Sensor + Large DATA: GE builds industrial Internet

With large data and sensors, GE hopes to create an industrial system of the Internet to further improve energy efficiency.

GE has a resounding name for future business--the industrial Internet. This is the terminology coined by GE's Research and development department, which hopes to gain massive amounts of data by adding more sensors to its products, and ultimately to help the company improve the energy efficiency of its locomotives, aircraft engines, and nuclear magnetic resonance equipment.

GE plans to invest $1.5 billion in "industrial Internet" projects over the next three years, some of which will be used to support research projects at the new software Research and Development Center in SANTA Lamon, Calif. For example, Anil Varma, a machine-learning expert at the research center, is experimenting with how to sift through the imperceptible warning signals in 20,000 of GE's production jets to predict which devices need to be maintained. For some model engines, the Varma algorithm can predict its maintenance requirements one months in advance, with a forecast accuracy of 70%, which can greatly reduce flight delays. (Editor's note: The engine fault has been notified more than once before boarding)

Large Data Engine

There are many such sensors, such as measuring temperature, pressure and voltage, which are rarely retained and studied in the past. In most flights, the engine retains only three averages, namely takeoff, cruising, and landing data.

According to Varma, GE's next-generation GEnx engine (equipped with a Boeing 787 aircraft) will retain all the underlying data for each flight, and will even be transmitted back to GE Analytics from the aircraft in real time. Such an engine would generate even more data in a year than all the data in GE's aviation business history.

While it is no longer a new concept for machines to communicate through sensors (such as the internet of Things), the size of GE's business can make this idea faster. "We have the largest industrial data sets because we've been running these devices for a long time," Varma said. "We also have historical data and monitor future data, which allows us to test the viability of any algorithm." ”

According to the head of the GE Industrial Internet project, former Cisco executive William Ruh revealed that GE's industrial Internet vision was born a few years ago during the financial crisis. As economic growth uncertainty increases, industrial customers are turning their attention from higher productivity to higher profit margins. The concept of big data is becoming more and more popular, and in the end, William Ruh's team began to consider whether it was time to develop a "data strategy" for GE products.

Researchers at GE Software Research and Development center demonstrate data visualization tools for power plants

Break through the physical limit of 1%

Venkat Venkatasubramanian, professor of data science and Engineering at Columbia University, argues that GE's use of large data technologies to solve industrial problems may not be straightforward. For a commercial retail company, it is enough to find a link between consumer data, such as the famous beer diaper theory. In this primary application, the current standard machine learning algorithm can be qualified. But for complex physical systems, the data model also needs to be able to explain the reasons behind the association.

At GE's Ramon Software Research and Development center, researchers are developing new user interfaces to help people visualize industrial data through maps, simulations, and social networks of devices like Twitter. One of the labs has many large-screen displays that connect to Microsoft's Kinect gaming devices, and workers in power plants can interact with gestures and data visualization to help make regional grid operations decisions.

GE also works with a Canadian power company to predict tree-trimmed hotspots by analyzing satellite imagery, weather maps, local blackout records, and other data (falling branches are one of the main causes of blackouts caused by lightning).

In addition, GE has worked with a medical center in New York to implant sensors in beds and medical equipment, reducing the rate of empty beds and improving the reception capacity of patients.

Ruh believes that even small progress can have a huge effect. GE's report this week estimates that the airline industry saves $2 billion a year for every 1% per cent increase in fuel efficiency, while the energy sector saves 4 billion of dollars. You know, GE's oil and gas pipelines and power plants carry 25% of the world's electricity every year.

"We can improve our operational efficiency by 1%, but that goal has not been achieved through better equipment because we have already put physics to the extreme." "Ruh said.

Obviously, GE already has the answer, that is the sensor + Big data "industrial Internet".

(Responsible editor: Schpeppen)