A Stochastic Conservative Field Transfer Method for Black-box Multiscale and Multiphysics Coupling

TL;DR

This paper presents a stochastic conservative field transfer method for black-box multiscale and multiphysics coupling that improves accuracy and conservation without requiring source discretization details, and is computationally efficient on GPUs.

Contribution

The paper introduces a novel stochastic approximation approach for field transfer in black-box coupling, enabling conservation and higher accuracy without source discretization information.

Findings

Outperforms radial basis functions in accuracy and conservation error

Compatible with black-box coupling scenarios

Achieves competitive performance on NVIDIA A100 GPU

Abstract

This paper introduces a new method for performing field transfer operations in black-box coupling, when source discretization information is not available. This approach uses a stochastic approximation of the Galerkin projection which leads to a method that asymptotically provides conservation. Error in the accuracy and conservation has been compared to the mesh intersection method and radial basis functions on a simple domain, as well as on meshes of the LTX fusion reactor. For all cases tested, our new method provides higher accuracy and less conservation error than radial basis functions and can be used for black-box coupling, unlike the mesh-intersection method. Additionally, we demonstrate the implementation and performance of our method on an NVIDIA A100 GPU, showing that the cost is competitive with the mesh intersection method.