Efficiency in Uncertainty: Uncovering the Physics of High-Efficiency, Low-Emissions Combustion

Wednesday, Aug 17, 2016

Turbulent combustion is just that. Jet engines and high-power furnaces produce a great amount of energy, along with its byproduct: emissions in the form of soot and other polluting materials. As we eye our future and governments demand ever-cleaner, more efficient performance, Michael Mueller, Assistant Professor of Mechanical and Aerospace Engineering, is creating mathematical models of uncertainty that may change the future of the field. 

One element of Mueller’s research centers on the quantification of uncertainty in turbulent combustion simulations. These simulations combine many different physical models, each with their own uncertainties, and the uncertainties that contribute most to predictions related to efficiency and emissions, such as soot and nitrogen oxides, remain unclear. To reduce the uncertainties in these predictions, the models that contribute most to the uncertainty need to be identified and improved. By building complex systems of models that capture the uncertainty inherent to combustion and turbulence, his calculations could help to shape a future in which increasingly complex computation drives rapid development of high-efficiency, low-emissions systems primed to meet the demands of tomorrow’s energy economy.

Using high-performance computing resources on campus and nationally, Mueller’s high-fidelity calculations can involve tens of thousands of processors working in concert to quantify multiple overlapping uncertainty models. The result: terabytes of raw data. As such, fast turnaround is essential, so Mueller’s team relies on the Tiger cluster, managed and supported by the Princeton Institute for Computational Science & Engineering (PICSciE) and OIT Research Computing. It boasts 10,304 cores and 200 GPUs. In addition, his team has leveraged the Cori supercomputer at the National Energy Research Science Computing Center.

In achieving a better understanding of the factors that influence efficiency of performance and pollutant generation, the ultimate aim is to “develop predictive computational models that can be used to design those systems computationally faster, cheaper, and better than relying on an expensive, empirical, and slow design-prototype-test experimentally driven process.”

It’s all about the physics: specifically, the complex interactions between chemistry, fluid mechanics, and other processes that affect the engineering process.

“HPC is absolutely critical in combustion science and technology,” Mueller explains. “On the science side, there are simply many things that cannot be measured at sufficient temporal and spatial resolution, and full-fidelity calculations for physics exploration are critical in filling the gap in our understanding.”

As for technology, Mueller’s models can be used to influence faster prototype development. “Building a prototype jet engine, for example, is incredibly expensive, so, the more computation that can be used, the more aggressive the design can be since a lot of designs can be rapidly evaluated.”

While many of Mueller’s high-fidelity calculations are still too time-intensive to become standard operating procedure in current industrial design cycles, they prove critical when engineers hit a computational wall. “I was working with a jet engine company who used some of our data to better understand a problem with soot emissions,” Mueller explains. 

Moreover, they are opening the door to a deeper understanding of soot and how it is formed. “We understand nitrogen oxides far better than soot,” Mueller says. “ The former are driven largely by chemistry and temperature. On the other hand, the latter are strongly driven by fine-scale dynamic interactions between turbulence, chemistry, and soot.” 

Significant opportunities also reside in the area of efficiency. “There are new high-efficiency ideas that rely on low-temperature combustion that produce essentially no emissions,” Mueller explains. “There are some interesting challenges here in terms of understanding and ultimately implementing such technologies, and computation will certainly play a critical role in their development.”

Mueller cites recent estimates of up to 50% improvement in automotive engine efficiency, and perhaps 20% for jet engines and stationary gas turbines. As such, it’s safe to say that the interest in quantifying uncertainty will remain high for the foreseeable future.

Learn more about PICSciE’s HPC resources, and visit Professor Mueller’s faculty profile to learn more about this and his other research projects.