19-12-2024 - Lecture Hall E

PostDoc presentation

Challenges in High-Fidelity Simulations of Two-Phase Flows Undergoing Phase Change

Jordi Poblador Ibanez

Multiphase flows undergoing phase change are present in many engineering applications. In particular, they play an important role in the energy transition landscape. Some examples include the injection of biofuels, the boiling of coolants in heat exchangers, or the formation of bubbles in water electrolysis. A common approach to study numerically these systems is based on the one-fluid formulation of the Navier-Stokes equations, which treats the phase discontinuity as a single fluid with varying properties and adds additional source terms to impose interfacial jump conditions, e.g., surface tension. Many legacy codes used to study non-evaporative conditions have been extended to phase-change simulations. However, the numerical discretization of the one-fluid governing equations under such conditions introduces errors that must be corrected to ensure the physical consistency of the results..

19-12-2024 - Lecture Hall E

PostDoc presentation

CaNS-Fizzy. A GPU-accelerated solver for two-phase turbulent flows

Giandomenico Lupo

Direct Numerical Simulations of two-phase flows with high levels of interface dispersion, such as bubble or droplet-laden turbulent flows, remain computationally expensive due to the high-resolution requirement imposed throughout the whole computational domain by the direct solution of the interface morphology and of the transport at the interface scale. The finite different solver CaNS-Fizzy has been developed to enable these simulations by exploiting the computational capacity of modern GPU clusters. The key features that keep the computational cost affordable are the one-fluid formulation of the two-phase flow, the use of a fast direct solver for the pressure Poisson equation, and the diffuse representation of the phase interface. The latter allows for continuous and smooth mapping of the physical fields and requires no explicit interface reconstruction, keeping the computational load constant regardless of the local interface topology and thus making the algorithm particularly suited for parallelization on GPU architecture.