Using near-infrared multivariate image regression to predict pulp properties.

APPLICATION: Product quality can be monitored efficiently by modeling data from NIR multi-spectral images of finished pulp.

Analytical chemistry is the time-honored way to monitor pulp quality, but laboratory testing is laborious, and online techniques for monitoring quality can often deliver the results more quickly to those who need the information. Multi-spectral NIR spectroscopy may soon offer an online technique that can speed up the process of evaluating four pulp quality indicators.

The researchers have developed statistical methods using multivariate image regression (MVIR) to evaluate pulp properties from multi-spectral NIR images of finished pulp. The technique can quickly determine the values of several properties from a single test sample. Preliminary tests have been promising for both off-line and at-line results.

The method evaluates four indicators of dissolving pulp quality S10, S18, DCM resin, and intrinsic viscosity. S10 and S18 describe the solubility of hemicelluloses, and DCM resin is a measure of extractable materials derived with dichloromethane solvent. Intrinsic viscosity is a measure of the average length of the molecular chains of polymers making up pulp fibers. The researchers also introduce a framework for studying pulp heterogeneity by deriving the spatial distribution of pulp across the imaged section of a pulp sample. View this paper online at http://www.tappi.org/index.asp?pid=29254

Manish H. Bharati and John F. MacGregor are with the Dept. of Chemical Engineering, McMaster University, Hamilton, Ontario L8S 4L7, Canada. Marc Champagne is with Tembec Inc, Temiscaming, Quebec JOZ 3R0, Canada. Email MacGregor at [e-mail].