Conference Agenda

The motivation of the study of fully-implicit scheme for two-fluid model is, comparing with semi-implicit scheme, the mainly advantage of fully-implicit method is the restriction of the time step on stability is very small, and large time step is allowed. However, fully-implicit scheme makes the degree of coupling among equations is strong, and the solution become more difficult. First, the calculation of Jacobian matrix is difficult and highly time- and memory- consuming due to matrix is large. The Broyden-type method is adopted because it only requires calculate Jacobian matrix one time during the iteration process. We use numerical difference to estimate Jacobian matrix for the first iteration, then calculate Jacobian matrix by Broyden method for the remaining iterations. Due to the sparsity of Jacobian matrix, the Schubert method is applied. This method makes full use of Jacobian matrix’s sparsity. Second, the convergence performance become poor especially for strong interfacial effect case. The reason is the non-linear interfacial models makes the Jacobian matrix highly ill-conditioned. The relationship between condition number and the degree of interfacial effect is studied, and we try to reduce condition number by modify governing equation. Finally, this solver is tested. The calculation performance under very larger time step is assessed.

High-resolution simulations of two-phase flows can augment existing experimental data used for thermal-hydraulic system code development. PHASTA is one such high resolution code that uses Direct Numerical Simulation of the Navier-Stokes equations with the Level Set method to resolve individual bubbles and droplets of the flow. PHASTA, when deployed on high performance computing systems, has shown remarkable performance at simulating turbulent bubbly flows which occur in prototypical reactor geometries and conditions. However, flows of high void fraction systems present a challenge to the Level Set method which is well known for its mass loss deficiencies. High void fraction annular flows are typical near the top of boiling water reactor fuel channels and steam separators. Annular flows generate a large amount of small, entrained droplets and bubbles which require a prohibitively fine mesh in order to resolve and conserve the mass of these smaller objects. New numerical methods or models must therefore be incorporated into PHASTA in order to accurately model these flows with economical grid sizes. The Conservative Level Set method is one such method which has shown superior mass conservation properties on a variety of complex two-phase flows. This work describes the development and testing of the Conservative Level Set method implemented in the PHASTA finite-element code. A simulation of a swirling annular flow in a BWR steam separator is conducted to test the method on a large engineering-scale problem.

An important industrial issue and a continuing challenge in computational fluid dynamics (CFD) is the simulation of gas-liquid flows when several two-phase regimes coexist.

To overcome the computational challenges associated with all-scale interface resolving approaches, all-flow-regime CFD models have been proposed. One such method is the Generalized Large Interface Model (GLIM), implemented in the NEPTUNE_CFD software. This framework allows a smooth modelling approach transition: the small-dispersed scales are modelled, whereas the large scale gas-liquid interfaces are explicitly treated. These methods can be useful for many nuclear thermal-hydraulic applications.

To achieve this goal, GLIM and other models in the literature require specific closure terms for configurations where the gas-liquid interface is recognized. In addition to the interfacial closure formulations, blending functions and interface recognition methods are essential to deal with the transition between scales.

The aim of the present work is to evaluate different formulations of the interfacial forces. The NEPTUNE_CFD code is used. A rising large bubble configuration where the interface can be well resolved or rather diffused over several mesh cells was considered as a preliminary validation test. Several large interface drag formulations and cell number over bubble diameter ratios have been tested, the results are discussed in this paper. Then, the model was applied to simulate an intermittent air-water cross-flow in a tube array, featuring at the same time dispersed bubbles/droplets and large gas-liquid interfaces, a configuration close to the one encountered in U-tube steam generators.

Pool scrubbing is an effective process to decontaminate radioactive aerosols as severe accidents happen in nuclear power plants. Bubble residence time is one of the key parameters to determine the aerosol decontamination factor (DF) which is defined to describe the efficiency of aerosol removal, especially in the swarm flow region which makes a significant contribution to the total aerosol removal. The Euler-Euler two-fluid method and the Lagrangian Particle Tracking (LPT) method are applied, the former is used to get the flow field information, and the LPT method is used to track the bubble movement to obtain the bubble residence time. Through the analysis of bubble residence time distribution, the model of probability density function for bubble residence time is preliminarily established. The probability density function obeys well an exponential decay behavior. The decay constant is fitted according to the CFD simulation results. In general, the developed model shows good potential in predicting bubble residence time. Furthermore, the effect of bubble diameter on the probability density function is investigated. The bubble diameter shows a strong effect on bubble residence time, which is because larger diameter bubbles experience greater buoyancy, resulting in a higher rising velocity.