Stimulation Initiative for European Neural Applications
Esprit Project 9811
HERACLAS - Quality Assurance by "listening"
Company background:
No information available
The problem
Neural networks are of growing significance as a means of quality
control. In the production of electric shavers, for example, occuring
defects reduce the products functioning. Until now, a proven method of
quality control has been acoustic detection of defects: Experts
classified the quality of electric shavers by listening to their
characteristic sounds.
Neural network application
A structure splitting the sounds of electric shavers into different
defect-classes has been chosen for the neural network. Then the neural
network has been trained with a certain amount of data, i.e. the
specific running sound of an electric shaver with a known defect. These
soundwaves were digitalized to make them accessible for a computer. The
neural net has been trained with this data until a perfect recognition
of defects was guaranteed.
Benefits
The application of neural networks in quality control of electric
shavers results in a reduction of rejections and production costs and
also guarantees high quality.
Generalization
Obviously, neural networks are highly recommended for quality assurance
in industrial productions. They help to automate and simplify the
process of control.
Contact Person
albit GmbH -
Hiltrud Verweyen-Frank.
Alexanderstraße 60,
45438 Mülheim.
Tel.: +49 208 495 630
Fax: +49 208 495 6350