viscosity curve as response in DOE

[#1]

Hi,

i am running an experiment based on DOE. one of the responses is viscosity cure that is related to formulation rheological behavior.

the plots are viscosity Vs shear rate. For different formulations, the plots appear to be a bit different. Is there any way to compare and treat it as response for DOE further analysis?

[#2]

In short, yes.  How might you quantify the differences in the curves?  You can look at slopes, max, min, and of course simple graphical plots.

[#3]

thanks @statman 

the curves are different - part of them with high viscosity in zero share rate and part with lower. and also the shape is a bit different for each run.

[#4]

JMP Pro has a platform called Functional Data Explorer that is designed for responses that are not a scalar values but instead are functions, profiles, or curves. It is an alternative approach to what @statman suggests through feature extraction.

FDE provides a feature called Functional DOE that treats each curve as a response to conditions defined by the runs in the DOE. You can fit a linear model and determine how factor settings affect the curve shape. You can optimize factor settings to achieve the desired shape. Actually, you can save the functional principal components and use them in analysis in the X or the Y role.

[#5]

Thank you @Mark_Bailey . How practically it can be done? for example, i two files

1. file with the DOE runs

2. viscosity file - file with one column including shear rate (x) and 10 columns (number of runs) with viscosity values (y). 

the target is as you described - fit a linear model and determine how factor settings affect the curve shape.

please advise

[#6 · Accepted solution]

Please read the chapter I previously provided. It explains how to set up FDE. First, though, you must join the two tables.

1. Stack the viscosity data so that all the responses are in a single column and the shear rate is duplicated for every row.
2. Add a new column to the data table with the viscosity that indicates the run number for each curve.
3. Join the two data tables matching on run number.
4. Save the new joined table and use it to start FDE.

[#7]

dear @Mark_Bailey thanks - i will try it 

[#8]

Check out this presentation:  What to Do When Your Data Is a Curve (2019-US-45MP-213)