CHEMICAL-ANALYSIS OF MULTICOMPONENT AQUEOUS-SOLUTIONS USING A SYSTEM OF NONSELECTIVE SENSORS AND ARTIFICIAL NEURAL NETWORKS

With the aim of creating a multisensor system for determining heavy-me tal cations (Cu2+, Pb2+, Cd2+, and Zn2+) and inorganic anions (Cl-, F- , and SO42-), measurements in mixed solutions were carried out with th e use of an array of sensors based on chalcogenide glass electrodes, a nd the possibility of using various methods of mathematical processing of the resulting intricate signals was studied. Three methods of data processing were used: multilinear regression, partial least-squares, and artificial neural networks. It was found that the multisensor syst em proposed was suitable for determining all of the analytes with a pr ecision of 1-10%. Because the responses of sensors in solutions of com plex composition deviated from Linearity, the lowest determination err ors were obtained with the use of an artificial neural network. As to the method of data securing (nonselective response of a sensor array) and processing (artificial neural network), the multisensor system dev eloped may be considered a prototype of a device of the ''electronic t ongue'' type.