Chemical analysis of wood chips in motion using thermal-emission mid-infrared spectroscopy with projection to latent structures regression.

The opacity and random surface orientation of process streams of solids complicates the application of spectroscopy to on-line analysis of solid streams. We have used a specialized form of thermal-emission mid-infrared spectroscopy, called transient infrared spectroscopy, combined with PLS regression to analyze a moving stream of wood chips for lignin, hemicellulose, glucan, and extractives content. Sixty-eight softwood samples from 6 species were analyzed using one regression, and 28 hardwood samples from 14 species were analyzed with another. Root-mean-square errors of prediction for the properties varied from 0.8 to 1.9 wt %, depending on property and wood type, with lignin and hemicellulose being the most accurately determined and glucan being the least.