基于SVM的混合气体红外光谱分析关键技术研究

为了解决海量混合气体光谱数据样本无法获取、混合气体组分气体特征吸收谱线重叠、混合气体组分浓度分布的随意性等问题,将支持向量机用于混合气体红外光谱分析中.提出了光谱数据样本特征选择、数据预处理、SVM校正模型参量优化及层次式混合气体光谱分析结构等关键技术.实验分析了上述4项关键技术对分析结果的影响.实验结果显示,采用关键技术的混合气体组分浓度分析的最大绝对误差为2.93%,最大平均绝对误差为0.73%.

Key Technologies Research of Mixture Gas Infrared Spectrum Analysis Based on SVM

In order to solve the difficulties that mass mixture gas spectrum data samples cannot be actually obtain by mixture gas component characteristic absorption lines seriously overlap, component gas concentration distribution is optional, support vector machine was used in mixture gas infrared spectrum analysis. The spectrum data sample characteristic choice, the data pretreatment, and the SVM Calibration model parameter optimization and mixture gas spectral analysis structure based on level were summarized and proposed. The influence between analysis result and the above 4 key technologies was analyzed by the method of experimental. The experimental result shows that the mixture gas component concentration analysis based on the key technologies max absolute error is 2.93%; the mean absolute error is 0.73%.