AI applications in diagnosing and modeling transport phenomena of complex fluid

Name Dr. Howon Jin
Affiliation Chemical Materials Solutions Center, KRICT

Abstract

Complex fluids, such as polymers and colloidal suspensions, are widely used across various industrial fields. These fluids exhibit unique characteristics in terms of flow behavior or mass transport, which differ significantly from those of simple fluids. The complexity arises from interactions between their components or the formation of microstructures within the fluid, making it essential to control or predict these behaviors to enhance the productivity and quality of chemical products.

Traditionally, research in this area has focused on modeling the physical properties of individual components. However, with recent advancements in artificial intelligence, machine learning technologies have also been applied to study the transport phenomena in complex fluids. This presentation will highlight several case studies that utilize AI to diagnose the state of complex fluids and model their flow characteristics.

First, we will explore research that analyzes various signals observed during the flow of complex fluids to diagnose their state. One example addresses clogging, a problem that arises when particulate suspensions cause particle deposition in microchannels. A method for predicting future clogging by analyzing particle deposition images during flow will be introduced. Additionally, research on diagnosing the state of battery slurry by analyzing pressure fluctuations during pipe flow will be presented. Chaotic fluctuation in these pressure signals, caused by microstructures within the suspension, are interpreted using artificial neural networks to extract information about the slurry’s dispersion state.

Beyond AI-driven classification in transport phenomena, data-driven approaches can also uncover underlying physical laws governing complex fluids. The final case study will focus on constructing constitutive equations that describe the flow behavior of complex fluids using machine learning. This research demonstrates a method that not only learns from data on deformation-stress relationships but also ensures the resulting equations satisfy physical constraints such as frame indifference. Through these examples, we will demonstrate how AI-driven pattern recognition can accelerate new discoveries in the transport phenomena of complex fluids.