Application of near Infrared Spectroscopy for Estimating Wood Mechanical Properties of Small Clear and Full Length Lumber Specimens

Abstract

Near infrared (NIR) spectroscopy, coupled with multivariate analytic statistical techniques, has been used to predict the mechanical properties of solid wood samples taken from small clear and full length lumber specimens of hybrid larch (Larix gmelinii var. japonica × Larix kaempferi). The specific mechanical characteristics evaluated were modulus of elasticity (MOE), modulus of rupture (MOR) in bending tests, maximum crushing strength in compression parallel to grain (CS), dynamic modulus of elasticity of air-dried lumbers (Efr), and wood density (DEN). Partial least squares (PLS) regression calibrations were developed for each wood property. The calibrations had relatively strong relationships between laboratory-measured and NIR-predicted values in small clear specimens, with coefficients of determination ranging from 0.61 to 0.89. The calibration models were applied to the prediction data sets and results suggested that NIR spectroscopy has the potential to predict mechanical properties of small clears with adequate accuracy (standardised prediction error=2.06-2.82). The PLS models based on spectra from the radial face ( R2 = 0.73-0.89) of wood were slightly superior to those from the tangential face ( R2 = 0.61-0.84). This might be due to the differences of the surface condition in terms of the anatomical structures and, thus, radial face better represents the sample. A reasonable predictive model for wood stiffness was also obtained from the full length lumber specimens, but the accuracy of the calibration for prediction was less than the small clear specimens ( R2 = 0.49-0.78). The regression coefficients obtained from the PLS models showed similar trends in all mechanical properties. It was suggested that the absorption bands due to the OH-groups in cellulose were the major contributors to building robust models for predicting the mechanical properties of wood