Using near-infrared multivariate image regression to predict pulp properties

Traditional laboratory testing of dissolving pulp is labor intensive and time consuming. Multivariate image regression (MVIR) offers a testing method that is much more rapid than the analytical chemistry approach. Pulp properties are predicted from multi-spectral, near-infrared (NIR) images of finished pulp. A single test sample can be used to predict four indicator variables for the quality of pulp - S10, S18, DCM resin, and intrinsic viscosity. The testing approach provides a framework for studying pulp heterogeneity through deriving spatial pulp property distribution across the imaged section of a pulp sample. Preliminary test results have been promising for both off-line and at-line studies.