A load balanced chemistry model with analytical Jacobian for faster reactive simulations in OpenFOAM

TL;DR

This paper presents a new open-source chemistry model for OpenFOAM that significantly accelerates reactive CFD simulations by dynamic load balancing and analytical Jacobian computation, achieving a 200-fold speed-up without losing accuracy.

Contribution

The paper introduces DLBFoam for load balancing and uses an analytical Jacobian with pyJac and LAPACK to enhance simulation speed in OpenFOAM.

Findings

Achieved 200x speed-up in large-eddy spray combustion simulations.

Maintained solution accuracy despite the speed improvements.

Implemented a dynamic load balancing model for reactive CFD in OpenFOAM.

Abstract

In this study, we introduce a novel open-source chemistry model for OpenFOAM to speed-up the reactive computational fluid dynamics (CFD) simulations using finite-rate chemistry. First, a dynamic load balancing model called DLBFoam is introduced to balance the chemistry load during runtime in parallel simulations. In addition, the solution of the cell-based chemistry problem is improved by utilizing an analytical Jacobian using an open-source library called pyJac and an efficient linear algebra library LAPACK. Combination of the aforementioned efforts yields a speed-up factor 200 for a high-fidelity large-eddy simulation spray combustion case compared to the standard OpenFOAM implementation. It is worth noting that the present implementation does not compromise the solution accuracy.