Construction of three-dimensional extrusion limit diagram for magnesium alloy using artificial neural network and its validation

Abstract

Conventional extrusion limit diagram (ELD) involves only two extrusion process variables and as such it does not account for the combined effects of multiple process parameters on the extrusion process with respect to pressure requirement and extrudate temperature. Attempts were made in the present research to construct three-dimensional (3D) ELD for a magnesium alloy in the space of initial billet temperature, extrusion ratio and extrusion speed. A method to build 3D ELD by integrating finite element (FE) simulations, extrusion experiments and artificial neural networks (ANN) was developed. In addition to initial billet temperature, extrusion ratio and extrusion speed, the temperature difference between the extrusion tooling and billet, the size of the billet and the shape complexity of the extrudate were taken as the additional process variables and integrated into the equivalent initial billet temperature, extrusion ratio and extrusion speed. The FE simulations, verified by performing extrusion experiments to produce magnesium alloy rods, were used to generate datasets for training the ANN. The ANN then predicted the peak values of extrusion pressure and extrudate temperature over a wider range of extrusion conditions, based on which a 3D ELD for the magnesium alloy was constructed. The 3D ELD was finally validated by performing extrusion experiments to produce magnesium alloy tubes. The results demonstrated that the constructed 3D ELD was reliable and able to provide guidelines for the selection of appropriate extrusion conditions.