Abstract
This article aims at designing a fuzzy logic-based classification and authentication system for commercially available tea and water brands. Electronic tongue instrumentation-based experimentally collected data had been used to design the classification and authentication system. Tea and water samples of six different types were evaluated using pulse voltammetric technique. However, this data generation and reporting are not under the scope of the current study. The classifier/authenticator design is a three-step procedure. At first, identification of the data in subspace was done using Sammon's non-linear mapping technique followed by entropy-based fuzzy clustering of water/tea data set formed and, finally, designing the expert system optimized using a particle swarm optimization to authenticate unknown water and tea samples with 100% efficiency. Thus, it deals with dimensionality reduction of the data set for visualization, clustering based on similarity, and development of a fuzzy logic-based expert system, as it is a powerful tool for dealing with imprecision and uncertainty.
TopBackground
Previously, a number of research had assessed the quality of beverages using various analytical techniques, such as high-performance liquid chromatography (HPLC), gas chromatography-mass spectrometry (GC-MS), capillary electrophoresis (CE) etc. (Ren et al., 2013; Yang et al., 2020; Zhou et al., 2022). However, all of the above-mentioned techniques are relatively time-consuming, complex, and expensive, which limits their practical applicability. Recently, electronic tongue (e-tongue)-based automated beverage and food quality monitoring has become quite prevalent (Banerjee et al., 2019; Calvini & Pigani, 2022; Garcia-Breijo et al., 2011; Hu et al., 2022; Moreno et al., 2006; Rifna et al., 2022; Sipos et al., 2012; Wang et al., 2019; Wu et al., 2022; Zeng et al., 2022).
An e-tongue consists of some non-specific solid-state ion sensors, different types of transducers, data collectors, and machine learning algorithms for data analysis aiming characterization of liquid samples (Al-Dayyeni et al., 2021; Apetrei & Apetrei, 2013; Escuder-Gilabert & Peris, 2010; Kirsanov et al., 2013; Marx et al., 2021; Ribeiro et al., 2021; Riul Jr. et al., 2010; Tahara & Toko, 2013). Various electrochemical methods, such as pulse voltammetry, potentiometry, amperometry, and stripping voltammetry were deployed in e-tongue to generate characteristic signatures (like current signatures due to pulse voltammetry) of the liquid sample to be analyzed. An e-tongue is not a taste sensor; the result of characterization is not necessarily correlated with human taste perception or compared with panelists. In dynamic interfacial techniques, it is not humanly possible to classify just by executing a visual inspection of the waveforms obtained as an output from an e-tongue. Cross selectivity (partial overlapping selectivity) of sensors employed and non-stationarities in the corresponding signals were the reasons, which demand the deployment of machine learning algorithms to formulate automated/computer-based systems for authentication purposes (Ciosek & Wróblewski, 2007; Vlasov et al., 2005). It might be appropriate to refer to a few research contributions regarding the applications of multivariate statistical, neural network-based, and fuzzy logic-based machine learning components practiced in e-tongue devices dedicated to quality monitoring.
Key Terms in this Chapter
Sammon’s Nonlinear Mapping: It is a distance-preserving tool used for the dimensionality reduction.
Electronic Tongue: It is an instrument that includes arrays of solid-state ion sensors, transducers, data collectors and data analysis tools.
Similarity: It is checked between two data points using their Euclidean distance.
Tea Sample: E-tongue signal-based data related to six different ISI certified grades of tea, namely Brookbond, Double-diamond, Godrej, Lipton, Lipton-Darjeeling, Marvel) are utilized.
Dimensionality Reduction: The higher dimensional data are mapped to the lower dimension for the purpose of visualization.
Fuzzy Clustering: It is done for a set of data points based on their similarity values.
Water Sample: E-tongue signal-based data set consisting of six number of water brands (Aquafina, Bisleri, Kingfisher, Oasis, Dolphin, and McDowell) is used.
Fuzzy Logic: It is a potential tool for reasoning after dealing with imprecision and uncertainty of data set.