A novel highly efficient Lagrangian model for massively multidomain simulations: parallel context

TL;DR

This paper presents a parallel MPI-based implementation of a novel 2D multi-domain simulation model that discretizes both interfaces and interiors, demonstrating promising accuracy and performance in grain growth simulations.

Contribution

It introduces a robust parallel implementation of a previously developed 2D multi-domain simulation model using MPI, enabling efficient large-scale simulations.

Findings

The parallel implementation shows high efficiency and scalability.

The model accurately simulates grain growth by curvature flow.

Performance comparisons indicate advantages over existing methods.

Abstract

A new method for the simulation of evolving multi-domains problems has been introduced in a previous work (RealIMotion), Florez et al. (2020). In this article further developments of the model will be presented. The main focus here is a robust parallel implementation using a distributed-memory approach with the Message Passing Interface (MPI) library OpenMPI. The original 2D sequential methodology consists in a modified front-tracking approach where the main originality is that not only interfaces between domains are discretized but their interiors are also meshed. The interfaces are tracked based on the topological degree of each node on the mesh and the remeshing and topological changes of the domains are driven by selective local operations performed over an element patch. The accuracy and the performance of the sequential method has proven very promising in Florez et al. (2020). In this article a parallel implementation will be discussed and tested in context of motion by curvature flow for polycrystals, i.e. by considering Grain Growth (GG) mechanism. Results of the performance of the model are given and comparisons with other approaches in the literature are discussed.