Research And Application On Parallelization Of Chemical Dynamical Simulation In PCs Cluster

Although the performance of the computer has increased thousands of times with the rapid development of computer science and technology, the people's demand of high-performance computing is increasing too. Therefore in practice, the single processor can not meet the demand of high-performance computing, which promotes the flourishing development of parallel processing technique in the recent twenty or thirty years.Chemical process simulation usually adopts accurate mathematic model which based on process mechanism and strict physical properties calculation. This kind of model often has large scale and nonlinear feature. The whole simultaneous equations of the typical chemical process simulation generally include tens of thousands equations, even some up to hundreds of thousands. To so enormous the whole simultaneous equations, single computer need to spend a large amount of calculation time solving them.Though parallel processing is an effective means to solve this kind of problems, the special parallel computer is very expensive. With the gradual popularization of the personal computer and decline of the price of network equipment, PC cluster which is established through interconnection network offers a scheme with well scalability and high usability. PC cluster adopts message passing mechanism to realize communication between processes. Now, popular parallel programming environment is MPI (Message Passing Interface) and PVM (Parallel Virtual Machine), which is based on the message passing. This paper regards MPI as parallel programming environment. Some researches about parallel processing have been achieved on self-built PC cluster. Dynamic simulation of distillation column is regarded as an example to research the parallelization of chemical dynamic simulation on PC cluster.This dissertation expatiates on the basic theory of the parallel processing and the basic situation of chemical dynamic simulation at first, and then introduces PC clusterand the message passing system of MPI. On the basis of these, a PC cluster experimental environment based on Windows 2000/XP and MPI is established. Through analyzing parallel performance of experimental environment, PC cluster is suitable for solving the large-scale problems. To certain scale problems, though the large-scale cluster can always shorten calculation time, the parallel performance of the whole cluster is not good. Here, on the contrary small-scale cluster can obtain higher parallel performance and utilize resources fully, but there is not much difference in both calculation time. Finally, conclusions of this dissertation and suggestions of future work are given.This thesis has five chapters.Chapter 1 mostly introduces the generation, the development and the classification of the parallel computer, the basic theory of the parallel processing and the basic situation of chemical dynamic simulation, the application actuality of parallel processing technique in the chemical dynamic simulation, and also introduces research effort of this thesis.Chapter 2 detailedly explains the composition and the feature of PC cluster, parallel programming on cluster and communication mode of MPI.Chapter 3 expatiates on the software and hardware environment and the communication system of self-built PC cluster. Then the realization of MPI on Windows and the method of realizing parallel programming with MPI are given. Lastly, the parallel performance of self-built PC cluster is analyzed.Chapter 4 describes dynamic mathematic model of distillation column, and puts forward the mathematic method for solving model and the parallel algorithm which is suitable for PC cluster, and divides subsystem according to the self feature of the distillation column.Chapter 5 summarizes this thesis and looks forward to the next job.