DNS of Turbulent Combustion

What is the DNS?

There are three levels of combustion CFD (computational fluid dynamics) approaches, depending on the ratio of the grid size to a Kolmogorov length scale of a given turbulent flow.

RANS (Reynolds averaged Navier-Stokes)
- Grid size >> Kolmogorov length scale
- Full range of scale modeling
- Most current engineering CFD codes
LES (Large eddy simulation)
- Grid size ~ 4-16 times Kolmogorov length scale
- Resolve ~ 80% of turbulent energy
- Need sub-grid modeling
DNS (Direc numerical simulation)
- Grid size ~ Kolmogorov length scale
- Resolve over 99% of turbulent energy
- No modeling of turbulence closure

그림4

Time and lengh scales of different CFD approaches in the continuum level.

DNS of turbulent reacting flow

Solve energy and species eqs. as well as Navier-Stokes eqs.
Resolve the smallest turbulent length scale and flame structures
Use realistic chemical kinetic mechanisms
- H₂/air detailed mechanism: 9 species, 21 elementary reactions
- C₂H₄/air reduced mechanism: 22 species, 18 global reactions
- C₇H₁₆/air reduced mechanism: 58 species
Have used the world’s most powerful supercomputer, Jaguar
- 3D DNS of turbulent lifted hydrogen jet flame in an autoignitive coflow (2008)
- 3D DNS of turbulnet lifted ethylene jet flame in a hot coflow (2009)
- DNS of ignition of n-heptane/air mixture under HCCI conditions (2010)

List of the World’s most powerful supercomputers as of Jun. 2015.

Growth of computational power and DNS capability

Fastest supercomputer broke 33.86 Pflop/s in 2015 (Tianhe-2, China’s National University of Defense Technology) and is expected to post 1 Eflop/s in ~ 2020.
DNS in early 1990
- 1 M grid points with 1 step chemistry
Today’s largest DNS
- Jet Reynolds number ~ 10,000
- 1 B grid points with ~ 10 species
- Total CPU hours ~ 10 M hrs
DNS in 2020
- Jet Reynolds number ~ 100,000
- Total CPU hours ~ 10 B hrs

Peformance of #1 (cyan), #500 (orange), and sum of all (purple), available from https://www.top500.org

DNS of Turbulent Lifted Hydrogen/Air Jet Flame in an Autoignitive Heated Coflow

To figure out the stabilization mechanisms of turbulent lifted jet flames in heated coflow, which is relevant in diesel engines and commercial boilers, 3D DNS was performed in a slot burner configuration with jet Reynolds number of 11,000 and over 940M grid points using detailed hydrogen/air chemistry and mixture averaged transport properties. To my best knowledge, it was the world lagest DNS run ever performed on spatially developing turbulent flame.

$mass fractions of HO_2 and OH$

3D volume rendering of scalar dissipation rate (left) and hydroxyl and perhydroxyl radicals (right) (Volume rendering was performed by Prof. Ma and Dr. Yu at UC Davis)

DNS of Turbulent Lifted Ethylene/Air Jet Flame in an Autoignitive Heated Coflow

This is a visualization of a turbulent lifted ethylene/air jet flame in an autoignitive coflow where the particles are colored by temperature and the volume rendered is the hydroperoxy field. Particle tracers carried along with the flow in the simulation of the ethylene jet originate in the coflow, and ignite shortly after being swept into the fuel stream. High concentrations of the hydroperoxy radical, which is most prevalent near the flame base before decreasing to its quasi-steady concentration along the fuel rich centerline of the simulation, marks the stabilization region. The increase in temperature of entrained fluid as the pathlines pass through the stabilization region is sustained only for the particles which are advected downstream in the core reaction zone: the particles making their way into the core of the jet are cooled by mixing with the cold fuel-stream.

Turbulent lifted ethylene_air jet flame visualization

3D volume rendering of perhydroxy radicals and fluid particles (This visualization was produced by Hongfeng Yu and Ray W. Grout and the simulation was performed by Chun Sang Yoo, Ed Richardson, Ramanan Sankaran, Jacqueline H. Chen)

DNS of Turbulent Lifted Hydrogen/Air Jet Flames in Heated Coflows near Augo-ignition Limit

This is a visualization of turbulent lifted hydrogen/air jet flames in heated coflows near auto-ignition limit where the volume rendereds are the hydroperoxyl and hydrogen peroxide fields.

3D volume rendering of hydroxyl and perhydroxyl radicals for three different coflow temperatures (750, 850, and 950 K) (Volume rendering was performed by Prof. Yu at Univ. of Nebraska-Lincoln)

DNS of Ammonia/Air and Ammonia/Hydrogen/Air Non-premixed Ignition in Turbulent flows

Basically, ignition delay time of non-premixed mixture is advanced by turbulent flow, comparing to laminar. However, it was observed that ignition delay time is retarded at the strong turbulent intensity. And this trend is changed by the hydrogen ratio. So, it will be valuable to investigate the turbulent effect on ignition delay of non-premixed mixture having the differential diffusivity.

오세영_1

The profile of ignition delay of non-premixed mixture varied by hydrogen ratio, as a function of turbulent intensity

2D Iso-contour of temperature and explosive mode

Chracteristics of the flame propagation represented by the blue color according to turbulent intensity