Nanometer ZnO Gas Sensor Based On The BP Neural Network For Gas Recognition

As the development of technology and enhancement of environmental protection, the requirements of the detection equipment, which has a good performance in mixture gases, are highly desired. But existing detection equipment cannot meet them. Therefore, the electronic nose based on gas sensor arrays and pattern recognition, such as artificial neural network, has become the one of the new tendency of mixture gases detection. In the present work, pure ZnO nanowires and Al, Ag, Ni doped ZnO nanowires are synthesized by improved thermal evaporation, and based on the difference sensitivities of these nanowires to the gas, the gas sensor arrays are built.. BP(Back Propagation) neural network are employed as a medium to identify the component and relative concentration of compound gases. The sensitive mechanism of ZnO nanowires has been discussed. The results are as following:1. The present work has prepared ZnO nanowires with high performance. There is no remarkable agglomeration in Ag doped ZnO nanowires, and each of these nanowires is same in size along the length. With increasing of the surface-area-to-volume ratio, the performance of these sensors become better and better.2. In the same testing parameters, the detection sensitivity of the ZnO nanowires doped with Ag is better than that from others.In these mixture gas(CO, H2, CH4), the best gas sensitivity property of ZnO nanowires doped with Ag is 32.060.3. Based on the difference sensitivities of these nanowires, the gas sensor arrays are built, and that will have the best output signal.4. In the present work, BP neural network can more effectively solve the problem such as cross-sensitivity of the ZnO gas sensors, helping to imrove the accuracy of mixture gases detection.The error of the network is less than 0.1%..5. The results show that it can identify the compnent and relative concentration of mixture gas(CO, H2, CH4) accurately by treating the output of gas sensor arrays and training it in the BP nueural network. In the present work the successful rate is 100%...