NCI is helping researchers from UNSW to delve into the mysteries of fuel combustion to reduce emissions and improve fuel efficiency.
"Almost 90% of Australia's total energy production relies on combustion," says Associate Professor Evatt Hawkes.
"It is very important that we move towards renewables, and the pace of developments of renewable energy is very impressive. Nonetheless, we cannot simply forget about combustion.
"Markets such as transportation and the provision of high temperature industrial process heat are difficult to address with other means. Even in the electricity market, in which solar and wind energy are making rapid inroads, a strong role for combustion should be anticipated as a cost-effective means of complementing high penetration renewable generation, such as via gas turbines, biofuels, and in a long term future, fuels made from solar energy."
A/Prof Hawkes and his team, including PhD student Shahram Karami and post-doctoral researcher Moshen Talei (now a lecturer at the University of Melbourne), are using Raijin to run the world's first model of a turbulent, lifted hydrocarbon flame, which is a type of flame that occurs in gas turbines and in many industrial furnaces.
Their model has revealed answers to long-standing combustion conundrums, and could lead to more fuel-efficient engines that emit less carbon dioxide and other pollutants.
"It's very difficult to design a new engine," explains A/Prof Hawkes.
"The engines we have today are the result of a long process of incremental improvements. To make any further improvements we're narrowed right down into a small corner of parameters. However, these improvements are urgently needed, so we need to improve our approach to engine design."
The only affordable way to design new engines is to first simulate the combustion process using computational models, says A/Prof Hawkes.
"You don't want to build a hundred or a thousand engines to find the best design. What we're aiming to do is develop computational models that are cheap enough that they can be used by industry."
But simulating combustion is not as simple as fuel + heat + oxygen = fire.
"Combustion is the oldest form of energy technology. But despite at least half a million years of trying to understand combustion, we're still not there yet.
"The first problem is that it involves a lot of chemical reactions. Even if you're burning a very simple fuel, say iso-octane which has just eight carbon atoms, a full chemical mechanism for that would involve about 1000 chemical species, and about 4000 chemical reactions.
"On top of that we have the problem of turbulence, and that's what our team is looking at.
"One of the key ways you can boost fuel efficiency and reduce emissions is by having very fast mixing," explains A/Prof Hawkes. "And the easiest way to do that is to increase the fuel injection velocity and push the flame further downstream.
"The downside of that, however, is that eventually you can push the flame so far downstream that it extinguishes. Obviously, this is not desirable in any combustion system – and least of all an aviation gas turbine!-"
Supported by ARC Discovery and LIEF Grants, A/Prof Hawkes and his team used their model to investigate the trade-off between faster fuel injection and flame stability.
"Despite decades of study of the mechanisms of flame stabilisation, there has not been enough evidence to fully support any one theory," he says. "One of the key problems is that the flame propagation speed was unavailable in experimental measurements since the experiments have been at best two-dimensional and lacked any information about out-of-plane flame propagation. Our simulations could assess this and what we found is that the fuel velocity coming in almost exactly balances the flame propagation speed. In other words the basic stabilisation mechanism is the flame propagation mechanism."
The complexity of turbulent combustion requires enormous modelling power and data processing capacity.
"One of these runs takes about one million hours on Raijin. We also generate a lot of data; almost 10 TB per run.
"The principal challenge is that turbulent combustion involves a massive range of length and time scales. In the automotive-sized diesel engine we model, the engine cylinder is about 100 mm in diameter and the smallest turbulent eddies would be in the order of tens of nanometres.
"Our modelling is a first-principles approach that directly resolves all of the relevant continuum scales. We target problems which are down-sized in terms of the range of scales, but which still capture the essential physics and chemistry.
"In our modelling of lifted flames, we used about one billion grid points and about 200,000 time steps, with seven variables per grid point and time step. And that was with the very simplest chemical model we can use to represent the simplest hydrocarbon, methane.
"We typically run jobs in the 4,000-8,000 core range, but in principle we could scale this right out to at least half of Raijin's 57,000 cores."