Reduced Order Modeling for Domain Decompositions with Non-Conforming Interfaces

Conceitos Básicos

The author proposes a reduced-order modeling strategy for two-way Dirichlet-Neumann parametric coupled problems solved with domain-decomposition sub-structuring methods.

Resumo

This paper introduces a novel approach to address complex problems using reduced order modeling techniques. By splitting the original coupled differential problem into sub-problems and applying reduced basis methods, the study focuses on solving steady and unsteady coupled problems with non-conforming finite element interfaces. The proposed method combines DEIM at the interface level to achieve a fully reduced-order representation of domain decomposition techniques implemented. The research emphasizes the importance of handling interface non-conformity issues in reducing computational complexity while maintaining accuracy.

Estatísticas

Reduced basis (RB) method is applied to obtain low-dimensional representation of solution.
Discrete empirical interpolation method (DEIM) used at interface level.
Reduced-order model (ROM) solved by sub-iterating between two reduced sub-problems.
Numerical verification conducted on both steady and unsteady coupled problems.

Citações

Principais Insights Extraídos De

A reduced order model for domain decompositions with non-conforming interfaces

by Elena Zappon... às arxiv.org 03-07-2024

https://arxiv.org/pdf/2206.09618.pdf

Perguntas Mais Profundas

How does this approach compare to traditional full-order models in terms of accuracy

The approach of using reduced-order models (ROMs) in domain decompositions with non-conforming interfaces offers a significant advantage over traditional full-order models in terms of computational efficiency. By reducing the dimensionality of the problem through techniques like Reduced Basis (RB) and Discrete Empirical Interpolation Method (DEIM), ROMs can provide solutions much faster than FOMs. This speedup allows for real-time decision-making and operational modeling, especially in complex coupled problems.
However, when it comes to accuracy, ROMs may not always match the precision of FOMs. The reduction in dimensionality can lead to some loss of information, particularly at the interface between subdomains. While efforts are made to capture essential features through basis functions and interpolation methods, there may still be some approximation errors compared to the detailed solutions obtained from FOMs.

What are the potential limitations or challenges faced when dealing with non-conforming interfaces

Dealing with non-conforming interfaces poses several challenges and limitations in reduced-order modeling for domain decompositions:

Interpolation Errors: When interpolating data across non-conforming grids using methods like DEIM or RB, there is a risk of introducing interpolation errors that could affect the accuracy of the results.

Complexity: Implementing techniques like INTERNODES or MORTAR for handling non-conforming interfaces adds complexity to the model setup and computation process.

Parameter Dependency: The parametric nature of interface conditions introduces additional complexity as these conditions need to be approximated accurately for various parameter instances.

Convergence Issues: Ensuring convergence between reduced sub-problems at non-conforming interfaces requires careful handling due to potential mismatches in discretization.

Computational Cost: Handling non-conforming interfaces may require more computational resources and time compared to conforming cases due to additional processing steps needed for data transfer.

How can this methodology be extended to handle more complex multi-physics problems

To extend this methodology to handle more complex multi-physics problems involving different physical phenomena or equations coupled together, several considerations need to be taken into account:

Multi-Physics Coupling: Incorporate multiple physics equations into separate sub-problems within each domain while ensuring proper coupling at shared boundaries/interfaces.

Extended Reduced Basis Methods: Develop advanced RB methods that can handle diverse sets of basis functions representing different physical quantities involved in multi-physics simulations.

Adaptive Mesh Refinement: Implement adaptive mesh refinement strategies that can dynamically adjust grid resolutions at interface regions based on solution requirements from different physics domains.

Dynamic Parameter Handling: Extend parameter-dependent approaches used for single physics problems by considering a broader range of parameters affecting multiple physics variables simultaneously.

By addressing these aspects systematically, it is possible to adapt the presented methodology for domain decomposition with non-conforming interfaces towards tackling more intricate multi-physics problems efficiently and accurately."