Deep learning-based study on fiber stress of basalt fiber grid reinforced wood-cement composite under blast loading

Abstract

Nowadays, energy and environmental issues are becoming increasingly severe, and green, low-carbon, and environmentally friendly building materials are gaining increasing favor. Moreover, to counter the threat of terrorist attacks, research in green civil air defense engineering is urgently required. Basalt fiber grid reinforced wood-concrete composite materials effectively combine the advantages of basalt fiber reinforced polymer (BFRP) and wood-cement, thereby meeting the requirements of mechanical performance and living comfort. To achieve automatic monitoring of fiber stress in basalt fiber grid reinforced wood-concrete composite materials under blast loading, ABAQUS software is used to simulate the blast response of the composite material. By integrating the advantages of Convolutional Neural Networks (CNN), Long Short-Term Memory networks (LSTM), and Kolmogorov-Arnold network (KAN) architecture, a deep fusion neural network model named ConvM-KAN is designed. This model establishes a mapping relationship between the displacement and stress of basalt fibers, enabling accurate fiber stress prediction. The code-based experimental results indicate that the basalt fiber grid can reduce the maximum tensile and compressive damage, as well as displacement deformation of wood-concrete under blast loading by 21.5%, 9.06%, and 24.8%, respectively. The ConvM-KAN model can accurately predict the stress conditions of the fibers, with the coefficient of determination R² for stress in the x and y directions reaching 0.998 and 0.992, correspondingly. Compared to standalone CNN and LSTM models, the proposed model exhibits higher accuracy, better generalization performance, and excellent robustness. In summary, this study promotes the use of green and renewable composite materials in civil air defense engineering.