Modelos para estimativas de propriedades físico- mecânicas de espécies de madeiras tropicais em função da densidade aparente e da frequência de vibração transversal

Abstract

The wood is an excellent material of constructive application, little inserted to the market of the national construction, but with great potential of exploration. This fact is justified not only by restrictions of a social-historical nature, but also by the ignorance of the species and their properties. The difficulties in characterization the wood, given the amount of test required and the need for specific machinery and specialized labor, contribute even more to the current situation. As an alternative to simplify the knowledge of the properties of wood, this study aims to adopt a new technological solution to estimate the physical-mechanical properties, non- desctrutive characterization, of tropical wood through the combination of apparent density and the natural frequency (first mode) of the transverse vibration test, based on regression models, given the lack of studies, at national and global levels, that use these two tests combined as estimators of all the physical-mechanical properties of wood. For this, 05 species of tropical wood were distributed in the resistance classes (C20 to C60) of the hardwood group. Sixty transverse vibration tests were be carried out on pieces of structural dimensions of square section (12 cm x12 cm) and variable length (2 m to 2.7 m) and another 720 of wood characterization, in specimens standardized according to ABNT NBR 7190:1997, totaling 780 experiments. For each species, and for the grouping, linear regression models were generated as a function of bulk density and transverse vibration frequency, as well as multivariable quadratic regression models using bulk density and vibration frequency as independent variables in order to estimate the physical-mechanical properties of the woods. From the regression models for the species in isolation, with the exception of the class C40 species, little correlation was noted between density and natural frequency with the other properties, resulting in estimates of poor precision. For the set of species, the best adjusted coefficients of determination (R²aj) were obtained when analyzing the quadratic multivariable regression models, obtaining values greater than 70% for 13 models out of the 16 evaluated. It is noteworthy that, for the main mechanical properties of the woods, and of greater importance for the structural calculation (fc0, fv0, fM, Ec0 and EM), the adjusted coefficients of determination were above 80%, indicating models with excellent fits for the prediction of properties.