Prediction of surface roughness of turned surfaces using neural networks

In this study, the prediction of surface roughness heights R_a and R_t of turned surfaces was carried out using neural networks with seven inputs, namely, tool insert grade, workpiece material, tool nose radius, rake angle, depth of cut, spindle rate, and feed rate. Coated carbide, polycrystalline and single crystal diamond inserts were used to conduct 304 turning experiments on a lathe, and surface roughness heights of the turned surfaces were measured. A systematic approach to obtain an optimal network was employed to consider the effects of network architecture and activation functions on the prediction accuracy of the neural network for this application. The reliability of the optimized neural network was further explored by predicting the roughness of surfaces turned on another lathe, and the results proved that the network was equally effective in predicting the R_a and R_t values of the surfaces machined on this lathe as well.