Distribution Of The Solution Of Partial Differential Equations Based On Agent Technology Research

With the development of compute performance, computers have been applied in more and more fields, and "science computation" has already become a kind of important science research methods. As a branch of science computation, the calculation of steady solutions of a PDE(Partial Differential Equation) is today of interest in various parts of industrial development. Multigrid algorithm is often used to get the steady solutions of PDE because the physical domain itself is too complex. The efficiency is well when the algorithm is implemented in the case of few blocks. However, when the number of blocks is increased, the efficiency deteriorates. The remedy is to add communication between the blocks, so that each block can get some necessary information of the whole system. However, too much communication would increase the cost of time, so how to keep the favorable ratio between communication and computation has become the focus of researches in recent years.To study whether Agent technology can be used in calculating steady solutions of PDE is the main task of this paper. We firstly introduce the conception of "science computation" and Agent. The IBC (Internal Boundary Condition) and communication strategies that are implied in multigrid multiblock method are analyzed, and we found their poor adaptability is the main reason that deteriorates the efficiency. To remedy it, we apply Agent technology in controlling communication so that the data is communicated in different models between the blocks in different cases. Then the amount of communications is limited and redundant communications is eliminated. To demonstrate the feasibility of our assumption, we developed a platform of computing the steady state solution of our PDE case by multigrid multiblock method based on the Agent technology.