DOE for HPLC column

Question

Hi All,

I am trying to make a model for an HPLC column (A-protein) to find out if the process can run faster/decreasing process time.

Parameters: Flow Eluent Inlet, pH of the eluent 3-4, temperature 3c-20c, cycle 4-10 cycles, purity 85-99.9% and pressure. As mentioned a model, the parameters can be fixed when working with a specific column.

I am in progress learning DOE, I have practiced some of the exercises but not at the level, where I can make a model.

If anyone can help, will be much appreciated. Thanks in advance.

David

P_Bartell · Answer

To add to @Mark_Bailey 's counsel, I suggest two things:

1. Consider, if you haven't done so already, taking the JMP Statistical Thinking for Industrial Problem solving online, free course found here:

https://www.jmp.com/en_us/online-statistics-course.html

The course is great both teaching you both the 'process' of statistical thinking AND 'how to in JMP'. Depending on your knowledge of statistical methods you may be able to just pick and choose which module. But if your knowledge level is immature or maybe just needs some brushing up...I suggest all of them. In the order they are listed on the home page for the course.

2. Find somebody with a sound knowledge of DOE at your disposal to ask questions and provide feedback on your work as it progresses...there are many helpful people here in the JMP User Community...but nothing beats face to face for a DOE newb.

buddrball · Answer

Hello! We're thinking of trying something similar. @david72st -- Did you try this? How did it work out for you.

Thanks!

Mark_Bailey · Answer

A few thoughts...

You can record and use several outcomes ('responses') in the same experiment. For example, you might measure peak height, peak separation, and so on.

JMP calls your HPLC parameters 'factors.' All of yours may be included in a Custom Design. I strongly suggest entering them as continuous factors, even if you are thinking about only a few levels. Also, considering using the widest possible range for each factor to elicit the larges possible effect. This practice dramatically increases the power of your statistical tests and gives you the most stable, precise estimates of the model parameters..

Define a model that includes terms for all of the possible effects: first-order or 'main' effects, interaction effects, and non-linear effects ('powers').

The Prediction Profiler will allow you to evaluate parameter settings once you select your model and fit the data.