Prediction of roughness surface in textured by electrical erosion using bayesian networks

Abstract

A model for prediction of parameters that defined roughness surface [Ra] when this texture is produced by Electro Discharge Texturing [EDT] is presented. The non-linearity, instabilities and expensive experimentation in this process, are main causes for use predictive techniques by means of robust and reliable algorithms, for study factors that present characterization hard. Series of experiments were conducted to produce plane surface textures using a modified EDM machine ALIC-1. The data collected in experimental phase were used for trained Bayesian models with Naïve Bayes and Tree Augmented Naïve Bayes [TAN] classifiers. Results show acceptable behavior within the operating range, consistent with the physical phenomena governing EDT process. Find a surface roughness with particular specifications is demonstrated.