An efficient implementation of the bidirectional buffer: towards laminar and turbulent open-boundary flows

TL;DR

This paper presents three enhancements to the buffer-based open boundary condition in SPH simulations, improving accuracy, independence, and consistency, validated through diverse laminar and turbulent flow cases, including complex geometries.

Contribution

The study introduces novel modifications to the buffer boundary condition in SPH, including continuum hypothesis application, region-constrained labeling, and first-order pressure boundary treatment, for better handling complex open-boundary flows.

Findings

Successful simulation of turbulent plane jet with strong backflow.

Significant performance improvements with minor code modifications.

Validation across multiple flow geometries and conditions.

Abstract

To effectively handle flows characterized by strong backflow and multiple open boundaries within particle-based frameworks, this study introduces three enhancements to improve the consistency, independence, and accuracy of the buffer-based open boundary condition in SPHinXsys. First, to improve the buffer consistency, the continuum hypothesis is introduced to prevent the excessive particle addition induced by strong backflow. Secondly, the independence of the bidirectional buffer is enhanced through region-constrained and independent labeling schemes, which effectively eliminate buffer interference and erroneous particle deletion in complex open-boundary flows. Thirdly, the original zeroth-order consistent pressure boundary condition is upgraded to first-order consistency by introducing a mirror boundary treatment for the correction matrix. The implementation is based on the rigorously validated weakly compressible smoothed particle hydrodynamics coupled with Reynolds-averaged Navier-Stokes (WCSPH-RANS) method, and both laminar and turbulent flow simulations are performed. Four test cases, including straight and U-shaped channel flows, a plane jet, and the flow in a 3D self-rotational micro-mixer, are conducted to comprehensively validate the proposed improvements. Among these cases, the turbulent plane jet is successfully simulated at a moderate resolution within a very compact computational domain involving strong backflow, a condition that is usually challenging for mesh-based methods. The three improvements require only minor modifications to the code framework, yet they yield significant performance gains.