Optimizing material selection using design of experiments and multivariate analysis < QSAR Optimizing Material Selection Using Design of Experiments / Multivariate Analysis >
background
The process of selecting raw materials typically requires a lot of time , as it requires testing all options using attribute data . A major drawback is that the resulting model , which tests all attributes , has difficulty predicting options with different physical and chemical properties . However , models built using the Quantitative Structure-Activity Relationship (QSAR) approach can be implemented with a smaller number of candidate material combinations and can be utilized more sustainably .
Problem statement
- How many options should be tested ?
- Do you have the manpower and time to conduct the necessary experiments ?
- What should we do when new materials come in ?
way out
- Use physical , chemical , and / or biological properties as variables instead of attributes .
- Use principal components for dimensionality reduction , not the traditional way .
- Design a covariate experiment using principal components to select a subset of options to test .
- Build predictive models using Partial Least Squares (PLS) or generalized regression models . This will allow you to obtain robust models of physical , chemical , and / or biological properties that do not change over time .
boon
- Selection of candidate substances using main components to eliminate guesswork
- Reducing the combination of substances to be tested
- Sustainable model can be utilized when additional materials need to be considered in the future.
(영상) DOE and Multivariate Modeling Techniques for Material Selection ( 2019-EU-45MP-092 )
This is material from the Explorers Guide seminar held at the Trade Tower on the afternoon of April 10, 2019.
The content presented in this seminar is summarized in PDF format, and the JMP Data Table used in the demo is also shared. 
