Extending electronic tongue calibration lifetime through mathematical drift correction: Case study of microcystin toxicity analysis in waters

Abstract

There are certain issues in e-tongue research which precludes wide adoption of these systems in routine analytical practice. An important problem relates to sensor readings’ drift which may invalidate corresponding multivariate calibration. Invalidation of established multivariate classification and regression models during a certain period of time leads to necessity of frequent e-tongue system recalibration requiring significant investment of time and efforts. An alternative approach can be based on mathematical sensor drift correction using sensor responses in certain standard solutions which are measured with required periodicity. In this study we show that application of univariate single sensor standardization approach (similar to single wavelength standardization suggested in spectroscopy) can significantly improve precision of both regression and classification models and can extend a calibration lifetime up to the period of over two months compared with two weeks for raw uncorrected data. As a case study analysis of microcystins in water samples with potentiometric e-tongue was addressed in two aspects: quantitative microcystins concentration evaluation and classification of samples into toxic and non-toxic ones.