Hi @ClusterFerret68,
Just to be sure, the three chemical compositions of the mobile phase are three independent continuous factors, not mixture (dependent) factors that sum up to a constant value ?
If all the factors are continuous and your goal is optimization, traditional "textbooks" DoEs commonly met for chromatography process optimization are Central Composite Design (CCD) and Box-Benhken. See more infos about it in one of my previous response : Solved: Re: Optimization task with 2 continuous and 1 discrete (2 levels) factors - JMP User Communi...
You can also use the Custom Design platform, specify your factors, and create a Response Surface Model by clicking on "RSM" in the "Model" panel : all 2-factors interactions and quadratic effects will automatically be added in the model, the design will be I-optimal (to minimize average prediction variance over the experimental space) for a recommended default number of runs equal to 21.
Here is the script to generate such design :
DOE(
Custom Design,
{Add Response( Maximize, "Y", ., ., . ),
Add Factor( Continuous, -1, 1, "X1", 0 ),
Add Factor( Continuous, -1, 1, "X2", 0 ),
Add Factor( Continuous, -1, 1, "X3", 0 ),
Add Factor( Continuous, -1, 1, "X4", 0 ), Set Random Seed( 1822801991 ),
Number of Starts( 8853 ), Add Term( {1, 0} ), Add Term( {1, 1} ),
Add Term( {2, 1} ), Add Term( {3, 1} ), Add Term( {4, 1} ), Add Term( {1, 2} ),
Add Term( {1, 1}, {2, 1} ), Add Term( {2, 2} ), Add Term( {1, 1}, {3, 1} ),
Add Term( {2, 1}, {3, 1} ), Add Term( {3, 2} ), Add Term( {1, 1}, {4, 1} ),
Add Term( {2, 1}, {4, 1} ), Add Term( {3, 1}, {4, 1} ), Add Term( {4, 2} ),
Set Sample Size( 21 ), Optimality Criterion( "Make I-Optimal Design" ),
Simulate Responses( 0 ), Save X Matrix( 0 ), Make Design}
)And in 3D the repartition of points look similar to a Central Composite Design (which would need 26 runs for 4 continuous factors) when you specify axial value of 1 (axial points "on face") :
I wouldn't add more terms or higher power in the model (except if domain expertise or previous experience give you strong hints about the usefulness of these terms in the model), and first check the results from this modeling on the measurements to see if the model is sufficient/reliable or if it needs an augmentation (and possibly include higher order terms).
I agree with previous answers about DSDs that this design is more adapted to a screening phase with large number of factors, not an optimization phase with a low number of factors.
DSDs are not recommended for less than 5+ factors, as you will need 13 runs (17 if you follow the default settings and recommendations from JMP with the 4 extra runs added) for up to 6 factors. So with a low number of factors, Custom Design would be less expensive in terms of experimental budget, with a greater flexibility to handle constraints and various factors types.
Numbers of runs needed for DSD with m factors are :
- 2m+1 for even numbers of factors
- 2m+3 for odd numbers of factors.
More infos on DSDs can be found in the Learning Center : Definitive Screening Design - JMP User Community
Similar presentation topics from the Community :
I hope this first answer will help you,
