Screening test in mixture design

[#1]

Hi Martin Y here:

I have a question about how to run screening in mixture design.

Below is an example:

Cement 1:  36%

Cement 2:  15%

gypsum:    5%

accelerator: 0.2%

retarder:   0.1%

additive 1: 0.2%

additive 2:0.3%

filler 1: 20%

filler 2: 23.2%

Total 100%

Suppose I want to run a screening test first, I want to pick 5 or 6 factors, the ingredients in the formulation all can not be zero. So all the ingredients have lower and upper limit.

So basically it will be a Etreme  vertices design.The problem is: Etreme Vertices for 6 factors is huge. 

Is there any more economical screening design for mixture design?  what I want is:

1. screening design ( 5 or 6 factors with lower and upper limit) without replicates.

2. Exteme design ( 3 or 4 factors with lower and upper limit) with replicates.

Thanks!

Martin Yang

[#2 · Accepted solution]

Hello, Martin.

Although Extreme Vertices designs are useful for some mixture situations, most of the time a Custom Design would be a more appropriate choice. Choose DOE > Custom Design. Specify all of your mixture components as Mixture factors. You can even specify the ranges you wish to use. Specify the model that you wish to fit. You can then specify the number of runs you wish to conduct (of course this will depend on the size of your model, too). JMP will create the design for the situation you specify.

[#3]

Hi Dan:

Thanks for your suggestion.

Is there any good training materials of custom design in formulation development area avaliable?

Thanks

Martin Yang

[#4]

Hello Martin,

Building on Dan's response, by using the Custom Design platform you also have the flexibility to investigate process factors in combination with mixture factors (this may not be applicable in your example but worth mentioning). In your example, you also have the opportunity to navigate to the red triangle>Advanced Options>Mixture Sum and designate in your case that the mixture sum = 57% assuming that your last two factors = filler were a constant = 43% for example. Finally you can opt for a I-optimal design which you may prefer for mixture experiments.

[#5]

Hi Louv:

Thanks for your suggestion.  I am very eager to try it.

Does JMP have any similar case study that I can go through, it will help me understand, I deal with Mixture design most of the time, have not touched Custom Design yet.

Thanks

Martin Yang

[#6]

Martin,

On our on-demand portion of our website I found this webinar which be helpful to you.

https://www.jmp.com/en_us/events/ondemand/mastering-jmp/mixture-designs.html

There is also a course offering from SAS Education

https://support.sas.com/edu/schedules.html?ctry=us&crs=JMIX

Lou