THICK-FILM PELLISTOR ARRAY WITH A NEURAL-NETWORK POSTTREATMENT

In pellistor gas sensors, the heat exhaust produced by the catalytic c ombustion of reducing gases increases the temperature of the device. A typical pellistor consists of a platinum wire supported in an alumina bead impregnated with a finely dispersed noble metal like palladium. The platinum wire serves as heater of the bead to its operating temper ature and as a thermometer. In reality, the temperature measured by th e resistance of the Pt wire is compared to that of a reference element which has a similar structure but without any catalytic activity. No selectivity of such a device has to be expected since the catalytic co mbustion of any combustible gas will lead to a temperature increase of the device. In order to try to achieve selectivity to methane, we hav e in a first step exploited the differential activity of palladium and platinum by using two screen-printed pellistors, one based on Pd and the other on Pt. At around 400 degrees C, all reducing gases including methane are oxidized by Pd whereas Pt oxidized all gases except metha ne, In order to extend the recognition process to combustible gases ot her than methane, that is to propane, and ethanol vapour, a small arra y of four pellistors with various percentages of Pd and Pt has been el aborated with thick film technology, which is very valuable for realiz ing series of similar sensors, required in arrays. The four microcalor imetric sensors are exposed to various gases and various concentration values. A recognition of methane, propane, and ethanol is obtained by neural network techniques. The network consists of three layers: an i nput layer; a hidden layer; and an output layer which permits gas iden tification. Back-propagation is used as the learning algorithm. In thi s case, the selectivity of the system is demonstrated.