproducibility? Ideally, these lines should fall on top of each other. This implies that different operators had problems measuring parts 2, 3, 6 & 9. However, the poor resolution is the probable cause of this.

Dear Miner ,

Thanks for the explanation, attached is the data, result in Excel and Mini tab (graph and session result). for you information.

Regarding R chart Bar, appreciate you explain more

a. it is a must to have minimum 5 possible range that show below the control limit. let say most of the point above the control limit or only 2 range was represented, it mean the study is invalid.

b. Appreciate you can explain more as why you say " resolution is so poor that any range other than zero fall outside the limit"

c. What can be done to correct the resolution, mesurement method or changing the gage.

d. What does it mean if the dot patten is not tight. is it we can conclude it is due to variation that create by operator for each trial

e. Regarding the operator * part interaction, what will be the judgment if the p value shows 0.00

Periannan

Regarding R chart Bar, appreciate you explain more

a. it is a must to have minimum 5 possible range that show below the control limit. let say most of the point above the control limit or only 2 range was represented, it mean the study is invalid. That is correct.

b. Appreciate you can explain more as why you say " resolution is so poor that any range other than zero fall outside the limit" Your gage is only reading to 0.01. The UCL is 0.00944. you can only have ranges of 0, 0.01, 0.02, ... Therefore any range other than zero is out of control. If your gage read to 0.005, you could have 0, 0.005, 0.010. This gives you two ranges within control limits. Still not good enough. A resolution of 0.001, gives 0, 0.001, 0.002, 0.003 ... This provides 10 possible ranges within control limits, which is good.

c. What can be done to correct the resolution, mesurement method or changing the gage. Your gage should be capable of a resolution of 0.001. In your data, your gage only reads out to 0.01 unless your operators are rounding off the third decimal. If they are rounding off, just repeat your study using all decimal places. If your calipers only display two decimals, try using a micrometer.

d. What does it mean if the dot patten is not tight. is it we can conclude it is due to variation that create by operator for each trial The dots are a combination of both repeatability and reproducibility. This chart is most useful for spotting individual parts that have a much higher measurement error. Maybe there was something wrong with a particular part that resulted in more variation such as a burr.

e. Regarding the operator * part interaction, what will be the judgment if the p value shows 0.00 This tells you that certain operators had trouble measuring specific parts. For example, if one operator measures a shaft diameter by averaging the min/max diameters while another measures the maximum diameter only, variations in the shaft roundness will have a greater impact on the second operator. This would show up in the interaction.


After reviewing your data, your gage is perfectly acceptable for measuring parts to tolerance for inspection (P/T Ratio = 7.07%). It is not suitable for use in measuring parts for SPC or for capability studies (%GRR = 59.92%, ndc = 1).

Look at selection of parts. Your max part measures 52.56, min is 52.52: that's a range of 0.04 against part tolerance of 0.6 ?

Look at selection of parts. Your max part measures 52.56, min is 52.52: that's a range of 0.04 against part tolerance of 0.6 ?

You are 100% correct. That is why I prefaced my first response with the following comment.

It also assumes that the parts selected adequately represent the actual process variation. If they do not, the study was not valid for assessing gage suitability for process control, and my comments are irrelevant.

Too many people overlook that very important requirement or deliberately select parts from the specification limits. That is okay if you are only using P/T Ratio as your indicator, but %GRR and ndc require parts that represent the actual process variation.