Set desirability - Maximize desirability VS Simulation experiments

[#1]

Hi there,

I need your support please to find the optimal conditions which allow having the highest Y1 (yield) while keeping Y2 (impurities) lower than 70.

Both responses have equal importance.

I’ve tried with “maximize desirability” after setting desirability. When we set desirability, we have 4 options (maximize, minimize, target and none).

• For Y1: I’ve selected “Maximize”
• For Y2: I am left with the option “Minimize”, but it is not really minimizing but not exceeding an Upper Limit (70). I’ve tried to select “minimize” and play with “Middle” and “low” values and was able to see that I am moving closer to my objective.                  

My Question n°1 would be how to set desirability for Y2 ?

My second question would be: Is there any way to maximize desirability while keeping a defect rate lower than 2% (the risk of exceeding the Upper Specification limit of 70 for Y2) ? Maybe maximizing desirability but also considering “simulation experiments” ?

Thanks a lot for your support!

Best regards,

@martindemel 

[#2]

Hi @JMPdiscoverer ,

My recommendation is to set:

• Y1 to maximize with the default values and desirability, and an importance of 1
• Y2 to minimize with the Low value at 70, the High value above the highest expected impurity, and the middle somewhere in between.  Start by setting the importance to 2, and increase it if the profiler doesn't lower it enough.

In the profiler, you should see the desirability function increase with lower impurity values, but only until it gets down to 70.

Note that you can set these using 'response limits' in column properties so they are auto-populated when you open the profiler. This way you only need to enter 70 as the minimum value for impurities and the profiler will fill in the others.

[#3]

There are many ways to approach this, depending on what you want to achieve.

I would actually set the desirability for Y2 similar to what you have, upper at 70 but would put the middle at something like 69.5 and the low at 69. This would imply that you must be lower than 70, but no additional benefit to go lower than 69 (or some suitable smaller value). 

Having said that, please keep in mind that there may be additional benefit to going much lower than 70. After all, Y2 is impurities! If you really don't want to go lower because of the cost involved in doing so, you may actually have a Y3 which should be cost as that is providing an additional constraint on your solution.