Modeling and interpretation of responses from e-noses in the detection of gases in air

A new method, called Shadow Clustering (SC), to build an e-nose net is proposed. The method, which is very appropriate to monitoring applications due its cheapness and low consumption, is applied to an array of sensors composed by a TiO2 thin film. A model describing the e-nose behavior is built and it is used to simulate the detection process of a mixture of Ethanol, Methanol and Propanol. The results obtained by the new method are compared with those obtained by other wellknown learning techniques: Neural Networks and Regression Trees.