(YIP) LEARNING PERIDYNAMIC (NONLOCAL) OPERATORS- A RELIABLE AND GENERALIZABLE APPROACH TO PREDICT MATERIAL DAMAGE

Abstract

Prediction and monitoring of heterogeneous material damage, where small-scale dynamics and interactions affect the global behavior, are ubiquitous in applications of interest to AF and DOD and to the broader engineering community. However, fundamental mathematical and computational challenges present, due to difficulties around computational scalability, variability, and data sparsity. Using peridynamic operators and machine-learning techniques, the PI proposes to develop a robust and reliable framework to address modeling challenges from above difficulties and predict the material-specific damage quantitatively, with the ultimate goal of providing robust predictions with error estimates and uncertainty quantification. To efficiently capture complex nonlinear modes of failure from the atomistic scale, the PI plans to employ datadriven peridynamic models. To address the variability in micromechanical parameters, a material-specific stochastic peridynamic model will be obtained through the nonlocal operator regression. To address the last challenge on data sparsity, measurements from different materials will be incorporated and a meta-learned nonlocal operator regression approach with better adaptability will be developed. Rigorous mathematical analysis will be developed and the obtained model in each step will be validated with experimental data.

Document Details

Document Type
DoD Grant Award
Publication Date
Mar 07, 2023
Source ID
FA95502210197

Entities

People

  • Yue Yu

Organizations

  • Air Force Office of Scientific Research
  • Lehigh University
  • United States Air Force

Tags

Readers

  • Computational Fluid Dynamics (CFD)
  • Distributed Systems and Data Platform Development
  • Structural Health Monitoring of Composite Structures.

Technology Areas

  • AI & ML
  • AI & ML - Bayesian Inference