Adaptive reliability analysis of regional infrastructure

Abstract eng:
This paper puts forward an approach for performance-based engineering and infrastructure risk analysis. The methodology is implemented in a new general-purpose software that is made freely available. Essentially, the software facilitates “plug-and-play” with multiple engineering analysis models in one reliability analysis. The developments comprise an analysis framework in three parts: the reliability/optimization analysis toolbox; the “bottom-up parameterized” model domain; and the graphical user interface. The model domain is in focus in this paper. The format, flexible organization, and ease-of-implementation of new models in the domain are key issues. Probabilistic and deterministic expressions and algorithms can be used as models, as well as external software applications, such as OpenSees. In this paper, particular focus is on probabilistic models for hazards, buildings, bridges, deterioration, damage, failure consequences, etc. This approach draws upon recent advances in performance-based earthquake engineering and probabilistic model development. The separation of aleatory (irreducible) and epistemic (reducible) uncertainty in the models is emphasized. This distinction is utilized to adaptively reduce the epistemic uncertainty by refining the models according to a new reliability importance vector. This vector ranks each epistemic random variable according to the change in reliability per “dollar” needed to change its variance. Consequently, reliabilities, risk, and loss curves are obtained with an optimized balance between model uncertainty and computational cost. The approach is applicable to regional risk analysis, as well the prediction of performance probabilities for individual buildings.

• Export as: BibTeX | MARC | MARCXML | DC | EndNote | NLM | RefWorks
• Add to your basket