Stimulation Initiative for European Neural Applications
Esprit Project 9811
Substitution of Analysers in Distillation Columns
Company background
No information available
The problem
Aside from sophisticated regulation technology, exact realtime
information about the process is necessary to regulate the purity of a
distillation column's product. The use of analysers would be the ideal
solution to capture this information but in practice their application
is hampered by high costs and deadlines up to 30 minutes.
Neural network application
An alternative method to expensive analysers is given by the application
of neural networks, learning the connection between the column's
pressure and temperature and the result of the analysis.
Benefits
If the initial data is both -free of errors and consistent- the error of
prediction achieved by the neural net is less than 0.2%: A result
surpassing the accuracy of analyzers or any other conventional method.
Contact person
Roland Kühne GmbH,
Hellersbergstraße 10b,
D-41460 Neuss.
Phone: +49 2132 163-3