Parameterized reduced order models from a single mesh using hyper-dual numbers

Abstract

In order to assess the predicted performance of a manufactured system, analysts must consider random variations (both geometric and material) in the development of a model, instead of a single deterministic model of an idealized geometry with idealized material properties. The incorporation of random geometric variations, however, potentially could necessitate the development of thousands of nearly identical solid geometries that must be meshed and separately analyzed, which would require an impractical number of man-hours to complete. This research advances a recent

Uncertainty modelling and computation in Bayesian operational modal analysis

“…Studies on aleatory uncertainty are widely reported. For instance, Brake et al (2016) assessed multiple parameterised reduced order models that introduce the aleatoric uncertainty for predicting the performance of a manufactured system, Sedehi et al (2019) proposed a hierarchical probabilistic framework to quantify the aleatory uncertainty in modal updating. It is often compared with epistemic uncertainty, typically to emphasise the importance of the latter.…”