Functional Data Analysis in Control of the Extrusion Process
Igor Duplancic, Branimir Lela, FESB, University of Split, Split, Croatia
Ante Musa and Oliver Zovko, FEAL, Siroki Brijeg, Bosnia and Hercegovina
Track: EXTRUSION/DIE THEORETICAL - Modeling, Microstructure and Simulation
ABSTRACT—Functional data analysis (FDA) is a relatively new method which has application
in various areas of human activities. Recently, it has been used in analysis of manufacturing
processes which are numerically controlled by PLC devices. Typical of such processes in the
aluminum industry are DC casting facilities, rolling mills, and extrusion presses. This method
allows the prediction of different variables based on input data that are recorded during each of
these processes. The present study shows how some of the important extrusion parameters, such
as extrusion force, section temperature, and others can be predicted. A new linear regression
mathematical model has been proposed for the prediction of these parameters. Regression
coefficients of the model are in the form of time dependent functions that have been determined
through the use of FDA methodology. A mathematical model has been developed using data for
technological extrusion parameters that were measured and recorded on a real extrusion press.
Simulation of the mathematical model shows that its predictions are in good in accordance with
the measured data on the extrusion press.
© 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.