Reliability Analysis of Data-Driven and Mechanics-Based Resistance Models for Headed Stud Connections in Recycled Aggregate Concrete Slabs

Specialized resistance models for headed stud connectors in recycled aggregate concrete (RAC) are missing for promoting RAC in steel-concrete composite structures. As load-bearing mechanisms of headed-stud connections using the respective natural aggregate concrete (NAC) and RAC are similar with potential differences resulting from concrete properties, it is of practical significance to know whether the models for NAC can predict the resistance of headed stud connections using RAC. Therefore, the present work evaluated the prediction performance of the existing mechanics-based and machine learning models. Besides, the authors built a novel computational model with the polynomial chaos expansion (PCE) for predicting headed stud connections. According to the Eurocode – Basis of structural design, reliability analyses were performed for these models based on the available test results of headed stud connectors in RAC, and the required partial factors were figured out for the respective models to determine the design resistance. As a result, the partial factor for the model in Eurocode 4 may be at least 1.94 instead of 1.25. This means the design resistance will be reduced by around 36% if NAC is replaced with RAC despite using the same concrete strength class. Additionally, the Eurocode 4 model leads to higher design resistance, i.e., more economical, than the other mechanics-based models. At the same reliability level, design resistance calculated by the PCE model is much higher than that by the machine learning models, even though they were derived from the same database of headed stud connectors in NAC. Notably, PCE-design resistance is the highest among all the models and is even 16% higher than the Eurocode 4-design resistance. It could be said that PCE is a promising method for designing structural components, taking the data-driven advantage, and reflecting the underlying mechanisms.