Abstract:

High-Fidelity physics based computational models enable the design and optimization of complex engineered processes. Moreover, important and strategic decisions might be taken relying on those computational models predic- tions. . Therefore, there is a need for improving their robustness and reliability. Towards those goals, understanding the impacts of uncertainties on input data and model structures on the predictions, often referred to as Uncertainty Quanti cation (UQ), has become a major issue. A key aspect in this context is the demand of a significant computational effort, what might be lessen by the use of reduced order models or any sort of surrogates. Here, we employ Gaussian Processes(GPs) as surrogates in a speci c application involving the computational modeling of Hydraulic Fracturing (HF). HF is a process used in Oil and Gas industries to artificially improve the permeability of the reser- voir. Tipically, the modeling of such a process involves scnearios that present uncertainties in the Geomechanics properties. The modeling of this process leads to multiscale, non-linear and free boundary mathematical problem that have to be solved through complex algorithms . A number of numerical examples is presented in order to prove the e ciency of GPs as surrogates for the original complex computational model.