Optimized punching shear design in steel fiber-reinforced slabs: Machine learning vs. evolutionary prediction models

This research paper focuses on utilizing Artificial Neural Networks (ANN), Multi-Objective Genetic Algorithm Evolutionary Polynomial Regression (MOGA-EPR), and Gene Expression Programming (GEP) to predict the punching shear strength of Steel Fibre-Reinforced Concrete (SFRC) slabs. In order to formulate predictions, research and analysis were carried out making use of a dataset, this dataset included several parameters that impact on punching shear strength, including SFRC slabs longitudinally and transversely, using ANN, GEP, and MOGA-EPR methods. The developed models exhibited very good performance, as the soft computing techniques (GEP and MOGA-EPR) achieved R ² values of 0.91 to 0.93, while the ANN technique was higher at 0.95. Furthermore, two case studies were incorporated to carry out cost analyses of the models in real-world applications. It was shown that the efficiency of the Machine Learning (ML) models in reducing the costs of materials is relatively high, as they were capable of better predictions than the standard methods employed by the codes. • The proposed system combines soft computing techniques with Machine learning protocols to generate practical punching shear design models. • The three new models offer designers powerful, more precise, straightforward tools for design. • Provide access to accurate predictions of the punching shear strength of SFRC for different design scenarios so engineers can make informed decisions and optimise their designs.