ROUGHNESS ESTIMATION OF A MILLED SURFACE BY USING NEURAL NETWORK BASED ON MINIMUM NUMBER OF EXPERIMENTAL MEASUREMENTS

Abstract

Artificial Neural Network is a powerful tool for prediction of parameter values, which presents a set of low input data, especially in terms of reducing costs and time for making measurements. The prediction of surface roughness according to the different tool trajectories of the finishing phase in milling process can be achieved both by unifying the results and by dividing the set of data into multi-classes. The paper presents, for the roughness parameter, how a set of low number of input data obtained by measurement is used for prediction as well as data set extension. The experimental tests were made for machining an aluminum 7075 part with plane surfaces at constant angle. The milling process was made without cooling.

Keywords

• prediction
• artificial neural networks
• roughness
• milling trajectory