Predictive 3D modelling of free oblique cutting introducing an ANN-based material flow law with experimental validation over a wide range of conditions
Résumé
Modelling of the cutting process needs to move from 2D to 3D configurations to get closer to industrial applications. This study introduces a predictive
3D finite element model of free orthogonal and oblique cutting with an Artificial Neural Network (ANN)-based material flow law and experimental validation in strictly the same conditions (cutting and geometrical). The flow law based on a neural network allows simulating the cutting process based on data coming from the material characterization tests without requiring any postulate concerning the expression of the flow law. The developments are applied to the formation of continuous chips for the titanium alloy Ti6Al4V and an unseen broad range of 36 cutting conditions is considered: 2 cutting edge inclinations, 3 uncut chip thicknesses and 6 cutting speeds. The predictive performance of the model (i.e., the evaluation of the trends of fundamental variables with the absence of tuning of both numerical parameters and model features when cutting conditions are significantly modified) is high for the forces, mainly cutting and passive, and the chip thickness ratio on all 36 cutting conditions. The accuracy of the main cutting
force is excellent: the average difference with the experiments is 4 %, within the experimental dispersion. No significant degradation of the results is brought by the apparition of the third, out-of-plane, force, which shows the ability of the model to handle orthogonal and oblique cutting configurations.