ニューラル・ネットワーク-紙パルプへの応用の可能性-

抄録

Artificial neural systems, also called neural networks, and neurocomputers, were spotlighted in Japan's recent announcement of their Sixth Generation project. This paper reviews these new systems involving the back-propagation algorithms and the adaptive resonance theory. Most of them have the advantages of self-learning and pattern recognition. There are so many cases in the paper industry where these advantages are required. If the hardware is only realized, we will be able to apply them to these cases, especially to process controls. It may herald the way for future thchnical break-through. The author has investigated applications of the neural system to evaluation of sheet formation and fiber analysis. These results led to the conclusion that the availability of conventional computer systems such as expert systems and statistical control systems would be an important factor in their use for the paper industry.