Estudos de modelos para predição de parâmetros de ensaios de cabos LAN

Abstract

This study consists of a comparison between two system identification modeling techniques applied to predict two important performance parameter involved on a LAN telecommunication cable, which are Insertion Loss and Attenuation Crosstalk Ratio Far End. This work focused on representing the relation between two different ways of measuring a cable by modeling techniques using Parametric Models, which are based on techniques of polynomial comparison performing output models comparison considering Akaike, Bestfit and Residuals analyses criteria and in sequence the Artificial Neural Network, which is an artificial intelligence technique inspired on nature that can identify standards and knowledge, performing a comparison between the real output and the model output. All of that using the Matlab system identification and neural networks toolboxes. On this work were presented the main concepts involved on modeling using parametric models and Neural Networks and was performed a comparison between the obtained results with the adopted criteria and the applied techniques pointing to each advantage and disadvantage for the solution of this problem.