DoE for Optimization of Chromatography Conditions

ClusterFerret68

Hi All,

 

I'm hoping for some guidance/support in setting up a DoE to optimize a few parameters in an HPLC separation.

In total I have 4 factors, all are continuous (three are chemical compositions of the mobile phase, 1 is column temperature).

I am anticipating cross effects between some of the conditions that will impact the responses in my study.

 

What is the best approach to DoE here?  Should I be running a definitive screening design or a custom design with higher-order interactions (4th)?  When I set up the custom design I'm not seeing any center point conditions for any of the factors at all.  Is this a problem?  Will I still see quadratic effects?  When I set up the definitive screening design I do see some 2-way cross effects indicated...but I'm concerned that I'll be missing other potential interactions.

 

Any tips or advice would be greatly appreciated....thanks in advance!

Chris

P_Bartell

When optimization is the goal, Definitive Screening designs aren't the first design I'd pick. Reason being, they are screening designs first and foremost with some unique properties, but screening designs nonetheless. So where are you in your depth, breadth, and maturity of the process/problem at hand? Are you screening factors or are past that? In other words, you know the key factors, levels of those factors which will surround the factor space optimal position, have a good handle on noise variables (controlled and uncontrolled), and a good idea of the model you'd like to fit?

 

If you are truly screening factors, a DSD isn't a bad way to go. It will give some insight for future, sequential experimentation. But if you are past screening, then a model driven optimal design approach is probably the way to go. If you want quadratic effects in the model, specify them and the coordinate exchange algorithm will provide the proper levels for the factors to estimate those effects as well as any interaction effects you seek.