Improvements And Applications Of A Hydrodynamic Model Based On MPS

The numerical simulation methods on hydrodynamics are reviewed in detail. They are classified into two categories--grid method and gridless method, corresponding to the Lagarangian and Eulerian approach, respectively, and the representative methods of each category are introduced. MPS method is a new approach of great advantage in dealing with problems of complex boundary shape and situations where violent deformation of free surface occurs. However,it is more or less deficient in calculation speed due to the large amount of operations. Thus the efficiency of MPS method must be improved for practical application.The concept of candidate particle list is introduced into MPS method to generate neighboring particle list. It is indicated by a numerical test that the time consumption is remarkably reduced. The relationship of the number of time step for updating the candidate list and the average time consumption for generating neighboring list is analysed theoretically, from which we obtain that the k exists, at which the minimum average time consumption is achieved.In coding the program of solving the pressure Poisson equation, a 1-D array is adopted to record only the nonzero elements of the coefficient matrix, by which the memory is greatly reduced and the operations involved in solving the matrices are significantly simplified. The Bi-CGSTAB algorithm, being more stable and fast-convergent comparing with the formerly-used ICCG method, is applied to solving the pressure Poisson equation. Owing to these two improvements, the simulation speed of the model is significantly accelerated.A numerical flume is designed in accordance with a laboratory experiment, in which the propagations of solitary waves are simulated using the improved MPS model. The modeled water surface elevations are compared with the experimental data which show good agreements.To overcome fluctuation of the modeled pressure, it is suggested that more appropriate model, such as model enhanced by the ALE model as an auxiliary computation procedure, is necessary.