reg taguchi method

[dhiyanes k]

hello sir

            i am currently working on a project "die casting process parameters  optimization".In that i have taken five factors at three levels.

the array i have choosed is L27 array with three interactions. My doubt is whether interactions are useful in industrial oriented experiments

since i am getting that there is some significance effect of interactions on the process say (AxB =2% ,BxC=3% AxC =2.5%). Also what are advantages of taguchi method compared to other doe methods like response surface methods and factorial designs.

thanks

dhiyanes

[Mr. Mahdyarfar, Mohammad]

Dear dhiyanes

The importance of interactions in your case is depend on the percent of error in your study. If the error is very low, so the interaction effect will be meaningful. The ANOVA is the best tools for obtaining the significance of each effect.
The most important advantage of Taguchi method over factorial designs or response surface designs is the need for less experiments. If you want to perform your study with full factorial, you will need to perform 243 experiments. The other advantage is that we are free to choose the study of desired interactions and forget the others.

Sincerely Yours
Mohammad Mahdyarfar

________________________________
From: tda...@googlegroups.com [tda...@googlegroups.com] On Behalf Of dhiyanes k [kdhi...@gmail.com]
Sent: Saturday, January 22, 2011 11:10 PM
To: tda...@googlegroups.com
Subject: reg taguchi method

[Rogier van Duin Duvaro]

For investigating five factors on three levels, I recommend using an L18 screening design.
In this case you can concentrate on the main factors via the ANOVA.

You then reduce in the next design only those factors which are significant and then address the interactions to them if needed.
By running a confirmation run on the paper champion and compare this with the prediction, you can have an indication whether you have forgotten something to include in your design, such as a factor.

In the L18 you can include 1 factor at 2 levels and seven factors at three.
And get the interaction between the first factor with two levels and second factor of three levels.
Or one factor at six levels and six of three levels.

If you take in your case five factors at three levels, you still can add more factors in the L18.
If you decide not to, you can analyze the empty levels and get an indication of the variance of the experiments.
If one of the empty levels is more significant, you probably have some influence from outside or forgot an important factor.

Good luck!

Met vriendelijke groet / Kind Regards / Mit freundlichem Gruß / Saludos

drs. Rogier van Duin NPDP

DUVARO CONSULTANCY

Tel. +31(0) 411 85 15 15
Mob. +31(0) 624 629 355
Email rogier....@duvaro.com
Skype rogier.van.duin
Internet www.duvaro.com

Akkerstraat 65
NL-5076HE
Haaren
the Netherlands

[lrsm...@peoplepc.com]

Dr. Taguchi's methods use factorial designs and could actually employ response surface methods. The difference with Dr. Taguchi is that he deliberately hits combinations of control factors with combinations of controlled noise factors (factors such as tolerances, environment variation, or factors that deteriorate over time) to find a combination of control factors to make your process and/or design robust or insensitive to noise.

If you are interested in interactions you can best study them with two-level factorial experiments. Three or more level factorial experiments are used to study response surface curvature or when you have more than two factors you want to explore.

Along with the other comments, I hope this helps or at least encourages you to get into a good book or class. Many people have had huge success using these techniques to improve die casting operations.

Larry

Sent from my BlackBerry® wireless handheld

---

From: dhiyanes k <kdhi...@gmail.com>

Sender: tda...@googlegroups.com

Date: Sun, 23 Jan 2011 01:10:56 +0530

To: <tda...@googlegroups.com>

ReplyTo: tda...@googlegroups.com

Subject: reg taguchi method

[Mohsen Ebrahimi]

dhiyanes   1. you can ignore intractions at the moment. 2. each method has its advantages depends on type and conditions of experiment. for example, taguchi method is usful for experiments containing a large number of variables with few levels.   sincerely M. Ebrahimi --- On Sat, 1/22/11, dhiyanes k <kdhi...@gmail.com> wrote:

[Mohsen Ebrahimi]

dhiyanes   1. you can ignore intractions at the moment. 2. each method has its advantages depends on type and conditions of experiment. for example, taguchi method is usful for experiments containing a large number of variables with few levels.   sincerely M. Ebrahimi --- On Sat, 1/22/11, dhiyanes k <kdhi...@gmail.com> wrote:

From: dhiyanes k <kdhi...@gmail.com> Subject: reg taguchi method To: tda...@googlegroups.com Date: Saturday, January 22, 2011, 7:40 PM

[dhiyanes k]

hello sir

             Thanks for your reply for my last questions that i posted to you.Now in my project i have taken five factors at three levels.

In that after analyzing with annova table ,for three factors f ratio calculated is less than f ratio tabulated.What does it mean?Is it necessary that f ratio calculated for the factors should be greater than f ratio tabulated.

Thanks

K.dhiyanes

[Mahbube Khorramnezhad]

hello

when Fcritical> F calculated is, that parameter is insignificant

---

From: dhiyanes k <kdhi...@gmail.com>

To: tda...@googlegroups.com
Sent: Thu, January 27, 2011 10:02:00 AM
Subject: reg taguchi method

[Mr. Mahdyarfar, Mohammad]

Hi

If you repeat the experiments, and have the a good DOF for the error, you should have greater f for a factor than tabulated f to conclude this factor have a meaningful effect.

Regards
M. Mahdyarfar

________________________________
From: tda...@googlegroups.com [tda...@googlegroups.com] On Behalf Of dhiyanes k [kdhi...@gmail.com]

Sent: Thursday, January 27, 2011 10:02 AM

To: tda...@googlegroups.com
Subject: reg taguchi method

hello sir