基于BP神经网络的气体传感器模型预测分析研究  

Research on Gas Sensor Model Prediction and Analysis Based on BP Neural Network

作　　者：盛佳慧 蔡勇[2] 

机构地区：[1]湖南大学物理与微电子科学学院,湖南长沙410082 [2]湖南大学半导体学院(集成电路学院),湖南长沙410082

出　　处：《工业控制计算机》2025年第2期19-20,24,共3页Industrial Control Computer

摘　　要：针对新型传感器研发中,电阻式半导体气体传感器电阻跨度大、输出特性不明确、测试精度要求高等问题,设计了一种高精度、宽量程传感器信号采集单元获取数据,引入BP神经网络进行数据拟合,并通过SA算法优化神经网络建立气体传感器模型,以MQ-8氢气传感器为例进行实验测试分析,结果表明测试样本的决定系数可达0.99985,整体平均误差为0.6%,实现了对气体浓度的高精度预测,为传感器研发中的数据分析提供了支撑。With the goal of addressing the issues with the resistive semiconductor gas sensor's large resistance span,ambiguous output characteristics,and high test accuracy in the process of researching and developing new sensors,this paper designs a wide-ranging,high-precision sensor signal acquisition unit to collect data,introduces BP neural network for data fitting,and then optimizes the neural network using the SA algorithm to create a gas sensor model.The experimental test analysis,using the MQ-8 hydrogen sensor as an example,demonstrates that the test samples'determination coefficients can reach 0.99985 and that the overall average error is 0.6%.This enables the high-precision prediction of gas concentration and supports data analysis in sensor development.

关 键 词：气体传感器 数据采集 BP神经网络 SA算法 模型预测 

分 类 号：TP2[自动化与计算机技术—检测技术与自动化装置]