Identification of organic solvents by a virtual multisensor system with hierarchical classification

To improve the selectivity of semiconductor gas sensors, temperature modulation is often used. We present a study showing the potential of this technique with information gained comparable to multisensor systems. A dynamic operating mode coupled with a low-complexity evaluation strategy allows the identification of six organic solvent vapors over a wide concentration range (2-200 ppm) with a single sensor, for example, for leak detection systems. The system features low false alarm rates; in addition, interference by other gases, such as CO or NO/sub 2/, can be suppressed. For even higher identification power, switching on-line between different temperature cycles was studied, which provides better information for critical decisions.