cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 
Try the Materials Informatics Toolkit, which is designed to easily handle SMILES data. This and other helpful add-ins are available in the JMP® Marketplace
Choose Language Hide Translation Bar

How do i design my experiments for optimization with CCD when i have lot of data?

Hi guys,

 

I am new to JMP and I am doing my Masters thesis on membrane filtration. I am trying to optimize and design a process. I screened several membranes for microfiltration (MF), ultrafiltration (UF) and nanofiltration (NF) and have lots of preliminary data. I decided to go for RSM-CCD as I saw in lots of research papers using it but i dont know how to approach it in JMP pro. I am trying to combine MF,UF and NF together for designing process. Also I tested membranes with different pore sizes (total 9 membranes) and did my experiments at different pressures MF (1, 2, 2.5, 3), UF (1, 2, 3, 4, 5) and NF (5, 10 and 15). I am trying to put factors-pressure, cross flow velocity and membrane with different pore sizes and response-flux, protein conc., RNA conc., and Volume reduction ratio. Should I go with RSM-CCD or try any another approach?

1 ACCEPTED SOLUTION

Accepted Solutions
MathStatChem
Level VI

Re: How do i design my experiments for optimization with CCD when i have lot of data?

Sounds like an interesting project.  Since this sounds like it is your first time designing and using DOE, my recommendation is to keep things simple by designing on each type of membrane separately.  The reason for this is that the membrane pore size range is dependent on the type of membrane, and designing and then analyzing experiments with those constraints can get complicated.  

 

Also, how much flexibility do you have to do interactive experimenation (that is, do some experiments, then adapt/re-design, and do more)?  If you have that flexibility, then I recommend a sequential approach:

  1. Screening:  initial screening of factors to determine which ones have the most impact, and to identify possible interactions.  You have already done some screening work, with the preliminary data you have, but you might benefit from doing more.  It looks like you have these factors:  membrane type, pore size, pressure, cross-flow velocity, response-flux, protein concentration, RNA concentration, volume reduction ratio.  
    Here is how you might setup the Custom DOE platform to start generating a design (in this case for the MF type filter).  Note that the Pore Size was added as a discrete numeric since it can only take on specific values.  
    MathStatChem_0-1738174739484.png

    If you carry this design generation through with the default settings, it will generate a 16 run design that allows for screening of the main effects and interactions between the factors.

  2. Optimization:  a more detailed design on the subset of factors that were identified in the screening experiments.  For these designs you typically use CCDs, but Optimal designs might also be useful.  You can "augment" a design (the screening design) to add in additional experiments that can let you estimate the curvature effects (the quadratic effects) if they are important, and further refine the estimation of the factor interactions.

View solution in original post

2 REPLIES 2
MathStatChem
Level VI

Re: How do i design my experiments for optimization with CCD when i have lot of data?

Sounds like an interesting project.  Since this sounds like it is your first time designing and using DOE, my recommendation is to keep things simple by designing on each type of membrane separately.  The reason for this is that the membrane pore size range is dependent on the type of membrane, and designing and then analyzing experiments with those constraints can get complicated.  

 

Also, how much flexibility do you have to do interactive experimenation (that is, do some experiments, then adapt/re-design, and do more)?  If you have that flexibility, then I recommend a sequential approach:

  1. Screening:  initial screening of factors to determine which ones have the most impact, and to identify possible interactions.  You have already done some screening work, with the preliminary data you have, but you might benefit from doing more.  It looks like you have these factors:  membrane type, pore size, pressure, cross-flow velocity, response-flux, protein concentration, RNA concentration, volume reduction ratio.  
    Here is how you might setup the Custom DOE platform to start generating a design (in this case for the MF type filter).  Note that the Pore Size was added as a discrete numeric since it can only take on specific values.  
    MathStatChem_0-1738174739484.png

    If you carry this design generation through with the default settings, it will generate a 16 run design that allows for screening of the main effects and interactions between the factors.

  2. Optimization:  a more detailed design on the subset of factors that were identified in the screening experiments.  For these designs you typically use CCDs, but Optimal designs might also be useful.  You can "augment" a design (the screening design) to add in additional experiments that can let you estimate the curvature effects (the quadratic effects) if they are important, and further refine the estimation of the factor interactions.

Re: How do i design my experiments for optimization with CCD when i have lot of data?

Thank you so much for the detailed explanation.