A commercial collaboration with GE Aerospace

Understanding turbulence is key to improving the operating efficiency in a wide range of applications, from aircraft engines to power-generating systems such as gas and wind turbines.

But turbulence is a highly chaotic process that is extremely hard to predict, particularly when trying to include operational wear of components, says Professor Richard Sandberg at the University of Melbourne’s Department of Mechanical Engineering.
He has been working on the issue in conjunction with GE Aerospace for the past ten years, focusing particularly on turbulence in aircraft engines.

Professor Sandberg says when it comes to designing new turbomachinery components, computational fluid dynamic (CFD) modelling is widely used. But these models are really only accurate in known situations, where the flow behaviour is well recognised and the models have already been calibrated.

If you want to do something different, to go off design, or think of more radical concepts, you can’t really trust these models anymore. And that is where we come in.

He and his team at the University of Melbourne have been developing highly detailed turbulence simulations able to account for micro-scale roughness caused by wear or additive manufacturing processes for GE Aerospace that are so complex they require supercomputers and many days to run them.

We try to reduce turbulence model uncertainty as much as possible, but this makes the simulations very costly. However, the reward is that we can really look at what happens in the flow and how it is impacted by operational wear. We get better understanding of the physics, which helps give GE Aerospace insight.

Each simulation also produces high-fidelity data that Professor Sandberg feeds into a machine learning method his group has pioneered to develop new, more accurate design tools. We mine that data to understand where existing models go wrong and try to design better models that GE can then use in its design loop.

He says all the models he and his team are developing will improve the prediction accuracy of turbulent flows in general and will be equally as useful for systems that generate thrust, such as aircraft engines, as for those that generate power, such as the wind turbine sector.