用神经网络提高SO_2传感器的检测速度

Enhancement of detection velocity of SO_2 sensor by neural network

制作了新型 IDC结构聚苯胺膜 SO2传感器，测得不同 SO2浓度的电导响应与恢复时间曲 线。从这些气敏特性曲线出发，以电导响应曲线的斜率作为网络的输入，对应的 SO2浓度作为 输出，建立了 BP网络预测推理模式，对四组数据的预测结果表明精度较高 (误差小于 3％），具 有很好的预测能力。这种方法不同于传统的标定方法，后者需要 4min才能达到稳定的响应，神经 网络样本检测不需要达到稳定的响应就可以预测 SO2的浓度，从而大大缩短了结果的响应时间 （缩短了 75％）。

： A kind of new sensor of simple structure, easy manufacture and good performance,IDC struc- ture with polyanline film has been studied, the gas-sensitive characteristics of polyanline film have been tested and the response and comeback curves of current conductance for different SO2 contents have been obtained. A forecasting model of BP neural network starting from gas-sensitive characteristics of polyan- line film with the curve of response current conductance as input nodes of network and the corresponding SO2 concentration as the output of the network has been advanced. By means of the network the four group data have been tested.The results show that the model possess good forecast characteristics with errors smal－ler than 3％ . This method is different from traditional calibration, the latter requires stable detection for4min, while neural network signal pattern detection does not necessarily require a steady response. And as a result, the response time can be greatly reduced (shortened by 75％ ).