| Description |
With funding from the Department of Energy (DOE), to further the use of government funded high-performance computing (HPC) software for engineering analysis of industrial problems, Reaction Engineering International (REI) is working with the University of Utah to leverage the Uintah Computational Framework (UCF) for commercial simulation of industrial flares. The Arches component of the UCF provides a reacting large eddy simulation (LES) capability, which is a more fundamentally accurate description of turbulent mixing and combustion than is obtained in conventional Reynolds Averaged Navier Stokes (RANS) approaches. Since the application of Arches to the simulation of commercial flares presents many challenges to a potential user, including software compilation, case definition, case setup, simulation and post-processing in an HPC facility, there is a need for streamlining this process to make Arches and commercial HPC facilities more accessible to flare designers and end-users. This paper provides an update on the results of our DOE program focusing on three areas: Evaluation of mesh sensitivity Evaluation of subgrid scale models Development of a web-based interface The computational demands associated with LES simulations performed on meshes that resolve 80% of the turbulent kinetic energy are extremely high. It is expected that in many potential industrial applications, the computational demands associated with this level of resolution may exceed available resources. Evaluation of the impacts of mesh refinement and the chosen subgrid scale model is an important element in hardening Arches for commercial simulation of industrial flares. Results in this paper show that quantitative predictions of combustion efficiency from a simple flare tip vary as the mesh size resolves up to 69% of the turbulent kinetic energy. Further work will be necessary to evaluate the predictive capability of Arches flare simulations that are completed on meshes of this size. Simulations have also been completed to evaluate impacts of two different subgrid scale models: 1) the commonly used dynamic Smagorinsky model, and 2) the novel Sigma model. The simulations indicate that the Sigma model provides quantitatively similar combustion efficiency predictions compared to the dynamic Smagorinskly model at a three-fold reduction in computer run-time. The paper also describes the front-end and back-end design of a web-based interface that is under development to streamline case definition, simulation, and post-processing of flare simulation results from a commercial HPC facility. |
