Shear connector design using data analytics approaches

The field of structural engineering can be augmented with advanced data analysis techniques. Structural engineering applications consist of datasets which may include experimental or/and computational data, and can be used to derive design provisions based on the measured data. Cluster analysis is a data exploration technique that involves identifying groups in a dataset and providing relationships between input parameters, which could supplement existing engineering intuition and knowledge. As a test case, this paper reanalyzes the existing test data for shear connectors using a cluster analysis. A database of push‐out tests was established from the literature, which was then sorted into subsets using two methods: (1) manual grouping based on engineering judgment and (2) Gaussian mixture models that detect clusters based on relationships between parameters in the data. The recommended data groupings based on the two methods were compared. Reliability analyses were conducted on each data subset to determine the recommended resistance factors. The results were compared to the resistance factors prescribed in the AISC 360‐22 Specification, which now permits a performance‐based alternative for the shear connector design.