Optimization of hot extrusion process using an artificial neural network

Extrusion of aluminium alloys is a complex process which depends on the characteristics of the material and on a process parameters (initial billet temperature, extrusion ratio, friction at the interfaces, die geometry etc.). The right choice of these parameters is fundamental to avoid surface damage of the extruded. One of the most important factors is the temperature profile. In the present work, a neural network has been implemented for optimizing the aluminium extrusion process, by determining the temperature profile of a Al 6060 alloy (UNI 9006/1) at the exit of the die. Parameters such as heating conicity at the entry of the die, permanence time in the induction heater and extrusion ratio have been varied. A three-layer neural network with back propagation (BP) algorithm has been trained with the experimental data from the industrial process. The experimental data refer to five different section bars, characterized by a specific extrusion ratio. For every section bar there is an optimal initial temperature profile of the billet, realized in the induction heater. The temperature has been controlled by five thermocouples allocated inside the induction heater to regular intervals. By means of optical pyrometry the instantaneous temperatures at the exit of the induction heater and of the die have been measured. It has been found that the temperature profile on the section bar, predicted by the neural network, closely agree whit experimental values. This indicates that the neural networks can be successfully used to predict the evolution of the extrusion process thus to control it.