A multiscale computational framework for the simulation of local features in large structures

Li, Haoyang

Permalink

https://hdl.handle.net/2142/108335 Copy

Description

Title

A multiscale computational framework for the simulation of local features in large structures

Author(s)

Li, Haoyang

Issue Date

2020-05-08

Director of Research (if dissertation) or Advisor (if thesis)

Duarte, Carlos Armando

Doctoral Committee Chair(s)

Duarte, Carlos Armando

Committee Member(s)

• Masud, Arif
• Olson, Luke
• Eason, Thomas, III G.

Department of Study

Civil & Environmental Eng

Discipline

Civil Engineering

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

Ph.D.

Degree Level

Dissertation

Keyword(s)

• Generalized Finite Element Method (GFEM)
• Extended Finite Element Method
• Multiscale
• Multiphysics
• Parallel
• Thermomechanical
• Plasticity
• Fracture mechanics
• Stress intensity factor
• Spot welds
• Iterative Global-Local
• Non-intrusive coupling
• Co-simulation

Abstract

In many engineering applications, it is necessary to account for interactions among multiple spatial scales through numerical simulations. Resolving fine-scale features such as geometrical details, cracks, and localized nonlinearities with high fidelity is crucial for the accurate prediction of the response and service life of structural components. Three-dimensional models with detailed meshes and advanced modeling techniques are usually required to accurately capture fine scale responses. However, adopting such models on the structural/global scale is computationally inefficient and sometimes unfeasible for problems involving a large number of local features. In most engineering applications, coarse 3-D or shell models are often sufficient for predicting the global response and hence are commonly adopted at the global scale. On the other hand, in critical regions, 3-D solid models and adaptive discretization methods are needed to capture fine-scale phenomena. Thus, the challenge lies in how to efficiently link models and discretizations at different scales. This work presents a multiscale computational framework based on the Generalized/eXtended Finite Element Method with global--local enrichments (GFEMgl) and the iterative global—local (IGL) method aiming at capturing localized responses with high accuracy and computational efficiency. In this framework, the GFEMgl is capable of capturing localized thermomechanical responses on coarse meshes via enrichment functions provided by parallel simulations of local boundary value problems where local features are accurately modeled and resolved. The IGL, in turn, provides a robust and non-intrusive approach to establishing a two-way coupling between global and fine-scale models. Additionally, as many closed- and open-source simulation tools are being developed nowadays, this framework enables engineers to perform co-simulation with multiple solvers based on the desired computational capabilities at different scales. A series of systematic numerical studies are conducted on a range of three-dimensional applications involving material nonlinearities, multiphysics responses, and fractures. Various attributes of the framework are assessed on benchmark problems and compared against reference numerical methods. The results show that the proposed methodology can deliver high fidelity solutions comparable to those obtained from a direct finite element analysis while offering promising computational efficiency. In addition, it is demonstrated that the user-time required for setting up models and discretizations for large representative problems can be noticeably reduced by utilizing the proposed framework.

Graduation Semester

2020-05

Type of Resource

Thesis

Permalink

http://hdl.handle.net/2142/108335

Copyright and License Information

Copyright 2020 Haoyang Li

Owning Collections