Electrochemistry

Abstract

Applications of pattern recognition methods to polarography were extensively investigated by Perone et. al.. The first study [294] deals with the qualitative analysis of solutions of Cd²⁺ , In³⁺ , and Sb³⁺ ions by stationary electrode polarography. Polarograms were digitized at 2 mV intervals to generate pattern vectors of synthetic mixtures. The learning machine was used to compute classifiers that recognize the presence or absence of individual metal ions. A good separability of the classes was found but predictive abilities were only moderate. Main problems are the (normal) fluctuation of electrochemical data and the sometimes severe peak overlap. Successful classification was obtained only for a peak separation of at least 40 mV and for maximum concentration ratios of 10:1.