Explainable machine learning in the design of hollow-core slabs: integration between predictive modeling and structural practice

Hollow-core slabs are widely used in large-scale structures, such as shopping malls and parking garages, due to their ability to span large distances and their reduced weight compared to solid slabs. However, their design must be carefully adapted to the specific characteristics of each project, such as the span magnitude, the loads to be supported, and the aesthetic requirements of the building. The thickness of the panels, the shape of the voids, and the necessary reinforcements are variables that directly influence the safety and economic efficiency of the application. The use of machine learning (ML) proves to be promising in this context, as it allows for the optimization of hollow-core slab design based on experimental data. Predictive modeling through techniques like Support Vector Regression (SVR) and explainability through SHAP provide more effective and transparent solutions, enabling engineers to understand the impact of each variable on the structural performance. By integrating synthetic data and machine learning algorithms, it is possible to optimize designs, considering the involved variables and minimizing uncertainties. The application of ML in hollow-core slab design not only increases the accuracy of calculations but also offers a more flexible and efficient approach to handling different design scenarios. By integrating explainable models, like SHAP, engineers can make more informed and secure decisions, adapting the slab design to the specific conditions of each project, optimizing costs, and increasing structural reliability. The methodology presented in this study offers a solid foundation for the use of computational tools in engineering practice, promoting innovation and sustainability in construction.