An initial investigation of structural reliability from sparse data
Abstract
Over the past several decades, major advances have been made in probabilistic methods for assessing structural reliability. They largely began with the widely-used First Order Reliability Method (FORM) and Second Order Reliability Method (SORM) and have moved forward in recent years to include more sophisticated simulation-based methods such as importance sampling, subset simulation, line sampling and an abundance of surrogate model based approaches. These methods have advanced such that probability of failure can be efficiently estimated for many systems very precisely with very low coefficient of variation.But, this precision does not necessary beget accuracy, although it is often implied (perhaps not justifiably so). A critical feature of most reliability methods is that distributions of random variables are known precisely, and in most cases are prescribed such that the reliability calculation can be performed using standard normal distributions. However, when dataare scarce it is often impossible to identify a unique probability distribution for the data.The effect of this is that the uncertainty in distribution causes uncertainty in the location of the limit surface in the probability space. The objective of the present proposal is to generalize the state-of-the-art methods in structural reliability to an imprecise probability framework that allows the assessment of probabilistic confidence measures in probability of failure estimates. This will be achieved by leveraging data-driven Bayesian multimodel inference coupled with the ?first order and second order reliability methods, importance sampling, and subset simulations. The long-term impact of this research is that these uncertainties stemming from data sparsity may one day be integrated into partial safety factors for the design of ship structures and enable a data-driven approach to harness maintenance, inspection, and sensor data to perform in-service structural safety and reliability assessments.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jul 27, 2018
- Source ID
- N000141812644
Entities
People
- Michael D Shields
Organizations
- Johns Hopkins University
- Office of Naval Research
- United States Navy