General Electric has a new name for where it thinks its business is headed: the “industrial Internet.”
The term, coined inside GE’s R&D division, reflects the company’s hope that adding more sensors to machinery will result in a deluge of data that will in turn let companies squeeze more efficiency out of locomotives, jet engines, MRI machines, and other equipment GE sells.
GE says it is investing $1.5 billion in the idea over a three-year period. Some of that money is being spent on research at a large new software R&D center the company has created in San Ramon, California.
At the San Ramon center, for instance, machine-learning researcher Anil Varma has been experimenting with ways to sift out subtle warning signs that can predict which of the 20,000 GE jet engines in service will need maintenance. For some types of engines, he says his algorithms can identify those needing attention with 70 percent accuracy and a month ahead of time, which could help prevent costly flight delays.
Most equipment sensors are still used reactively—a light on a dashboard might turn red if something’s wrong. And GE’s older jet engines do plenty of such sensing—measuring things like temperature, pressure, and voltage. Historically, however, not much of the information tracked that way has been retained and studied. For most flights, Varma says, three averages—one each for takeoff, cruising, and landing—are the only engine data kept.
With products such as its next generation of GEnX engines (which will be used on the Boeing 787) the idea is to eventually retain all the original readings from every flight, and possibly transmit them from an airplane in real time, Varma says. He says those engines will produce more data in a single year than GE has collected in the history of its aviation business.
Although the idea of ubiquitous sensors that communicate information from machinery isn’t new (it’s sometimes also called the “Internet of things”), the huge scale of GE’s business could help speed that vision to reality. “We have some of the biggest industrial data sets, because we’ve been operating this equipment for a long time,” Varma says. “We have the before and after and can test any algorithm and see how it works.”
GE’s idea for the industrial Internet took root during the economic crisis, says William Ruh, the former Cisco executive hired to lead the effort (and who GE says coined the term). With economic growth uncertain, industrial clients began looking to productivity gains to boost their profitability. They were also hearing more and more about “big data” and had begun asking GE whether it had a “data strategy” for its equipment.
GE may find it’s not always easy to apply big-data techniques to industrial problems, says Venkat Venkatasubramanian, a Columbia University professor affiliated with the school’s Institute for Data Sciences and Engineering. For a commercial retailer, he says, it might be enough to discover a correlation in customer data—say, that a person who buys beer also buys diapers. “There is some low-hanging fruit, which the standard machine-learning algorithms will help you detect and solve,” he says. But in more complex physical systems, models will also need to explain why such associations exist.
At GE’s San Ramon center, researchers are developing new user interfaces that can help people visualize industrial data with the help of maps, simulations, and Twitter-like social networks for equipment. One room is filled with large display screens hooked up to a Microsoft Kinect, a video-game controller that senses a person’s movement (see “Microsoft’s Plan to Bring About the Era of Gesture Control”). Researchers there demonstrated how a utility worker could use hand gestures to swim through data visualizations to help make decisions about his section of the power grid. GE says it is also working with a Canadian utility company to predict tree-trimming hot spots using satellite images combined with maps of weather conditions and locations of past outages. Falling tree limbs are a major cause of storm-related power failures.
It’s not only infrastructure getting the software treatment. New York City’s Mount Sinai Medical Center is working with GE to put sensors and transmitters on hospital beds and equipment, to keep track of which are in use. GE claims the 1,100-bed hospital could admit 10,000 more patients a year if it had better information.
Ruh believes that even small improvements can have a big impact. GE released a report this week estimating that a 1 percent gain in fuel efficiency could be worth $2 billion a year to the aviation industry and twice that to the power industry. GE’s gas turbines and other utility equipment are involved in delivering 25 percent of the world’s electricity.
“We know, operationally, that we can change 1 percent,” says Ruh. But, he adds, “it’s not going to be done anymore through better devices, because we are reaching the end of what physics can give us.”
This new data poisoning tool lets artists fight back against generative AI
The tool, called Nightshade, messes up training data in ways that could cause serious damage to image-generating AI models.
Rogue superintelligence and merging with machines: Inside the mind of OpenAI’s chief scientist
An exclusive conversation with Ilya Sutskever on his fears for the future of AI and why they’ve made him change the focus of his life’s work.
Data analytics reveal real business value
Sophisticated analytics tools mine insights from data, optimizing operational processes across the enterprise.
Driving companywide efficiencies with AI
Advanced AI and ML capabilities revolutionize how administrative and operations tasks are done.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.