A Novel Approach to Information Fusion for Multicomponent Gas Quantitative Analysis with Fourier Transform Infrared Spectrometer

摘要

Fourier transform infrared Spectroscopy (FT-IR) has been widely used for multi-component gas analysis in many fields because of its many advantages. In order to improve analysis accuracy, information fusion is sometimes used. In this paper, "PRESS method", an information fusion method used currently for multi-component gas analysis with FT-IR spectrometer, was introduced at first. Then, its logicality in application was analyzed. And a novel information fusion method based on noise level was proposed. Compared with current used approach, noise level of transmittance rather than the testing samples' residual errors was used to determine the models' fusion weights. Finally, several sets of testing samples were analyzed to verify the method proposed here. The analysis results showed that the mean relative error or absolute error of the analysis results obtained with the novel approach were no more than that calculated with "PRESS method". Therefore, the novel fusion method is a well-proven and practical method.