Gas identification system using graded temperature sensor and neural net interpretation

Abstract: "An innovative approach is described for enhancing the selectivity of an integrated multi-element thick film gas sensor. A temperature gradient maintained along the sensor surface induces spatial sensitivity and selectivity gradients. The response map of the sensor is examined in the vapor of several organic compounds and their binary mixtures. Subtle but recognizable and separable signatures are observed. Emphasis is on identification, since quantitation of identified mixtures is straightforward. An effective and robust classification technique using a neural network trained via the back propagation method is described."