Structural collapse risk assessment of steel tube coupler scaffold based on BP neural network and sensor optimization

Abstract This study proposes a novel framework integrating real-time monitoring and machine learning for structural progressive collapse risk assessment of steel tube-coupler scaffolds (STCS). Laboratory tests, including slip tests, rotational tests, and progressive collapse tests, were conducted to quantify joint stiffness and structural failure behaviour. Finite element method (FEM), which established based on beam element, was further adopted to establish a dataset. By fusing axial forces and biaxial displacements from strategically placed sensors, a Back propagation (BP) neural network achieves 84.13% accuracy in localizing failure points and classifying collapse risks into four levels. Principal Component Analysis (PCA) reduces required monitoring parameters by 51.8%, maintaining 97% variance contribution with only 13 critical sensors. This paper introduces an approach for cost-effective safety monitoring in STCS system, combining measurement efficiency with predictive fidelity through optimized IoT-enabled data acquisition.