Modelling Surface Roughness of EN8 Turned Component using Artificial Neural Network

Shafts demand different surface roughness's at different sections along its length to mount different components. This requires setting machining parameters at various levels and need operator skills. Sometimes finishing processes are to be implemented that add to the machining time and cost. In the present work, neural network has been implemented to predict the surface roughness of the turned component of EN8 shaft material using tungsten carbide tool on CNC machine. The surface roughness data for training and testing was generated by measuring surface roughness values on turned components using recommended values of machining parameters-spindle speed, feed rate, depth of cut and tool nose radius from Machining Data Hand book using L27 orthogonal array (OA). Neural networks of different configurations i.e with one and two hidden layers with different number of neurons in each were tested and compared by calculating MSE for each during training and it was found that the network with 4-8-1 configuration was found to be best network. Three additional practical data of surface roughness were generated by turning the material using different combinations of input parameters to serve as validation data. The neural network predicted the roughness values in close approximation to the practical values which established the robustness of the network.