Computational Models and Uncertainties: Estimation of Reliability and Risk

Abstract

Great advances in computational mechanics and numerical methods have permitted the solution of complex, previously intractable nonlinear problems either in solid or fluid mechanics.
Computers have now sufficient memory and efficient operational systems which allow the solution of problems with a great number of degrees of freedom in a reasonable time. On the other hand, the
theoretical models, and their numerical representation, implement relationships between variables which are not necessarily known with precision. These uncertain variables introduce randomness in the results and this, in turn, affects conclusions regarding reliability and quantification of risks. This presentation discusses the need to integrate the deterministic calculations from numerical models with
methods to assess the probabilistic nature of the predictions. At the same time, the presentation discusses briefly different strategies to implement the estimation of probabilities, as well as
probabilistic design procedures based on the reliable satisfaction of different performance requirements. The discussion is illustrated with two applications: the collision force between an ice mass and an offshore exploration platform, and the collision force between vessels and the piers of a bridge.