Experimental investigation and predictive modeling of shear performance for concrete-encased steel beams using artificial neural networks

This manuscript employs a highly efficient artificial intelligence (AI) technique for machine learning (ML) through artificial neural networks (ANNs) and introduces a novel numerical predictive model capable of accurately forecasting the shear capacity of concrete-encased steel (CES) beams. The research begins by conducting shear tests on nine CES beams with high-strength steel, which addresses a significant gap in the shear performance of high-strength steel in CES beams. Subsequently, a comprehensive database of CES beam shear tests is established to train and validate the ANN model. The database consists of 242 sets of test data, compiled from published literature, encompassing a wide range of geometrical and material properties. A sensitivity analysis of the proposed model is then performed using a Pearson chi-square test to determine the relative importance of each input parameter on shear strength. Furthermore, a thorough examination is conducted to assess the impact of each parameter. Finally, the proposed predictive model is compared against current design codes, including ANSI/AISC 360-16 (USA, North America), BS EN 1994-1-1:2004 (Europe), and JGJ 138-2016 (China, Asia). The comparison reveals that ML technology exhibits higher accuracy and robustness in predicting shear bearing capacity.