Abstract:

The mathematical treatment of the gas-liquid interface has presented a formidable challenge since the inception of multiphase flow research. In the 1970s, the field began with the homogeneous flow model, which simplified the system by treating the two phases as a single mixture. This was followed by the separated flow model, which incorporated slip ratios to account for relative velocity; however, this approach proved numerically unstable as liquid velocity approached zero, and negligence of local physics.

To address these limitations, Zuber introduced the drift-flux model, which considers the relative velocity between phases in relation to their spatial distribution. While this allowed for the development of constitutive relations based on observed flow regimes—providing macroscopic data with reasonable accuracy for system analysis codes—the explicit mathematical treatment of interfacial geometry remained absent.A significant shift occurred with the introduction of the two-fluid model, which utilized the macroscopic jump condition derived from the surface Leibniz rule. Despite this advancement, the model still requires constitutive relations for interfacial area concentration (IAC) to achieve closure. Since the 1990s, researchers have pursued two primary paths: developing flow-regime-dependent IAC correlations and seeking methods for the dynamic treatment of interfacial area. While diagnostic instrumentation has advanced significantly to measure these concentrations, fundamental challenges persist.

As we enter the age of Artificial Intelligence, integrating machine learning offers a promising frontier for overcoming these long-standing hurdles. This lecture provides a historical overview of two-phase flow modeling, examines ongoing technical challenges, and discusses the potential of AI to deepen our understanding of complex two-phase flow structures.

About the speaker:

Shuichiro Miwa is an Associate Professor at the University of Tokyo. His research encompasses both fundamental and applied aspects of gas-liquid two-phase flows, with a focus on interfacial transport phenomena, thermal hydraulics, and the development of advanced modeling frameworks for nuclear reactor systems. His work addresses persistent challenges in the mathematical treatment of the gas-liquid interface—from early drift-flux and two-fluid models to modern closure relations for interfacial area concentration. More recently, he has integrated machine learning techniques to better characterize complex two-phase flow structures and overcome limitations inherent in conventional constitutive models.