Quantitative analysis of granules moisture content within a fluidized bed drying process using simultaneously...

Monitoring granule moisture content (MC) and timely determination of the process endpoint are essential for product quality control in fluidized bed drying (FBD) process. This study compares two methods for measuring granule moisture using Near-infrared (NIR) and Raman spectra during the large-scale commercial manufacturing of hydroxychloroquine sulfate (HCQ). Although Raman spectra exhibit relatively weak water scattering, moisture variability indirectly affects model performance. Partial least squares (PLS) and support vector machine (SVM) regression models were employed to construct calibration models. In combination with chemometrics strategies, spectra preprocessing methods, such as standard normal variate (SNV), multiplicative scatter correction (MSC), Savitzky-Golay (SG), normalization difference (ND), derivatives, and orthogonal signal correction (OSC) were applied to enhance model accuracy in specific spectral regions. In the comparative analysis, NIR spectra calibrated with the PLS model demonstrated superior predictive performance, with a validation set correlation coefficient (R²_p) of 0.9990, a root mean square error of prediction (RMSEP) of 0.105, and a low prediction residual error sum of squares (PRESS) of 6.216. Conversely, Raman spectra calibrated with the SVM model achieved improved predictive performance, with an R²_p of 0.9600, an RMSEP of 0.474, and a low bias of 0.003. External verification indicated that the accuracy of MC analysis by Raman was slightly lower than that of NIR. However, at-line Raman spectra provide external verification to corroborate NIR results. The findings demonstrate that humidity and temperature data loggers positioned near the bottom of the fluidized bed, together with an online NIR spectroscopy tool, can effectively deliver real-time process understanding. This integrated approach enables accurate and fast determination of the drying endpoint, typically between 1.5% and 2.5%.