Online Raman Analysis Technology For Gasoline Production And Its Application

Process analytical technology,analysis of critical-to-quality attributes in the process through the use of an inline or online analyzer,has rapidly been developed in past years.It can offer timely and accurate analytical data which helps to stabilize production,optimize operation and save energy.Spectral analysis weighs most in the technology.And recently,online Raman spectroscopy technology plays an important role due to its own spectral character.In this thesis,several key technologies have been studied and applied in both catalytic reforming and blending units for gasoline production.These key technologies are as follows:1.A novel method is proposed for the gain calibration of online Raman spectra acquired from a CCD array spectrometer.Any kind of pure substance with standard Raman spectrum can be used to establish a gain calibration curve based on spectra measured by offline and online instruments.The online spectra are corrected according to the gain calibration curve.The correction method can eliminate the influence of different lasers,Raman probes,and spectrometers.As a result,it greatly increases the spectral versatility between offline and online instruments.More importantly,no complicated calibration model is used.2.An updated background subtraction algorithm is proposed which can automatically recover Raman signal.This algorithm is based on an iterative polynomial smoothing method which highly reduces the intervention of experience and priori knowledge.Firstly,a polynomial filter is applied to smooth the input spectrum.The output curve of the filter divides the original spectrum into two parts,top one and bottom one.Secondly,a proportion is calculated between the lowest point of the signal in the bottom and the highest point of the signal in the top.The proportion is a key index which decides whether a new iteration is need.If so,the minimum value between the output curve and the original spectrum forms a new curve which goes into the same filter in the first step and starts another iteration.If not,the algorithm stops.Finally,the output of the filter is the background which the algorithm gets from the original spectrum.Results from the simulation experiments not only show that the iterative polynomial smoothing algorithm achieves good performance at time consumption and accuracy of recovery,but also prove that the algorithm adapts to different types of background and a large range of signal-to-noise ratio(1:200-1:10).Furthermore,real measured Raman spectra of some mixture samples are presented to demonstrate the performance of the algorithm.3.The feature of gasoline Raman spectra is extracted to predict its research octane number.Three methods are compared,including interval partial least squares(iPLS),forward interval partial least squares(FiPLS)and backward interval partial least squares(BiPLS).In the experiment,the SPXY(Sample Set Partitioning Based on Joint X-Y Distances)method is used to get the training set,the cross validation set,and the test set.Also the robust regression algorithm is introduced to remove abnormal samples in the training set.Compared with PLS model without feature selection,BiPLS algorithm can reduce input dimension of PLS model by 50.00%,root mean square error of cross validation(RMSECV)by 18.92%,and root mean square error of prediction(RMSEP)by 13.86%.4.A new set of online Raman analytical system,which measures the research octane number of gasoline,is developed for optimization of a catalytic reforming unit.Application results show that the system can rapidly correlate the change of reaction temperature and other operating conditions.The repeatability error of research octane number is 0.015 and the mean absolute error is 0.22.5.A completed system including a Raman analyzer is developed to monitor research octane number of blended gasoline.Application results show that the repeatability error and the mean absolute error and of research octane number are 0.05 and 0.26 respectively.The system responses fast and correlates with the change of operating conditions of the blending unit.Comparing with the catalytic reforming gasoline,the blended gasoline has components that are more complex and its research octane number can be affected by multiple conditions.However,the methods proposed in this thesis have successfully applied in both catalytic reforming gasoline and blended gasoline.Moreover,the model keeps high accuracy without frequently recalibration.