Analysis of Mechanical Properties of 2xxx T3 and T8 Aluminum Extruded Rods by Means of Neural Networks
Igor Duplancic, Branimir Lela and Frane Vlak, University of Split, Split, Croatia
Jere Prgin, TLM Sibenik, Sibenik, Croatia
Track: EXTRUSION/DIE, THEORETICAL - Neural Networks
ABSTRACT --- Production of 2xxx-aluminum-alloys solid sections is related
with several mutually connected metal-forming and thermal-treatment processes.
Experience accumulated in extrusion plants through the years of the production
of these sections usually results with databases. These databases offer
possibilities to relate section properties with process parameters. This paper
describes the procedure for analysis of mechanical properties of 2011 bars in
both T3 and T8 tempers. For this purpose, a number of extrusion processes on a
horizontal 50-MN extrusion press were analyzed. Depending on the bar
dimensions, solid dies having one, two, or more holes were employed. After
extrusion, all bars were cold drawn with various reductions. All bars were
thermally treated to appropriate temper. In this way, a database was established.
An artificial neural network (ANN) model has been developed to relate
mechanical properties to bar dimensions and cold-drawing reduction for both
tempers. After the ANN training and testing, an analysis was conducted in order
to show how the mechanical properties of extruded bars depend on the particular
aforementioned parameters. Optimization was conducted for the purpose of
finding the optimal value of cold drawing reduction for achieving the best
mechanical properties. Result of these investigations can be useful for industrial
© Extrusion Technology for Aluminum Profiles Foundation (ET Foundation). All rights reserved. No part of The Proceedings may be reproduced in any form without the express written permission of the ET Foundation.