ANALYSIS OF FOODSTUFFS WITH SEMICONDUCTOR GAS SENSORS OPERATING WITH AMBIANT AIR - PRODUCT DISCRIMINATION USING DIFFERENT APPROACHES OF DATA-ANALYSIS

The aim of this study was to demonstrate that ambiant air can be used for rapid discrimination of food products using six different semicond uctor gas sensors. The headspace of six groups of dried-sausages were analyzed during 3 days, to test the time-robustness of the system. We demonstrated that product classification must be done using the inform ation contained in the dynamic part of the signals. Analysis of varian ce showed that the level of the steady-state is influenced by the day of analysis. To reduce the number of variables without loosing the dis criminant information we used non-linear modeling and backward selecti on of variables. Successfull classification was obtained with both dat a processing (100% and 97%) and results were validated by cross-valida tion. We demonstrated that using ambiant air anti a limited number of gas sensors, it is possible to perform a vapid classification of foods tuffs.