Bayesian model selection and parameter estimation for fatigue damage progression models in composites

Abstract

A Bayesian approach is presented for selecting the most probable model class among a set of damage mechanics models for fatigue damage progression in composites. Candidate models, that are first parameterized through a Global Sensitivity Analysis, are ranked based on estimated probabilities that measure the extent of agreement of their predictions with observed data. A case study is presented using multi-scale fatigue damage data from a cross-ply carbon–epoxy laminate. The results show that, for this case, the most probable model class among the competing candidates is the one that involves the simplest damage mechanics. The principle of Ockham’s razor seems to hold true for the composite materials investigated here since the data-fit of more complex models is penalized, as they extract more information from the data.