Reliability Prediction based on Family of Models

Complex engineering structures are usually analyzed with finite element models (FEM) that are calibrated using monitoring data. To represent civil structures better with such models, different types of uncertainties such as those due to modeling and measurement should be incorporated. The uncertainties in the data collected by means of intermittent testing or monitoring, limitations of the models, and the non-stationary nature of behavior need to be considered. The objective of this paper is to illustrate the use of family of models and to make predictions based on the calibrated FEM of a movable bridge. First, two calibrated parent FEMs will be created to represent the undamaged and damaged condition of the structure. Then, the modeling and measurement uncertainties will be incorporated by generating the off-springs from the parent models. Afterwards, the family of models will be employed to estimate the system reliability for different possible system models and different levels of correlation of the random variables. Finally, the reliability of the movable bridge will be computed for the undamaged and damaged conditions.