Multi-sensor Data Fusion Research Based On SF₆ Decomposition Product

As an important branch of sensor technology, the gas sensor has more important practical significance in the study of the data fusion of the gas sensor in the historical background of everything. Cross interference of gas sensors, resulting in a single gas sensor for quantitative qualitative detection of mixed gases distortion. The pattern recognition method of the sensor array and the modern control theory can eliminate the cross interference between the sensor array and the modern control theory, and realize the qualitative and quantitative accurate detection of the mixed gas.In this paper, the multi-sensor data fusion algorithm is studied, and the gas sensor array is used to detect the gas mixture. The principle of electrochemical gas sensor is introduced, and the causes of cross disturbing of electrochemical gas sensor are analyzed in terms of sensitivity, selectivity, temperature and humidity response and detection principle. In the GIS fault diagnosis, the SF₆ decomposition product mixture gas?CO, SO₂, H₂ S, H₂? is detected as an example, the improved linear neural network algorithm can achieve very good detection results. Thesis in turn tells the development status of gas detection technology at home and abroad, through the multiple regression analysis, BP neural network algorithm research, theoretical analysis of the advantages and disadvantages of them, finally carries on the introduction to the linear neural network, which is the combination of the first two algorithms improve the linear neural network, design three layer data structure learning of multi sensor data fusion. According to different kinds of electrochemical gas sensors, the signal conditioning circuit of the three electrode and the four electrode gas sensor was designed. And continue to introduce the SF₆ based LabVIEW decomposition products to monitor the experimental system of the hardware foundation.By building a SF₆ based LabVIEW decomposition products monitoring system, using RA601 four in a mixed gas detection chamber, to achieve the SF₆ decomposition products CO, H₂ S, SO₂, H₂ testing experiments. The experimental system using Ni pci-6251 signal acquisition card of sensor signal conditioning circuit signal in signal acquisition and PC using LabVIEW modular calibration display and algorithm matlab programming realization pattern recognition are combined to realize PC software. PC software mainly realizes the function of data storage, human-computer interaction, different pattern recognition algorithms and so on.Finally, through the SF₆ decomposition products monitoring experimental system, respectively by using multiple regression analysis, BP neural network and improved neural network algorithm of CO, H₂ S, SO₂, consisting of two H₂ gas mixtures were test. The results of the experiments show that using BP neural network algorithm in PC to a good data fusion accuracy, transplantation and improvement of linear neural network in the embedded instrument testing, in degrees are better than the front two common existing data fusion algorithm experimental results and algorithm complex was. The system is able to realize the analysis of composition and content of multi mixture gases. Compared with the traditional detection method, the system realizes the fast and accurate on-line monitoring of mixed gas.The above research work has a good theoretical guidance and reference value for the SF₆ fault diagnosis of GIS decomposition products.