Application of sensor array and artificial neural network for discrimination and qualification of benzene and ethylbenzene

This paper describes a method for discrimination and quantification of benzene and ethylbenzene in different mixtures of these compounds. Data from Taguchi gas sensor arrays were obtained. A neural network was applied to sensor response analysis. The paper presents research on neural network application for quantitative analysis of organic solvent mixtures, revealing significant possibilities of improving the selectivity of low-cost gas sensors by combining them into arrays. Investigation of several neural network structures show that with appropriate data pre-processing, the 6-12-12-2 network provides satisfying results, and may be implemented in low-cost universal measurement systems.