RAPID PREDICTION OF WOOD EXTRACTIVES CONTENT USING NEAR INFRARED SPECTROSCOPY

摘要

The feasibility of Fourier transform near infrared spectroscopy (FT-NIR) to rapidly determine extractives content of wood was investigated. The hot water, 1% NaOH and benzene-ethanol extractives content of 106 samples were analyzed with traditional wet chemical technique, the FT-NIR spectra of which were collected from wood meal using integrating sphere. Partial Least Square (PLS) regression analyses were carried out to establish relationships between the data sets of wet laboratory chemical data and the FT-NIR spectra. The prediction models for the extractives content of wood samples were established using full cross validation in the wavenumbers ranges of 7500~4598 cm~1, 6102~4247 cm~1, and 7500~5446 cm~1 by the spectral data preprocessing methods of the l.DerMSC and l.Der. The high coefficient of determination (R~2) between the predicted NIR values and those obtained from traditional wet chemical methods were 0.9800, 0.9823, and 0.9648, respectively. The low root mean square errors of cross validation (RMSECV) values were 0.29%, 0.48%, and 0.24%, respectively. Residual prediction deviation (RPD) values were 7.07, 7.51, and 5.33, respectively. The models derived from wood meal spectra were very precise. The predictions were very good, and standard errors of prediction varied from 0.20%~0.18%, -0.36%~0.42%, and -0.25%~0.24%, respectively. Finally, one can find that NIR spectroscopy technique has many advantages, such as portability, good reliability, and high accuracy and is good for the content measurement of nondestructive wood hot water, 1% NaOH and benzeneethanol extractives.