Electronic Nose And Its Application

An electronic nose (EN) is an instrument, which comprises an array of chemical sensors with partial specificity and an appropriate pattern-recognition system,capable of recognizing simple or complex odors. EN finds promising applications in food industry aviation and spaceflight, environment inspection; thus this research area is of great importance and become an international research focus.In this dissertation, the principle and the strucfure of the EN were analyzed by comparing with those of olfaction organism. The construction of sensor array, the design of signal pretreatment circuit and the choice of pattern-recognition (PARC)methods were discussed. A precise mathematic modeling of sensor array was established for the first time based on random process theory The effects of the dimension of the sensor array, signal condition and the resolution of A/D converter upon sensor array performance were studied.Based on the work thereinbefore, an electronic nose system was designed and prepared using single chip microprocessor (SCM), which has been used successfully to classify and recognize three kinds of gasoline and five kinds of distilled spirit within 20min with injection of less than l00μl gaseous sample.The EN hardware system is an SCM application system, utilizing all of interior resource and lots of exterior interface techniques of the SCM. A new signal condition circuit was used so that sensitivity can be held constant over wider dynamic range of the gas sensor resistance values. A new flowing injection system was used, keeping the reproducibility of the sample injection, which facilitates the record and utility of dynamic response of sensor arrayThe EN software system comprises software for SCM and software for PC. The former, written in C language, is responsible for the system controlling, processing,analyzing and displaying when EN works in offline mode, while it is responsible for sampling and data transfer when EN works in online mode. The latter, written in Visual C++ language, provides a friendly interface when EN works in online mode,through which user can control the system easily realizing data process and results display in real time, or modifying database of the SCM system.EN feature extraction and PARC algorithms were studied. Several methods of feature extraction were used, which facilitate classification and recognition of distilled spirit and gasoline. A new feature extraction method is adopted in classification and recognition of distilled spirit using principle component analysis (PCA), which provides a solution to the construction of transform matrix for dimensionality reduction.