Session: Student Poster Competition

Submission Number: 185055

Gpu-Accelerated Density-Based Flamelet Combustion Solver for Compressible Reacting Flows 

The numerical simulation of combustion processes in propulsion systems remains challenging due to the strong coupling between flow compressibility, heat release, and chemical kinetics, as well as the high computational cost associated with industrially relevant configurations. To accelerate the aero-engine design cycle and support the further development of rocket engine and scramjet, an innovative GPU-accelerated density-based flamelet combustion solver is developed in this work. The density-based framework is adopted to provide a unified and conservative treatment of compressible flows, allowing the same solver to be applied consistently to compressor, turbine, and combustor, while also enabling extension to high-speed reacting flow regimes such as detonation. The governing flow equations are solved using a finite volume method with a matrix-free second-order MUSCL scheme, coupled with a flamelet-generated manifold (FGM) model for efficient combustion modeling. Two reaction scalars, mixture fraction and progress variable, are employed to retrieve local species compositions from a pre-tabulated two-dimensional flame manifold, while thermodynamic properties are evaluated using JANAF polynomial. A total-enthalpy-based energy equation is adopted, and a simplified expression for the diffusive enthalpy flux is derived under the unity Lewis number assumption. To achieve high computational efficiency, the solver is ported to heterogeneous hardware using a hybrid parallelization strategy combining OpenACC, CUDA, and MPI, enabling substantial acceleration on GPUs. The proposed solver is validated against the Sydney bluff-body flame, where improved agreement with experimental data is achieved compared to a conventional one-dimensional flamelet model. In addition, simulations of the Sandia D flame case are performed to demonstrate the solver’s capability in modeling partially premixed combustion regimes, which are not captured by flamelet model. GPU performance tests are implemented on counter flow diffusion flame and show an order-of-magnitude speedup relative to CPU-based simulations. Furthermore, RANS simulations of the KJ66 gas-turbine combustor are conducted, demonstrating the solver's robustness, efficiency, and suitability for realistic engineering applications.

Presenting Author: Yifeng Wang Shanghai Jiao Tong University

Presenting Author Biography: Yifeng Wang is a PhD candidate at Shanghai Jiao Tong University. His research is focus on computational reacting flow.

Authors:

Yifeng Wang Shanghai Jiao Tong University
Lin Shi Shanghai Jiao Tong University
Rui Wang Shanghai Jiao Tong University
Junjie Wang Shanghai Jiao Tong University
Feng Wang Shanghai Jiao Tong University
Hui Xu Shanghai Jiao Tong University