Estimation of wood failure percentage under shear stress in bamboo-wood composite bonded by adhesive using a deep learning and entropy weight method

Wood failure percentage (WFP) is one of most important indices to evaluate the shear properties for wood-based composite bonded by adhesive, especially for cross-laminated timber. A new WFP measurement method for wood-bamboo composite bonded by phenolic resin (PF) and methylene diphenyl diisocyanate (MDI) separately under shear stress was proposed by combing image processing and deep learning (DL): its accuracy was corrected by the entropy weight method. The results show that the WFP measured by the DL method is in good agreement with the experimental value, however, the experimental conditions still exert a significant influence on the accuracy of measurement attained using the DL method, in which the inherent color of the adhesive and the change in the substrate of the specimens are the main factors affecting measurement accuracy. The relative error between the DL method and the experimental value can be reduced to no more than 14.29% after correction by the entropy weight method.