i want to do an design of experiment for regression modelling and optimization with some factors and one of them is categorical factor or variable which is a design of friction stir welding tool design, one tool design is totally different from another like cylindrical tool, threaded cylindrical tool, square tool etc,and other variables has numeric values how to do the design of experiment.......is it possible to follow the taguchi method.
I think yes. As far as it is a variable you can use it in Taguchi Design. For example:
Factors Level 1 Level 2
Tool Design Square Cylindrical
Rotation speed rpm 500 700
Welding Speed mm/min 200 300
How many levels do you have per factor? How many factors do you have in total?
If most of your factors are continuous/numerical, you can use something called a Definitive Screening Design. This method will test your numerical values at 3 levels and test for quadratic terms. You can also add some factors to accommodate 2-level categorical variables. If this sounds good to you, I can guarantee its a smaller design than the Taguchi Design.
I have total 3 factors.one of them is tool design.two are numeric variable.each has 5 level. Mr. Andrew Ekstrom if you can kindly give me some reference paper of definitive screening design it will be more clear to me.
If you perform a web search for Definitive Screening Designs you should be able to find the original paper. (I don't have it. I usually just read it on-line if I need it, sorry).
If you have 5 levels for the categorical variable Tool Design, you can also look at an "Optimal Response Surface". If you have 5 levels for a continuous variable, you can fit a quartic model. Which is a bit much. The Optimal Response Surface will be able to accommodate you well.
In general, you only need 3 levels per continuous variable.
Mr. Shivraj Yeole can you give me some reference paper that has used taguchi method for categorical factor
Hey Abhijit,
I uploaded an Excel file with a Taguchi Design and an Optimal Response Surface. I decided to look at the linear, quadratic, 2-way interactions and 3-way interactions. The Taguchi Design has some issues with confounding of 2-way and 3-way interactions. The Optimal Response Surface does not. Both designs are 25 runs. If you do not want to see 3-way interactions, the Optimal Response Surface requires only 20 runs. Between the 2 methods, I would go with the Optimal Response Surface.
Peter Goos and Brad Jones wrote a book called. "Optimal Design of Experiments: A Case Study Approach". It is highly read able and understandable. It reads almost like play.
Good Luck
Dear Abhijit Banik
please use design expertise software .
Factors Level 1 Level 2
Tool Design Square Cylindrical
Rotation speed rpm 500 700
Welding Speed mm/min 200 300
give number of factors 3
after that select numeric factor 2
text factor 1
after entering two factors
the third factor u should select text and enter the factor
use response surface methodology optimization technique
since u r doing in welding u can increase the factors 4 or 5
I have implemented definitive screening design and analysis steps are also provided. It is one of the efficient and effective technique. The design and analysis can be performed with JMP or Minitab software. You can refer to the article: https://www.researchgate.net/publication/309383755_Multi_Response_Simulation_and_Optimization_of_Gas_Tungsten_Arc_Welding
- Badges
- Science topic
- Similar topics
- Medicine
- Epidemiology