A hybrid finite element–machine learning framework for predicting the ultimate moment capacity of stiffened cellular steel beams under bending test

Information from the abstract

Cellular steel beams (CSBs) provide structural efficiency through circular web openings, but these openings also induce stress redistribution and local instability. This study investigates simply supported stiffened CSBs under a central point load within the applicability limits of EN 13 January 1993. An ABAQUS finite element model was developed and validated against experimental results from the literature using load–deflection response, ultimate load, and failure mode. The validated model was then used to generate 200 nonlinear FEM specimens under three-point bending for parametric investigation. To reduce reliance on costly full-scale testing and repeated nonlinear simulations, two symbolic-regression-based surrogate models, MOGA-EPR and GEP, were developed to predict the maximum moment capacity (Mu). MOGA-EPR achieved the best testing performance (R2 = 0.944), MAPE = 7.657%, RMSE = 23.546, and MAE = 10.976), outperforming GEP (R2 = 0.827), MAPE = 24.858%, RMSE = 51.833, and MAE = 34.132). PRCC analysis identified web thickness (tw) as the most influential parameter governing (Mu). The results suggest that MOGA-EPR offers a reliable and interpretable surrogate modelling approach for rapid estimation of the moment capacity of stiffened CSBs within the considered design domain.

Why this record is monitored

This record has an Impact Signal of 71/100 based on recency, source, collaboration, and bibliographic signals. It prioritizes monitoring and is not a judgment of research quality.

Related topics: Structural Load-Bearing Analysis · Topology Optimization in Engineering · Structural Health Monitoring Techniques