An EMI Source Finding Method Based on a Neural Network
Jesús Manuel González Bueno · 2003
This paper presents a technique based on neural networks for prediction of radiated emission sources from electric and electronic equipments. The goal is to estimate the position of the sources of EMI (modeled as elemental isotropic antennas) from measurements made with a CISPR measurement system, which uses only amplitude data (electric field strength). As a basic approach, a feed-forward neural network trained with back-propagation algorithms is proposed for “learning” the EMI to source parameters mapping. The learning process is accomplished by using computed data from a number of sets of isotropic sources having different configurations. The applicability of this method is theoretically verified by evaluating whether given unknown sources can be found correctly.