Using sensor arrays and pattern recognition to identify organic compounds

The recent development of chemical sensor arrays promises to solve some of the industrial, military, and domestic problems of gas detection and monitoring, but there are many problem areas to be addressed before these types of devices become readily available. The work presented here represents initial expeditions into a new, hybrid discipline for chemical analysis, which combines materials science, chemical sensing techniques, and the application of pattern recognition for automatic information extraction. Specifically, two kinds of chemical sensor array design and construction are discussed. The nature of the outputs from these sensor arrays is examined for qualities such as information content, stability, reliability, and accuracy. Several methods of pattern recognition are explored for their ability to classify sensor array information. Preliminary results indicate much promise in the use of neural networks for the analysis of mixtures, which is a vexing problem. It is found that the most appropriate pattern recognition technique depends to a large degree on the complexity of the sensing problem.