Sensitivity Analysis in Reliability-Based Lifetime Performance Prediction Using Simulation

Abstract

The time-varying performance of deteriorating materials and structures can be simulated using probabilistic methods. The propagation of performance uncertainties is affected by the characteristics of input random variables. In general, the number of samples and/or their quality do not satisfy the rigorous statistical constraints required to obtain reliable probability distributions. This deficiency encountered at the beginning of assessment process can be alleviated later by investigating the importance of variables through variability and sensitivity analyses. These analyses help the identification of significant variables in a quantitative way. In this paper, evaluation of the time-varying performance of deteriorating materials and its effect on structures measured by the reliability index profile is addressed based on three aspects: (1) acquisition of statistical characteristics of input random variables from imperfect samples; (2) model-based simulation; and (3) evaluation of variability and sensitivity measures of the nondeterministic performance model. Examples are presented to illustrate these aspects.