A Generalized Probabilistic Learning Approach for Multi-Fidelity Uncertainty Propagation in Complex Physical Simulations

Abstract

Two of the most significant challenges in uncertainty propagation pertain to the high computational cost for the simulation of complex physical models and the high dimension of the random inputs. In applications of practical interest both of these problems are encountered and standard methods for uncertainty quantification either fail or are not feasible. To overcome the current limitations, we propose a probabilistic multi-fidelity framework that can exploit lower-fidelity model versions of the original problem in a small data regime. The approach circumvents the curse of dimensionality by learning dependencies between the outputs of high-fidelity models and lower-fidelity models instead of explicitly accounting for the high-dimensional inputs. We complement the information provided by a low-fidelity model with a lowdimensional set of informative features of the stochastic input, which are discovered by employing a combination of supervised and unsupervised dimensionality reduction techniques. The goal of our analysis is an efficient and accurate estimation of the full probabilistic response for a high-fidelity model. Despite the incomplete and noisy information that low-fidelity predictors provide, we demonstrate that accurate and certifiable estimates for the quantities of interest can be obtained in the small data regime, i.e., with significantly fewer high-fidelity model runs than state-ofthe-art methods for uncertainty propagation. We illustrate our approach by applying it to challenging numerical examples such as Navier-Stokes flow simulations and monolithic fluid-structure interaction problems.

A Novel Physics-Based and Data-Supported Microstructure Model for Part-Scale Simulation of Laser Powder Bed Fusion of Ti-6Al-4V

“…In future research work, the proposed microstructure model shall be employed to inform a nonlinear elasto-plastic constitutive model, thus contributing to the long-term vision of achieving accurate thermo-mechanical simulations of selective laser melting processes on part-scale. Furthermore, SLM process parameters shall be inversely adjusted to yield specific microstructural distributions and hence desired mechanical properties by deploying novel efficient multi-fidelity approaches for (inverse) uncertainty propagation [79].…”

Multi-Fidelity Simulation Modeling for Discrete Event Simulation: An Optimization Perspective

“…Current methods construct multi-fidelity simulation models by manual simplification. For example, MFSM uses numerical relaxation [6] to This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/ build multi-fidelity models in computational fluid dynamics simulation.…”

Transfer learning based multi-fidelity physics informed deep neural network

“…The primary challenge behind the application of the DNN for engineering applications is the need for data. It is a well-acknowledged fact that DNNs are data-hungry tools [36]. Unfortunately, for the current work, the focus is on problems where one has access to very few high-fidelity data.…”

“…Unfortunately, these methods only work for cases where the low-fidelity data is able to capture the trend and the models of different fidelities have a strong linear correlation Both co-Kriging motivated approaches and MLMC fails when the low-fidelity and high-fidelity data have a space-dependent, complex and nonlinear correlations. To address this issue, researchers have recently proposed methods that are rooted in Bayesian statistics [36] and nonlinear auto-regressive algorithm [37].…”