Stimulation Initiative for European Neural Applications
Esprit Project 9811
Classification of Defects in Pipelines
Company background
Pipetronix GmbH is a subsidiary of PREUSSAG AG, based in Lingen,
Germany.
The problem
The detection of defects in pipelines is made difficult by the great
number of control images which have to be taken over very long distances
and to be evaluated by experienced personnel.
Neural network application
Images of the interior of oil- or gas- pipelines taken by ultrasound
technology are classified by means of neural networks which are trained
by examples of defects. The system operates on two-dimensional images
from which anomalies are extracted and assigned to main classes of
defects.
Benefits
The simple architecture of neural networks allows for the classification
of defects; a problem having no solution in rule-based systems.
Therefore a faster evaluation than by customary methods of
defect-detection can be achieved. The long-term goal is a completely
automatic analysis of pipelines.
Contact person
Pipetronix GmbH -
Gerhard Kopp.
Friedrich-Ebert Straße 131,
D-49811 Lingen.
Tel +49 591 710 20.