Accelerated Testing in Structural Reliability Using CAE Models

• Activity Number: 313
• Type: Contributed
• Date/Time: Wednesday, August 14, 2002 : 10:30 AM to 12:20 PM
• Sponsor: Section on Physical & Engineering Sciences*
• Author(s): Vijay Nair*+ and David Mease+
• Affiliation(s): University of Michigan and University of Michigan
• Keywords: reliability ; computer aided engineering ; accelerated testing

Abstract

Traditional reliability assessment methods based on physical testing can require prohibitively large sample sizes in many applications. This has led manufacturers to employ virtual testing using CAE models in place of physical testing. However, when the CAE models are not valid, the resulting reliability assessment may be unreliable. In this talk, we develop theory and methodology in which traditional physical testing can be used in conjunction with CAE models to create a new type of accelerated testing that requires smaller sample sizes than traditional test plans, while at the same time exhibits robustness with respect to inaccuracies in the CAE models. These test plans are implemented by physically testing a biased sample of products and employing a variance reduction technique such as importance sampling. The CAE model is used as a prior belief for failure probability from which we then derive the sampling plan which minimizes the variance. An example is given illustrating the advantages of this methodology in terms of both sample size reduction, as well as robustness to inaccuracies in CAE models.