 |
Center for Nonlinear Studies |
Surrogate models are powerful tools used to approximate expensive functions in design optimization and uncertainty quantification. Multi-fidelity methods fuse data from various sources to reduce reliance on high-fidelity simulations. For cases where experimental or computational uncertainties can be estimated, we introduce an extended recursive co-kriging model that enables the surrogate modelling of data from multiple sources with generally heteroscedastic noise. This data fusion method is combined with adaptive sampling and model management to present a framework for non-deterministic multi-fidelity surrogate construction.
MS Teams: Join the meeting now
Host: Svetlana Tokareva (T-5)