Prediction model for progressive collapse resistance of concrete beam-column subassemblies considering tensile catenary action using machine learning algorithms

• A database of 114 beam-column subassemblies tested between 2009 and 2024 was compiled • Three machine learning technique-based models were developed • An explicit and interpretable prediction model was proposed and validated Following the gradual crushing and spalling of concrete at large deformations, tensile catenary action (TCA) usually develops in axially restrained reinforced concrete beam-column subassemblies and provides a crucial resistance mechanism to avoid progressive collapse under abnormal scenarios. However, existing analytical formulations for TCA have limitations such as oversimplified assumptions of uniform yielding, reliance on empirical displacement functions and involvement of complex mechanics, which restrict their accuracy and applicability. The emerging machine learning (ML) techniques provide possible paths to overcome these limitations. In this study, a comprehensive database of 114 concrete beam-column subassemblies tested between 2009 and 2024 was compiled covering a wide range of configurations. Then three ML techniques i.e., backpropagation neural networks (BPNN), support vector regression (SVR) and gene expression programming (GEP), were developed and systematically evaluated. Comparative analysis results indicate that BPNN model exhibits strong fitting ability but limited generalization, while SVR model provides unstable predictions. Results demonstrated that the GEP model not only outperforms both by attaining the highest predictive accuracy and lowest error indices but also yields an explicit, interpretable predictive expression. Further validation against existing analytical model demonstrates that the proposed GEP-based model avoids the deficiencies of conventional formulations and provides an accurate estimate of TCA resistance. Overall, the developed model offers a practical and efficient tool for robustness assessment and progressive collapse design of concrete structures.