Application of NIR Imaging and multivariate data analysis for pharmaceutical products

In Process Control (IPC) and Process Analytical Technology (PAT) are gaining more and more importance on pharmaceutical production sites for process and product quality monitoring. Near-infrared spectroscopy (NIRS) proved to be useful as a PAT-tool in the pharmaceutical industry. NIRS, with its noninvasive character is suitable for raw material qualification, quality determination and process monitoring. Low sample preparation efforts and the access to chemical and physical parameter makes NIRS ideal for process analysis. This diploma thesis deals with the evaluation of an NIR camera system (HE-LIOS), based on Hyper Spectral Imaging (HSI) as a PAT-tool. The system's specific characteristics are the push-broom technology and a Free Programmable Gate Array (FPGA). The specific electronics and optics allow fast quantitative and qualitative in-line monitoring. The spectral data obtained with the HELIOS-system were analyzed by Multi-variate Data Analysis (MVDA), containing the Principal Component Analysis (PCA) and Partial Least Squares Regression (PLS). Spectral data filtering and preprocessing routines prior to data analysis, were discussed and evaluated in terms of statistical model performance. Here, different preprocessings were applied in order to transform spectra to pseudo-absorbance, standard normal variate (SNV), multiplicative scatter correction (MSC), spectral averaging, etc. The readiness of the HELIOS for instant product quality or process monitoring , was evaluated by analysis of different adjustments and detection limits of common pharmaceutical products and materials. Therefore, different ratios of binary and ternary powder blends, coating thickness and moisture of tablets and concentrations of polyethylene glycol (PEG 6000) in water were investigated. This diploma thesis highlights the high potential of the NIR camera system HE-LIOS as a PAT-tool for the pharmaceutical industry. Especially the high sample rate of up to 333 Hz enables real-time monitoring of fast continuous manufacturing processes on the basis of different chemical and physical parameters.