Prediction Of Surface Roughness and Tool Wear from The Machined Surface by Machine Vision Approach

Abstract

Tool wear show a clear relationship with the surface roughness as the condition of the cutting tools shows the potential for a better control of surface roughness. The quality of the finished surface is affected by the frictional resistance, creep life, machine vibration or fatigue strength of the machine. In this study, the recurrent neural network (RNN) model is developed to predict the surface roughness and tool wear based on the machined surface texture. The images of surface of the machined product were captured using CMOS camera. The images are converted into a greyscale images and undergo image enhancement to produce a better-quality image. The tool flank wear was measured after each set of the cutting parameter provides. The features such as average grey level, standard deviation and entropy were then extracted. The value obtained for the surface roughness and tool wear were trained together with the features extracted into the MATLAB software. The findings shows that the propose machine vision system can be implements for automated prediction of tool wear and surface roughness with accuracy of 92.75% and 64.59% for feature extracted. The features extracted also shows a better prediction accuracy compare to the cutting parameters by using the RNN prediction model.