Study on the Approach of Assembling Models

周建军,郭延祥
2011-01-01
Abstract:Traditional method of modeling that aims at a single physical process and simple domain,is difficult in simulating a complicated large system,falling behind the computer technology trend toward multi-cores and networks.Parallel computation with multiple models and processes is a direction in the future.This paper proposes a method of assembling different time-dependent model that work together to simulate complicated processes using multiprocessing and networks technologies.In this method,a model is changed into a model component by simply inserting a probe function into the main time loop of its source code,and the cross-model communication information is represented uniformly by a quantity and a mesh series number and organized in four levels.The engaged models are managed in framework with a central controller,and their running process and sequence are controlled via a series of instruction code.Coupling of models is implemented by time-scheduling operation or different interpolation operation.A case study of assembling 1-D and 2-D hydrodynamic models verified the method's applicability and revealed some problems.The performance and efficiency of parallel computing by the method are discussed by means of assembling 1-D heat exchange models in different sub-domains.The presented method provides an efficient approach to solve complicated problems.
What problem does this paper attempt to address?