CLASSIFICATION OF VOLTAMMETRIC DATA USING COMPUTERIZED PATTERN RECOGNITION

STEVEN DALE SCHACHTERLE, Purdue University

Abstract

Computerized pattern recognition was used for discriminating between complicated and uncomplicated electrode processes, and for identifying the mechanism of the electrode process using the shape information contained in the data from a single voltammetric experiment. The voltammograms are represented by their Fourier transforms and are classified using a training set composed of curves based on the theory for cyclic linear sweep voltammetry (CLSV). Features selected to give the best classification accuracy with theoretical CLSV data could then be used for the classification of experimental CLSV or cyclic staircase voltammetry (CSCV) data. The use of theoretical curves in the training set allows broad coverage of both the types of electrode processes and values of fundamental parameters, such as rate constants, n-values, etc. Several feature sets based on the Fourier transform were tested, involving either scaling of the features or combinations of features which emphasize changes in curve shape with changing scan rate. Of the six feature sets tried, the unscaled Fourier transform coefficients were found to give the best classification results on the experimental data. The theoretical data were classified with an accuracy of 97%. The experimental data were correctly classified 93% of the time when the shape of the voltammogram was determined by a single mechanism which was included in the theoretical training set. Experimental systems included Fe(III), Cd(II), Ti(IV), Cr(VI), Pb(II), Eu(III), Tl(I), In(III), Iron(III)-protoporphyrin chloride, and Benzil, and were selected to cover as many of the theoretical mechanism types as possible. These experimental systems were classified equally well from CLSV and CSCV data.

Degree

Ph.D.

Subject Area

Analytical chemistry

COinS