基于自适应变异粒子群算法的混合核ε-SVM在混合气体定量分析中的应用

摘要

针对利用不分光红外吸收法（NDIR）的多组分气体传感器对汽车尾气进行同时测量时，红外光谱特征吸收谱线重叠较为严重，所测气体浓度是交叉吸收干扰后的结果，造成测量误差大，分析精度低的问题，提出了一种自适应变异粒子群算法的混合核ε-SVM方法，建立三组分混合气体定量分析模型，已消除混合气体之间相互干扰产生的误差问题。实验中，采集CO2、CO、C3H8的浓度信号，作为模型输入，通过模型回归分析，得到对应的混合气体组分浓度，通过实验数据对模型性能进行分析，结果表明，该模型的平均误差相比于传统模型明显减低，取得较好的精度。%Due to the simultaneous measurements of automobile exhaust gas by using the multi-component gases sensor based on the dispersion of light infrared method(NDIR),the text is the result of the cross absorption and in⁃terference,resulting in the large measurement error and low accuracy. To solve this problem,a kind of mixed kernel functionε-SVM based on adaptive mutation particle swarm optimization algorithm is put forword to establish a mod⁃el for the quantitative analysis of three component mixture gases. Collect the concentration signals of CO 2,CO and C3H8 as the model inputs,through the model regression analysis,the outputs are corresponding mixed gases concen⁃trations. thus,the problem of mutual interference can be solved. Finally,the performance of the model is analyzed through the experimental data,the result shows that the average error of the model is significantly reduced com⁃pared to the traditional model.