Stimulation Initiative for European Neural Applications
Esprit Project 9811
Control of Tempering Furnaces
Company background
No information available
The problem
Many industrial processes are still carried out by human operators, as
automation technology can not provide the industry with robust and
efficient means of control. The tempering of rollers is one example of
such process: The heating has to be regulated in order to achieve a
steady tempering along the roller and to keep it on the necessary
temperature for a certain time. However, until now radiation and
conduction of heat lead to certain feedbacks which prevent any
mathematical modelling of the process.
Neural network application
In co-operation with a tempering company a system for
temperature-regulation has been developed which does not require any
mathematical modelling. The core of this regulation system is a neural
network which predicts -and if necessary corrects- the development of
the heating process. The neural network takes into account all nonlinear
influences and is trained by data of already performed heating
processes.
Benefits
The advantages of this system are the high quality of control resulting
from automated optimizing as well as an avoidance of stoppages in the
production process. Furthermore, an adaption of the regulation system
for any nonlinear process involving several parameters is possible.
Generalization
Replacing inaccurate and costly manual control processes by neural
network-technology is one of the main areas of neural network
applications. Neural networks simplifiy the control of processes in
which mathematical methods can not supply a complete or exact
description by taking all nonlinear influences into consideration.
Contact person
Dr. Stefan Gehlen -
Zentrum für Neuroinformatik GmbH,
Universitätsstraße 160,
D-44801 Bochum.
Phone: +49 234 9787-53
Fax: +49 234 9787-77