OLCF 25: Detailing Combustion
In 2017, the Oak Ridge Leadership Computing Facility celebrated 25 years of leadership in high-performance computing. This article is part of a series summarizing a dozen significant contributions to science enabled by OLCF resources. The full report is available here.
A team led by Jacqueline Chen at Sandia National Laboratories (Sandia) has used OLCF resources since 2005 to conduct simulations of turbulent combustion, which can help engineers develop better predictive models for fuel-efficient engines.
Using the OLCF’s Cray XT4 and XT5 Jaguar and Cray XK7 Titan, Chen’s team has performed direct numerical simulations (DNS) of combustion phenomena, including fuel-air mixing, autoignition, and flame dynamics at multiple scales. Chen is particularly interested in combustion processes that more efficiently distribute heat and lower nitrogen oxide emissions.
By detailing the important mechanisms that can contribute to more efficient combustion processes, the team is helping pave the way for automobiles that could use 25 to 50 percent less fuel than those on the road today.
During diesel engine combustion, air and fuel mix violently and react, generating heat that causes the mixture to spontaneously ignite (autoignition).
Finding the right conditions for efficient combustion is tricky, and simulating these conditions on multiple scales is impossible without a leadership-class resource. Multiple factors—such as how much fuel and air is added, how much mixing takes place, and whether combustion occurs as spontaneous ignition or propagating flames—contribute to this complex process.
Using the DNS code S3D, which solves for compressible, reacting flows using Navier–Stokes equations, Chen’s team explored these factors, capturing length scales ranging from micrometers—the size of some molecular interactions—to centimeters, which allows for the nuances of fuel-air mixing and chemical reactions to be studied with unprecedented fidelity.
In 2006 Chen’s team used Jaguar to simulate a 3D turbulent lean methane–air premixed flame. Lean premixed combustion is of interest to the engineering community because lean burning distributes heat efficiently and generates lower nitrogen oxide emissions. The team discovered that intense small turbulent eddies can cause a broadening of the flame’s preheat zone, an area that continuously heats a flame so that it does not die once ignited. This same turbulence, though, does not disrupt the thin areas where chemical reactions are occurring. The DNS data from this study were used to assess models of the degree of flame wrinkling, which affects the overall burning rates.
Since autoignition can occur simultaneously with turbulent mixing of cold gaseous fuel and heated oxidant, the team also simulated a 3D lifted flame to study how a flame stabilizes above a burner. By capturing the full range of turbulence scales from the largest to the smallest eddies, the team provided the first fully-resolved simulation of a lifted turbulent hydrogen flame stabilized by autoignition, concluding that autoignition was the key mechanism responsible for flame stabilization.
The simulation of a 3D lifted hydrogen flame provided a large dataset for engineers to use as they develop the predictive models that will optimize the internal combustion engines and gas turbine designs that could reduce emissions and increase efficiency. The team’s most recent work continues to uncover new details about combustion processes, thanks to the wide range of scales—both in space and time—made available by leadership-class computing.
Chen is the founding director of the Center for Exascale Simulation of Combustion in Turbulence—an interdisciplinary center focused on codesigning the software and tools necessary for the future of combustion modeling and simulation in the exascale era. Her work is setting the stage for combustion simulation on exascale computing systems.
Related Publication: Sankaran, et al. (2007), Structure of a Spatially Developing Turbulent Lean Methane–Air Bunsen Flame, Proceedings of the Combustion Institute, Volume: 31, no. 1. DOI: 10.1016/j.proci.2006.08.025.