Research Of Parallel Discrete Element Method Simulation Software

In nature, there are lots of granular or objects that can be abstracted into particles. The Phenomenon of interaction between these objects in the industrial and agricultural production is very common. Our projects team, based on the research of the interaction of these particles, adding mechanical parts, according to the basic theory of discrete element method, research the interaction between particles and mechanical. For example, the corn, soybeans and others planting process in the agricultural production, the mechanization of spreading fertilizer process, pharmaceutical production and packaging process, and even the interaction between wheel and soil particles, etc., can be simulated using discrete element method on the computer. Through computer simulation, we can get related mechanical data, these data can help mechanical designer designing mechanical.According to the basic theory of discrete element method, the project team developed a three-dimensional discrete element method simulation software TDEM, achieved good simulation results, and play an important role in actual production. However, in the course of using, we find that it often requires a very large amount of computation to obtain good simulation results, the procedure often requires days or even weeks to get the whole simulation results, therefore, it is high time to improve the simulation software Computing speed. On the other hand, the traditional single-core processor has been gradually fading out of sight of people, multi-core processors have become mainstream. However, traditional serial program cannot make full use of the multi-core processor, we need to modify the past program, take full advantage of parallel performance of multi-core processors to increase computing speed. Meanwhile, the mainstream high-performance computing is still carried out in the distributed environment, and therefore, for the large-scale simulation, we need to explore the distributed discrete element method simulation algorithms to further improve the scale and speed of simulation. In addition, there is new distributed environment composed by the multi-core processors. We also need to reconsider the parallel algorithm, making the algorithm more suitable for running on the cluster composed by multi-core processors node in order to achieve better computational efficiency.The main job of this thesis is to optimize and improve the discrete element method simulation software TDEM which was developed by our project team, to program the parallel simulation algorithm which is suitable for running on multi-core processors, to enable them to take full advantage of multi-core processor computing ability, and to explore method for task balance and decide the scheduling method of parallel task. meanwhile, in order to further improve the size and speed of discrete element method simulation, based on the original software, we also explored the distributed parallel algorithm can be used in the cluster and calculated simulation using space division method.The thesis first introduces the basic principles of discrete element method and the calculation process, describes some method of increasing compute speed of the discrete element method and the method which our software selected. Then the thesis introduced OpenMP and MPI that using on multi-core processors and in distributed environment, and according to the different characteristics of multi-core processors and distributed cluster system, we designed the different algorithms and combined the two algorithms on SMP cluster. Each multi-core SMP node complete the assigned tasks using multi-core algorithms, the whole SMP cluster system adopt the algorithm for distributed environments. Ordinary users can use multi-core version of the software which has no other additional requirements to calculate, they can run parallel simulation of discrete element method by using this version on ordinary multi-core computers as same as using an ordinary software; And professional users can consider using the distributed algorithm or the fusion algorithm for a more large-scale simulation. Finally, we made the initial testing for the parallel software, including the software concurrency, task balancing, parallel performance.After the actual test, the parallel simulation software greatly improves the speed and scale of computation, has a good parallel performance, and with the development of modern multi-core processors, can improve the calculation speed of the software itself, which play an important role for practical application of the software. Parallel simulation greatly reduced the time required, improved the efficiency of mechanical design, and provided a powerful help to improve the social productive forces.