Machine learning analysis for the rheological mechanism of polysaccharide colloids

Natural polysaccharide colloids such as konjac glucomannan (KGM) colloids have attracted a lot of attention due to their exceptional water absorption, water retention, and thickening properties. However, investigating the rheological mechanism of polysaccharide colloids remains challenging due to the complex dynamic behaviors of bio-macromolecular clusters in shear flow fields, which are difficult to observe experimentally, and the intricate coupling effects of various parameters on rheological performance. In this study, multi-scale simulations combining molecular dynamics simulation and Brownian dynamics simulations revealed the dynamical structures of KGM clusters and single molecule chains. The comparison between experimental and numerical results of viscosity and storage modulus validated the effectiveness of the multi-scale simulation in analyzing the dynamical behavior of large KGM clusters (∼500 nm) over a long time (∼2 ms). The relationships between numerical viscosity and parameters aligned with the experimental results. The total spring force of the bead-spring models of KGM clusters was an appropriate numerical indicator of the colloid storage modulus. The results of the exclusion of volume forces indicated that the interaction between KGM segments was an important origin of colloid viscosity. Based on the comprehensive dataset from experiments and simulations, integrated machine learning models were used to predict colloid viscosity and analyze the importance of factors. The results indicated that the extreme gradient boosting (XGBoost) model exhibited the best predictive performance with an R² value of 0.90. This study provides the rheological properties of KGM colloids by experiments, the dynamical behaviors of KGM clusters and individual molecular chains via multi-scale simulations, and integrated machine learning models for predicting colloidal viscosity and analyzing the main factors influencing KGM colloid viscosity. It contributes to a deeper understanding of the rheological mechanisms of polysaccharide colloids and enriches the research methods for polysaccharide sol systems.