Detection of wet-pockets on the surface of Tsuga heterophylla (Raf.) Sarg. by near infrared (NIR) spectroscopy

Abstract

Abstract Multivariate statistical models capable of rapidly and accurately predicting the surface moisture content in Tsuga heterophylla were developed based on near infrared (NIR) spectra of small specimens precisely conditioned within the hygroscopic range. In an initial research phase, the applicability of NIR as predictor of surface moisture was investigated. The first derivative spectra in the range of 1300–2100 nm yielded the best results. The partial least squares regression (PLS-1) model had coefficient of determination (R2) of 0.98, root mean square error of cross validation (RMSECV) of 0.97%, root mean square error of prediction (RMSEP) of 1.05%, and ratio of performance to deviation (RPD) of 7.25. In a subsequent phase, an inline pilot-plant NIR system combined with this PLS-1 model was constructed. The prediction ability of the NIR system was tested with line speeds of 0, 100, 200, and 400 mm s-1 on kiln-dried full-length lamination boards classified as “wets” after conventional kiln drying. In a calibrated range of moisture content (0–25.4%), the NIR system demonstrated R2 values of 0.79 and 0.74, RMSEP values of 3.13 and 3.28, and RPD values of 2.18 and 1.67 at a line speed of 0 and 100 mm s-1, respectively, regardless of the presence of knots and surface roughness. These results demonstrate that the NIR system at a line speed of 0–100 mm s-1 could be used to provide entire surface moisture distribution and to detect local moisture peaks that indicate surface wet-pockets in kiln-dried lumber destined for lamination.