Chemometric modelling of thermally modified wood using Fourier-transform infrared spectroscopy

This study investigates the potential of Fourier-transform infrared spectroscopy combined with chemometric modelling to predict mechanical properties of thermally modified Pinus elliottii wood. Specimens were subjected to three thermal treatments (160 °C, 200 °C and 240 °C) and analysed using second-derivative Fourier-transform infrared spectra. Principal component analysis revealed clear spectral separation among treatments, while partial least squares regression partial least squares regression was employed to model bending modulus, bending strength, and hardness (radial and tangential). In addition to regression modelling, spectral–mechanical relationships were examined through Pearson correlation analysis across all wavenumbers, enabling the identification of highly predictive spectral regions and their correspondence with structural wood compounds. Fourier-transform infrared spectra captured consistent chemical changes, especially in regions associated with cellulose (∼1030 cm −1 ), lignin (∼1230 cm −1 ) and hemicelluloses (∼1730 cm −1 ). Partial least squares regression models exhibited high predictive accuracy for bending modulus ( R 2 = 0.93; RPD = 2.18) and tangential hardness ( R 2 = 0.91; RPD = 2.05), indicating excellent robustness. In contrast, bending strength and radial hardness presented lower performance ( R 2 = 0.81 and 0.74, respectively), attributed to greater mechanical variability and weaker spectral contrast. Therefore, this study demonstrates that Fourier-transform infrared spectroscopy, supported by chemometric tools such as variable importance in projection scoring and spectral correlation, enables fast, non-destructive and accurate prediction of wood mechanical behaviour of thermally modified pine woods. This approach represents a viable strategy for sustainable quality assessment of engineered wood products.