Monitoramento da dessorção de água na madeira por espectroscopia no infravermelho próximo

Abstract

Moisture is inside as the most important properties of wood, because of its influential variation, such as strength and density, consequently it is necessary to monitor it. In general, a determination of the moisture of the wood is performed by gravimetric method, which is often unfeasible to be carried out in industrial production processes, and a near infrared (NIR) spectroscopy is a viable and reliable alternative to estimate the properties of lignocellulosic materials. In this context, the objective of this study was to apply spectroscopy in the NIR in the development of models for the monitoring of wood desorption by means of cross-validations and independent, as well as verify the quality of products. In this study, 40 specimens of 25 x 25 x 50 mm (R x T x L) of wood from Eucalyptus sp., 6 and 6.5 years of age were evaluated. During a drying of the previously saturated wood samples, the specimens had their masses determined and non-NIR spectra measured in 10 drying steps, from a saturated condition to an anhydrous condition. The spectra were obtained by integrating sphere and optical fiber on the tangential and radial faces, besides the transverse face of the test pieces machined by circular saw and tape. The moisture values of the wood were correlated with the spectra in the NIR. Principal component analysis (PCA), partial least squares regression (PLS-R) and partial least squares discriminant analysis (PLS-DA) were performed using information. The best global model for wood moisture estimation was generated from spectra obtained on the transverse face machined by a band saw in the integration sphere (R²cv = 0.96 and RMSECV = 8.64%) and transverse sawing via optical fiber (R²cv = 0.94 and RMSECV = 10.69%). From the best global model, obtained by way of acquisition sphere of integration in the cross-face machined by sawing tape the moisture range that presented the best non-NIR performance in both the correct classification percentage and in the determination of moisture to 0 to 40%. An integrating sphere and a transverse or radial face are the most suitable for the number of NIR spectra, especially in woods with 0 to 40% humidity of models generated to estimate the humidity of Eucalyptus wood.