Surface Roughness Prediction in High Speed Turning of Ti-6Al-4V: A Comparison of Techniques

Abstract

Surface finish of machined products is important and plays an important role in ascertaining its quality and other attributes. Surface roughness of difficult to machine materials like titanium alloys are difficult to model due to several factors influencing it. This study makes an attempt to compare the performance of a statistical technique, Response Surface Methodology (RSM) and two Artificial Neural Network (ANN) techniques namely Multi Layered Perceptron (MLP) and Radial Basis Function Neural Network (RBFNN) to model and predict the surface roughness parameters Ra and Rt in high speed turning of Ti-6Al-4V. Experiments have been carried out using uncoated carbide inserts under dry condition. The input parameters for the modeling studies include cutting speed, feed rate and depth of cut. This work also makes use of tool wear and cutting tool vibration (Vy) which are uncontrollable parameters as the inputs for modeling studies. The ANOVA analysis has revealed that feed rate and cutting tool vibrations are the most significant parameters affecting Rt and cutting speed and vibrations affect Ra. A comparison between the modeling techniques revealed that RBFNN performed better in terms of prediction accuracy when compared to MLP and RSM.