Wood shrinkage prediction using NIR spectroscopy

Abstract

The ability to predict wood shrinkage could help manufacturers avoid lumber with abnormal dimensional stability or match pieces with similar properties in glued assemblies. Near infrared (NIR) spectroscopy is a rapid, nondestructive technique that has been used to predict various wood properties, including extractive content and density. Fifty-seven mahogany (Swietenia macrophylla) blocks were scanned using an NIR spectrometer, and were measured for specific gravity, extractives content, and total volumetric swelling. Models were created to predict the wood properties using the NIR data. These models could provide reasonable predictions of shrinkage, density, and extractives content. The use of nonlinear kernel and wavelet statistical techniques improved model performance. It may be possible to use NIR spectroscopy for the on-line sorting of wood according to dimensional stability.