MS Title:
Physics-informed surrogate modeling for uncertainty quantification and optimization

Description:
Uncertainty quantification (UQ) aims at quantifying and understanding the influence of ubiquitous uncertainties arising in science and engineering. For design optimization of engineering systems, managing the impact of uncertainties is pivotal, and a UQ engine should be incorporated to drive the optimization process. Surrogate modeling is a general UQ approach that seeks to replace the original expensive computational model with a cheap substitute. The data-fitting surrogate modeling approaches, e.g., Gaussian process metamodeling, even combined with dimensionality reduction techniques, still face difficulties in handling high-dimensional uncertainties. The physics-informed surrogate modeling can address this challenge by injecting domain/problem-specific prior knowledge. Multi-fidelity UQ and scientific machine learning are two emerging paradigms for physics-informed surrogate modeling, but there also exist numerous specialized physics-informed metamodeling approaches for specific applications.

This mini-symposium aims to highlight new research trends in physics-informed surrogate modeling. Contributions to the methodology of physics-informed surrogate modeling, as well as applications in probabilistic analysis and optimization of engineering systems, are welcome. 

Session Chairs:
Cheng Su, South China University of Technology, Guangzhou, China. E-mail: cvchsu@scut.edu.cn
Ziqi Wang, University of California, Berkeley, United States. E-mail: ziqiwang@berkeley.edu
Jianhua Xian, University of California, Berkeley, United States. E-mail: jianhua.xian@berkeley.edu