A Machine Learning-Based Model of Non-Newtonian Hydrodynamics with Molecular Fidelity

Watch a one-hour seminar presentation by Professor Huan Lei from Michigan State University exploring an innovative machine-learning approach for modeling non-Newtonian fluid dynamics. Discover how to construct continuum models while maintaining molecular fidelity through micro-macro correspondence, utilizing encoders for micro-scale polymer configurations and macro-scale nonlinear conformation tensors. Learn about the development of the deep non-Newtonian model (DeePN2), which introduces a new form of objective tensor derivative while maintaining conventional non-Newtonian fluid dynamics model structure. Examine numerical results demonstrating the model's accuracy and effectiveness in bridging microscopic and macroscopic descriptions of non-Newtonian fluids. Part of the Statistical Physics in Living Matter series focusing on non-equilibrium states under adaptive control at the Isaac Newton Institute.