Big Data Meets Additive Manufacturing 

When you think of the aerospace technologies, the term “cutting-edge” doesn't often lag far behind. Even so, GE Aviation's new jet engine fuel nozzle is helping to take that association one step further: rather than being made from 20 different parts, it's manufactured in a single step with a 3D printer, which has made it 25 percent lighter.

But because the nozzle's base material is made in a way unlike anything else in the industry, engineers are in a pinch to establish effective measures of quality control. Put simply, because the parts are made by the relatively untested process of superheating metal powders, the product's microstructure along with its physical properties are uncertain.

As lasers heat the alloy of cobalt, chrome and molybdenum to over 2,250 degrees Fahrenheit, there are plenty of opportunities for error.

“We are dealing with a microscopic weld pool that's moving at hundreds of millimeters per second,” says Todd Rockstroh, a mechanical engineer at GE Aviation. “Every cubic millimeter is a chance for a defect.”

And when you're 30,000 feet above ground, you can't afford to have a defect.

So coming to the rescue is big data, along with complex analytics that can help to better monitor and analyze parts during the laser sintering process. Perhaps most importantly, the technology will be able to monitor any temperature abnormalities that could structurally compromise the nozzle. “When the weld pool is too small, things could be colder than they should be. When it's too big, it could be too hot,” Rockstroh says.

In the past, welders have monitored their weld pools through shaded glasses by actually listening to the sound of the sizzling metal. More recently, advanced techniques involving infrared sensors, cameras and pyrometers have overtaken these methods, which GE is now taking advantage of in order to create a holistic view of what's happening during the laser sintering process. But as the terabytes accumulate, having a sophisticated analytics program to sift out the useful data is the trick.

“We are talking about monitoring large parts that take anywhere between 10 and 100 hours to produce,” Rockstroh says. “That's when it gets real tricky. It is critical to know how each cubic millimeter is being built and not trust that you are good enough at process control.”

GE has already estimated that the big-data-enabled “in-process” inspection could increase production speeds by 25 percent, while cutting down on inspection after the building process is complete by that same 25 percent. As orders for GE's next-generation LEAP jet engine come in, the company plans to print 100,000 components to meet industry needs.

Full story at the Txchnologist