There are different methods for design of experiments (as per my knowledge). These include Full factorial, fractional factorial, Taguchi and lastly response surface in which there are CCD and Box Benkhen. I just need to know basic difference between them and which to apply where.
Usually I have heard that Response Surface is preferred and would like to know why this is so? What advantage does it has over other techniques?
I think that this is misunderstanding. RSM is a methodology for optimization (modeling response surface and looking for optimal solution based on results from DOE. it si the second step) whereas for example Box-Behnken (we can say one particular design which we can use) is specific design for situation in which we have indentify significant and influencing factors (we can say for screening purpose). But I am not expert :-)
As per my knowledge when there is sufficient experimental material full factorial is used. When experimental naterial is insufficient fractional factorial is used and one or more main effects/interaction need not to be estimated. Taguchi method is used to reduce the runs i.e. number of treatment combinations. RSM is a methodology for optimization as stated by Dr. Michel Rost..
While considering a designed experiment for use in any situation one must keep in mind the main objective of the experiment. The objective may be treatment comparison, factor screening, response optimization or fitting a regression model. The basic designs like crd, rcbd, latin square designs, incomplete block designs and their variants are used basically for treatment comparisons. The factorial experiments are used for selecting important factors from a large set of factors. Fractional factorial designs are considered when we are not able to run a factorial experiment in full due to lack of experimental material. Response surface methodology is a mixture of regression analysis and experimental designs meant for response optimization. Different designs are used in this methodology according to the situation. Basically we may fit regression model by using data generated through any design but, the design specially developed for fitting regression model providing certain nice statistical properties usually called optimal designs are preferred over other designs for this purpose. Taguchi method requires less number of runs.
Majorly the Box benkhen design (BBD) in response surface methodology gives designers the opportunity of optimizing processes using 3 factors and 3 responses. other designs may not allow a designer this luxury and those that allow such do so with much more experimental runs than ypou will find on the BBD.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics