Based on the temperature modulation, the quantification of CO and H2 is carried out to solve the problem of poor selectivity for metal oxide semiconductor gas sensors. In this work, the effect of the synergistic modification as Cu2+ doping and Au loading of SnO2 gas-sensing material on sensitivity and selectivity of CO and H2 was investigated. The stable resistance value as the static characteristics and the following-up time as the dynamic characteristics were used by an algorithm of BP neural network to establish a model for quantifying CO and H2 concentrations. The result shows that the operating temperature were reduced to 250 °C from 350 °C by Au loading of 3.0 mol%Cu2+ doping SnO2 gas-sensing material, the response value was increased to 58.3 from 9.3 at 1000 ppm of CO, and the selectivity (RCO/RH2) was increased to 1.55 from 0.33 due to the grain size decrease of SnO2 gas-sensing material and the catalytic effect of Au. And these data including the resistance value and the following-up time were used to obtain WijT, Vjq, ΘjT and ΓqT matrices including the weight values wij, vjq and threshold values θj, γq for the model by training through the algorithm of BP neural network the weight values and threshold values matrices, where the relative error of the quantitative results of CO and H2 concentrations was less than 2.0%, which provides a new way for the quantification of mixed gas.
Read full abstract