Forecasting the Ultimate Load Capacity of Flat Slabs with Artificial Neural Networks

This chapter delves into the challenges of predicting the ultimate load capacity and shear punching resistance of fibre-reinforced flat slabs, which are increasingly used in civil and transportation construction. Traditional methods often fall short due to their reliance on empirical formulas and simplifying assumptions. The study introduces an Artificial Neural Network (ANN) model trained on a dataset of 232 fibre-reinforced flat slab samples, collected from 30 studies. The model's architecture includes four hidden layers with ReLU activation functions, Batch Normalization, and Dropout regularization, optimized using the Adam algorithm. Key findings include the identification of effective depth and concrete compressive strength as the most influential variables affecting punching resistance. The model demonstrates strong predictive performance with metrics such as MAE of 25.36 kN, RMSE of 37.57 kN, R² of 0.936, and MAPE of 11.88%. The results highlight the potential of ANN as a practical tool for designing flat slabs, contributing to more efficient and data-driven structural solutions.