Simultaneous determination of relative humidity and ammonia in air employing an optical fibre sensor and artificial neural network

The simultaneous determination of ammonia and relative humidity (RH) in air employing an optical fibre chemical sensor based on a nafion-crystal violet composite and a multivariate calibration based on and artificial neural network (ANN) has been described. Each and every measurement was made with a new film of 5 μm thickness, prepared from a 1.0×10−3 mol l−1 crystal violet and 1.0×10−2 mol l−1 nafion (as sulphonate groups) solution. Studies were performed in the ranges of 30–70% relative humidity and 0–25 ppm ammonia. A feed-forward ANN, with error back propagation training algorithm, was employed for treatment of data. Input neurons corresponding to reflectance intensities measured at 562, 583, 605, 626 and 660 nm were employed. The optimised ANN provided standard errors of prediction (SEP) of 28.4 and 7.3% for NH3 and RH, respectively, when fed with spectral data recorded after 2 min of exposure (5 input neurons). The generalisation capability of the ANN was improved when it was fed with spectra data recorded after time intervals of 30, 60, 90 and 120 s (20 input neurons), providing SEPs equal to 9.9 and 4.5% for NH3 and RH, respectively. This improvement can be explained considering that water vapour reacts faster than NH3 with the film (a time interval <30 s is enough for the reaction to reach the equilibrium state) while the reaction rate for NH3 is dependent on the RH (higher RH inhibits ammonia reaction).