ABSTRACT

A hybrid approach is presented here, based on the Design of Experiments (DOE) methodology and on Artificial Neural Networks (ANNs) for the modeling of the three component cutting force system in longitudinal turning of AISI D6 tool steel specimens. Twenty seven experiments were conducted having all different combinations of cutting parameter values. The experimental results were analyzed using ANOVA (Analysis of Variance) method. Signal-to-noise (S/N) ratio as well as interaction charts were produced in order to examine the interactions between process parameters. Finally, a Feed-Forward Back-Propagation Neural Network (FFBP-NN) was developed to simulate machining data. The selected inputs for the ANN model are the cutting speed, the feed rate and the depth of cut. The outputs are the three components of the cutting force, namely the feed force (Fz), the radial thrust force (Fx) and the tangential (main) cutting force (Fy). The neural network toolbox of MATLAB® software was used to develop, train, and test the ANN. The results obtained indicate that the proposed modeling approach could be effectively used to predict the three component cutting force system during turning of AISI D6 tool steel, thus supporting decision making during process planning and providing a possible way to avoid expensive and time consuming experiments.