Neural network-based metamodel for scaffolding behavior: Application for structural load analysis and safety enhancement

Abstract

Scaffolding safety remains a critical challenge in construction, contributing significantly to site accidents and injuries. This study addresses the issue by developing a metamodel to simulate scaffolding behavior under dynamic loads, such as wind and worker activities, to prevent failure and enhance structural reliability. A finite element model (FEM) was developed to analyze 30,000 scenarios, significantly reducing the time required for structural assessments. To optimize efficiency further, a neural network was trained to accurately predict scaffolding responses, achieving an impressive R${}^2$ value of 0.9996, thus minimizing reliance on time-intensive FEM simulations. While not a full digital twin implementation, this research establishes a strong foundation for integrating such technology in the future. By demonstrating the potential of metamodeling for improving safety and efficiency, the study offers valuable insights for advancing digital solutions in construction safety and sets the stage for further exploration of digital twin systems in the industry.