The Study Of Parallel Method Of Boussinesq Equation For Nonlinear Shallow Water Wave Simulation

Boussinesq equation is a nonlinear dispersion wave equation in shallow water. The numerical models based on Boussinesq equations could simulate the sitution of waves propagate on uneven bottom and the influence between waves and structure. It could simulate refraction, diffraction, reflection, shoaling etc. Therefore, Boussinesq equation models have been used widely. Along with continuous improvement of Boussinesq equation, it could simulate deep water region and great area field, however, computing only by one processor could not statisfy the projects'actual need because of the limitation of the memory's size and CPU's speed of computer. In this thesis, it designs parallel program of Boussinesq equation in order to solve this problem, and utilizes high performance parallel machine cluster system to Boussinesq equation's wave simulation.In this thesis, the improvements of dispersive property, shoaling, the application of parallel metods of Boussinesq equations are reviewed. Then Parallel machine classification and performance analysis method are introduced. It adopts the Boussinesq equation set up by Madsen and Sch?ffer(1995). This equation has better dispersive property than ever; it can be used on terrain of changed depth of water and slope. The center compact difference in space and center difference in time are applied in Boussinesq equation. As a result, it can obtain higher precision than the center difference in space while the number of difference points is not changed. ADI method is applied while solving this difference equation. It ensures that the solution process is stable.MPI+Fortran parallel program language is used in this paper. The program has good portability. Adopting region decomposition method to design parallel program, the thesis uses different decomposition region at the different computation direction. It considers the load balance, avoiding dead lock, setting dummy processor and so on. The data's exchange among processors uses"send and receive"method. At last, this parallel program is confirmed by models, the result of calculation tallies with the serial result. Paralle program reduces the computing time in a certain degree.