基于支持向量机的红外光谱天然气分析系统

针对天然气组分浓度分析过程中海量训练样本无法实现、天然气组分气体的主次特征吸收谱线重叠严重、气相色谱分析法分析速度慢、无法进行实时在线分析等问题，将支持向量机（SVM）这一新的信息处理方法与红外光谱分析法结合，应用于天然气组分浓度分析中，利用SVM处理小样本的能力解决海量样本问题；通过建立反映天然气光谱数据样本与天然气组分浓度关系的SVM校正模型，解决天然气组分气体主、次特征吸收谱线重叠严重的问题；利用红外光谱分析速度快的优点，缩短分析时间。设计研制了基于SVM和红外光谱的天然气分析系统。该系统使用傅立叶红外光谱仪获取天然气红外光谱数据样本，对红外光谱数据样本进行数据预处理后，通过SVM校正模型进行计算分析，得出天然气组分浓度。实验结果表明，该方法的最大偏差为3.95%，与气相色谱分析法相比，具有分析速度快、可实时在线分析等优点。

The Infrared Spectrum Natural Gas Composition Analysis System Based on Support Vector Machine(SVM)

One of the most important indexes for the measurement of natural gas property is its composition analysis. However, there are still various problems in natural gas composition analysis as follows: (1) great huge numbers of gas samples difficult and impossible to handle; (2) both primary and secondary characteristic absorption lines seriously overlapped; (3) the previous gas phase chromatographic analysis in common use works too slowly to perform real time analysis on line. Combined the infrared spectrum analysis with the Support Vector Machine (SVM), a new data processing method, this study designed a completely new gas composition analysis system. The working principle of this new system was that huge number of samples could be dealt with by the SVM; the overlapping absorption lines could be solved by the SVM correction model established on the relationship between light spectrum data and gas composition concentration; and the infrared spectrum could shorten the analysis time because of its high speed. Compared to the previously used chromatographic analysis, this new analysis system was proved after practical uses to be characterized by advantages such as (1) being able to realize real time analysis on line, fit for the field use; (2) being able to perform analysis off line also, fit for the lab test; (3) storing those light spectrum data into database for the convenience of analysis from historic data; and (4) displaying data by graphs, tables, and spectrograms and preserving the analysis results as well.