Fully Resolved Autoigniting Transient Jet Flame Simulation Gauss Centre for Supercomputing e.V.

COMPUTATIONAL AND SCIENTIFIC ENGINEERING

Fully Resolved Autoigniting Transient Jet Flame Simulation

Principal Investigator:
Andreas Kempf

Affiliation:
Institute for Combustion and Gas Dynamics, Chair of Fluid Dynamics, University of Duisburg-Essen

Local Project ID:
GCS-JFLA

HPC Platform used:
Hazel Hen of HLRS

Date published:

Combustion processes are of paramount importance for transportation, power generation and processes of the chemical and materials industries. As the world energy demand grows rapidly, non-fossil fuels and even fossil fuels remain to be of great importance – in particular, due to their high energy density, their ease of storage, and the existing, costly infrastructure that is globally available. And yet, given the toxic gaseous and particulate emissions and the great danger of climate change, it is of great importance to make combustion as clean and efficient as possible – wherever combustion of fossil fuels cannot be avoided or replaced by non-fossil fuels.

The present work analyses the ignition behaviour of the cleanest fossil fuels (natural gas, essentially methane CH4, with a very small carbon content) and develops methods that can help the improvement of the combustion systems for more complex, non-fossil biofuels. Transient mixing and ignition play a significant role in many combustion systems, where combustion efficiency and emissions are controlled by ignition and mixing dynamics – in particular in cases with pulsed fuel injection. In such setups, even small changes in temperature, pressure, turbulence and mixing affect ignition, and hence, combustion. This means that ignition dynamics determine the efficiency of the combustor and its pollutant emissions from it.

As the chemical state of the flame is linked to the mixing of fuel, oxidizer and products, the complex dynamics of the transient jet during an injection pulse affect ignition, and thus combustion. The relation between transient jet dynamics and chemistry due to ignition needs to be understood – with the help of experiments and simulations.

Recent experiments by DLR (German Aerospace Center) successfully determined the ignition properties of a transient jet, relying on advanced high-speed imaging techniques. These techniques, however, cannot provide a “complete picture” in four dimensions (space and time) or for all relevant quantities, which necessitates suitable computer simulations.

In the present project, the DLR-experiments are reproduced by state of the art numerical tools with detailed descriptions of chemistry, turbulence and mixing. The aim is to create a database of the time-resolved evolution of an ignition kernel in a transient jet, which can be used to analyse the process in great detail – by ourselves and by other researchers. Such a large-scale simulation with detailed chemistry of a pulsed jet in a very large computational domain has not been attempted before – neither by us or by any other group.

This project required significant supercomputing resources (53 million core hours) since the ignition of a transient jet can only be resolved in the smallest time and length scales. This necessitated simulations using more than 650 Million computational points in space – and at each point, the complicated differential equations describing flow and combustion had to be solved hundreds of thousands of times. This enabled to describe the flow structures occurring in a domain of 65x35x35 mm3 only – such is the complexity of turbulent reactive flows! As a result, a major challenge of this project was of a computer-scientific nature – on coordinating and orchestrating more than 90 thousand processor cores, working together for over twenty days; to store and to backup many Terabytes of simulation data; and to mine these data to gain scientific insights.

First, a good agreement between the computations and the experiments was demonstrated, confirming that the simulations were providing meaningful data for the subsequent, very detailed analysis. The resulting numerical database consisted of detailed information to be further analyzed, containing all parameters that could possibly be part of the ignition dynamics.

The database was mined and important observations were extracted, however, we expect that this dataset will be analysed further by other researchers in the years to come.

Overall, this work demonstrates the necessity of HPC systems in order to understand the complex physics involved in common devices – and how even the biggest industrial players have to depend on cheaper, much-simplified descriptions to design their products. This also implies how much room for further optimisation is left – optimisations which can be achieved in the future by conducting very detailed simulations presented in this project.

As the principal investigator Prof. Dr.-Ing. Andreas Kempf and project member Eray Inanc of the Institute for Combustion and Gas Dynamics, Chair of Fluid Dynamics, at the University Duisburg-Essen, we acknowledge the grant of high-performance computing resources at HLRS Stuttgart (HPC system Hazel Hen) with grant number 44141 and name GCS-JFLA.

Project Member and Scientific Contact

MSc. Eray Inanc
Institute for Combustion and Gas Dynamics
Chair of Fluid Dynamics
University Duisburg-Essen
Carl-Benz-Straße 199, D-47057 Duisburg (Germany)
e-mail: eray.inanc [@] uni-due.de

Local project ID: GCS-JFLA

October 2019

Tags: HLRS CSE Universität Duisburg-Essen