Uso da espectroscopia no infravermelho próximo para monitorar o gradiente de umidade da madeira durante a secagem

Abstract

Wood humidity is among the characteristics that most affect the performance and applications of wood in the industry and that is why it is important to monitor it. However, the most used methods do not allow monitoring the humidity gradient in wooden pieces. The aim of this study was to develop models to estimate the wood humidity gradient during drying based on Near Infrared (NIR) spectra and monitor the spatial variation of humidity during drying process. Newly felled Eucalyptus tree wood rafters were processed and resulted in ten prismatic specimens with dimensions of 50 (radial) x 50 (tangential) x 150 mm (longitudinal) used to measure the NIR spectra and the masses during drying. Samples adjacent to the rafters were obtained to determine the initial humidity of the pieces by gravimetry and to estimate the humidity value of the pieces during the monitoring of mass loss and spectral acquisition. Half of the pieces were submitted to natural drying in a controlled environment while the rest of the wood samples were dried in an oven at 60ºC. The pieces were analyzed in 5 steps during drying at each 20% loss of water mass. Spectra were measured at 100 points spatially distributed within each piece of wood and at each drying stage. Initially, 25 spectra were collected at equidistant points covering the entire original longitudinal x radial surface of the specimen. To access the interior of the wood and collect spectra, the pieces were flattened to a certain depth. Thus, the specimen had its initial height of 50 mm reduced to 41.67 mm in the radial direction (reduction of 16.67%) by means of an electric planer. Immediately after planing, 25 spectra were collected on the newly produced surface. Thus, other planing operations were performed on the specimen until accessing the central region of the part (50% of the original height). After spatial monitoring of the first part, the other parts remained in drying and were processed and analyzed by losing 20%, 40%, 60% and 80% of the initial water mass. The spectra collected on the surfaces of each piece at each drying stage were applied to a pre-established partial least squares regression to estimate the wood humidity from the spectral signatures. The predictive model presented R² of 0.90 and RMSEV of 11.55% and allowed to monitor the spatial variation of the moisture gradient inside the pieces of wood during drying.