Paper Number
DR7                         My Program 

Session
Data-Driven Rheology

Title
Graph neural network in prediction of force chain networks in dense suspensions. Part 1: scalability and methods

Presentation Date and Time
October 16, 2024 (Wednesday) 1:30

Track / Room
Track 6 / Room 501

Authors (Click on name to view author profile)

1. Aminimajd, Armin (Case Western Reserve University, Macromolecular Science and Engineering)
2. Maia, Joao (Case Western Reserve University, Macromolecular Science and Engineering)
3. Singh, Abhinendra (Case Western Reserve University, Macromolecular Science and Engineering)

Author and Affiliation Lines (in printed abstract book)
Armin Aminimajd, Joao Maia and Abhinendra Singh
Macromolecular Science and Engineering, Case Western Reserve University, Cleveland, OH 44106

Speaker / Presenter 
Aminimajd, Armin

Keywords
computational methods; data-driven rheology; dense systems; networks; particles; suspensions

Text of Abstract
The mechanism of shear thickening in dense suspensions has been recently linked to a transition from an unconstrained lubrication-dominated rheology to a constrained pairwise relative motion rheology state. In this work, we consider the constraints to originate from frictional contacts from static friction between particles. The particle simulation scheme considered here, lubrication flow discrete element modeling (LF-DEM), has been successful in quantitatively reproducing the non-Newtonian shear rheology of dense suspensions. The frictional forces stabilize the buckling of the force chain network under external deformation, leading to enhanced viscosity. In this work, we ask if we can predict the occurrence and structure of the force chain network (FCN). While the traditional particulate simulation methods are expensive and time-consuming, recent deep learning techniques have emerged as a powerful tool to simulate and predict properties of soft matter systems. Here, we use the graph neural network (GNN) to represent particles and their interactions as nodes and edges. Herein, we train the GNN model using datasets from our simulation data to predict FCN in suspensions at different stresses, packing fractions, system sizes, and bidispersities. Our machine learning model is accurate and interpolates and extrapolates to conditions far from its control parameters. The method used in this study can be used to predict the future rheological and characterization of complex particulate systems.