Classifying surface roughness for turning process with neural network

Abstract

using stylus measurement technique. The drawbacks of this technique are the contact between the stylus tip and the measured surface will produce scratches and the measuring speed is slow. A new system for controlling surface quality for turning process has been developed that applies the concept of machine vision technique. It uses a DCT (Discrete Cosine Transformation) technique for image processing and Self Organizing Map (SOM) as learning architecture in neural network. This system is able to classify the surface roughness of turning part into “WORN” and “UNWORN” category