Factor-based analysis of voltammetric data for the qualitative and quantitative evaluation of complex liquids

Abstract

Factor-based techniques were used for the qualitative and quantitative analysis of voltammograms of complex liquids. Both normal pulse and square wave voltammetry were evaluated, at a variety of conditions, and with different electrode materials. Square wave voltammetry at a platinum electrode had the greatest variance in the resultant scores plots, so this techniques was used in all subsequent experiments.

Initially, individual scores plots were generated of fruit juices, beers, wines, coffees, and milks. For qualitative analysis, principal components analysis was used to differentiate sub-populations on scores plots. Enhanced differentiation of sub-populations was achieved by selecting out a portion of the voltammogram, and regenerating the pattern recognition plot.

For quantitative analysis, principal components regression and partial least squares algorithms were compared for their predictive ability. Square wave voltammograms of lactate, pyruvate, glucose and ethanol in beer where generated, over a known concentration range of the individual species, and prediction of the individual concentrations was done using PCR and PLS. To assess the accuracy and precision of PCR and PLS, correlation coefficients, and mean square errors were calculated. Further, PCR and PLS were used to predict individual concentrations of pyruvate, glucose and ethanol in a ternary mixture.