On the prediction of surface roughness in turning using artificial neural networks

The aim of the present work is to use the technique of artificial neural networks, with backpropagation routine, for the prediction of surface roughness of workpieces produced by turning process. To provide the data required to train the networks, specimens were turned at different cutting conditions using tools having different pre-machining tool-flank wear. The vibration level during machining, and the after-machining tool-flank wear were measured and the surface roughness parameters of turned specimens were assessed. Several neural networks were trained with changing both of the network structure and the number of training samples. The best of these networks were selected to be used for the prediction of surface roughness parameters, vibration level and tool wear. From the results obtained it may be concluded that the developed neural networks permit acceptable outputs for the prediction of the investigated surface roughness parameters, vibration lever and tool-flank wear. Also, based on these networks, an integrated system for continuous prediction of surface roughness was developed, and a computer program was designed for selecting the appropriate machining conditions required to achieve desired values of surface roughness parameters.