Characterization of a Voltammetric Electronic Tongue

Sammanfattning: Electronic tongues were developed some ten years ago. These systems consist of an array of nonspecific sensors and a signal processing unit. The sensor signals are processed by pattern recognition methods which makes possible the extraction of specific properties from the sample. Depending on the calibration, attributes such as quality or taste can be determined. These systems also detect changes in the sample that they are not calibrated for which makes possible the detection of anomalous occurrences. Such sensor systems are suitable for process control and surveillance. Important factors are the sensitivity and stability, specifically the sensor's ability to respond to small changes and to provide true and reproducible readings over time. Problems with sensor stability are commonly referred to as drift. The major topic of this thesis is the improvement of long term stability for electronic tongues used in liquid process applications. Drift counteractions, such as renewal of the electrode surface by polishing, was compared with mathematical correction methods. Since drift is induced by the environment of the sensor, mathematical correctional actions must include reference samples and the induced drift must be identical between measurements. These conditions restrict and complicate the use of mathematical drift counteractions. It was found that mechanical polishing renewed the electrode surfaces, and that the induced drift was unique for each sample. The sample induced drift pattern can be treated as information from the sample, but only if the sensors are renewable in a repeatable way. Applications where polishing the electrode surfaces are necessary to obtain repeatable analyses are described, such as the detection of urea and measurements in corrosive environments such as wine. Electrochemical oxidation of urea in water is difficult to use for analytical purposes because of residues left on the electrode surface. An important result from this thesis is that mechanical cleaning of the electrodes between samples gives sensor signals that are both repeatable and proportional to the concentrations of urea and glucose. An experimental design was employed for optimal effect of the calibration of urea in the presence of glucose as a disturbance in the sample. The goal was to minimize the correlation between the two analytes. This made possible the prediction of both analyte concentrations. Wine is a complex sample to analyze with many sources of disturbances for electrochemical measurements. The carefully planned experiments and calibrations reported in this thesis minimized covariance and background effects. A method for prediction of bisulfite, histamine and ascorbic acid concentrations in wine was developed. The method was tested with spiked samples of white-, red-, rose-wines and even apple juice. The reproducibility of the measurements was excellent. Since polishing renewed the electrodes between measurements, a validation performed one month after the calibration was also predicted with good results. This demonstrated that the renewal of the electrodes eliminated special requirements for maintenance and storage of the sensor. Drinking water surveillance has been performed with an electronic tongue. The potential of using a voltammetric electronic tongue for multicomponent analysis of compounds in drinking water has been evaluated. By using such a non-selective sensor it was possible to detect anomalies without the need of a specific sensor for each type of event. The device can be calibrated for the most likely events, and it can also be used for sensing and alarm when exceptional events occur. The detection of surface active species like detergents is normally done by titration. An in-line sensor that could control the washing process by detecting the concentrations of detergents during the different steps in a wash cycle could enhance the performance of washing machines. An electronic tongue was used to predict concentrations of detergents in samples from different stages in the washing procedure. The tongue was compared with a much simpler conductivity meter, and due to the covariation of supporting electrolyte, both sensors were able to predict concentrations of ionic surfactants. The electronic tongue showed promising results in predicting also non ionic surfactants where the conductivity meter failed. The detection mechanism was probably due to shielding the electrode surface from electro active species in the samples.