Parallel Acceleration Of Particle Method And Its Application On The Study Of Liquid Sloshing

With the rapid development of the world economy, the demand of energy for the world continues to climb. Currently, the exploration of offshore oil and gas is one of the most important energy sources. So it makes the number of vessels which are used to storage or transport the liquefied gas and oil increase rapidly each year. It is almost impossible for the liquid tank to keep full at all time. Therefore, liquid sloshing becomes an unavoidable problem. Liquid sloshing have a tremendous impact pressure on cabin structure and may cause serious damage to it. Since the liquid sloshing is inevitable, how to suppress the liquid sloshing has become a very important and urgent research topic.Particle method is used to study the liquid sloshing and the structure for restraining sloshing in this work. And considering the large amount of computation problem of particle method, parallel computation is adopted to accelerate the computation. The specific contents are as follows:First of all. this paper introduces the basic ideas, theories and numerical implemen-tation methods of two kinds of particle method. In this part, the basic theories of smooth kernel function approximation and particle approximation are discussed. And the particle approximation of hydrodynamics control equations and other equations used in this work are also introduced. The last part of this section descripts the numerical implementation of two particle methods, including the selection of kernel function, the treatment of the boundary and the strategy of neighboring particle searching.Secondly, due to the large amount of computation of particle method, two different kinds of parallel method are used respectively to accelerate the computation of two kinds of particle method. The parallel implementation of neighboring particle searching based on CPU by Message Passing Interface (MPI) is discussed firstly in this section. Through the analysis of parallel performance, it is pointed out that the data transmission is very time consuming for the strategy of MPI based on CPU. Therefore, the data transmission between different CPUs should be reduced as few as possible when designing the program. Parallel computation based on Graphics Processing Units (GPU) is also studied in this section. The comparison shows that parallel computation on GPU has the same precision as it on CPU. Then, according to the analysis of parallel computation, it is concluded that using multiple threads to access the global memory on GPU and the reasonable pre-treatment of program are both can significantly improve the computation of parallel program.Thirdly, the numerical simulation of liquid sloshing in two dimensional square box is researched. The study shows that the free surface and the pressure on the side of box is most violent when the excitation of surge and pitch is in the same phase. Based on it, the effect of reducing the liquid sloshing inside the box with baffle "I", "T" and "--" is respectively studied. The amplitude of the elevation of free surface and the pressure of side wall is analyzed under different size of each baffle. The sloshing can be suppressed very well by adding three kinds of baffle if choosing the appropriate size of each baffle.Finally, a preliminary study of three dimensional liquid sloshing is considered. After the analysis of the elevation of free surface, it is found that the elevation at the corner of box increases significantly under coupled surge and sway excitation. Furthermore, the effect of reducing the liquid sloshing for the box with baffle "+" is considered. It shows that the liquid sloshing is suppressed very well. The study of three dimensional liquid sloshing in this section can be the foundation for the further research.