Marie Sourroubille, Isaac Y. Miranda-Valdez, Tero Mäkinen, Juha Koivisto, Mikko J. Alava
Abstract: The rheological characterization of thermoresponsive polymers typically demands extensive experimental observations across multiple parameters. The present work demonstrates how Gaussian Process Regression (GPR) can expedite this process by efficiently characterizing the thermogelation of methylcellulose fluids. By employing GPR as a surrogate model for Active Learning, we capture the effects of multiple parameters on the gelation dynamics of methylcellulose, requiring fewer experimental observations than traditional factorial design experiments. Additionally, we leverage the common rheological practice of frequency sweep analysis at step-increasing temperatures using GPR. Our work shows how GPR models can be trained to predict rheological behaviors at both short and long timescales, and even to predict rheological material functions. These findings suggest that GPR is a powerful tool for enhancing the efficiency and depth of rheological characterization, making it highly valuable for both research and industrial applications.
