Raman Spectrum On-line Monitoring In Aspirin Synthesis Process

Strong fluorescence interference and baseline drift phenomenon exist in Ramanspectra, especially in Raman information of the flowwing material in the reaction systemso application of Raman analysis technique in the chemical synthesis process on-linemonitoring is not easy to achieve. Taking aspirin synthesis system as the research objectby using multivariate statistical methods, quantitative analysis of a variety of material inthe reaction system was performed to realize the process monitoring. The research resultsmainly has the following several aspects:(1) Using Savitzky-Golay convolution smoothing first-order derivation for locatinglocal extrema, piecewise linear fitting was proposed to eliminate baseline drift in Ramanspectra. Compared methods of piecewise linear fitting, polynomial fitting detrending andwavelet high-pass filtering algorithm, correlation coefficient between corrected spectraby the three algorithms and original spectra and the was0.9377,0.8604and0.8691respectively. Piecewise linear fitting method of baseline correction based on subspacevector angle determination is better than the former.(2) Obtain Raman information of flowwing samples in eaction system, the on-lineRaman spectra exist baseline drift, baseline drift cover up of spectrum differences of thebeing measured samples, which affect determinate the reaction end-point; Baselinecorrection for spectrum processing, using subspace pattern recognition principle, selectingthe Raman spectra of reactants (salicylic acid, acetic anhydride) and catalyst (sulfamicacid) as the vector-subspace, the Raman spectra of reaction time as a concerned vector,radian values between concerned vector and the vector-subspace were obtained, reactionend-point can be determinated by analysing changes in radian values with time seriesanalysis model.(3) The mixed system contains three components which are salicylic acid, aceticanhydride and acetylsalicylic acid (aspirin). According to the amount of material changesbefore and after reaction made concentration sample build correction and validation,using Raman spectra combined with partial least squares establish and salicylic acid,acetic anhydride determination and acetyl salicylic acid quantitative analysis model andrealize synthetic acetylsalicylic acid system of components concentration prediction.RMSEP of the three components were0.008150,0.010134and0.011101respectively viavector normalization+Savitzky-Go1ay smoothing (polynomial=2, points=5) firstderivative+a centralized. Robust fitting effect was obtained in training model by usingderivative spectrometry.(4) After concentration in calibration were converted into proper dimensionless, PLSanalysis model was established, with which number of PLS factors was4and Cross-ValMethod was "leave one out". Raman spectrum pretreatment was selected baselinecorrection and the Savitzky-Golay smooth first-order derivative combined with normalizion (1–norm). Percent conversion of salicylic acid was use to determine spannumber of Savitzky-Golay convolution smoothing first-order derivation. Withpretreatment of baseline correction and Savitzky-Go1ay smoothing (polynomial=2,points=3) first derivative combined with normalizion (1–norm), the results ofdetermination coefficient R2, RMSEP and MD were0.9776,1.6462and-1.37, for modelcalibration, the results of R2, RMSEP and MD were0.9691,0.8003and-0.4921. Thedetection accuracy was improved by model calibration, and monitoring requirement canbe achieved with this analysis model.