Middle Surface Approximtaion For Parallel And Distributed Substrate Coupling Analysis

Abstract

In this paper a new method for accelerating substrate coupling modeling and analysis is introduced. This method is based on separating substrate to two different parts for simulation. The separation achieved by using a multi layer perceptron neural network, by approximating the middle surface of substrate for finding other points potentials. By separating the substrate, it could be simulated by different parallel processors and time of simulation in this modeling method, decreased typically by 57%. The simulation procedure distributes by applying different boundary conditions resulted from neural network approximation