Data-Driven Approaches for Predicting Polymer Solution Phase Behavior: Integrating High-Throughput Experimentation and Machine Learning
Author(s)
Amrihesari, Mona
Advisor(s)
Editor(s)
Collections
Supplementary to:
Permanent Link
Abstract
Advancements in artificial intelligence (AI) and machine learning (ML) have enabled data-driven discovery across scientific disciplines, yet their application in polymer science remains limited by the scarcity and inconsistency of high-quality experimental data. This thesis addresses this gap by developing a standardized, high-throughput method for measuring polymer solubility using a parallel crystallizer. The method captures key experimental variables—such as temperature, concentration, and mixing conditions—to generate a comprehensive turbidity dataset comprising over 1,000 polymer–solvent combinations. ML models trained on this dataset predict transmission percentage as a function of concentration and temperature, offering improved accuracy and granularity over traditional binary classifications. Additionally, a data-driven extraction model is introduced to quantify precipitation kinetics by deriving experimentally defined parameters from raw turbidity profiles. Together, these contributions advance polymer informatics by integrating structured experimentation with predictive modeling, offering a scalable framework for solubility prediction and a novel data-centric approach to studying precipitation kinetics.
Sponsor
Date
2025-04-15
Extent
Resource Type
Text
Resource Subtype
Dissertation