Machine-learning based non-Newtonian hydrodynamic model with molecular fidelity

Publisher:闻天明Release Time:2022-04-19Number of visits:10

Speaker:    Huan Lei, Michigan State University

Time:         9:00-10:00 , Apr.22

Location:   Tencent Meeting

    Meeting ID600-920-452


Host:          Qifeng Liao&Shixiao Jiang



A long standing problem in the modeling of non-Newtonian hydrodynamics of polymeric flows is the availability of reliable and interpretable hydrodynamic models that faithfully encode the underlying micro-scale polymer dynamics. The main complication arises from the long polymer relaxation time, the complex molecular structure and heterogeneous interaction. We developed a deep learning-based non-Newtonian hydrodynamic model, DeePN$^2$, that enables us to systematically pass the micro-scale structural mechanics information to the macro-scale hydrodynamics for polymer suspensions. The model retains a multi-scaled nature with clear physical interpretation, and strictly preserves the frame indifference constraints. The construction is end-to-end, strictly preserves physical symmetries, and retains the physics interpretation. Unlike the empirical closures based on the dumbbell structure, DeePN$^2$ can faithfully capture the broadly overlooked viscoelastic differences arising from the specific molecular structural mechanics without human intervention. Different from the common machine-learning-based approaches, the learning of constitutive dynamics does not rely on the expensive time-series sampling from the micro-scale simulations. The present learning framework is general and applicable to other challenging multiscale dynamical systems, where direct learning using the micro-scale time-series samples may have limitations.



Dr. Huan Lei received his Ph.D. in applied mathematics from Brown University. Prior to joining the MSU faculty in 2019, he worked as a research scientist and postdoctoral research associate at the Pacific Northwest National Laboratory. His research work mainly focuses on multiscale modeling and stochastic simuations, with applications to multiscale fluid and soft matter systems, where empirical models based on conventional physical laws show limitations. The main research goal is to develop numerical algorithms to learn reliable computational models of mesoscale transport, assembly, and transition processes directly from the micro-scale first-principle-based descriptions. He is the recipient of the NSF CAREER Award.