Online Process Analytical Technology (PAT) enables real‑time measurement of critical quality attributes, enabling deeper process insight and immediate control decisions. Producing value‑added products from biobased materials supports a more sustainable chemical industry, but the complexity of biomass requires compact, robust PAT tools for continuous monitoring. LED‑based flow spectrophotometric devices offer low power use, fast warm‑up, and stable output, though narrow bandwidth can limit performance. This work introduces multi‑wavelength, multi‑LED platforms for direct analysis of challenging samples from biomass pyrolysis and pulp‑and‑paper processes.
For the first time, advanced statistical techniques based on moving window partial least squares regression (PLS) are utilised in the selection of most suitable LEDs for particularly process streams. Four LEDs, coupled with a CCD (charge-coupled device) spectrometer were selected for the characterisation of specific target analytes within pyrolysis process streams. By applying a PLS prediction model to the recorded signals, the developed analytical device showed great accuracy during on-site testing, with the predicted values, for five process analytes, following the trend of the values obtained using a reference GC-FID method.
Following a similar LED selection approach, using LED and photodiode technologies, a portable analytical device for biosolvent purity monitoring was developed. Applying advanced statistical data manipulation enabled the real time delivery of highly accurate results, within 0.15%, simultaneously for solvent purity and concentrations of several main impurities, allowing for real-time process monitoring and decision making.
Building on the developed technologies, a portable, multi-wavelength LED system was developed to monitor wood extractives in the pulp and paper process. Interval‑selection PLS was applied to determine which LEDs provided the best predictions of total and dissolved colloidal soluble (DCS) fractions of resin acids, fatty acids, and triglycerides. The predicted profiles showed strong agreement with reference measurements obtained through liquid–liquid extraction, using tert‑butyl methyl ether, followed by GC‑FID analysis.