The simultaneous flow of a mixture with two or more phases occurs in several industrial and energy conversion processes. The reliable prediction of multi-phase flows is therefore of particular interest for a wide range of applications. Multiphase fl ows can be identi ed as dispersed (e.g. bubbles, droplets, parti-cles...), separated (e.g. annular fl ow, free-surface flow...) or a combination of the two (e.g. droplet annular flow...). They can also be classi ed by the state of matters of the phases:
- Gas-Liquid flows
- Gas-Solid flows
- Liquid-Solid flows
- Immiscible-liquid flows
- Three-phase flows
The numerical approaches used to model dispersed multiphase flows are often identi ed by the formulation applied to the different phases. The dispersed phase is defi ned as the phase composed of bubbles, droplets or particles, and the continuous phase is defi ned as the fl uid in which these dispersed element are generally immersed. In Computational Multi-Fluid Dynamics, the continuous phase is usually modeled using a fixed Eulerian frame and the following formulations are often applied to the dispersed phase:
- Euler-Lagrange, the dispersed phase is tracked on a Lagrangian frame (Dispersed-Phase-Model...)
- Euler-Euler, the dispersed phase is tracked on an Eulerian frame (Drift Flux model, Eulerian Two or Multi-Fluid model...)
- Euler One-Fluid, the dispersed phase is tracked using the same Eulerian frame as the continuous phase (Volume-Of-Fluid technique, Level-Set technique...)
The various approaches are then often classi ed in two groups, depending on the averaging procedure of the governing equations which results in different computational costs per dispersed element:
- Interface tracking (e.g. VOF, LS) where the interaction between the phases is modeled at the spatial scale of the dispersed phase (high CPU per dispersed element)
- Non-interface tracking (e.g. DPM, DF, Eulerian T- M-F) where a macro-scopic model of the interaction between the phases is employed (low CPU per dispersed element)
In general, the first group of models offers the possibility to obtained detailed solutions of some dispersed elements without experimental calibration of the numerical model. However, most of the techniques developed and available nowadays are still subject to numerical instabilities, especially when dealing with complex phenomena. Related with high computational cost per dispersed elements this approach is mainly used in applications where the physics near the interface between the phases is particulary relevant. The use in industrial applications is often limited to speci c local phenomena or cases of separated flow, e.g. channel fl ow, free-surface, where the interface scale is comparable with the problem scale. The second group of models usually provides macro-scopic descriptions of the multiphase fl ow, allowing the study of small to large scale engineering systems. The main drawback of these models is the necessity of experimental calibration which is further complicated by the coupling between various flow phenomena (turbulence, dispersed-continous phase interaction, dispersed-dispersed phase interaction...).
Although the advances made in the numerical modeling of multiphase flows have permitted to consider Computational Multi-Fluid Dynamics as a viable tool, the research field is still very active in order to improve the accuracy and reliability of multiphase modeling techniques as well as to increase their range of application.
Modeling of two-phase heat and mass transfer represents a challenging issue in Computational Fluid Dynamics. The main approaches to model multiphase flows are using either statistical or direct methods to estimate the interaction and exchange between the phases. The general objective of the current activities of the CFDLab is to develop quali ed numerical approaches, supported by proper reference data, to increase the range of application, the accuracy and the reliability of multiphase modeling techniques.