DoE in formulationHi Team,
I have a customer working with formulation and they wish to test the effect of 12 different molecules in a DoE. They only wish to look at the combination of four molecules at one given time. So my thought was to add all the 12 molecules as categorical factors and then adding the constraint of only four.
Does anyone have experience in setting up a similar design and have some data on this?
Thanks in advance
Best Christian
Re: DoE in formulationAre the molecules entered as a mixture component and the amount varies or are they entered independently at the same level?
Are the combinations randomly selected or are combinations based on molecular properties?
How much budget do they have? That is, are they limited by the number of runs?
Re: DoE in formulationHi mark, thanks for the response
When a molecule is entered it is a categorical factor so either added/not added. My first question to them is whether this can be converted to a continuous factor. I am visiting them this coming Monday.
They measure the amount of bacteria left after adding different molecules. I have tried to answer your questions below:
Are the molecules entered as a mixture component and the amount varies or are they entered independently at the same level? They enter independently at the same level and they will add up to four molecules.
Are the combinations randomly selected or are combinations based on molecular properties?
Randomly selected.
How much budget do they have? That is, are they limited by the number of runs?
They wish to see what would be the minimum number of runs they can perform. But I believe that runs are relatively cheap
Re: DoE in formulationIf it were only two at a time, then you could use a covering array!
This study is an unusual screening experiment. They are probably not interested in modeling at this point, just finding the best combination to take forward into optimization. But they still need to test all possible combinations of four molecules.
I made a full factorial for 10 two-level categorical factors. That design is 1024 runs. Then I added a column formula to count the number of times "present" occurred in the Pattern. I selected the rows where the count = 4. There are 210 runs.
I attached the design data table.
Re: DoE in formulationI've solved this "given n choose m" type problem with cosmetic formulations using Balanced Incomplete Block Designs. You create the BIDB table for the treatment combinations (given 12 possible treatments choose 4) and then include that design as a covariate factor in a custom DOE (if they want to look at other process terms as well). If they are just doing screening on reagents then you might be able to the away with the BIDB by itself.
M
