Simple and clear explanation. I found you at the right time. I have a question. Can you please tell me which DOE should I select If I want to conduct an experiment in which the factors are increasing at a regular interval. For example, cutting speed, Range= 50 mm/s to 150 mm/s with an increment of 10 mm.( 50,60,70,80,90....)
Frans - good to hear from you. Here's another video to watch - th-cam.com/video/_qS1wanpB7M/w-d-xo.html It uses different software, but you'll get the idea of how to do a 3 level DOE and which one to use. There is no need to test at more than 3 points. You just need enough points for the maths to construct a curve and 3 points will do that...
This is a very good example on full factorial on Minitab. I have one comment and one question please. Comment: with this setup we could not use the standard deviation as we would have if the arrays were placed across the sheet. So, we miss looking at the graph plots for the standard deviation. Question: had we used the other setup where the arrays were across the sheet and calculated two columns for the mean and the standard deviation, would we get the same results?
Hi Mohammad - good to hear from you. The using the method in the video you can get analysis of Ln S. Which will essentially give the same answer as if you calculated across the row Std Dev. The numbers for the across the row method are more understandable. But if you just want to know which factor has the biggest effect on SD and how to minimise SD, both methods will work almost identically....
Thanks for this amazing video. How exactly can Minitab specify an optimum formula? or does it only suggest the way to go as in the values that will likely yield the optimum formula?
Great example Paul, I really enjoyed going through this! May I ask what is your opinion on removing a main effect with no significance when it has a significant interaction? Like here the factor C (Infill) seems like a not important thing (you also said to remove it) but it interacts with Nozzle (AC) which has a significant effect on the response. Removing it would cause a non-hierarchical model (minitab suggests to keep it). Is it that important to have it? I haven't found any clear rule for this so far. What would you do with your practical approach?
Zoltan - good to hear from you. When you're using software, it's best to keep a main effect if it is part of an interaction, but not important itself. The only reason you do this is so the software can work out where to set the interaction. So if the AB interaction is important, the software needs to know where A is set and where B is set to work out the interaction setting. If you remove A from the model the software now cannot see A and B and cannot work correctly. My main software is DOE Pro and it;s work if you remove main effects like that...so it;s a software issue, not a maths issue...
Hi Brandon - the reponse is the output data from your test that you're interested in improving. So if you weld, maybe the weld strength is you output reponse..
Great to hear from you, in minitab there is an optimiser function on the DOE menu. So it's really just a case of telling the software what the target is and letting it come up with a suggested setting...If DOE is your thing, I have a new book Design of Experiments for 21st century engineers you can get it from LULU.COM....
Hi Brandon - not necesarily, if you collect 2 data points per run set up, then you're doing 2 replications. As a good rule you should try to collect minimum of 30 data points in total in the experiment...
Thank you Paul, as always good video
No worries....
Thanks!
Simple and clear explanation. I found you at the right time.
I have a question.
Can you please tell me which DOE should I select If I want to conduct an experiment in which the factors are increasing at a regular interval. For example, cutting speed, Range= 50 mm/s to 150 mm/s with an increment of 10 mm.( 50,60,70,80,90....)
Frans - good to hear from you. Here's another video to watch -
th-cam.com/video/_qS1wanpB7M/w-d-xo.html
It uses different software, but you'll get the idea of how to do a 3 level DOE and which one to use. There is no need to test at more than 3 points. You just need enough points for the maths to construct a curve and 3 points will do that...
This is a very good example on full factorial on Minitab. I have one comment and one question please.
Comment: with this setup we could not use the standard deviation as we would have if the arrays were placed across the sheet. So, we miss looking at the graph plots for the standard deviation.
Question: had we used the other setup where the arrays were across the sheet and calculated two columns for the mean and the standard deviation, would we get the same results?
Hi Mohammad - good to hear from you. The using the method in the video you can get analysis of Ln S. Which will essentially give the same answer as if you calculated across the row Std Dev. The numbers for the across the row method are more understandable. But if you just want to know which factor has the biggest effect on SD and how to minimise SD, both methods will work almost identically....
ear paul, please upload video reg- gasification process analysis in Minitab
Thanks for this amazing video. How exactly can Minitab specify an optimum formula? or does it only suggest the way to go as in the values that will likely yield the optimum formula?
yes that is exactly what minitab is doing...
Great example Paul, I really enjoyed going through this! May I ask what is your opinion on removing a main effect with no significance when it has a significant interaction? Like here the factor C (Infill) seems like a not important thing (you also said to remove it) but it interacts with Nozzle (AC) which has a significant effect on the response. Removing it would cause a non-hierarchical model (minitab suggests to keep it). Is it that important to have it? I haven't found any clear rule for this so far. What would you do with your practical approach?
Zoltan - good to hear from you. When you're using software, it's best to keep a main effect if it is part of an interaction, but not important itself. The only reason you do this is so the software can work out where to set the interaction. So if the AB interaction is important, the software needs to know where A is set and where B is set to work out the interaction setting. If you remove A from the model the software now cannot see A and B and cannot work correctly. My main software is DOE Pro and it;s work if you remove main effects like that...so it;s a software issue, not a maths issue...
@@paulallen5321 Thank you for your swift answer. I just received your DoE book, can't wait to start reading it. Best regards from Hungary!
In this video how did you get your response? i am not sure what that means in doe experiment.
Hi Brandon - the reponse is the output data from your test that you're interested in improving. So if you weld, maybe the weld strength is you output reponse..
Great tutorial Paul.
Can you please tell me how to find an optimized point?
Great to hear from you, in minitab there is an optimiser function on the DOE menu. So it's really just a case of telling the software what the target is and letting it come up with a suggested setting...If DOE is your thing, I have a new book Design of Experiments for 21st century engineers you can get it from LULU.COM....
instablaster.
Would the number of replications be equal to the number of runs?
Hi Brandon - not necesarily, if you collect 2 data points per run set up, then you're doing 2 replications. As a good rule you should try to collect minimum of 30 data points in total in the experiment...