Raman Spectral Decomposition Based On Multivariate Curve Resolution And Lorentz Function Constraint

Liu Wei,Dai Lian-Kui

DOI: https://doi.org/10.19756/j.issn.0253-3820.171505

IF: 1.193

2019-01-01

Chinese Journal of Analytical Chemistry

Abstract:To overcome the shortcoming of traditional Raman spectral decomposition method that requires a priori information of pure spectra, a new Raman spectral decomposition algorithm based on multivariate curve resolution and Lorentz function constraint was proposed. This method was based on the bilinear model for mixture spectra, and only the mixture spectral matrix was needed to perform the analysis. It first decomposes the mixture spectra with principal component resolution, and then the evolving factor analysis was performed to find the distribution of the mixture spectral eigenvalues evolves along with the pixel direction, so that it could determine the number of components in the mixture, the range of their peaks and the initial estimation 14 their spectra. Finally an alternative least squares method with Lorentz function constraint was proposed to optimize the pure spectra, during the process, it first fits the pure spectra with Lorentz function, and then based on the fitted spectra, it regresses the spectra and the coefficient matrix with non-negative alternative least squares method. The above optimization process was repeatedly performed until the fit error for the mixture spectra was within the threshold. In the experimental part, the proposed algorithm was adopted to decompose the Raman spectra of the circulating fluid in an adsorption tower of a p-xylene unit, with which the relative pure spectra was obtained in the mixture fluid. Finally, with the repeated quantitative analysis experiments for its main component of p-xylene, we prove the effectiveness for the proposed algorithm.