Non-intrusive coupling of codes for multi-physics simulations using the immersed approach
Loading...
Authors
Nanal, Narendra
Issue Date
2023-05
Type
Electronic thesis
Thesis
Thesis
Language
en_US
Keywords
Aeronautical engineering
Alternative Title
Abstract
Complex systems encountered in almost all modern-day engineering applications involve multiple distinct physical processes. Traditional ways of simulating multi-physics interactions, e.g., fluid-structure interactions, often involve a single but adaptable mesh with monolithic sets of coupled equations or involve intrusive ways of coupling two or more simulation codes. Immersed approach, such as the modified Immersed Finite Element Method (mIFEM) is a volume-based method that can efficiently and effectively couple multiple physics-based solvers with independent meshes where each representing a physics model. The mIFEM algorithm is modularly implemented as an open-source software called OpenIFEM. It is well-suited for coupling two solvers non-intrusively, mitigating the implementation cost and the computational cost involving mesh updates, thus offering the versatility of keeping the codes’ original formulation and implementation. This study focuses on developing a framework to couple pre-existing codes and expand their capabilities to perform multi-physics simulations. For this work, the immersed approach, specifically OpenIFEM, an open-source implementation of mIFEM, is considered the most suitable option. The contributions of this work can be divided into two steps. The first step involves creating a method to couple an external code with OpenIFEM, and the second step involves redesigning the existing mIFEM algorithm to produce more precise outcomes for the specific application chosen. Based on the accessibility of external codes, two different coupling strategies are developed. The first coupling strategy involves building OpenIFEM with the external code as a shared library. An alternative unique coupling strategy, which utilizes MPI communications, is also developed. Here, the two codes are built and launched independently. Separate communicators are maintained for each code to retain the independent MPI communications within each code. Different codes involved in the simulation interact only by exchanging necessary quantities. These key quantities are exchanged at each time step via synchronized MPI communications. Both of the proposed strategies facilitate a non-intrusive coupling, i.e., a coupling without any changes to the governing equations or data structures of the individual codes involved. The effectiveness of the proposed framework is demonstrated through two distinct multi-physics applications. The first coupling demonstrates interactions of thin shells with the surrounding fluids. A shell solver which represents thin solids using its mid-surface is coupled with an Eulerian fluid solver from OpenIFEM. To address fluid-shell interaction, a new extension of the mIFEM algorithm is proposed. This procedure facilitates robust, accurate, and realistic interfacial loading as well as immersed geometry representation during the interaction with the surrounding fluid. For the second multi-physics coupling, a Lagrangian solid solver from OpenIFEM is coupled with SABLE (Eulerian solid mechanics shock physics code provided by Sandia National Laboratories). The second application in this work provides a novel approach to simulate high-velocity impacts. Compared to the existing coupled Eulerian-Lagrangian techniques, this approach is more straightforward to implement since it only requires the exchange of forces and boundary conditions between the two domains. The ``immersed'' aspect of the method simplifies the utilization of a computational domain that consists of multiple materials, including a combination of solids and fluids. Several numerical tests demonstrate the validity and effectiveness of the immersed approach for the chosen applications. The techniques developed for the two multi-physics applications exhibit novelty and address certain limitations of the current methods. Furthermore, the proposed framework is extremely flexible and can be conveniently employed to couple diverse codes in the future, accelerating the advancement of different multi-physics simulations.
Description
May2023
School of Engineering
School of Engineering
Full Citation
Publisher
Rensselaer Polytechnic Institute, Troy, NY