| The predicted growth of computational power and network bandwith suggests that science and engineering computation will continue to develop in distributed paralleling environment. In this scenario, computational fluid dynamics (CFD) as a typical science computation domain will shift to distributed computation. Such a distributed CFD system exhibits autonomy,interactivity, heterogeneous, dynamic scalibility. It requires the development of new software methodology for manage the complexity of the system and harvesting the power of the expected autonomous computing and data communication. Intelligent agents, as a new paradigm for developing software applications, satisfies the demands of the CFD system mentioned because of agents'autonomy, intelligence, cooperation, interactivity, and adaptivity.So far, in agent-based scicence and engineering computation domain the researches focus on using agent conception to analyse the complex software system , in which, agents are software object just interacting to each other to complete overall goal. The paper proposed an agent-based autonomy CFD solver system based on analysis of the potential talent of the agent oriented technology in this domain. The design and the implementation of a protype of an agent-based autonomy CFD solver are presented.The paper describes how the system works in solving the given problem by splitting the solving domain into sub-domain, switching numerical scheme timely and deciding when to appropriately interact the ghost cell values at run time. The improved BDI (Belief-Desire-Intention) agent architecture is adopted for the rational requirement of the system. In such a system intelligent agents are used to a) independently calculate in subdomain, b) control the execution of individual solvers on each sub-domain and coordinate the execution of the staff, c) mediate rationality between adjacent sub-domains, and d) switch different numerical scheme during the solution. At last the example results are put forward. |
PDF Full Text Request