Structural Reliability Assessment Using Probabilistic Models and Engineering Data

Abstract

Structural reliability assesses the probability that a structure will satisfy specified safety and serviceability requirements. This assessment is challenging because loads, material properties, deterioration, and measurements are subject to uncertainty. This review explains how probability models and engineering data can address that problem. The review follows a decision-first framework. It starts with the limit state and the time period. It then links each source of uncertainty to a suitable method. By synthesizing recent engineering studies within this framework, the review also identifies recurring gaps in failure data, model-error treatment, time-varying uncertainty, surrogate-model uncertainty, and reporting consistency. The main methods include the first-order reliability method (FORM), the second-order reliability method (SORM), Monte Carlo simulation, Bayesian updating, surrogate models, and active learning. Recent studies show that field data can improve bridge and building models and that surrogate and active-learning methods can reduce the number of costly high-fidelity simulations. Each method also has limitations that depend on the structure of the reliability problem. A local method can miss another failure region. A data update can fit noise. A surrogate can be accurate on average and still be wrong near failure. Therefore, engineers need checks at every step. They should state the decision, test the input model, compare methods, and report model error. No one method fits every case. A useful result must be accurate enough, clear enough, and tied to a real engineering choice.