Abstract.A fixed-mesh approach has been developed to facilitate predicting a certain class of gas-liquid-rigid interaction problems. All the basic equations are discretized on a Cartesian grid in a finite-difference manner. The Volume-Of-Fluid and Boundary Data Immersion methods are employed to treat the gas-liquid and fluid-rigid interfaces, respectively. A hybrid OpenMP-MPI approach is adopted for parallel computing. The developed code is validated through comparisons with experiments of an oil-air flow driven by a rotating disk with holes. The simulated results demonstrate the capability in capturing the velocity distributions.
IntroductionGas-liquid flows driven by rotating rigid objects appear in many engineering applications. The interplay among centrifugal, buoyancy, viscous and surface tension forces gives rise to a rich behaviour involving the gas-liquid and fluid-rigid interface motions. Large-scale simulations of multiphase flows are expected to be useful for complementing experiments and gaining insight into the dynamics. It is desirable that the numerical method is efficient and robust for massively parallel computing.Direct numerical simulation of interfacial flows is often formidable mainly due to the spatiotemporal change in moving and deforming interface and the high degree of freedom. There are currently several approaches classified with respect to the numerical treatment how the kinematic and dynamic interactions are coupled on the interface. When addressing fluid engineering problems of practical applications with complex boundary, one has preferably employed a Lagrangian method using a finite element mesh. The numerical methods include Arbitrary Lagrangian Eulerian [1] and Deforming-Spatial-Domain/Stabilized Space-Time [2] approaches. If the body-fitted mesh is provided, the state-of-the-art Lagrangian approaches are satisfactory for achieving accurate predictions. However, for a system involving complex boundary, it requires considerable efforts to generate the high quality mesh and to reconstruct the mesh topology when the mesh element is added or deleted. Moreover, in parallel computing, one often encounters nontrivial issues how to keep the computational-load balance of each compute node in the domain decomposition, how to optimize the data communication between the Eulerian and Lagrangian frames, and so on. In the present study, we