Super Computing, Advanced Manufacturing, and the Industrial Internet

December 20, 2013 | Author: GE Software |

II_post_122013

The problem: for the past two years, icy weather in New Brunswick, Canada, has wreaked havoc on Caribou Wind Farm’s wind turbines. The accumulation of ice on the surface of the blades has ground operations to a halt. Waiting on a sunny days just isn’t an option.

Enter the Titan Cray XK7, a top-notch supercomputer at Oak Ridge National Laboratory. Scientists at GE Global Research are using the immense power of the XK7 to model how ice molecules form on a surface. It’s capable of modeling molecules that are only 50 nanometers in size and disappear in one femtosecond (one quadrillionth of a second) . By dynamically simulating how molecules behave individually on a surface and with each other, the supercomputer can help guide research and engineering–in this case scientists will try to figure out what makes some surfaces friendly to ice formation versus others. This level of detail is impossible to achieve in real-life experiments. The idea is that supercomputing simulations will reveal a way to deter ice formation on the surface of turbine blades that could lead to significant cost reductions throughout the energy industry.

But supercomputing’s impact goes beyond solving large-scale break-fix issues. Modeling highly complex systems also helps optimize and innovate on existing designs. Wind turbines in general have a problem called aerodynamic blade noise. It’s a limiting factor in designing efficient wind turbines, constraining rotor diameter, speed, and the overall number of turbines in a farm. Reducing blade noise even slightly, by let’s say 1 decibel per turbine, would result in a two-percent annual energy yield increase per wind turbine. According to GE Global Research, that could result in an extra 5GW of additional power capacity, enough to power several large cities combined.

The GE Global Research team set out to tackle this blade noise problem by partnering with Sandia National Laboratories in Albuquerque, New Mexico. The powerful Red Mesa supercomputer allowed the team to use high-performance computing (HPC) to create models that dynamically predict blade noise in the presence of turbulent air flow. The end result could lead to new turbine design informed by the adjustment of complex and multivariate factors that are normally too difficult to control in the lab. Wind tunnels and field measurements, which are ultimately used to test a turbine’s design, would be greatly enhanced by prior HPC simulation.

For advancing industrial research through the use of supercomputing, GE has won awards from HPCWire and International Data Corp (IDC). But even more important is the trend-setting case that supercomputing can provide actual, tangible benefits to industry by giving engineers the tools to model complex and realistic environmental conditions in a controlled space. Ultimately, more informed design will equal much better products.