MPI 병렬기법을 이용한 상선주위의 난류유동장 해석

Abstract

The finite volume based multi-block RANS code, WAVIS[1-2] developed at MOERI is converted into MPI-parallelized code with a linux PC-clusterimg machine. The linux PC-clustering machine has 48 node with a Intel Pentium IV 3.2 GHz processor for each node. About a parallel method, a structured grid block is decomposed into several multi blocks and 1 cpu for each block is allocated[3]. Examining the parallel performance, the speedup test is performed with a simple ship hull, series60. The fully non-linear free-surface flow around a series60 hull is simulated and consistence/convergence of parallelized solution is investigated. The parallelized WAVIS code is applied to a self-propelled LNG carrier ship. WAVIS uses the cell-centered finite volume method for discretization of the governing equations. Convection terms are discretized using the third order QUICK scheme and the central difference scheme is utilized for diffusion terms. To ensure divergence-free velocity field, the SIMPLEC method is employed. The realizable k-ε turbulence model with a wall function is employed for the turbulence closure. The free surface is captured with the two-phase level set method[4] and body forces are used to model the effects of a propulsor without resolving the detail blade flow. The propeller forces are obtained using an unsteady lifting surface method based on potential flow theory[5]. The numerical procedure followed the self-propulsion model experiment based on the 1978 ITTC (International Towing Tank Committee) performance prediction method. The self-propulsion point is obtained iteratively through balancing the propeller thrust, the ship hull resistance and towing force that is correction for Reynolds number difference between the model and full scale. The unsteady lifting surface code is also iterated until the propeller induced velocity is converged in order to obtain the propeller force. The self-propulsion characteristics such as thrust deduction, wake fraction, propelle