Production of high-quality forest wood biomass using artificial intelligence to control thermal modification

Abstract

Fast-growing wood plantations have been widely used as an alternative to the suppression of native vegetation. Forest wood can be enhanced by thermal modifications, which improve the wood’s natural durability and dimensional stability. In addition, the wood’s appearance is also enhanced, as its color is made darker, thus more similar to those of tropical woods, which increases its commercial value. However, such heat treatment needs precise process control tools, since it improves the physical properties of wood but reduces its mechanical resistance. Therefore, the use of artificial intelligence, as an artificial neural network (ANN), to conduct the thermal process, can be an alternative to overcome the laborious and detailed work needed to perform the treatment. The purpose of this research was to evaluate the physical, mechanical, and colorimetric properties of the ANN-assisted thermally modified woods of Eucalyptus urophylla and Pinus oocarpa, in order to predict their structural and esthetic properties. The thermal modification improved the physical properties but reduced the mechanical strength of the wood, from 150 ºC and above. Such treatment also promoted color changes. The ANN showed high precision in evaluating and monitoring the properties of the thermally modified woods, with a correlation coefficient for validation higher than 91%, coefficient of determination above 83%, and mean absolute percentage error below 12%. Furthermore, the root mean squared errors and mean absolute percentage error were lower than 13% for all evaluated parameters, showing high accuracy and normal distribution of errors. Therefore, ANN seems a promising alternative for the classification and characterization of thermally modified wood and can be implemented for final quality control of the product.