Just amazing, you made me understand this chapter in 3 videos, me who was pulling my hair out on manuals. Thank you very much for the quality of your work.
Hey Andy! I've been watching your videos for a long time and love the content. You have been a tremendous help to me and my I appreciate the work you do. One small correction to your statement about 6 Sigma. 6 Sigma is achieving at least 6 standard deviations between your mean and either specification limit. ±3 sigma is only a 3 Sigma process. Edit: I realized you are just speaking about the standard deviation in terms of its description of a normal distribution (6 StDev is expected to capture 99.7% of all data), not a "6 Sigma process". My bad! Keep up the amazing content!
Great question, and I remember having this exact same question come up when I first looked at a Gauge R&R report. It doesn't make sense, because as Engineers we've been told that variation within our parts is a bad thing! But in a gauge R&R, we can the majority of our variation to come from our parts, and not from our measurement system. So it is a good thing when that part-to-part variation is a big number, and that essentially means that our measurement system contributed a small part of the overall (total) variation that we observed in our study.
@@greenbeltacademy Oh okay, I wouldn't have thought of it like that. Thank you for your explanation! Your content is amazing, and I wish I had found it months ago so I wasn't trying to absorb so much before this test 😵💫
How different is this process when you do an attribute MSA rather than a variable MSA? For example I have a camera that tells me if a part is good or no good. The problem is the camera is giving false rejects so my boss asked me to do an MSA/Gauge R&R on it. I feel like a lot of the calculations in this video wouldn't work on the binary data I am going to get but I could be wrong.
Great question Brad! So the process is very different when you're dealing with attribute data. Google "Attribute Agreement Analysis" to understand how to quantify the variation in a measurement system when you're dealing with attribute data.
Great video!! It helped me understand the concept. I'm still confused about a specific scenario when you have a fixture with different stations, should I run my GRR placing the part always on the same stations and randomly? Would that add more variation in my GRR results? Thanks in advance.
Great question!! Okay, so the pure definition of repeatability is the variation that is unique to the measurement equipment, and that should also capture any variation in the equipment fixture. So when you run your GR&R you can randomize the fixture that you're using when you take your replicate measurements, but definitely know that this will increase your repeatability. If your repeatability ends up being unacceptable, then you can look at your fixture specifically as a source of variation that could be improved to reduce your repeatability.
Hello! This was an amazing series and I learned so much for an upcoming project I have for my work so thank you for the information! I do have one point of posssible correction for the excel sheet template we get from the link. I noticed that the final equation for the AV cell might have missed a set of parentheses for the max - min cause I think it is dividing the min average by d2 and then subtracting that from the max average. It was causing my AV % to skyrocket to like 90%. Either way the template saved my life for my project and I extremely appreciate the work you put into educating people on this topic!
This is a great video. One question, how do I interpret the result if %study variation exceeds 30%, while %tolerance is smaller than 10%? Is this measurement capable or not?
Are you using that measurement system for product control or process control? This result could occur is you have a very capable process which would result in very low part-to-part variation. Also if you don't select 10 samples that span your entire design tolerance, you can also get this result.
@@greenbeltacademy I am using it for product control. The AIAG MSA manual states that those 10 samples must be chosen in order to cover the entire process spread, which is true in theory but not in practice, especially in the NPI stage of a project, because not many products were produced. But we have to assess the measurement system at the beginning. So there is a lot of argument on this.
@@gjfan8990I completely agree, it can sometimes be difficult to get parts that span your entire tolerance range, and even your entire process range if you're working with a new process or a new design!! I've experienced that challenge on many occasions! The good news is, you're using your measurement equipment for product control, so % Tolerance is more meaningful for you, AND you can explain why your %study variation is high due to the fact that your samples are limited to a narrow range.
If you are looking to reject or accept product against a tolerance what range of parts should I choose for my study? Spread the tolerance or spread over the known historical process variation? I’ve been told tolerance but as the process is capable I need to misprocess to generate the arts that cover the tolerance.
It depends on which analysis method you choose! If you're going to use the % of Process Variation metric, then having parts that span your entire process range is a must!
@@greenbeltacademy shouldn’t I be using the tolerance % for inspection? If it depends then on what grounds so you not have to have parts that span the spec?
@@moosemoss2645 great question, I think in a perfect world, you'd attempt to have parts that span both the design tolerance and the known historical process variation. Also, yes, if you're using the gauge for conformance (product control) then yes getting parts that span the design tolerance is important to demonstrate that your precision is acceptable across the entire design tolerance.
@@greenbeltacademy thanks. Part of me doesn’t doubt that measuring over the design tolerance is the correct way but on the other hand if you have to misprocess to make the parts to fill this spread then use the historical deviation in your calcs it feels kind of strange.
Because when adding up variance (in standard deviation) we don't simply add the percentages, we must convert them to variance first, then add them up. Also, the total percentages do not add up to 100% for this same reason.
Hi Andy. Alan here. I been doing quite a bit of research on Gage R&Rs. A few articles mention you should also ensure individual categories (repeatability, reproducibility, part-to-part) meet criteria as well. In particular the R&R numbers should be less the 9% for each and your part-to-part variation is very high. What say you? I recently picked up the MSA book by the AIAG group and was wondering if they talk about that within it.
Hey Alan! yes the reproducibility for that data set is very high, and would be a focus area if I wanted to improve my GR&R. In terms of 9% for each element (Repeatability and Reproducibility), I think that's a very tight requirement. If each element was less than 9%, then the resulting overall GR&R would be way less than 30%, which would essentially be like moving the goal-post for GR&R requirements. This can be very difficult to achieve, especially as processes become more capable, and as design tolerance are shrunk.
Do you take the same percentages (less then 10, between 10 and 30, greater than 30) and use them as a criteria for the R&R plus part-to-part variation percentages too just like the percent to tolerance ratio?
Hey @@alanarnott4465! Okay so that criteria (30%,etc) does not apply to the part-to-part variation. You want the part-to-part variation to be very high, which essentially means that the majority of the variation within your gauge R&R came from the parts/process - which is what we want! That criteria only applies to the Gauge R&R value.
@@greenbeltacademy gotcha. If I got results of 20.23 for repeatability, 16.94 for reproducibility and part-to-part of 62.83 would that be considered good or bad results?
Dear Sir, I am trying to find online how to do Gage R n R using Specification tolerance method when NDC is less than 5. I could not find any. Not sure whether you could provide any step by step guide on this method.
Do you answer questions by email? I have a very challenging Gage R&R study my employer is demanding of me, and it's been an endless headache. I really could use some advice from someone knowledgeable.
Hey Jack, yes I do. Sign up for my email newsletter and just hit reply. To be honest, I always priorities emails from existing students, but I will try to reply as quickly as possible
@greenbeltacademy Thank you Andy. I found the newsletter sign-up page, but after filling it out & hitting the Submit button, I get a message that says: "Not found, error 404"
I love your videos, but this one has left my original question unanswered, although I only watched the 1st and 3rd videos in this series. I have Gage R&R chart (from somewhere else) for appraisers to complete, but I'm stuck on how to interpret the results as nothing jumps out at me to show whether the results are good or bad.
Wonderful set of 3 videos. Keep making these video
Thank you!
best 3 tutorial videos iv seen in my life👍
Wow, thanks so much!
Just amazing, you made me understand this chapter in 3 videos, me who was pulling my hair out on manuals. Thank you very much for the quality of your work.
You're absolutely welcome, I'm happy to help!
Thanks for this series! you laid out the information in a easily digestible way.
Your absolutely welcome, and thank you for the positive comment!
Great video Andy, thank you for sharing.
You're absolutely welcome John!
Thanks very much. A great resource!
You're absolutely welcome!
Watched all 3 videos. Very informative and Thank you so much for your effort to make this topic easy understandable to everyone. Much appreciation..🎉
You’re absolutely welcome!
I’m glad you liked those videos!
Best set of videos explaining the GRR concepts!
Glad it was helpful!
Thank you for sharing .... Awesome !! 🤩💚💯
You’re absolutely welcome!!
Hey Andy! I've been watching your videos for a long time and love the content. You have been a tremendous help to me and my I appreciate the work you do.
One small correction to your statement about 6 Sigma. 6 Sigma is achieving at least 6 standard deviations between your mean and either specification limit. ±3 sigma is only a 3 Sigma process.
Edit: I realized you are just speaking about the standard deviation in terms of its description of a normal distribution (6 StDev is expected to capture 99.7% of all data), not a "6 Sigma process". My bad!
Keep up the amazing content!
No worries, and thanks so much for the awesome feedback!!
Very informative 3 videos.... thank you!
Thanks so much!!! I"m glad you enjoyed that video!
Glad you like them!
Excellent series! Thank you.
You're welcome!
Great job, thanks!
You're welcome!
The best video I've seen so far ! thank you! XD
Wow, thank you so much!
Great Video, thanks for your sharing...
You're absolutely welcome!
Great series, thanks
Thank you so much, I"m glad you enjoyed it!
Hey Andy - Why is it good that the part-to-part variation is high in the final table and graph of this video?
Great question, and I remember having this exact same question come up when I first looked at a Gauge R&R report.
It doesn't make sense, because as Engineers we've been told that variation within our parts is a bad thing!
But in a gauge R&R, we can the majority of our variation to come from our parts, and not from our measurement system.
So it is a good thing when that part-to-part variation is a big number, and that essentially means that our measurement system contributed a small part of the overall (total) variation that we observed in our study.
@@greenbeltacademy Oh okay, I wouldn't have thought of it like that. Thank you for your explanation! Your content is amazing, and I wish I had found it months ago so I wasn't trying to absorb so much before this test 😵💫
@@catch-my-drift You're absolutely welcome, and I"m happy to help!!!
Hey Andy, It was a great material regarding GRR. I wonder if Green Belt Academy has video about GRR using ANOVA?
Thanks! I teach ANOVA for GR&R in my course, but I wanted to focus just on the basics here on youtube.
Very nice interpretation. Thanks
Thank you!
Thank for the videos! What about destructive tests?
Great question!
The method I've described here only applies to non-destructive testing. Destructive testing requires a completely different approach.
Does the interpretation of GRR results percentages (
Yes it does!
What AIAG standard is the d2 chart located? Looking to get the complete chart. Thanks.
It's their 4th edition of the MSA (measurement system analysis) Book
How different is this process when you do an attribute MSA rather than a variable MSA?
For example I have a camera that tells me if a part is good or no good. The problem is the camera is giving false rejects so my boss asked me to do an MSA/Gauge R&R on it.
I feel like a lot of the calculations in this video wouldn't work on the binary data I am going to get but I could be wrong.
Great question Brad! So the process is very different when you're dealing with attribute data.
Google "Attribute Agreement Analysis" to understand how to quantify the variation in a measurement system when you're dealing with attribute data.
Great video!! It helped me understand the concept. I'm still confused about a specific scenario when you have a fixture with different stations, should I run my GRR placing the part always on the same stations and randomly? Would that add more variation in my GRR results? Thanks in advance.
Great question!! Okay, so the pure definition of repeatability is the variation that is unique to the measurement equipment, and that should also capture any variation in the equipment fixture.
So when you run your GR&R you can randomize the fixture that you're using when you take your replicate measurements, but definitely know that this will increase your repeatability.
If your repeatability ends up being unacceptable, then you can look at your fixture specifically as a source of variation that could be improved to reduce your repeatability.
@@CQEAcademy Thanks for the taking the time to reply
You’re welcome
Hello! This was an amazing series and I learned so much for an upcoming project I have for my work so thank you for the information!
I do have one point of posssible correction for the excel sheet template we get from the link. I noticed that the final equation for the AV cell might have missed a set of parentheses for the max - min cause I think it is dividing the min average by d2 and then subtracting that from the max average. It was causing my AV % to skyrocket to like 90%.
Either way the template saved my life for my project and I extremely appreciate the work you put into educating people on this topic!
Wow thanks for the detailed feedback! It's great to hear that it's helping with your project.
This is a great video. One question, how do I interpret the result if %study variation exceeds 30%, while %tolerance is smaller than 10%? Is this measurement capable or not?
Are you using that measurement system for product control or process control?
This result could occur is you have a very capable process which would result in very low part-to-part variation.
Also if you don't select 10 samples that span your entire design tolerance, you can also get this result.
@@greenbeltacademy I am using it for product control. The AIAG MSA manual states that those 10 samples must be chosen in order to cover the entire process spread, which is true in theory but not in practice, especially in the NPI stage of a project, because not many products were produced. But we have to assess the measurement system at the beginning. So there is a lot of argument on this.
@@gjfan8990I completely agree, it can sometimes be difficult to get parts that span your entire tolerance range, and even your entire process range if you're working with a new process or a new design!!
I've experienced that challenge on many occasions!
The good news is, you're using your measurement equipment for product control, so % Tolerance is more meaningful for you, AND you can explain why your %study variation is high due to the fact that your samples are limited to a narrow range.
@@CQEAcademy Great, thanks for the feedback.
@@gjfan8990You're welcome!
If you are looking to reject or accept product against a tolerance what range of parts should I choose for my study? Spread the tolerance or spread over the known historical process variation? I’ve been told tolerance but as the process is capable I need to misprocess to generate the arts that cover the tolerance.
It depends on which analysis method you choose! If you're going to use the % of Process Variation metric, then having parts that span your entire process range is a must!
@@greenbeltacademy shouldn’t I be using the tolerance % for inspection?
If it depends then on what grounds so you not have to have parts that span the spec?
@@moosemoss2645 great question, I think in a perfect world, you'd attempt to have parts that span both the design tolerance and the known historical process variation.
Also, yes, if you're using the gauge for conformance (product control) then yes getting parts that span the design tolerance is important to demonstrate that your precision is acceptable across the entire design tolerance.
@@greenbeltacademy thanks. Part of me doesn’t doubt that measuring over the design tolerance is the correct way but on the other hand if you have to misprocess to make the parts to fill this spread then use the historical deviation in your calcs it feels kind of strange.
Hi Andy, how repeatability 12.10% + reproducibility 25.19% = Total GRNR 27.95%? I thought the sum would be 37.29%?
Because when adding up variance (in standard deviation) we don't simply add the percentages, we must convert them to variance first, then add them up. Also, the total percentages do not add up to 100% for this same reason.
Hi Andy. Alan here. I been doing quite a bit of research on Gage R&Rs. A few articles mention you should also ensure individual categories (repeatability, reproducibility, part-to-part) meet criteria as well. In particular the R&R numbers should be less the 9% for each and your part-to-part variation is very high. What say you? I recently picked up the MSA book by the AIAG group and was wondering if they talk about that within it.
Hey Alan! yes the reproducibility for that data set is very high, and would be a focus area if I wanted to improve my GR&R.
In terms of 9% for each element (Repeatability and Reproducibility), I think that's a very tight requirement. If each element was less than 9%, then the resulting overall GR&R would be way less than 30%, which would essentially be like moving the goal-post for GR&R requirements. This can be very difficult to achieve, especially as processes become more capable, and as design tolerance are shrunk.
Do you take the same percentages (less then 10, between 10 and 30, greater than 30) and use them as a criteria for the R&R plus part-to-part variation percentages too just like the percent to tolerance ratio?
Hey @@alanarnott4465!
Okay so that criteria (30%,etc) does not apply to the part-to-part variation.
You want the part-to-part variation to be very high, which essentially means that the majority of the variation within your gauge R&R came from the parts/process - which is what we want!
That criteria only applies to the Gauge R&R value.
@@greenbeltacademy gotcha. If I got results of 20.23 for repeatability, 16.94 for reproducibility and part-to-part of 62.83 would that be considered good or bad results?
Thanks for all of your insight by the way.
Dear Sir, I am trying to find online how to do Gage R n R using Specification tolerance method when NDC is less than 5. I could not find any. Not sure whether you could provide any step by step guide on this method.
Hey there, that's a tough topic. I don't think I have any good resources to share on that topic
@@greenbeltacademy Ha, thank you very much and I am proud that I ask a tough question.
Did you remove the Gauge R&R spreadsheet?
No you should be able to find it here: www.Greenbeltacademy.com/GaugeR&R
Articulated very well.
Thanks!
Do you answer questions by email? I have a very challenging Gage R&R study my employer is demanding of me, and it's been an endless headache. I really could use some advice from someone knowledgeable.
Hey Jack, yes I do. Sign up for my email newsletter and just hit reply.
To be honest, I always priorities emails from existing students, but I will try to reply as quickly as possible
@greenbeltacademy Thank you Andy. I found the newsletter sign-up page, but after filling it out & hitting the Submit button, I get a message that says: "Not found, error 404"
I love your videos, but this one has left my original question unanswered, although I only watched the 1st and 3rd videos in this series. I have Gage R&R chart (from somewhere else) for appraisers to complete, but I'm stuck on how to interpret the results as nothing jumps out at me to show whether the results are good or bad.
Hey Paul, man that's a huge bummer. Are you on my email list? If so, hit reply and send over those results and we can walk through them together.
what if our data is discrete (true or false)
Great question!
That’s a completely different mathematical operation.
Google attribute agreement analysis to learn more!