A physics and data co-driven surrogate modeling method for high-dimensional rare event simulation

TL;DR

This paper introduces a co-driven surrogate modeling approach combining physical models and data-driven corrections, enabling efficient and accurate rare event simulation in high-dimensional uncertain systems.

Contribution

It develops an adaptive physics-data driven surrogate model with active learning and importance sampling for high-dimensional rare event probability estimation.

Findings

Effective in static and dynamic problems with high-dimensional uncertainties

Achieves high correlation and low bias in surrogate predictions

Improves rare event probability estimation accuracy

Abstract

This paper presents a physics and data co-driven surrogate modeling method for efficient rare event simulation of civil and mechanical systems with high-dimensional input uncertainties. The method fuses interpretable low-fidelity physical models with data-driven error corrections. The hypothesis is that a well-designed and well-trained simplified physical model can preserve salient features of the original model, while data-fitting techniques can fill the remaining gaps between the surrogate and original model predictions. The coupled physics-data-driven surrogate model is adaptively trained using active learning, aiming to achieve a high correlation and small bias between the surrogate and original model responses in the critical parametric region of a rare event. A final importance sampling step is introduced to correct the surrogate model-based probability estimations. Static and dynamic problems with input uncertainties modeled by random field and stochastic process are studied to demonstrate the proposed method.