An intelligent gas sensor system for the identification of hazardous airborne compounds using an array of semiconductor gas sensors and Kohonen feature map neural networks

An intelligent gas sensor system for identification and quantification of hazardous airborne compounds has been developed. As gas detecting devices semiconductor gas sensors with partially overlapping selectivity have been used. Because of the low selectivity of semiconductor gas sensors, one gas sensor alone can give no accurate statement concerning type of gas and the actual concentration. Applying an array of semiconductor gas sensors in combination with Kohonen feature map (KFM) neural networks, unknown species of gases can be identified and quantified if the gas sensor system has been calibrated with this sort of gas earlier. The influence of different network parameters, e.g. the number of nodes in the network or the number of pattern vectors used to train the KFM have been studied. It has been found, that the KFM is able to identify all compounds which have been used for calibrating the gas sensor array and for training the KFM