No abstract

The effect of data preprocessing on a calibration model based on attenuated total reflectance-Fourier transform infrared spectroscopy was investigated for in situ measurements of the solution concentration during cooling crystallization. L-Glutamic acid cooling crystallization was taken as a case study. It was found that smoothing and derivative is a superior preprocessing method for model calibration, which can effectively remove spectral nonlinearity, improve the prediction accuracy of the calibration model, and enhance model robustness in detecting faults. Verification experiments demonstrated the effectiveness of the established calibration models. These models also show better performance in detecting the generation of crystal nuclei in the probe window, which is of great significance to eliminate the fatal probe fouling.