Deviations from bilinearity in multivariate voltammetric calibration models

Abstract

This work considers the problem of lack of bilinearity in multivariate calibration. In voltammetry this issue especially relies on the analysis of overlapping signals, which change the shape, sensitivity or shift along the potential axis, causing a significant loss of linearity. It limits the quality of many chemometric models designed for linear data. Improvement of the predictive ability of multivariate calibration models is achieved by pre-processing of the raw data. In this work we proposed the application of a technique called orthogonal signal correction (OSC). We demonstrated that orthogonal correction enables the removal of almost all non-linear effects, disturbing voltammetric signals that impede the building of effective PLS models. The methodology was presented using simulated signals, and also in determination of the nanomolar concentration of scandium in the presence of a high and changing excess of nickel.

MATLAB in electrochemistry: A review

“…Bilinearity is a property assumed by multivariate linear calibration algorithms. This requires that a data matrix can be expressed as a sum of single component (analyte) data matrices, where each of them decomposes into the product of two vectors containing the concentrations and currents [79]. However, in many analytical situations, slight deviations from these assumptions have to be considered, such as, for example, in the presence of interactions between different individual components.…”

“…Any of these effects strongly hinder the direct application of bilinear data models. Such problems become more complicated when signals generated by successive analytes or interferences overlap [79].…”

Derivative Linear Sweep Voltammetry and Discrete Wavelet Transform for the Simultaneous Determination of Codeine and Thebaine by Artificial Neural Networks

“…In this sense, various linear and non-linear chemometrics techniques have been employed, including artificial neural networks (ANNs), [26][27][28][29][30][31][32][33][34] PLS, [35][36][37][38][39] MCR-ALS, [40][41][42] and wavelet transform. [43,44] A problem arising in the application of linear sweep voltammetry for the simultaneous determination of multiple species, that influences the bilinearity of the voltammograms and the accuracy of the analytical quantification, is the effect of the background interference on the shape of the voltammogram for the latter components that is itself affected by the presence and the concentration of the former electroactive species, [45] which finally distorts the observed voltammograms due to the overlapping of the voltammetric signals. It has been proven that the use of derivative techniques in these situations can be helpful for quantitative determinations because depending on the order of derivatization, they allow the linear or higher-order interference of the background, noise, and other interferents to be removed and the signal sensitivity and the resolution power to be increased.…”

“…Artificial neural networks (ANNs), due to their ability to model both linear and non-linear relationships between dependent and independent variables, are being used increasingly in electroanalytical determinations. [45,51,52] The preprocessing of the voltammetric data for the analysis by ANNs is often required to reduce their high dimensionality and complexity, to gain advantages in training time, to avoid redundancy in input data, to obtain a model with better generalization ability, and may simplify the model representation. In this sense, Discrete Wavelet Transform (DWT), developed from the Fourier Transform (FT) during the late 1980s, is an interesting choice because of its ability to compress and denoise the data simultaneously.…”

Principal Components – Based Techniques in Voltammetric Determination of Caffeic, Syringic and Vanillic Acids

“…Application of automatic baseline correction was extensively studied. linear modelling basedo nl atent variables,w ith data obtained by the means of linears weep voltammetry [13], anodic stripping voltammetry [14,15],a dsorptive stripping voltammetry [16][17][18][19],d ifferential pulse polarography (or voltammetry) [20][21][22][23],d ifferentialp ulses tripping voltammetry [24,25] and square-wavev oltammetry [26,27]. Additionally,a pplication of non-linear methods such as polynomial PLS and artificial neural networks( ANN) is also possible [28][29].…”