Inference of mechanical properties and structural grades of bamboo by machine learning methods

Features such as fast–growth rate, high strength–to–weight ratio, high carbon sequestering capability amongst others, make bamboo an excellent alternative environmental–friendly construction material. Therefore, it is very important to establish the most appropriate geometrical and/or physical properties that can be used to infer capacities as well as structural grades for bamboo such as Guadua angustifolia Kunth (GAK). Thus, an extensive experimental characterization of physical and mechanical properties of GAK was conducted by two independent laboratories –with samples from the same plantation in Colombia. Pooling of the two datasets were performed in order to create a larger data and undertake a more rigorous statistical analysis using machine learning (ML) methods. In addition, regression equations of mean and characteristic values for parallel–to–fiber compression, shear and bending capacities, and flexural stiffness were determined based on ML methods employing geometrical and physical properties. Finally, ML methods were used to propose a classification method based on four capacity classes that could enable a simpler grading process for structural bamboo species such as GAK.