Failure mode identification in reinforced concrete flat slabs using advanced ensemble neural networks

Reinforced concrete (RC) flat slabs without transverse reinforcement are commonly used in RC buildings. Despite their appeal and widespread use, these slabs are susceptible to brittle shear failure. While most previous research has focused on estimating the punching shear strength (PSS) of RC flat slabs, accurately identifying their failure modes is crucial for effective design and reinforcement. This paper presents an analysis of ensemble neural network and ensemble deep neural network models, including bagging neural network (BaggingNN), model averaging (MA), separate stacking (SS), and integrated stacking (IS) algorithms, to develop a predictive model for failure mode identification. The results of this new model are compared with those of earlier studies. To evaluate how variables such as concrete strength and reinforcement ratio impact the failure modes of RC flat slabs, the model's prediction process is examined using the SHapley Additive exPlanation (SHAP) method. The findings indicate that the IS algorithm outperformed the BaggingNN, MA, and SS algorithms, as well as models from prior research.