Quantitative analysis of solid and liquid contents in reactive crystallization by in-situ Raman with support vector regression

Quantitative analysis of crystallization processes remains challenging due to the complicated relationships among the multiple phases and their spectra. In this work, in situ Raman spectroscopy was chosen to quantitatively analyze the solution concentration and suspension density simultaneously in a crystallization process. A model based on support vector regression (SVR) was developed to separate the information of solution and solid from the Raman spectra. The simultaneous correlations of the spectra with solution concentration and suspension density were achieved and the results were evaluated. It demonstrated that the SVR-based method proposed in this work gave satisfactory fitting and prediction performance. Furthermore, the model was validated in the reactive crystallization of the ternary NaHCO3-Na2CO3-H2O system. It suggested that baseline removal is not necessary when using this model. The effect of bubbles should be included for the clear solution but could be neglected in solid suspensions. The results of this work provide a simple and rapid option for the quantitative analysis of crystallization processes and similar suspension systems.