Safa Jamali

Rheology-Informed Neural Networks (RhINNs)
for complex fluid modelling

Safa Jamali, Assistant Professor
Department of Mechanical and Industrial Engineering
Northeastern University

Abstract: Reliable and accurate prediction of complex fluids’ response under flow is of great interest across many disciplines, from biological systems to virtually all soft materials. The challenges include solving non-trivial time and rate dependent constitutive equations to describe these structured fluids under various flow protocols, and in correlating constituents of a complex material to its rheological behavior. On the other hand, advances in data-driven approaches to material design and discovery promise a leap in accelerated design cycles for new materials. I will present Rheology-Informed Neural Networks (RhINNs) as a general platform for prediction of rheological behavior in complex fluids. This includes a neural network architecture capable of solving Ordinary Differential Equations (ODEs) adopted for complex fluids, in forward and reverse problems, as well as a multi-fidelity approach in which scarcity of experimental data is compensated by readily-available model predictions to train the machine learning platform. The proposed RhINNs are employed to solve the constitutive models with multiple ODEs by benefiting from Automatic Differentiation in neural networks. We present direct and inverse solutions of a series of constitutive equations for different flow protocols by employing our RhINNs methodology. I will show that the RhINNs framework is capable of learning complex model parameters by training on a series of limited experimental data. We show that the model can be extended to various models by including different systems of ODEs, solved for arbitrary geometries, and recover complex kymographs of kinematic heterogeneities and transient shear banding of thixotropic fluids. Finally, I will show the results of our hybrid RhINNs platforms as the next generation of accurate constitutive models for complex fluids and as “digital rheometer twins” that can be used in place of a physical rheometer.

Bio: I am an Assistant Professor of Mechanical and Industrial Engineering at Northeastern University. I received my PhD from Case Western Reserve University’s Macromolecular Science department, followed by a period of postdoctoral training at MIT’s Chemical Engineering, Mechanical Engineering and Energy Initiative. I joined Northeastern University in 2017, and my research group is focused on a number of data-driven and computational methods for physics and rheology of particulate systems and complex materials. These areas include hemorheology and biophysics of cell suspensions, science-based data-driven methods and machine-learning platforms, network science of complex fluids, and physics of colloidal systems amongst other topics.