Approximation of Phenol Concentration Using Computational Intelligence Methods Based on Signals From the Metal-Oxide Sensor Array

Computational intelligence methods achieve high efficiency in the analysis of multidimensional data from e-nose, the equivalent of the human sense of smell. This paper presents and compares selected and applied to approximations of five concentration levels of phenol algorithms. The measured responses of an array of 18 semiconductor gas sensors formed input vectors used for further analysis. The initial data processing consisted of standardization, principal component analysis, data normalization, and reduction. Nine systems based on soft computing can be divided into single method systems using neural networks, fuzzy systems, and hybrid systems like evolutionary-neural, neuro-fuzzy, and evolutionary-fuzzy. All the presented systems were evaluated based on accuracy (errors generated) and complexity (number of parameters and training time) criteria. A method of forming input data vector by aggregation of the first three principal components is also presented. The key contribution is applying and comparing nine CI techniques for estimating phenol concentration based on signals from metal-oxide sensor array.