(490c) Learning Hidden Rheology Using Rheology-Informed Graph Neural Networks (RhIGNets)

Accurate and reliable property prediction of structured fluids is of great interest across many disciplines, especially in designing new soft materials and processes involving them. The challenge is to solve non-trivial time and rate dependent constitutive equations to describe these complex fluids under various flow protocols. We present Rheology-Informed Graph Neural Networks (RhIGNets) for solving systems of Ordinary Differential Equations (ODEs) adopted for complex fluids. The proposed RhIGNets are employed to learn the hidden rheology of complex constitutive models with multiple coupled ODEs by benefiting from graph mode of neural networks. From a practical point of view, an exhaustive list of experimental tests is often required to identify model parameters for a multi-variant thixotropic or viscoelastic constitutive model. RhIGNets are found to learn these non-trivial model parameters for a complex material using only a limited number of data points from a single flow protocol, enabling accurate modeling with limited number of experiments and at an unprecedented rate. We also show that the RhIGNets are not limited to a specific model and can be extended to include various constitutive relations of choice and recover complex manifestations of thixotropic-elasto-visco-plastic fluids.