| Raman spectral technology, with its fast speed of analysis, low sample requirements and high accuracy of prediction, has been increasingly used in the field of on-line analysis. To meet the need of a p-xylene (PX) adsorption separation process, this thesis applies a Raman spectral decomposition method to component analysis. The method has been proved that it can be used for the quantitative analysis of the circulating fluid and the feed of the PX separation process. Furthermore, an on-line analyzer for the feed is developed and has been successfully applied in an industrial PX unit. This thesis includes the following contents.1) A Raman spectral decomposition method for on-line quantitative analysis of circulating fluid in the above process is proposed. The components of circulating fluid includes methylbenzene (referred to as MB), ethylbenzene (EB), p-xylene, m-xylene (MX), o-xylene (OX), p-diethylbenzene (PDEB). Because of the wide change range of each component concentration in the circulating fluid, the Raman spectral technology is most appropriate for on-line analysis. Each component in the circulating fluid has its own special Raman peaks, but the Raman spectra of pure components overlap together. The common used method based on the peak area and concentration is no longer feasible; a recent method, indirect hard modeling, needs complex calculation. Therefore, this thesis proposes a Raman spectral decomposition algorithm. It is assumed that a spectrum of mixture can be decomposed into the sum of pure component spectra, whose decomposition coefficients are determined by each component concentration for the mixture. To verify the effectiveness of this algorithm,10 typical samples of circulating fluid are formulated. Experimental results draw the following three conclusions. First of all, a spectrum of mixture can be decomposed into the sum of pure component spectra and the average deviation of the error spectrum is less than 0.05 to the maximum of the original spectrum. Secondly, there exists linear relationship between relative decomposition coefficients and relative concentration to a reference component. Thirdly, the model built by the algorithm can precisely predict the composition concentration of testing samples. The standard errors of prediction (SEP) for the concentration of MB, EB, PX, MX, OX and PDEB are 0.30%,0.09%,0.56%,0.38%,0.37% and 0.54% respectively.2) An on-line Raman analysis system is developed for the feed of an adsorption separation process. The quantitative analysis model is built based on the above spectral decomposition algorithm, and the SEP for the concentration of EB, PX, MX and OX in the feed are 0.21%,0.12%,0.16% and 0.22% respectively. Then, a novel quantitative analysis module is developed based on the existing framework of on-line analysis software. Combined with the existing hardware system, an on-line Raman analysis system for the feed is developed.3) The above system has been applied to the on-line analysis for the feed of the separation process. The system has been working for more than 1 year. The running results show that the system has good repeatability, whose signal to noise ratio of the measured spectrum is about 283:1, and the standard errors of EB, PX, MX and OX are 0.03%,0.02%,0.05% and 0.06% respectively. Compared to the off-line gas chromatographic analysis data, the SEP of the on-line analyzer for the concentration of EB, PX, MX and OX are 0.33%、0.11%、0.30% and 0.20% respectively. The Raman system can meet the daily need of industrial processes because of its high reliability and accurate prediction. |
PDF Full Text Request