I have encountered a DOE aspect issue as below:
Background:
1. A new plastic mold with 20-cavity, that means each shot/trial will yield 20 parts
2. each part has two critical dimensions.
3. there are 11 factors/variables (injection molding machine parameters) which will have impact on these 2 critical dimension
Purpose of DOE
1. By DOE, custom design, to detect which factors/parameters are RED factor or significant factors
2. Once significant factors is determined, what range of parameter can be used to control these significant factors? and achieve design specification.
what have I done?
1. in JMP, I open a custom design
2. in Response dialog, I input all 40 response variables and their upper and lower limit
3. in factor dialog, I input all 11 variables with 2-level, and change necessary into if possible
4. click on "continue", then "make table"
5. come a DOE plan table with 8-trial
6. done all 8-trial and come a result with 40-response
7. Analysis> Fit model, and select 40-response and click Y; select 11-factor and add to construct model effect; hy here I concern that
--If I follow multivariate analysis principle, I fail to get what I want? which descripted in purpose of DOE
- If I select standard least square from personality, I can get what I want? find significant factor, and use profiler to define their range, but I think it is unreasonable to multivariate analysis.
in addition, in JMP guide, I have yet found any DOE example is related to multivariate, but in real world, there are many case like that especially in industrial. I also have put that case in
http://www.jmpforum.net/forum.php/forum.php?mod=viewthread&tid=1599
but by now, no one can help me.
