The application of neural networks to a condition monitoring system for the optical fibre drawing process
This source preferred by Siamak Noroozi
This data was imported from Scopus:
Authors: Shi, H., Tite, C., Hope, T., Noroozi, S., Poyntz-Wright, L. and Marshall, A.
Journal: Intelligent Engineering Systems Through Artificial Neural Networks
It is well known that optical fibre quality can be affected by the conditions present during fibre drawing, particularly in the drawing furnace. Early furnace decay can lead to problems with fibre production and thus can be detrimental in terms of fibre properties. Currently, a furnace rebuild is carried out regularly, based on a scheduled maintenance plan. A monitoring system which could recognise the need for a rebuild would be beneficial since (i) furnace life could be extended which could reduce changing time and increase machine utilisation or (ii) reduction in fibre yield could be avoided through recognition of early furnace decay. The drawing tension and speed were found to be good indicators of furnace decay and these have been reported previously (Shi, 1996). However, drawing tension is also affected by some other drawing parameters such as furnace power, speed and preform diameter. In this paper, neural networks are used to integrate information from different drawing parameters in order to recognise the furnace state during the optical fibre drawing process and thus decide on the necessity of the furnace rebuild. Different structures of neural networks for application in a furnace condition monitoring system were investigated and the results are presented.