Estimativa da umidade de toras de Eucalyptus sp. em campo a partir de espectrometros no NIR portáteis

Abstract

Moisture affects operational performance during industrial wood processing. Although time-consuming, the gravimetric method has been widely used to determine wood moisture. Near-infrared (NIR) spectroscopy has been used to quickly and reliably estimate the physical and chemical properties of organic materials and has been identified as a promising alternative for wood characterization. The objective of this study was to develop a methodology for applying the NIR technique to monitor wood moisture variations in logs stacked in the field using portable spectrometers. Using a portable drill with a drill bit and a chainsaw, sawdust samples were collected from freshly felled Eucalyptus sp. logs stacked in the drying yard. The sawdust was classified into different particle sizes, stored in falcon tubes, and used for spectral collection and moisture determination in the laboratory. Furthermore, wooden discs were removed from the logs, sectioned into wedges to determine the actual wood moisture content, and obtained NIR spectra. These data were used to calibrate and validate models to estimate sawdust moisture content based on spectral signatures. Log drying in the yard was monitored through spectral collections and moisture content determination of wedge wood and sawdust produced by drills and chainsaws using the gravimetric method during log drying at 15-day intervals. Spectra were collected with portable microNIR and TrinamiX equipment on wood sampled from saturated to equilibrium moisture content. The information was evaluated by principal component analysis (PCA), and the moisture values ​​obtained from sawdust were associated with the NIR spectra using partial least squares regression (PLS-R) to fit real-time moisture predictive models. The results show that the predictive models performed well, with MicroNIR presenting R²cv = 0.9370 and RMSE = 4.3% for the global model, while the model developed by TrinamiX presented R²cv = 0.8735 and RMSE = 6.1%. Therefore, a strong relationship was observed between the moisture content of the wedge and sawdust, indicating that these parameters can be used as reliable indicators of wood moisture content. Furthermore, the NIR spectra acquired by both MicroNIR and TrinamiX proved to be efficient in estimating wood moisture content in real time from sawdust. This approach can be used in large-scale industrial optimization, generating higher-quality products by monitoring material moisture content prior to the production process.