624119_en-US

liliconcom

How to define couples of responses from a mixture DoE?

One of my DOE users has defined a mixture design with 4 factors. To be more precise it is a mixture of mixtures.

During the tests it estimates 2 responses: pH and conductivity.
These two responses work as a couple. They do not necessarily want to define an optimal setting but couples of responses and therefore the optimal setting per couple.

 

I did a small example (with no constraints). Let's say that you have about an error of +/-0.01 for y1 and +/-0.02 for y2.

So, if you want to achieve (y1,y2)=(1.25,0.1), your limits are ([1.24, 1.26] & [0.08,0.12]) and you have the following design space:

let's change the couple (y1,y2)=(0.75,0.3)

and so on, you can do itcouple by couple, but if the number of couple isgrowing , does anyone have a way to do it efficiently?

 

Thank you for your suggestions,

Florence

Re: How to define couples of responses from a mixture DoE?

Idea just out of my head without thinking  

How about having a script, changing the response limit values for target, low, high to the desired couple and optimize the response, this could be done in a loop and build a report for all different scenarios. (alternatively, have that many response columns as you have couples and add the specifiv response limits with match target to each column. Then fit a model using all those responses and check fit separately. 

 

 

Re: How to define couples of responses from a mixture DoE?

Yes, Martin, you are right, a script is the good answer and I created it few weeks ago. 

The user accepted today that as a nice workaround, so I share this example here. 

You have two data set. One related to the DOE (I use the tiretread data) and one with the couple of expected responses. For each variable, we have the accepted variation.

 

Nota: due to the name of report section names created by the Fit Model platform, this script works only if your JMP is in English!

Next step would be to generalized that....