Abstract:

This paper investigates the use of Fourier transform infrared spectroscopy (FT-IR) for quantitative analysis of bio-crudes from hydrothermal liquefaction (HTL) of biomass. HTL is a versatile process rendering virtually all biomasses suitable for conversion into bio-crude and side-streams. However, continuous processes require rapid analytical methods applicable to highly diverse bio-crudes. Bio-crudes were obtained from two different continuous HTL reactors (lab scale and pilot scale) and in some cases with recirculation of water. The bio-crudes originated from a diverse range of feedstocks including lignocellulosics (pine, Miscanthus), microalgae (Spirulina, Chlorella vulgaris), and residues (sludge, dried distiller grains with solubles). Quantitative analysis of water content, total acid number, and total content of phenolics was performed using FT-IR. Principal component analysis indicated a potential correlation between quantitative measurements and FT-IR. Partial least-squares regression was used to develop predictive models that performed well considering the high diversity of the bio-crudes. The content of phenolics was in the range of 83.1–254.6 mg g–1 (gallic acid equivalent), and the model calibration was good (Root Mean Square Error, RMSE = 19.7, slope = 0.81, y-exp = 81.2%). A diverse set of test samples were subjected to the models. The relative difference for measured and predicted phenolic content was generally <15%. Total acid numbers (TAN) were 7–98 mgKOH g–1, and the model calibration was found to be satisfactory considering the titration method used (RMSE = 18.5, slope = 0.53, y-exp = 52.6%). The relative difference for measured and predicted TAN was generally <20%. The water content (Karl Fischer titration) was 1–24%, and the model calibration was very good (RMSE = 2.0, slope = 0.93, y-exp = 92.6%). The water content was generally predicted within 1.5%, and the relative difference for measured and predicted water content was large (2.7–16.6%) due to the small values. All models included samples that deviated and could be considered outliers; however, their deviations were explained from their composition and were retained in the models. Overall, the results show the potential of FT-IR as a universal technique to obtain rapid quantitative results from a variety of bio-crudes processed using different reactors.