Inter-System EMC Prediction with CG-GRBF Networks

Abstract

With the development of electronic technology, electromagnetic compatibility (EMC) becomes more and more important, and the EMC prediction is one of the most crucial EMC procedures. However, the EMC prediction is strongly non-linear and uneven, so it is tremendously hard to use traditional methods to conduct such prediction. To solve problems above, intelligent algorithms are introduced in our work to make EMC prediction. A general CG-GRBF Network is raised in this essay with a combination of Generalized RBF Networks (Radial Basis Function Networks) and the Conjugate Gradient Method (CG), which optimizes a selection of the standard deviation of the radial basis function. A contrast experiment among CG-GRBF, BP and GRBF networks is conducted and it turns out that the CG-GRBF Network is much better than the other two networks. Thus, the CG-GRBF Network can be one of the best choices to conduct the EMC prediction.