Deixis Online

“Our simulations are targeted at understanding the basic flame structure and how it is modulated by turbulence,” Bell explains. “Some of our work is based on simulation of actual experiments but overcomes the difficulty of extracting and interpreting diagnostic information from an experiment.”

Standard combustion experimental techniques, Bell says, use lasers to excite chemical species – a technique called induced fluorescence. The excited species emit light at a certain frequency that helps elucidate details of the flame structure. However, for these lean systems, measured fluorescence signals are extremely weak and the flames themselves are highly dynamic. Moreover, few species lend themselves to this approach, “and the interpretation requires a great deal of analysis.”

In contrast, computer simulation, Day says, “incorporates very detailed models, both for the flame dynamics (and) the turbulent flow field.” With access to supercomputing resources, “simulations can provide a substitute for some of the data that is too difficult to measure.”

Lean and clean

Combustion requires three simultaneous elements: fuel, oxidizer and heat – for example, natural gas, oxygen and a pilot light.

Traditional fuels such as natural gas and ethane are usually burned in a so-called diffusion flame whose location on a burner “is determined by where the fuel and oxidizer are brought together,” Day says.  But diffusion flames burn hot enough to disassociate nitrogen gas and exhaust nitric acid, a standard greenhouse gas that isn’t breathable.

Lean-burning flames can dramatically lower exhaust temperature – and greenhouse gas emissions. To achieve a lean burn, “the fuel and oxidizer are mixed ahead of time so that only the source of heat (such as a pilot or spark) determines the flame location,” Day explains. But this makes the flames a bit harder to control.

Consistency is important. Heat at the flame expands the air and fuel as it moves in a turbulent stream. But for hydrogen-based flames, small jitters and wrinkles produced by the turbulence are unstable. Wrinkles grow quickly and tear the flame apart by the so-called thermo-diffusive (TD) instability. Such flames “burn intensely and non-uniformly, with enhanced emissions production,” Day says. TD-unstable flames are difficult to predict and model.

“The classical flame structure can be thought of as balanced and one-dimensional,” Day says. “TD-unstable flames are inherently multidimensional with strong variability. In extreme cases, the flame becomes disconnected into islands of reaction and can extinguish locally or globally.”

Like medicine, which relied solely on understanding the macroscopic human body before molecular biology and genetics, “traditional engineering models for combustion systems are predicated on the idea of a simple one-dimensional flame structure,” Day says. “Much of the vast commercial machinery for designing and optimizing combustion devices cannot be used to predict the behavior of systems that burn as cellular flames.”

Enter Jaguar, which Bell says has allowed his team to conduct first-of-a-kind simulations that include detailed models for the kinetics, molecular transport and turbulence properties of lean hydrogen flames, at the full scale of the laboratory experiment.

Images and animations generated from Day and Bell’s simulations illustrate a generic conceptual model of how flames burn. The images depict flames, wrinkled and dancing, their colors indicating the constituent mixtures, chemical emissions such as nitric oxide, and turbulent vortices but with detail that real-life experiments would be hard-pressed to reveal so clearly.

The simulation’s realistic character reflects, Day says, “some nice mathematical analysis, two decades of software and algorithm development and petascale (1 quadrillion bytes) computing resources.”