AFRL selects XG Sciences to develop graphene nanoplatelet-based high-energy ultracapacitors; potential benefit for automotive applications
DOE to award up to $9M for Small Business programs in broad range of technologies including vehicles, biomass and hydrogen and fuel cells

GE Global Research, ASU and Cornell to partner with Lawrence Livermore National Lab on development of next-generation fuel injectors

Computer simulation of liquid spray from a test fuel injector. Click to enlarge.

Lawrence Livermore National Laboratory (LLNL) has selected GE Global Research to participate in an incubator program that will use high-performance computing (HPC) in an effort to accelerate development of next-generation liquid fuel injectors. Global Research will collaborate with Arizona State University (ASU) and Cornell University on the project “Improving Models of Spray Breakup in Liquid Fuels Combustion”.

The scientists hope to gain a better understanding of critical unsteady spray phenomena observed in fuel injectors used in today’s liquid-fueled engines. These unsteady spray phenomena are sometimes inaccessible to experimental measurements. Computer simulations can provide much needed insight into the origin of the unsteadiness, but doing this requires supercomputers to accurately capture the underlying physics.

Within the project, GE mechanical engineer Madhu Pai, from the Computational Combustion Lab (ATMS) will have six months of dedicated access to a portion of the LLNL Sierra supercomputer to study the physics behind the working of the fuel injector to optimize its design.

Currently fuel injectors are designed after lengthy optimization trials, partly because today’s fuel injectors have complex geometries that challenge conventional wisdom on how these injectors work. High-fidelity computer simulations can significantly reduce the number of trials and can provide insight into why a fuel injector behaves the way it does.

Using the supercomputer, we will apply a methodology called Large Eddy Simulation (LES) to model the fuel injector. The supercomputer will give us a 360 degree view of the inside of the injector, so that we can better understand the physics behind the design. Having a better understanding of how the fuel/air mixture combusts will help us ultimately build more powerful engines that consume less fuel and have lower emissions. HPC will ultimately help in reducing development time and cost of the fuel injector.

—Madhu Pai

Aircraft fuel injectors are being studied in this trial, but successful testing of this computer simulation methodology could yield new insights that benefit other GE products, including the fuel injectors used in locomotives and land-based gas turbines. The methodology can potentially be applied to study nebulizers for aerosol delivery.

Access to LLNL’s supercomputing pilot program, known as “hpc4energy”, was highly sought after. More than 30 companies applied; GE Global Research was one of six selected. The goal of the program is to facilitate more R&D engagement between the National Labs and energy companies to help increase America’s economic competitiveness.

The supercomputing project will begin in April at the LLNL’s facility in California.



Oh boy. Being in simulations myself, this is a bunch of lies. In project applications to be approved it is always written that computer simulations can reduce costly trial and error. This is a lie. Computer models are rarely capable of simulating all physical conditions and processes involved without simplifications (raising the issue of fidelity so you need validation by experimental setup anyway) and are mostly useful for qualitative analysis. They need to be set up by experts and take enormous amounts of time to run, so you can only test so many setups. With well designed experimental trial and error you could probably test a lot more cases in a lot less time. The wages of programmers and analysts are rarely taken into account, neither is the very limited number of cases that can be evaluated in reasonable amount of time nor the expenses of the supercomputers involved. But it will get you funding.


Fuel injector issues are overcome by injecting fuel into the throat of a variable area Venturi. The high air velocity and low pressure at the throat nebulizes and vaporizes the fuel. Concurrently, throttling losses are reduced because the air flow in unrestricted until sonic velocity is reached, at the thoat, near the bottom dead center of the piston cycle.

James Meyer
[email protected]


I agree with SimonDM.

Models, for even complex systems are very valuable and affordable for testing multiple "difficult" cases, if they are based on known physics.

Models for even simple systems with poorly understood physical phenomena must be built up from esoteric small elements (if they are understood, and they are not here) or an often large assortment of small scale tests (very interesting for LLNL but very slow and expensive for us) or empirical testing (in which case the unsophisticated empirical model is the result, not the tool).


interesting you should write that. I wasted several years of my youth also, studying mathemetics. A total of 4 years studying differential calculus. None of it ever used - come to that I doubt if I even needed Pythagorus LOL.

I understand the justification for this huge waste of my time was that the ability to solve min max solutions would aid my future employer in making the most efficient products possible. etc etc.

In reality there are very few engineering jobs that need math for that or any other computation. I venture to suggest perhaps less than one in a thousand. Hardly justifies all this math training.

However they do make many thousands of jobs for the maths teachers with their handsome salaries secured by membership in public service unions please note.

The thing is in real life very few things have sharp efficiency curves and if you are 20% off the mark with a process design you may only end up paying a 5% penalty in output. By the time administrative fixed costs are factored in, the financials and labor the penalty may show up as being less than 1%.

Consequently when I hear journalists spouting about the need for more science and math training in schools I have to laugh.

I am one of the few who happen to enjoy mathematics, however I respect that it is not popular with most people who despite seeing it not having a cost benefit to their career path nevertheless end up having it forced down their throats by academia. Can't get a degree in scientific disciplines without it, although stating that math may be the latin of the 21st century may not be far from the truth.
Math may provide practice with analytical skills so they say but if you need to analyze a situation with a team of people then previous expertise with the Craft of War video game may provide more useful life skills than math, my opinion.

Regarding gas turbines the biggest power improvement over the years have come from the ceramic blade coatings being able to withstand higher temps. When discussing the electronic interface to the aviation turbine with an insider, it appears increased telemetry to the pilots now plays a greater role. Disappointingly I learned that control to the engine is done solely with conventional positioning devices within the injector. The rest is up to the thermodynamics.


TT. I think that, as you say, the importance of higher math applies to fewer and fewer people all the time.

But for science, and therefore society, the need actually gets greater - if we are to continue to advance.

For most of us higher math is but a brief (and all too often forgotten) view of the importance of exact numbers over generalizations.

Higher math is the core of simulations and advanced models; it enables the design of high efficiency turbine engine compressors by iterating theory and experimentation.

A few % gain in compressor efficiency is just as useful as higher turbine inlet temperature.

The comments to this entry are closed.