2024 AIChE Annual Meeting

(693g) A Human-Readable Polymer Representation Method for Rapid Analysis of Polymerization Modeling Results

Authors

Arturo, S. G. - Presenter, The Dow Chemical Company
Arora, A., University of Minnesota
Mwasame, P. M., University of Delaware
The generation of novel structure-property relationships for polymeric materials with modern machine learning and artificial intelligence methods demands a proper representation of the polymer. Determining the best way to represent polymers is an area of active research. Representation methods applicable to molecules are insufficient due to the composition- and size-distribution of polymer chains within a material. Averaged polymer structure is insufficient, as two materials with the same averaged structure may have different distributions set by polymerization process conditions. An exhaustive way to represent a polymer to obtain the best structure-property relationships would work but would not be human-readable. A human-readable method to represent polymers that still retains some level of detail for decision-making and for structure-property relationships is presented. Use of the method on the outputs of a kinetic Monte Carlo polymerization model shows the impact process conditions have on distributions of polymer chain compositions and sizes in a rapid manner.