Assessment of the Bearing Capacity of Double Shear-Bolted Connections in Structural Steel

Abstract

This research explored the accurate prediction of bearing capacity in double shear-bolted connections constructed from structural steel using the CatBoost machine learning (ML) algorithm. The study incorporates the Shapley additive explanations (SHAP) method to enhance interpretability, elucidating the significance of impact of each variable on predictions at both local and global scales. The results demonstrate that SHAP provides valuable insights into the decision-making process of the ML model. They identify pivotal variables for specific scenarios, enriching the comprehension of the predictions. Crucially, the alignment between feature importance derived from the ML model and SHAP underscores the criticality of certain factors in bearing capacity estimation. This study advances the precision of bearing capacity forecasts for double shear-bolted connections and underscores the advantages of employing interpretable ML approaches to unravel complex predictive models.