A novel multipoint luminescent coated ultra violet fibre sensor utilising artificial neural network pattern recognition techniques

Abstract

A novel multipoint luminescent-coated optical fibre sensor system, which is capable of detecting ultra violet (UV) radiation and utilises artificial neural network (ANN) pattern recognition techniques to interpret its output is presented. The sensors of this investigation were fabricated on a 1mm core diameter polycarbonate fibre by removing sections of the fibre cladding and replacing it with a phosphor doped epoxy that emits visible light when stimulated with ultra violet radiation. The various sensing locations utilise phosphors that have different emission wavelengths resulting in individual spectral peaks associated with each of the sensing points. The system described is therefore capable of measuring levels of UV exposure at multiple points on a single fibre loop. The optical sensor system is interrogated using a fibre optic spectrometer, and the signal analysis of the data arising from the sensors is performed using artificial neural networks pattern recognition. This allows the interpretation and classification of the resulting complex spectral patterns, which contain spectral cross-coupling due to the phosphor emission overlap from the separate sensing points. Initial measurements were trained using a feed forward ANN, and the system was 100% successful in classification of all subsequent test samples analysed on the trained network, some of which are presented.