Towards effective decision support for structural design and risk management: An information-dependent probabilistic system representation enhanced with support vector machine and unfair sampling

Structural design and risk management typically involve uncertainties related to structural performance and loading conditions, which must be effectively managed to ensure compliance with safety requirements. Additionally, the relationships among parameters influencing structural performance are often complex and not easily discernible, thereby complicating the decision-making process. To address these challenges, this paper proposes a decision support framework based on the concept of information-dependent probabilistic system representation. The framework aims to identify unacceptable design parameters in structural design and enhance risk management by updating probabilistic models of uncertain parameters for similar structures when new observational information becomes available. To overcome the computational challenges of structural reliability analysis, a support vector machine (SVM) is employed as a surrogate model for the finite element analysis typically used to evaluate the performance of engineering structures. Additionally, to handle the imbalance issue in the SVM training dataset, an unfair sampling method is introduced. An illustrative example involving a reinforced concrete structure subjected to earthquake loading is presented to demonstrate the effectiveness of the proposed framework.