Use of a simple tin oxide sensor array to identify five malodours collected in the field

A laboratory-made malodour sensing system including 12 commercial tin oxide gas sensors (Figaro Engineering) is used to identify five typical sources of olfactive annoyance: printing houses, paint shop in a coachbuilding, wastewater treatment plant, urban waste composting facilities and rendering plant. In this work, all the samples are collected in the field from real malodours in uncontrollable conditions. The ability of the system to predict the origin of unknowns odoriferous samples is investigated. The test of various pre-processing data algorithms shows that the best classification results are obtained with a parameter free of the sensor base-line. The differences in sensor responses among the five odours are shown by icon plots and confirmed by principal component analysis, which highlights four representative clusters. Classification models calibrated by discriminant analysis and artificial neural network are validated on unknowns samples. Chemical relationships between the sensors and the classification results proves that the recognition is not fortuitous. In spite of the influence of environmental parameters, results demonstrate the ability of a simple system to detect and identify typical olfactive annoyances