基于神经网络技术提高红外气体分析器的选择性

Study on Increasing Selectivity Index of Infrared Gas Analyzer with Neural Network

应用前馈神经网络算法消除非目标参量对主传感器的干扰,从而提高了红外气体分析器的选择性.以检测甲烷为例,在干扰气体乙烯的体积分数变化了7600×10-6时,经神经网络融合处理后,分析器的选择性系数从3.17提高到422,主传感器输出的引用误差从58%降为0.65%,实现了对甲烷的准确识别.实验结果表明,该方法具有实际应用前景.

To increase the selectivity of infrared gas analyzer, a method is proposed by applying the back-propagation neural network can eliminate the cross-interference between non-aim parameter and aim parameter adopting nonlinear approach ability and generalized function of neural network, so the selectivity is improved appreciably. The analysis of typical examples shows that the selectivity magnitude of IGA can be increased from 3.17 to 422 with the aid of this method and consequently gas is detected accurately. The method provides an effective approach for improving selectivity of IGA and exhibits practical prospect.