A novel quantitative analysis method of multi-component mixture gas concentration based on support vector machine (SVM) and spectroscopy is proposed. Through transformation of the kernel function, the seriously overlapped and nonlinear spectrum data are transformed in high-dimensional space, but the high-dimensional data can be processed in the original space. Some factors, such as kernel function, range of the wavelength, and penalty coefficient, are discussed. This method is applied to the quantitative analysis of natural gas components concentration, and the component concentration maximal deviation is 2.28%.
© 2005 Chinese Optics LettersPDF Article