Nonlinear multivariate modeling of strand mechanical properties with near-infrared spectroscopy

Abstract

The purpose of this research was to determine if nonlinear calibrations of near-infrared spectra could improve the prediction of strand mechanical properties. Strands similar in dimension to strands utilized in an oriented strand board composite process were prepared and tested for mechanical properties in three-point bending and then calibrated to near-infrared reflectance spectroscopy. It was found that an additional 7% to 16% of the variation in mechanical properties could be accounted for when second- and third-order terms were applied. Interpretation of models identified the magnitude of importance that various wood polymers play on mechanical properties and this interpretation was validated through wet chemistry. This work is significant because it demonstrates the potential of using near-infrared spectroscopy to monitor shifts in strand mechanical properties prior to wood composite manufacture and it helps to provide the fundamental relationship between near-infrared models, wood chemistry, and the prediction of mechanical properties.