Application of Multivariate Analysis to Gas Detection with Semiconductor Sensors

Abstract Today, tin dioxide sensors are largely used in many applications. Their main disadvantage is their poor selectivity, especially when detecting one gas in a gas mixture. A special automatic testing bench has been developed to measure sensor conductances at different temperatures and under controlled gas concentration (ethanol, carbon monoxide or methane diluted in air). The partial least square method use these responses. Thanks to about fifty experiments for calibration and prediction steps, it allows the determination of three gas concentrations in the range of 10 to 1000 ppm1. The performances depend on the stability and the reproducibility of sensors.