SVM在混合气体光谱分析中的应用

针对混合气体多维光谱定性和定量分析中组分气体吸收谱线重叠、定性和定量分析无法使用同一方法、训练样本数目有限及输入光谱的维数等难题,将支持向量机应用于混合气体多维光谱分析中,利用核函数将重叠的多维光谱数据进行高维空间变换后求得SVM回归模型,可同时进行混合气体组分浓度的定量分析和组分种类的定性分析.在混合气体为天然气的组分浓度和组分种类分析实验中,组分浓度的最大误差为1.74%;组分种类的识别准确率大于94.87%,效果明显优于其他方法,为混合气体多维光谱分析提供了新的方法.

Application of SVM to mixed gas spectrum analysis

Multidimensional spectrum qualitative and quantitative analyses of mixed gas have the following difficulties: overlapping of component gas feature spectrums,qualitative analysis can't use the same method with quantitative analysis,limit number of training samples and dimension of input spectrum.A novel method for mixed gas analysis is presented which is based on support vector machine.After transformation of kernel function,overlapping multidimensional spectrum is mapped into high dimension space,so that mixed gas analysis can be carried out based on SVM.The quantitative analyses of component concentration and the qualitative analyses of gas composition can be carried out.The proposed method was applied in the analysis of natural gas.The experimental results show that the maximum deviation of the component concentration is 1.74% and the component recognition accuracy is more than 94.87%.The proposed method has evident advantage over classical methods such as ANN.