Glubam Roof Truss with Riveted Glubam Connections Adopting Thin-Walled Steel Tube: Experiment, Modeling, and Model-Updating
Da Shi,Yongjia Xu,Cristoforo Demartino,Guoshan Lu,Zhi Li
DOI: https://doi.org/10.1016/j.jobe.2024.110550
IF: 7.144
2024-01-01
Journal of Building Engineering
Abstract:Bio-based laminated structures offer a timely solution to meet the pressing demand for resource-efficient and environmentally responsible construction practices in civil engineering. Riveted connections are a kind of commonly used joint and significantly impact the mechanical behaviors of global structure. However, existing studies on the complex behavior of such joints for bamboo-based structures are insufficient. This study investigated the hysteretic behavior of riveted Glued Laminated Bamboo (glubam) connections using thin-walled steel tube and corresponding planar roof truss with thin-walled steel tube connections by experimental and numerical methods. Firstly, cyclic tests were conducted on the riveted glubam connections with thin-walled steel tube to investigate their hysteretic behavior. The test matrix encompasses four distinct connection configurations, carefully selected to assess the connection behavior of glubam. After the investigation of connection behavior, cyclic tests were further conducted on glubam planar roof truss structures to evaluate their global and local hysteretic behavior. Subsequently, the numerical hysteretic models for joints and hybrid roof trusses were developed within the OpenSeesPy framework. To accurately and efficiently calibrate the model parameters, a novel parallel genetic algorithm (PGA) was proposed to carry out an initial calibration on model parameters. Based on PGA, a more comprehensive model updating framework combining the neural network, prior knowledge and PGA was developed to obtain optimal parameters. Finally, the numerical models were successfully validated against the experimental data, demonstrating the effectiveness of the established numerical model and model-updating framework.