Process Analytical Technology (PAT), both offline and online, enables real-time measurement and analysis of critical quality attributes throughout manufacturing, facilitating enhanced process understanding and immediate control decisions [1]. Herein, two spectroscopy-based PAT approaches were developed for application to sustainable production processes, namely a biosolvent production process and a paper manufacturing process, where the use of biomass as raw material and the complex nature of products generated in manufacturing from lignocellulosic biomass require sophisticated PAT strategies.
Firstly, a chemometric-assisted NIR spectroscopy-based method was developed to quantify furfural, furans, formaldehyde, formic acid, acetic acid and water in the Furacell™ process used to produce the biosolvent Cyrene™. Quantification of the targeted analytes was performed by GC-MS, GC-FID and Karl Fischer titration, and the obtained values were used as reference concentrations. Interval PLS (iPLS) outperformed full-spectrum PLS through optimised spectral preprocessing and interval selection. Excellent predictions were achieved, with water showing the highest performance (R²=0.971), followed by formic acid (R²=0.965), furans (R²=0.901), and furfural (R²=0.836). REP values ranged from 8.1 to 22.0% across all analytes.
During paper manufacturing using the thermo-mechanical pulping (TMP) process, significant amounts of wood extractives (resin acids, fatty acids, triglycerides) are released, causing pitch deposition and operational downtime. Therefore, a rapid, solvent-free UV–Vis spectroscopy approach combined with machine and deep learning techniques was developed for real-time extractives monitoring in papermaking process streams. Regression models (MLR, PLS, Random Forest, SVM, and XGBoost) were applied comparing with reference values obtained by solvent extraction followed by GC-FID analysis. Strong predictive performance was achieved, with MLR yielding an REP of 24.4% and R² of 0.781 for total resin acids, and XGBoost an REP of 33.1% and R² of 0.727 for total fatty acids.
Both spectroscopic frameworks provide a foundation for developing portable devices for on-site, real-time process monitoring in sustainable manufacturing.
References: