Stimulation Initiative for European Neural Applications
Esprit Project 9811
Optical Positioning in Industrial Production
Company background
Kratzer Automatisierung GmbH is a company specializing in the automation
of industrial processes and is based in Munich, Germany.
The problem
In the producton of test tubes for the analysis of gases the scale's
zero has to be positioned on the division between two substances which
varies in dependence of manufacturing parameters. Until now, the test
tubes were positioned manually before being labelled with scales.
Therefore, the labelling machine's clock frequency was limited to
1Hz.
Neural network application
The user defines a test tube segment bearing the relevant information
for the position of the scale. A neural network -a
multi-layer-perceptron consisting of eight neurons- is trained to
register the division between the substances inside the test tube.
Afterwards, an accuracy of 0.1 mm is achieved, only taking less then 0.5
seconds per tube.
Benefits
The application of the neural network-based optical system for
positioning purposes improves the product's quality and enhances the
machine's clock frequency. Furthermore, the number of rejects is
reduced.
Generalization
Tasks of optical measurement and positioning can be dealt with by the
application of neural network-technology, even those causing great
problems to classical methods of image processing.
Contact person
Klaus Eder -
Kratzer Automatisierung GmbH.
Carl-von-Linde-Straße 38,
D-85716 Unterschleißheim / München.
Phone: +49 89 32152-0
Fax: +49 89 32152-109