Explainable artificial intelligence framework for structure's limit load extimation

The recent advancements in machine learning (ML) and deep learning (DL) have significantly expanded opportunities across various fields. While ML is a powerful tool applicable to numerous disciplines, its direct implementation in civil engineering poses challenges. ML models often fail to perform reliably in real-world scenarios due to lack of transparency and explainability during the decision-making process of the algorithm. To address this, physics-based ML models integrate data obtained through a finite element procedure based on the lower bound theorem of limit analysis, ensuring compliance with physical laws described by general nonlinear equations. These models are designed to handle supervised learning tasks while mitigating the effects of data shift. Widely recognized for their applications in disciplines such as fluid dynamics, quantum mechanics, computational resources, and data storage, physics-based ML is increasingly being explored in civil engineering. In this work, a novel methodology that combines machine learning and computational mechanics to evaluate the seismic vulnerability of existing buildings is proposed. Interesting and affordable results are reported in the paper concerning the predictability of limit load of structure through ML approaches. The aim is to provide a practical tool for professionals, enabling efficient maintenance of the built environment and facilitating the organization of interventions in response to natural disasters such as earthquakes.