Portal de Periódicos da CAPES

EMBEDDED MULTILEVEL MONTE CARLO FOR UNCERTAINTY QUANTIFICATION IN RANDOM DOMAINS

2021; Begell House; Volume: 11; Issue: 1 Linguagem: Inglês

10.1615/int.j.uncertaintyquantification.2021032984

2152-5099

Santiago Badia, Jerrad Hampton, Javier Príncipe,

Machine Learning in Materials Science

The multilevel Monte Carlo (MLMC) method has proven to be an effective variance-reduction statistical method for uncertainty quantification (UQ) in partial differential equation (PDE) models. It combines approximations at different levels of accuracy using a hierarchy of meshes whose generation is only possible for simple geometries. On top of that, MLMC and Monte Carlo (MC) for random domains involve the generation of a mesh for every sample. Here we consider the use of embedded methods which make use of simple background meshes of an artificial domain (a bounding-box) for which it is easy to define a mesh hierarchy. We use the recent aggregated finite element method (AgFEM) method, which permits to avoid ill-conditioning due to small cuts, to design an embedded MLMC (EMLMC) framework for (geometrically and topologically) random domains implicitly defined through a random level-set function. Predictions from existing theory are verified in numerical experiments and the use of AgFEM is statistically demonstrated to be crucial for complex and uncertain geometries in terms of robustness and computational cost.