Efficient Implementation of Global-Local Iterative Coupling and Acceleration Techniques in Abaqus for Nonlinear Structural Analysis

Główne pojęcia

This paper presents the implementation and convergence study of the Global-Local Iterative Coupling (GLIC) technique in the commercial software Abaqus, using its co-simulation engine. The GLIC approach allows for a non-intrusive multi-scale modeling of complex structures by coupling a global coarse model with local refined models, while maintaining the global model unchanged. The authors explore the use of three different convergence acceleration techniques (Aitken's relaxation, Anderson acceleration, and Broyden's quasi-Newton method) to improve the computational efficiency of the GLIC approach.

Streszczenie

The paper presents the implementation and evaluation of the Global-Local Iterative Coupling (GLIC) technique in the commercial software Abaqus, using its co-simulation engine. The GLIC approach allows for a non-intrusive multi-scale modeling of complex structures by coupling a global coarse model with local refined models, while maintaining the global model unchanged.
The key highlights and insights are:

The GLIC approach involves iteratively exchanging displacements and reaction forces between the global and local models until convergence is achieved. This allows for a more accurate coupling compared to the traditional sub-modeling technique, which introduces errors due to the unidirectional transfer of information.

Three different convergence acceleration techniques are explored to improve the computational efficiency of the GLIC approach: Aitken's relaxation, Anderson acceleration, and Broyden's quasi-Newton method. These techniques aim to reduce the number of global-local iterations required for convergence.

The performance and accuracy of the GLIC approach with the different acceleration techniques are evaluated through two challenging use cases: a holed plate with localized elasto-plasticity and a bolted joint connecting two plates.

The results show that the GLIC approach with the acceleration techniques can provide accurate solutions compared to the reference monolithic finite element model, while offering the convenience of a non-intrusive multi-scale modeling approach.

An inexact solver strategy is proposed to further improve the computational performance of the GLIC approach by relaxing the convergence criteria for the global and local solvers.

Overall, the paper demonstrates the potential of the GLIC approach with acceleration techniques for efficient and accurate multi-scale modeling of complex nonlinear structural problems in the context of the Abaqus software.

Statystyki

The holed plate use case showed the following performance metrics:

GLIC-Aitken: 32 global increments, 163 global Newton-Raphson iterations, 338 local Newton-Raphson iterations, 65 global-local iterations
GLIC-Anderson: 41 global increments, 256 global Newton-Raphson iterations, 513 local Newton-Raphson iterations, 99 global-local iterations
GLIC-Broyden: 41 global increments, 256 global Newton-Raphson iterations, 513 local Newton-Raphson iterations, 99 global-local iterations
The bolted joint use case showed the following performance metrics:

GLIC-Aitken: 14 global increments, 434 global Newton-Raphson iterations, 1093 local Newton-Raphson iterations, 151 global-local iterations
GLIC-Anderson: 14 global increments, 647 global Newton-Raphson iterations, 1533 local Newton-Raphson iterations, 220 global-local iterations
GLIC-Broyden: 14 global increments, 691 global Newton-Raphson iterations, 1648 local Newton-Raphson iterations, 234 global-local iterations

Cytaty

"The results show promising results for the coupling technology and constitute a step forward in the availability of non-intrusive multi-scale modeling capabilities for complex structures and assemblies."
"The convenience of a global-local approach resides in the mitigation of the modeling burdens for creating multiple complex finite element models for multiple load cases or geometry configurations."

Kluczowe wnioski z

On the implementation in Abaqus of the global--local iterative coupling and acceleration techniques

by Omar... o arxiv.org 04-25-2024

https://arxiv.org/pdf/2404.15299.pdf

Głębsze pytania

How can the GLIC approach be extended to handle more complex multi-physics problems, such as fluid-structure interaction or thermo-mechanical coupling

The Global-Local Iterative Coupling (GLIC) approach can be extended to handle more complex multi-physics problems by incorporating additional coupling techniques and advanced algorithms. For fluid-structure interaction problems, the GLIC approach can be enhanced by integrating fluid dynamics solvers with structural mechanics solvers. This integration allows for the exchange of information between the fluid and structural domains, enabling a more comprehensive analysis of the interaction between the two systems. By implementing co-simulation techniques, the GLIC approach can facilitate the concurrent simulation of fluid and structural responses, leading to a more accurate representation of the coupled physics.
In the case of thermo-mechanical coupling, the GLIC approach can be extended to include thermal analysis alongside structural analysis. By incorporating thermal solvers into the global and local models, the GLIC approach can capture the effects of temperature variations on the mechanical behavior of the system. This integration enables the simulation of complex thermo-mechanical interactions, providing insights into how temperature changes affect the structural response.
Overall, extending the GLIC approach to handle more complex multi-physics problems involves integrating specialized solvers, implementing advanced coupling techniques, and optimizing the iterative process to ensure accurate and efficient simulations of coupled phenomena.

What are the potential limitations or challenges in applying the GLIC approach to large-scale industrial problems with a large number of local patches

When applying the GLIC approach to large-scale industrial problems with a significant number of local patches, several potential limitations and challenges may arise. Some of these include:

Computational Complexity: As the number of local patches increases, the computational cost of the GLIC approach also escalates. Managing a large number of global-local iterations and coordinating data exchange between multiple patches can lead to significant computational overhead.

Modeling Complexity: Handling a large number of local patches requires meticulous modeling and meshing of each patch, which can be time-consuming and prone to errors. Ensuring consistency and accuracy across all patches becomes challenging as the complexity of the system grows.

Convergence Issues: With a large number of local patches, ensuring convergence of the global-local iterations becomes more challenging. Variations in material properties, boundary conditions, and mesh densities across patches can impact the convergence behavior of the iterative process.

Communication Overhead: Coordinating communication between numerous local patches and the global model can introduce communication overhead, potentially leading to delays in data exchange and synchronization issues.

To address these challenges, efficient parallel computing techniques, advanced convergence acceleration methods, and robust data management strategies can be implemented. Additionally, optimizing the meshing strategy, refining convergence criteria, and streamlining the communication process can help mitigate the limitations associated with applying the GLIC approach to large-scale industrial problems.

Can the GLIC approach be integrated with other advanced modeling techniques, such as the Generalized Finite Element Method or phase-field methods, to capture the propagation of fracture or other complex phenomena

The GLIC approach can be integrated with other advanced modeling techniques, such as the Generalized Finite Element Method (GFEM) or phase-field methods, to enhance its capabilities in capturing complex phenomena like fracture propagation. By combining the GLIC approach with these advanced methods, the modeling of intricate fracture processes and other phenomena can be significantly improved. Here are some ways in which the GLIC approach can be integrated with these techniques:

Generalized Finite Element Method (GFEM): By incorporating GFEM into the local models within the GLIC framework, the simulation accuracy can be enhanced, especially in regions with complex geometries or material behaviors. GFEM allows for the enrichment of the finite element basis functions, enabling a more accurate representation of localized phenomena such as crack propagation or material interfaces.

Phase-Field Methods: Integrating phase-field methods into the local models of the GLIC approach enables the simulation of fracture propagation and damage evolution in a continuous and cohesive manner. Phase-field methods provide a numerical framework for modeling crack initiation, growth, and coalescence, allowing for a more realistic representation of fracture behavior in structural components.

Non-Local Damage Models: Incorporating non-local damage models within the local patches of the GLIC approach can improve the prediction of damage evolution and failure mechanisms in materials. These models consider the influence of material properties beyond the immediate vicinity of a point, allowing for a more comprehensive analysis of fracture propagation and structural integrity.

By combining the GLIC approach with advanced modeling techniques like GFEM, phase-field methods, and non-local damage models, engineers and researchers can achieve a more accurate and detailed representation of complex phenomena in structural analysis and design. This integration enhances the predictive capabilities of the GLIC approach and enables the simulation of a wide range of challenging engineering problems.