You are here

Robust and Verifiable Data-Driven Analysis and Design Using Machine Learning

Akil Narayan, University of Utah

Vahid Keshavarzzadeh, University of Utah

Xueyu Zhu, University of Iowa

Ramakrishna Tipireddy, Pacific Northwest National Laboratory

Advances in computational science have enabled researchers to analyze new complex systems ranging from small scale materials to large scale infrastructure systems. Robust analysis and design of innovative systems require development of advanced yet efficient (and often data-driven) mathematical models with certifiable convergence, which address a wide range of complicated phenomena such as nonlinear, nonlocal and non-deterministic effects. Statistical learning paradigms have, in particular, shown tremendous potential to catalyze with existing computational strategies to increase predictive power and design quality.


This mini-symposium aims to highlight recent developments in advanced computational techniques, and in particular machine learning approaches with verifiable accuracy, for efficient analysis and design optimization of broad range of systems pertinent to computational mechanics. Approaches of interest include but are not limited to multifidelity-multilevel analysis for fast uncertainty quantification, simulation-based design, design optimization under uncertainty including shape and topology optimization, multiscale topology optimization with the focus on computational efficiency and robustness, novel robust optimization and control formulation, advanced sampling techniques for forward and inverse problems, high-dimensional design space exploration, physics-based neural networks and explainable artificial intelligence (AI). We expect this minisymposium provides a forum that encourages interaction of interdisciplinary researchers in the areas of uncertainty quantification, design optimization, numerical methods and machine learning.