基于改进BP神经网络的矿井CO检测方法的研究

Research on Detection of CO in Mine Based on Improved BP Neural Network

采用催化传感器和电化学式气体传感器配合使用的传感器阵列。为了解决2种传感器对矿井CO和CH4气体的交叉敏感问题,提出了一种基于改进BP神经网络的矿井CO检测方法。通过MATLAB仿真可以看出,基于神经网络的传感器阵列方法可以明显提高CO检测精度。实际输出值和期望输出的绝对误差平均值为3.43 ppm,相对误差平均值为1.43%。

In order to increase the precision of CO detection, electrochemical sensor was used with a catalytic sensor. Because there was the cross sensitivity of two sensors to CO and methane gas in mine, a kind of CO detection method based on improved BP neural network is presented. According to the results simulated by MATLAB, the precision based on BP neural network and gas sensor array can increase the measurement precision of CO. The average absolute error of actual output and desired output is 3.43ppm, and the average relative error is 1.43%.