Processing the signals from an optical fibre pH sensor by using an artificial neural network

The optimisation of response range of an optical fibre pH sensor based on immobilized bromophenol blue (BPB) in silicone rubber membrane has been studied by using an artificial neural network (ANN). The absorbance spectra of the immobilized BPB were scanned out by using an optical fibre spectrophotometer at different pH values. The-absorbance data were used as an input data for the ANN training. A feedforward ANN with backpropagation error algorithm was employed. Eight input neurons, which correspond to different absorbance intensities measured at 625 - 800 nm, were employed. The trained network with 11 hidden neurons was accurate in predicting the response of the sensor with an average prediction error of 0.09 pH when the network was used for measuring unknown buffer solutions measurements. The application of feedforward ANN allows the expansion of the pH response range of the sensor from its limited linear range (pH 2.50 - 4.50) to an extensive calibration response range of pH 2.00 - 12.00.