Antoine - Thanks for posting a comment, it help with the channel - If want more information on my DOE approach I've just released a new book... ' Design of experiments for 21st century engineers' you can get it from LULU.COM
Gerard - If want more information on my DOE approach I've just released a new book... ' Design of experiments for 21st century engineers' you can get it from LULU.COM
Tafi - thanks for your positive comments. If DOE is your thing you might want to look at my book 'Design of Experiments for 21st Century Engineers' you can get it from LULU.COM.
Very good lecture, thanks. one question is you concluded from L12 that Angle, Stop, and Pin are most influential factors, and you said we could do another DOE. What the DOE can be, full factorial?
Lohit - please send me your data in an excel sheet. Tell me how many variables you have the high and low values and the target you want to hit with the model you're creating. Paul.allen@allenp.co.uk
Thanks for an excellent demonstration. I have a question about doing a DOE with both numerical and categorical factors. Is that only possible with a Taguchi type experiment or could the categories be accommodated bin other types of DOE?
@@paulallen5321 Thanks so much Paul. I am looking to find optimal parameters for a process. Some of the variables I identified can be varied on a continuous basis, while others are distinct non-numerical categories e.g. type of chemical/ additive. The final aim is to maximize the % of component A in the A/B mixture. Your description of the Taguchi method makes it seem like an excellent screening tool to see which variables have the dominant effect and then go into a factorial design to optimize one or two of these dominant variables. Is Taguchi the best way to do this first screening or would you recommend a different type of DOE?
Hello Paul, my question might be very silly as I am new to DoE and I am using Design-expert software, I wonder why are you doing a L12 Taguchi model for 7 factors? the software suggest me a L8 for 7 factors at 2 levels. I am very confused.... Thank you for your videos.
Hi LVBS - It's tricky to explain but your software is giving you bad advice. I would chose the L12. The L8 was used 50 years ago to save calculation time. But you are taking risks with your model, that 50 years ago could save 3 weeks of hand calculations!! so was worth a risk. With software there is no calculation time so no benefit and lots of risks...
@@paulallen5321 thank you Paul for your responses. It really helps to clarify why the discrepancy, I just need to find how to talk the software to give me a L12 instead :)
Hi Paul, I use Design-Expert software. The software help page states that "Taguchi orthogonal array designs are not recommended. Modern designs along with ever improving and available computer power have superseded the need for these “calculate by hand” designs. Use optimal designs with measured responses instead." Would you agree with this statement? Thanks!
Personally I wouldn't agree with that. If you want to test 11 variables there is no software in the world going to get you out of that problem, you need a good DOE pattern, the L12 is a great choice to do this. I know they are simple, but that's the point...
@@paulallen5321 Min res IV design, two levels with 22 runs. I think Design Expert suggests this if there are interactions between factors as such in chemical systems. In Taguchi L12 , are main factors confounded with interactions if there are any?
That's not true in the L12, the confounding is spread very evenly through the pattern and there is no 2 way aliasing. It is also an all pairs test, which is a highly powerful test to do on any system...
Thanks for your thorough Explanation! However, I'm wondering if my experiment require a long time to finish (around five hours). I will have to invest a vast time to finish a full Taguchi design of experiment. Is it able to Omit duplicate experiment? if it is possible then how can I do the following data processing without S/N ratio?
Thanks for getting in touch - If you do not replicate then the analysis cannot find a robust setting in this case. Replication is one of the most important things you can do in an experiment. It should always be done if money permits...
In taguchi method l9 Please could you tell me why i have got , the sn ratio plot and mean plot were inversely The value of sn ratio were negative Smaller is better formula was used
Hayat - good to hear from you, personally I never use the S/N ratio I get a mean plot and a Std Deviation plot. This keeps these 2 aprt and easy to understand. If you mix them in the S/N ratio you lose sight of what is happening...
@@paulallen5321 I use sn ratio because i didn't completely understand why the mean plot is used for ? so if i want to depend on rank in table respons in order to find the most significant or effective parameter should i use the rank in table response of means or of sn ratio?
And if you dont mind could you tell me please if it correct or not , that the plot and table response of sn ratio used only to find the optimum parameters while plot and table response mean are used to determine the most effective parameters using rank value ?
