Original title: Comparative Analysis of Shear Strength Prediction Models for Reinforced Concrete Slab-Column Connections
Authors: Sarmed Wahab, Nasim Shakouri Mahmoudabadi, Sarmad Waqas, Nouman Herl, Muhammad Iqbal, Khurshid Alam, Afaq Ahmad
The article merges machine learning, design codes, and Finite Element Analysis to predict shear strength in slab-column connections. It compares various methods—ACI 318-19, Eurocode 2, Compressive Force Path, and different neural networks, including PSOFNN and BATFNN. Validating results with FEA, it optimizes weights for predictive accuracy. Among the models, PSOFNN emerges as the most effective, boasting a 99.37% R value and the lowest MSE and MAE. Even compared to the best FNN model, PSOFNN triumphs, showcasing superior results for SCS=1. This study illuminates the PSOFNN’s prowess in predicting shear strength, outperforming traditional methods and other neural networks in accuracy and reliability, offering promising applications in concrete design and analysis.
Original article: https://arxiv.org/abs/2311.12824