Complex multiphase gas-liquid flows, including boiling, are usually encountered in safety related nuclear applications. For CFD purposes, modeling the transition from low to high void fraction regimes represents a non-trivial challenge due to the increasing complexity of its interface. For example, churn-turbulent and slug flows, which are typically encountered for these gas volume fraction ranges, are dominated by highly deformable bubbles. Multiphase CFD has been so far relying on an averaged Euler-Euler simulation approach to model a wide range of two-phase applications. While this methodology has shown to date demonstrated reasonable results (Montoya et al., 2013), it is evidently highly dependable on the accuracy and validity of the mechanistic models for interfacial forces, which are necessary to recover information lost during the averaging process. Unfortunately existing closures, which have been derived from experimental as well as DNS data, are hardly applicable to high void fraction highly-deformable gas structures. An alternative approach for representing the physics behind the high void fraction phenomena, is to consider a multi-scale method. Based on the structure of the gas-liquid interfaces, different gaseous morphologies should be described by different CFD approaches, such as interface tracking methods for larger than the grid size interfacial-scales, or the averaged Euler-Euler approach for smaller than grid size scales, such as bubbly or droplet flow. A novel concept for considering flow regimes where both, dispersed and continuous interfacial structures, could occur has been developed in the past (Hänsch et al., 2012), and has been further advanced and validated for pipe flows under high void fraction regimes (Montoya et al., 2014) and other relevant cases, such as the dam-break with an obstacle (Hänsch et al., 2013). Still, various short-comings have been shown in this approach associated mostly to the descriptive models utilized to obtain the continuous gas morphology from within the averaged Eulerian simulations. This paper presents improvements on both concepts as well as direct comparison between the two approaches, based on newly obtained experimental data. Both models are based on the bubble populations balance approach known as the inhomogeneous MUltiple SIze Group or MUSIG (Krepper et al., 2008) in order to define an adequate number of bubble size groups with its own velocity fields. The numerical calculations have been performed with the commercially available ANSYS CFX 14.5 software, and the results have been validated using experimental data from the MT-Loop and TOPFLOW facilities from the Helmholtz-Zentrum Dresden-Rossendorf in Germany (Prasser et al., 2007).

This content is only available via PDF.
You do not currently have access to this content.