| The rapid non-destructive detection of trace components in complex systems is one of the challenges in analytical chemistry. Currently, Raman spectroscopy is one of the popular of rapid trace analyses due to its advantages of rapidness and non-destruction. However, the model of Raman spectra collected from complex system is not only determined by the concentrations of analyses, but also influenced by other factors including the matrix interference, nonlinear effects and sample fluctuation presented in complex systems. In addition, Raman spectroscopy is also affected by huge interference of fluorescent signal. These factors make it far more difficult to analysis Raman spectrum signal than other spectroscopy, which block the development of quantitative analysis techniques of Raman spectroscopy in recent years. Therefore, it is urgently needed to develop new algorithms to accurately extract the Raman spectra of weak nonlinear intrinsic information to achieve rapid non-destructive testing of complex multi-component systems.In this paper, we proposed a new method for developing multi-scale modeling, and tried to extract the information of the complex changes in the Raman spectrum of the trace components accurately, which solved the problems in Raman spectroscopy modeling. To verify the effectiveness of the algorithm, we applied the algorithm to the conventional serum component analysis and the model transfer of corn spectrum. The results show that multi-scale modeling approach significantly improves the accuracy and robustness of the model. This method is not only applicable to data processing of Raman spectroscopy, but also to the data processing of near-infrared spectroscopy. It is indicated that the method possesses good adaptability. The thesis is divided into the following four sections:1. Based on the summary of the researches domestic and overseas and the analysis of complex systems, the importance of quantitative analysis methods was elaborated. Raman spectroscopy and multi-scale modeling techniques were reviewed, and then the advantages of combining both technologies were proposed.2. We analyzed the basic concepts and principles of multi-scale modeling techniques in detail, and deeply discussed the signal characteristic of Raman spectra. Then we defined the joint point of the multi-scale modeling and Raman signal, and then theoretically elaborated multi-scale modeling of specific processes. At last, we identified their weight in the Raman spectroscopic analysis, thus providing theoretical support for subsequent applications.3. In order to investigate the role of multi-scale modeling in improving the model itself, we used the laboratory-constructed Raman spectroscopy platform to collect Raman spectroscopy of serum. Then we applied the method of multi-scale modeling for the detection of 10 component in serum for quantitative analysis, and then the modeling results were evaluated.4. In order to clarify the multi-scale modeling applications in the model transfer, we applied this new method to modeling transfer among different corn spectrum collected from three instruments, and then the results were evaluated.This study showed that the multi-scale modeling approach can take advantage of the spectral signal of time / frequency characteristics of multi-scale, and integrate data preprocessing and multivariate calibration operations in order to avoid the loss of important information, with high adaptive features. In our study, we successfully applied it to the conventional serum component analysis and the modeling transfer, obtained good results and widened its application. We will combine it with Raman spectroscopy analysis technology, which is expected to provide a new method for nondestructive rapid detection of complex systems.. |
PDF Full Text Request