Search

This paper reports the results for the quantitative determination of ammonia (NH3) in aqueous solution by a UV-vis spectrophotometric method and artificial neural network (ANN) intelligence tool. Quantitation of NH3was based on the chemical reaction of NH3with cobalt(ii) (Co2+) ions in basic medium to form a blue hexamminecobaltate(ii) ([Co(NH3)6]2+) complex. Characterizations of Co2+ion in solution included photostability, pH effect, response time, Co2+ion concentration effect, dynamic linear range and reproducibility, which were performed using a UV-vis spectrophotometer. The pink cobalt species gradually changed to blue with increasing NH3concentration. The absorption calibration curve was linear over the NH3concentration range of 0.6–3.5 mM at optimum pH 8 with a reproducibility relative standard deviation (RSD) of <4.0%. The interference effect was found to be negligible for a number of foreign ions present in the reaction medium during NH3determination in an aqueous environment. A set of absorbance data for the [Co(NH3)6]2+complex at selected wavelengths was input for ANN training using a back-propagation algorithm. The trained network with 22 hidden neurons, a 28 500 epoch number and 0.001% learning rate has extended the dynamic NH3concentration range to 0.6–5.9 mM with a calibration error as low as 0.0649 × 10−3. The proposed ANN electronic sensor shows promise for NH3estimation in unknown water samples based on pattern recognition.