An electronic nose using time reduced modelling parameters for a reliable discrimination of Forane 134a

For lots of applications, new systems based on gas sensor array and called electronic noses are in investigations. To be used for the toxic gas detect ion, these systems need to be accurate and must provide a rapid evaluation of the target gas. In this application, we use metal oxide sensor array to detect, in presence of several humidity rates (18-85%), a refrigerant gas (Forane 134a) which can cause health troubles and greenhouse effects. In this aim, several fami lies of representative variables as the transient or modelling parameters a re extracted from the time sensor responses. These variables are then group ed in learning databases and tested with recognition methods to select the best association between database and method. Next, unknown gas samples wer e taken to validate this choice. The results show the possibility of a good discrimination and also identification of the target gas in presence of di fferent humidity rate. (C) 2001 Elsevier Science B.V. All rights reserved.