基于传感器阵列及神经网络算法的NH_(3)和NO_(2)混合气体体积分数识别

Volume fraction identification of NH_(3) and NO_(2) mixture gases based on sensor array and neural network algorithm

针对电阻型气体传感器具有的交叉敏感性,开发了基于WO_(3)传感器阵列及神经网络算法的NH_(3),NO_(2)混合气体体积分数预测技术。采用火焰合成法合成了La掺杂的WO_(3)敏感材料并制备了气体传感器,与商用MQ—137电阻型气体传感器组成阵列。通过提取特征值、神经网络训练,构建传感器阵列输出与气体体积分数的映射模型,并使用该模型由传感器阵列的响应结果对NH_(3),NO_(2)混合气体进行体积分数预测。实验结果表明:经训练后的神经网络能对NH_(3),NO_(2)混合气体中各组分体积分数进行有效预测,平均预测误差分别为3.64%和2.48%。本文所开发的传感器阵列及神经网络算法有效避免了电阻型传感器选择性差的局限,实现了对NH_(3)和NO_(2)混合气体的高效识别和体积分数测量。

Aiming at the cross sensitivity of resistive gas sensors,volume fraction prediction technology of NH_(3) and NO_(2) mixture gases based on WO_(3) sensor array and neural network algorithm is developed.La-doped WO_(3) sensitive material synthesized by flame synthesis method and gas sensor is prepared,and constituent array with commercial MQ—137 resistive gas sensor.By extracting eigen value,neural network training,construct mapping model for sensor array output and gas volume fraction,and use this model to predict volume fraction of mixed gas of NH_(3),NO_(2) by the response result of sensor array.The experimental results illustrate that trained neural network can effectively predict the volume fraction of each component of mixed gas of NH_(3),NO_(2),the average prediction errors are 3.64% and 2.48%,respectively.The developed sensor array and neural network algorithm effectively avoid limitation of poor selectivity of resistive sensor,realize efficient identification and volume fraction measurement of mixed gas of NH_(3) and NO_(2).