How to insert factors that are linked with each other in a DOE

[#1]

Hi all,

I want to design an experiment on biomass pyrolysis in JMP 18. I have 6 target values (e.g., mass yields of the different pyrolysis products) and I want to include the biomass composition and the pyrolysis temperature as experimental factors.

For the temperature, I know how I have to include it (steady factor with lower and upper boundary).

The biomass composition should be divided into 4 different parameters, i.e., lignin, cellulose, hemicellulose and ash content. With that, I want to generate a model that can predict e.g., biochar yields based on the biomass composition and pyrolysis temperature. I will use four different biomasses, i.e. the contents of these different parameters are predefined and also linked with each other (i.e., Biomass 1 has a lignin content that is linked with its cellulose, hemicellulose and ash content). How should I include the biomass composition in the model as a factor? I thought about a 4-staged discrete numerical factor for each biomass parameter. Would that be valid? That means I would end up with 4 individual discrete numerical factors for the biomass composition alone. Is there a way to tell JMP that the individual stages of each of these factors are linked to each other (so that I don't end up with a higher number of experiments than actually needed)?

Thanks a lot for your help.

Best regards

Jannis

[#2]

..so it would look like that :

Each column for the individual contents of lignin, cellulose, hemicellulose and ash would stand for one biomass.

[#3 · Accepted solution]

Hi @jgrafmu2,

Welcome in the Community !

What you're describing for the composition of the biomass seems to be a mixture situation, as the 4 components add up to a specific fixed sum, and you're interested in the ratios of each components in this mixture (instead of the individual quantities).
So it seems a Custom design with 4 mixture factors (for biomass composition, expressed as ratios and not quantities) and 1 numeric continuous or discrete factor for temperature could be used in your experimental setup. You can find an example of this type of design here : https://www.jmp.com/support/help/en/18.1/#page/jmp/mixture-design-with-nonmixture-factors.shtml#

If you need more guidance in the design building, please provide more details about the objective, factors, model assumed, and experimental budget available.

I hope this first answer will guide you,

[#4]

Dear Victor,

thanks a lot! That is exactly what I was looking for.

Best regards

Jannis