Suitability of CAE neural networks and FEM for prediction of wear on die radii in hot forging

Prediction of tool wear in hot die forging along the entire arbor radius by wear models known so far is a very difficult task. On these parts of tools significant changes of contact pressure and sliding lengths occur along the die curvature during the plastic flow of material formed. A new approach presented in the paper combines the use of a conditional average estimator neural network (CAE NN) with the exploitation of results obtained by the finite element method (FEM) and also data from other sources. Consequently new parameters as well as the results of experimental work can be taken into account. In this paper a brief overview of models for prediction of tool (die) wear are discussed. The theoretical background of CAE NN, as well as its application to the modeling of the tool wear phenomenon, is presented. Some results of FEM analysis of the hot forging process that serve as input parameters in the CAE NN model are also briefly discussed. Two relevant practical applications are shown. In the first example, tool wear was modeled at a higher number of strokes (blows), by knowing wear at a lower number of strokes. In the second example, the number of strokes was the output parameter—the number of strokes causing predetermined wear at any point of the tool engraving curvature (arbor radius) was predicted. A comparison between the measured and predicted values of wear demonstrated good agreement that was assessed by a corresponding coefficient of determination.