Finite element method and artificial neural network modeling at estimation of extrusion temperature during the stroke

The extrusion process is to produce the desired geometry profile by passing the work piece through the die. The work piece placed into the container is forced out of the die opening which gives the profile form with the force exerted by the stamp. Factors; such as extrusion speed, temperature, extrusion rate, friction conditions and design of die which are influence the extrusion process. From these factors, temperature is one of the most important variant in extrusion process. In practice, there is a very complicated thermal change during the extrusion process. The heating of the container is provided before the hot billet entering to container, thus the temperature can be stabilized. Due to the deformation and friction of the work piece in the die channel, there is usually an increasing temperature change during the process. The temperature changes during this time are particularly dependent on the heat transfer to the container from the billet, the deformation conditions and the heat occurring with the friction. The maximum temperature is usually seen in the die corners where the work piece is exposed to intense deformation and where the work piece in the die channel. As the increased temperature generate the risk, it is important to determine the upper limit of the temperature. The experimental studies, which is applied with experimentation and experience, can be measured the temperature only at one point, and at the same time it is quite expensive and laborious. However, with computer-assisted analysis programs, the extrusion temperature can be estimated and the consistency of the results can be compared. The aim of this study is by showing the applicability of the Finite Element Method and Artificial Neural Network, to make an efficient prediction of the temperature at various points, determine limit of increased temperature which is the critical level during the stroke. Key Words : Extrusion, FEM, artificial neural network, modeling