Modelling and prediction of surface roughness in Ti-6Al-4V turned surfaces: use of DTCWT image fusion and GLCM

Abstract

In manufacturing process the quality of the surface roughness is one of the most important requirements. Nowadays, evaluations of surface roughness using machine vision methods are widely used. In this work, an attempt has been made to develop a feature extraction method which includes Dual Tree Complex Wavelet Transform (DTCWT) based image fusion and Gray-Level Co-Occurrence Matrix (GLCM) for surface roughness modeling of turned Ti-6Al-4V surfaces. Coated carbide tool inserts have been used for turning Ti-6Al-4V bars. A simple computer vision system has been developed to capture the different turned surface images. The captured images are subjected to image pre processing. Pre processed turned images have been subjected to DTCWT based image fusion. These fused image coefficients are converted into GLCM and the second order textural features like Energy, Entropy, Contrast and Homogeneity have been extracted. These second order statistical features along with cutting conditions namely, speed, feed rate and depth of cut and flank wear are given as inputs to a Radial basis function neural network (RBFNN) for modeling and prediction of surface roughness parameter R_a from the turned surface images. Finally, a comparison has been made with regard to prediction accuracy of surface roughness obtained using DTCWT based image fusion features and feature extracted using DTCWT without image fusion.