Custom DOE with Disallowed Combinations Scr

[#1]

Hi,

I have 5 factors for a custom DOE and I need two linear constraints:

Mean(CAP + RAP + OAP + EAP) + 10 < RRP
Max(CAP, RAP, OAP, EAP) - Min(CAP, RAP, OAP, EAP) < 150

It is ok for CAP, RAP, OAP, EAP and RRP to vary significantly between 100 and 500, but if they vary, I want them to be somewhat close together. For example 450,400,350,450,500 is ok, but 100,500,100,500,500 is not.

I've tried the below constraint. The design calculates, but the constraint is not applied.

AND(
Mean(CAP + RAP + OAP + EAP)+10 < RRP,
Max( CAP, RAP, OAP, EAP )- Min( CAP, RAP, OAP, EAP )<150)

Thanks for your help

[#2]

I think the syntax for the call to the Maximum() function is incorrect. Try this code instead:

And( Mean( CAP, RAP, OAP, EAP ) + 10 < RRP, Max( CAP, RAP, OAP, EAP ) - Min( CAP, RAP, OAP, EAP ) < 150 )

[#3]

Thanks, but it did not work. In fact I cannot spot the difference between both formulas - did you maybe copy the wrong one?

[#4]

The difference is that your version had a single argument for the Mean() function (the sum of the factor levels) and my version has four arguments for the same Mean() function (the individual factor levels).

I don't know why the expression did not work. Remember that when the expression evaluates true, the current treatment is disallowed. Do you want to change your logic?

[#5]

Oh yes, that plus was nonsense, thanks. This condition works now, but the Max-Min does not. Also I got the logic wrong, I changed the <>.
I am trying it with a very simple condition such as

Mean( CAP, RAP, OAP, EAP ) + 10 > RRP & CAP - RAP > 10

but still this one is not applied. I get DOE suggestions with CAP=350 and RAP=150.

[#6]

You might need parentheses to make sure that the difference is calculated before the comparison.

BTW, what is the reason to constrain your factors so?

[#7]

No, parenthesis did not work either. In fact JMP removes them when I click back after I generated a design.

Those values are pressure zones of a polishing tool. It is meaningful to vary the pressure within the desired ranges, but it is not very meaningful to have two out of the four zones at very high pressure while the other two are at very low. Still I want variation to a certain degree between the zones to optimize the polishing homogeneity.

[#8]

Constraints are used to avoid treatments that might produce no response or are physically impossible. (Safety or cost might be other considerations.) They are not intended to avoid treatments that you suspect produce undesirable responses (failures). The treatments are intended to provide sufficient data to fit the model that can make unbiased predictions of the response regardless of whether the response is judged to be desirable. I am just trying to clarify your use of 'not very meaningful.'

I will try to get to the issue of the correct JSL expression for the disallowed combination later.

[#9 · Accepted solution]

I have used all the methods for specifying constraints. I never encountered a situation before where I did not get a design with the desired runs. I do not know why it is not working now. I got much better results, though, by using Or() instead of And() to consider both constraints.

I attached my design data table and a script for the custom design. I suggest that you contact JMP Technical Support ( support@jmp.com ) for more help. Please report back here if you learn the nature of the problem and the resolution.

[#10]

Thanks a lot for the help. Thinking about it again, in fact, only OR() makes sense.

First a run is created/suggested. If it violates condition 1, it is disallowed. OR if it violates condition 2, it is disallowed too.

The AND would only disallow runs that violate both conditions at the same time.

I need to work on my logics