MERGING NEURAL NETWORK MATERIAL RHEOLOGICAL BEHAVIOR MODELLING WITH FEM SIMULATION OF ORTHOGONAL METAL CUTTING

The fully integrated use of a neural network (N) paradigm for material rheological behaviour modelling and a FEM code for the simulation of orthogonal metal cutting is illustrated. A NN is trained to reconstruct the stress-strain curve of the work material on the basis of experimental tensile tests in a wide range of temperature and strain rate values. The learned NN is capable of predicting work material properties in the whole range of temperature and strain rate values utilized for training. The material rheological behaviour, modelled by the NN, is used in the FEM simulation to provide the work material properties to the FEM code for each node of the workpiece mesh during the simulation. To achieve this result, a continuous information exchange between the learned NN and the FEM code during each iteration is devised.