topic Mixture Design 14 factors or more... some needs to be equal to each other... in Discussions

Mixture Design 14 factors or more... some needs to be equal to each other...

Annelies — Mon, 13 Jan 2020 15:55:08 GMT

I'm new with JMP and I was hoping to set up a basis for our experiments for economical and time efficiency reasons.

I'm working with Epoxy and Polymers. For our products we use a component 1 and a component 2 system. When these components are mixed it get cured over time. Those component 1 and component 2, excist out of two parts. Part 1 is now called Liquid. These ingredients need to be equal to each other(reacting to each other in a ratio) or can hardly change. Part 2 is now called solid. These ingredients can easily change and influence the responses on their own.

Component 1 Component 2

GGREEN003A GRED0007A

GPURPLE009A GRED0008A (liquid part, total 7 ingredients)

GBROWN018A GPURPLE007B

GBLUE003A

GGREY019A GBLACK001A

GGREY025A GGREY031A (solid part, total 7 ingredients)

GGREY013A GGREY006A

GGREY006A

Everything together is 100%

In Custom Design you can Specify a factor with Hard-to-Change levels. But my factor is part of a mixture, so I'm not sure how to set this up with a Mixture Design.

Can someone help me with this?

If you need more explanation, please let me know.

Re: Mixture Design 14 factors or more... some needs to be equal to each other...

HadleyMyers — Mon, 13 Jan 2020 16:31:03 GMT

Hi,

This sounds a bit like a "Mixture of Mixtures" design (see help files: http://www.jmp.com/support/help/14-2/mixture-of-mixtures-design.shtml). I've set something up based on the example you described in the table below. I assumed equal portions of Component 1 and Component 2. Please take a look and let me know if it makes sense if it is something you can work with.

Re: Mixture Design 14 factors or more... some needs to be equal to each other...

Annelies — Tue, 14 Jan 2020 11:39:59 GMT

I have looked into it and it has answered a couple of my issues! :)

My only issue right now is that my C1 liquid has 1 value and my C2 liquid has 1 value. They need to react to a specific ratio for all the functional groups to react.

I have attached a file with the minimum and maximum as an example.

Unfortunately you can't open it as a "Load Factor". It gives the alarm "The current JMP Data table is not a valid factor table."

C1 Liquid is 0.4468 nothing lower and nothing higher

C2 Liquid is 0.3385 nothing lower and nothing higher

So they are part of the mixture, but they are also a constant factor.

Do I need to exclude them from the mixture factore then?

If I ad the constrain of:

1 C1 liquid + 1 C2 liquid </= 0.7853 (0.4468+0.3385)

-1 C1 liquid + -1 C2 liquid </= 0.7853

It changes back the minimum and maximum values as 0 and 1.

I'm sorry if it is standing obvious in the online handbook. All the infromation is kind of overwelming at the first glance.

The hardest part is to understand what for factors do I have and how do I place it into JMP?

I have read quiet a bit, but I figured it would be faster to just ask. :)

Re: Mixture Design 14 factors or more... some needs to be equal to each other...

Mark_Bailey — Tue, 14 Jan 2020 12:19:54 GMT

You can easily exclude the constant components. First, do not explicitly define them as mixture factors. Next, click the red triangle and select Advanced Options > Mixture Sum. Enter the value the represents the portion of the mixture that contains the non-constant components. I think that is 1.0000 - (0.4468 + 0.3385) = 0.2147 in your case. That is, only 21.47% of the total mixture is changing from run to run.

Re: Mixture Design 14 factors or more... some needs to be equal to each other...

Annelies — Tue, 14 Jan 2020 13:02:53 GMT

Thanks, I understand!

I was thinking too difficult. Due to the fact that all my factors were ingredrients, I was too focused on the mixture part.

Not realising I have constant factors and that it is ok to exclude them, even if they are part of the mixture.

Thank you Mark and Hadley :) it is solved.

Re: Mixture Design 14 factors or more... some needs to be equal to each other...

statman — Tue, 14 Jan 2020 18:56:13 GMT

It sounds like you've got what you want, but for posterity...mixture designs are intended for number of variables typically ≤ 4. I recommend reading "Tips for JMPing into Mixture Experimentation", Daniel J. Obermiller and, of course papers and THE book by John Cornell.

Re: Mixture Design 14 factors or more... some needs to be equal to each other...

Dan_Obermiller — Tue, 14 Jan 2020 19:35:57 GMT

Thanks for the plug for an old paper, @statman ! Hopefully there is some usable information in that paper, as I did cover more than just software. But I will have to say that the paper was written based on version 3 of JMP, so many of the software specific tips have been incorporated into the software by now. The software specific information may not be as useful.

Re: Mixture Design 14 factors or more... some needs to be equal to each other...

Annelies — Wed, 15 Jan 2020 10:04:07 GMT

I was already afraid of that. Due to the fact I hadn't found an example in the Handbook for that the Mixture Design section with more then 5 factors(ingredrients).

Thanks for the recommendations. I will look into that!

"papers and THE book by John Cornell", you mean the writer John A. Cornell, who has written:

Experiments with Mixutres Designs, Models and the Analysis of Mixture Data.

A Primer on Experiments with Mixtures.

And then I guess you mean the first book as THE book. :)

I just want to make sure.

Re: Mixture Design 14 factors or more... some needs to be equal to each other...

Mark_Bailey — Wed, 15 Jan 2020 14:14:03 GMT

If you are referring to the comment that mixture experiments are generally restricted to less than five components, I disagree. I have designed many experiments for mixtures with dozens of components and many linear constraints. The design methods, especially custom design, are not limited by the number of components. I will say, though, that the more components you have in an experiment, the more complicated it is. But that should not limit your experimentation.

I just (finally) bought the third edition of Cornell's book. It is one of the best DOE books in my opinion, but it is very technical and mathematical.There is another book written by Snee and Hoerl, "Strategies for Formulations Development: A Step-by-Step Guide Using JMP," that might be more useful at this time. It is also rigorous but more on the practical side. I don't want to discourage you from reading Cornell's book. I just want to set realistic expectations.

Re: Mixture Design 14 factors or more... some needs to be equal to each other...

Annelies — Wed, 15 Jan 2020 15:30:55 GMT

So I need just to play with it a little bit more to get used to it.

Yeah I noticed the techinal and mathematical terms in a preview.

@Mark_Bailey thanks for you recommendation. I will look into it aswell.

Now that people are still using this subject! A different question but with the same scenario.

I can open another discussion if that is more easier....

Same scenario but this time the ratio is the only constant value.

C1 liquid reacts with C2 liquid in a ratio 100:75

So the formula will be: 4*C1 liquid = 3 C2 liquid

As far as I understand from https://www.jmp.com/en_in/events/ondemand/mastering-jmp/mixture-designs.html

The constrain will become:

4*C1 liquid -3 C2 liquid </= 0

4*C1 liquid -3 C2 liquid >/= 0

The other constrain is just like last time, that all ingredients for component 1 is 50% and all the ingredients for component 2 is 50% of the total blend.

The picture is a little bit small now... sorry. (just to be sure it can also be found in the appendix.

For some weird reason I get now "Optimal designer failed to converge."

Have I filled in something wrong here?

Re: Mixture Design 14 factors or more... some needs to be equal to each other...

Dan_Obermiller — Wed, 15 Jan 2020 15:48:31 GMT

When you start constraining your components this much and JMP fails to find a design, my first question is always: do you even have a feasible experimental region left? Look at your component constraints along with these additional constraints. Is there a viable region to experiment over?

Re: Mixture Design 14 factors or more... some needs to be equal to each other...

Annelies — Thu, 16 Jan 2020 09:02:08 GMT

I understand, I had something in my mind and I probably need to rewrite the components values.

I'm sure it is possible, I just need to check my factors again.

I have simplify it and then it was possible. See picture down here.

Due to the ratio there is not really a plane, but more like a line were the optimal can be reached.

That line is in an area of 0 by 0.5.

C1 liquid and C2 liquid are reacting with a ratio of 100:75. In this example in a range of 0 bij 0.5.

(0 is not really what I want, because it needs to react, but I wanted to see if it was possible.)

The optimal will be checked with every Filler ingredients. So GGREY006A... etc.

So how I see it, JMP can check a line area.

And yes I noticed my ratio fault here. :)

It's now C1 liquid : C2 liquid

75 : 100

And I wanted C1 liquid: C2 liquid

100 : 75

I need to swap the 4 and the 3 around.

I'm going to play a little bit more with JMP.

Re: Mixture Design 14 factors or more... some needs to be equal to each other...

Annelies — Thu, 16 Jan 2020 10:18:08 GMT

Made it! :) Sorry for the spam... Thanks for all the help and replies!

Re: Mixture Design 14 factors or more... some needs to be equal to each other...

statman — Fri, 17 Jan 2020 17:18:54 GMT

Yes, John A. Cornell is recognized as one THE experts on mixture designs. I recommend his papers " Embedding Mixture Designs inside Factorial Experiments", "Mixture Experiment Approaches: Examples, Discussion, and Recommendations". Also Ron Snee has some good papers "Screening Concepts and Designs for Experiments with Mixtures" and Design and Analysis of Mixture Experiments".

Enjoy!

Re: Mixture Design 14 factors or more... some needs to be equal to each other...

statman — Fri, 17 Jan 2020 17:21:20 GMT

Hey Dan, I still find that paper very useful for folks just getting started with thinking about mixtures! Well written.

Re: Mixture Design 14 factors or more... some needs to be equal to each other...

statman — Sat, 18 Jan 2020 01:30:21 GMT

Generally doesn't mean your n = 1 example...LOL. Really most papers and texts you read describe mixture designs as having few variables. But there are many approaches to getting things done. One of my favorite quotes is. “Two equally competent investigators presented with the same problem would typically begin from different starting points, proceed by different routes, and yet could reach the same answer. What is sought is not uniformity but convergence." (Box, Hunter, Hunter).
Cheers.

Re: Mixture Design 14 factors or more... some needs to be equal to each other...

loganshawn — Fri, 14 Oct 2022 20:08:51 GMT

Hi, statman, if I have a formulation which contain 100 components, can I use mixture design?

Re: Mixture Design 14 factors or more... some needs to be equal to each other...

loganshawn — Fri, 14 Oct 2022 20:15:58 GMT

Thanks, Mark, I was so discouraged by reading John Cornell's book. It is full of math formula and I gave it up after 40 pages. My work is related to formulation which contain 100 ingredients. Is it possible to have a design with 100 factors?