I am planning to do an experiment regarding the instantaneous and continuous firing of certain inflammable liquids. The plan of the experiment is as follows, where I use following parameters namely
(i) four types of bases (concrete, asphalt, vinyl and plywood)
(ii) the base is titled for four different angles in degrees (0,1,2 and 5)
(iii) three flammable liquids (petrol, kerosene and diesel) According the above factors, I need to do the experiment for 48 combinations. Kindly suggest me how to optimize and reduce the number of combinations with the aid of taguchi method.
Hi Akilan,
Please find attached one option for your experiment. This design consists of 24 runs only. In the excel file I also included the alias matrix between main effect contrast vectors and two-factor interaction contrast vectors.
If you assume that two-factor interactions are negligible, then the maximum correlation between main effect contrast vectors is just 0.19, which it's not really bad. Is the price that you have to pay for using a smaller design. The main effects contrast vectors and the two-factor interaction contrast vectors are also partially correlated; the maximum value is 0.5 but other correlations are also small. So, you could use a model selection method to also include two-factor interactions in your model and get more information.
To know more about the analysis and construction of mixed-level designs with partial correlation I recommend you the book of Wu and Hamada, Experiments: Planning, Analysis and Optimization, 2nd Ed, 2009.
I generated the design using JMP. I suggest you to try it, it is really good. You can get a 30-day trial version. With JMP you can get all the characteristics of the design I sent you, e.g. D-Efficiency, etc. Moreover, you can do the analysis using JMP; for instance, forward selection regression.
I hope it helps.
The only thing that I couldn't figure it out was, what kind of contrasts does JMP use to compute the correlation. I hope someone can answer such question or, if you find it out, please let me know. Sorry for that.
Hi,
You can use Minitab software for this purpose. By using Taguchi method, you can reduce experiments to 8, 9 and 16 runs.
Follow this order in Minitab:
Stat >>> DOE >>> Taguchi >>> Create Taguchi design
- Badges
- Science topic
- Similar topics
- Higher Education
- Universities