Interrogation of multipoint optical fibre sensor signals based on artificial neural network pattern recognition techniques

Abstract An optical fibre multipoint sensor system incorporating multiple (3) U-bend sensors is presented which is capable of detecting contaminants in water. The sensors are based on 62.5 μm core diameter polymer clad silicone (PCS) fibre which has had its cladding removed in the sensing regions. The addressing of the fibre is achieved using an optical time domain reflectometer (OTDR) and is, thus, capable of spatially resolving power loss (along the fibre’s length). The signal analysis is performed using artificial neural networks (ANN) pattern recognition, which allows classification of the samples under test, thus, allowing the true measurand to be recognised. The system described is capable of both measurement at multiple points on a single fibre loop, and of recognising cross-sensitivity from interfering parameters such as lime scale coating in hard water and the presence of other species, e.g. alcohol in the water. Experimental results and the suitability of the ANN for their interpretation and classification are reported.