Stimulation Initiative for European Neural Applications
Esprit Project 9811
Optimizing Facilities for Polymerization
Company background
No information available
The problem
Operation and controlling of facilities for polymerization involve
certain difficulties: High standards in the quality of products and
reproducability. Certain parameters call for an optimal control of
production to run the process constantly and reproducable. On the other
hand, the process is hard to control because of its nonlinear
development and its relevant parameters depending on a great number of
measurements. Besides, there are physical factors not covered by
measurement technology. The manual regulation does only lead to
insufficiency and variation of product quality.
Neural network application
In this application, the production of polyethylene has been examined.
In this process, density and the melting index are highly relevant
factors of product quality. Until now, these factors could not be
measured online but only with a delay of time - or even less accurately
in a laboratory. Neural networks have been trained with the relevant
factors in order to predict the melting index and density in real time.
Benefits
This example shows that trained neural networks are capable of
predicting the processes melting point and density. The facility's
operation can be improved by this application which provides the
relevant data almost thirty minutes earlier than laboratory measurements
and therefore guarantees timely interventions in process control.
Generalizations
Non-linear processes, for example in the chemical industry, call for the
use of forecasting facilities to improve process control. Therefore they
open up a vast area of applications for neural networks suited for
processing online measurements.
Contact Person
ATLAN-TEC KG.
Obere Färberstraße 11,
D-41334 Nettetal,
Tel.: +49 2153 800 237,
Fax: +49 2153 891 32.
![[ ]](../logo_small.gif){kind=logo}
SIENA