An Electronic Nose For Monitoring Hazardous Chemical Goods

As the continuous improvement of petroleum, chemical industry, automobile, the requirements of hazardous chemicals are increasing. That makes the production, storage and using of hazardous chemicals as the important part of the life cycal of hazardous chemicals. Based on the combustible gas, this paper studies the portable electronic nose (Enose) system for monitoring hazardous chemicals. A new method of sensor array optimization and the new hybrid architecture for gas mixture recognition have been proposed for Enose. The topic of this thesis is part of the national's 863 program funded project named as "Hazardous chemicals monitoring sensor signal processing and network interface technology" (No.2006AA040102).Firstly, take the binary gas mixture of methane and ethylene as the identify object, choose the Figaro gas sensors as the sensor array, variance analysis and principal component analysis are combined to optimize the sensor array of Enose system for decreasing the hardware costs, but this method has some limitations. In order to make a good suitable for different conditions, the principal component analysis was used to optimize the sensor arrays, which make full use of the relation between sensors and the contribution of sensor. The identification effect of the pre-optimized array and the optimized array is compared, the optimized array occupies less gas sensor in the array, reduces the hard ware and algorithms costs, and improves the network convergence speed, and meanwhile, it has the similar gas identification properties as the original large sensor array.Secondly, in oder to improve the accuracy of gas mixture recognizing, this paper employ the characteristic of principal component analysis (PCA), independent component analysis (ICA) and back propagation neural network (BPNN) sub networks, proposing a new hybrid architecture with two main stages for gas mixture recognizing. The first stages was used to confirm the type of gas by using PCA and BPNN, the second stage is used to calculate gas concentration which base on ICA and BPNN sub networks. The proposed hybrid architecture network is used to identify the binary gas mixture of methane and ethylene, and the identification result are remarkably better than other signal processing method for gas sensor array. Besides, the hybrid architecture network is used to identify the ternary gas mixture of methane, ethylene and propane and obtain the good recognization effects.Last, a new Enose system is set up for the Identification of Binary Gas Mixtures based on the mentioned array optimization technique and hybrid architecture network. The data real-time acquisition, signal process, display and control are realized by MSP430. And also, the wireless transmission platform is founded between Enose and computer, which can monitor a lot of Enose simultaneously. The master system Human-Machine Interface and data storage is accomplished by using LabVIEW8.6. Through analyzing the experimental result, the Enose evaluate the leakage of hazardous chemicals correcctly. What's more, the Enose is installed at Dalian guangming specialty Gas Company to monitoring the state of hazardous chemicals, the system works well after debugging in the job site.