基于神经网络的矿尘浓度传感器输出特性研究

Study of Mine Dust Density Sensor Output Characteristic Based on Artificial Neutral Network

通过对矿尘浓度传感器输出特性研究,提出运用人工神经网络的方法对其输出数值进行校正。通过样本数值对神经网络进行训练,得到非线性校正模型。实现将传感器原始显示数值输入校正模型,经过模型处理后得到一个新的输出数据,该数据比原始数据更精确。做到在不改变传感器硬件结构的基础上实现对输出数据的校正,以提高显示数值的精度。

The characteristic of mine dust density sensor output was studied. It puts forward a proposal that use the algorithm of artificial neutral network to calibrate the output date of sensor. Choose an appropriate kind of network model first and then use some sample value to train the neutral network to obtain a nonlinear corrected model. Put the primitive displayed value of sensor into the model. It will process this value by the model and then give a corresponding output value. Those output value are closer to the real mine dust density value than the primary value. So it will calibrate the output value and enhance the precision of displayed value without changing the hardware architecture of sensor.