GE, one of the UK's largest manufacturers, is using big data analytics to predict maintenance needs.
GE manufactures jet engines, turbines and medical scanners. It is using operational data from sensors on its machinery and engines for pattern analysis.
GE is using the analysis to provide services tied to its products, designed to minimise downtime caused by parts failures. Real-time analytics also enables machines to adapt continually and improve efficiency.
Bill Ruh (pictured) is vice-president for software at GE Research. Recruited from Cisco two years ago, Ruh has a customer-facing role, looking at the business opportunities enabled through the industrial internet.
Ruh said: "The airline industry spends $200bn on fuel per year so a 2% saving is $4bn. GE provides software that enables airline pilots to manage fuel efficiency."
Another product, Movement Planner, is a cruise control system for train drivers. The technology assesses the terrain and the location of the train to calculate the optimal speed to run the locomotive for fuel economy.
Software developed at GE is now being used by a Canadian electricity supplier to prune back trees cost-effectively along its electricity distribution lines. Vegetation falling on power lines is the biggest cause of electricity outages, according to Ruh.
One third of GE's business is about servicing equipment. He said: "Most customers do planning in the rear-view mirror, which is almost always wrong." He said heavy industry can make considerable savings moving to zero downtime and this is the main focus of his work.
"We invested $1.5bn over four years to develop services and create new software. We are working on making devices more intelligent using sensors; and controllers that can be configured in real time," he said.
For instance, in a wind farm the wind turbines at the front affect the turbines behind them. This may lead to vibrations causing a failure due to a stress fracture on the turbine's blades. "We can adjust turbines based on the wind. We adjust the blade in real time to avoid vibration." As a result he claims GE is able to deliver a 2 to 5% improvement on the efficiency of the wind farm.
He said the amount of data generated by sensor networks on heavy equipment is astounding. A day's worth of real-time feeds on Twitter amounts to 80GB. "One sensor on a blade of a gas turbine engine generates 520GB per day, and you have 20 of them."
Predictive analytics is something IT systems management tools aspire to through self-healing and autonomous computing. Ruh said: "The industrial world never picked up on datacentre automation." However, now there is a greater demand for such technology as people with many years of experience retire.
Behind the scenes
In Ruh's experience, the internet of sensors is very different to the internet used by humans. He said: "The internet is optimised for transactions, but in machine-to-machine communications there is a greater need for real time and much larger datasets." Cloud computing in its present form is not wholly suitable for these kinds of machine-to-machine (M2M) interactions. "We are seeing more processing on machines, due to latency," he explains.
The technology is based around in-memory database systems. Since the datasets are extremely large he said GE uses NoSQL and Hadoop. "We have also developed our own database for time series analysis." He said GE is also working with Microsoft Azure and Amazon Web Services to investigate how to offload processing to the cloud.
The machine, like a human, need to know about its location and is capable of posting status updates. Some may even be "friends" in the social networking sense. According to Ruh, a GE engine on an aircraft could alert ground engineers when the lane has landed, through its social network of engineer friends. Potentially, this would allow the engineers to be alerted that the aircraft is on the ground, so the ground team can carry out any maintenance that the engine has alerted them about through status updates.
Given the potential for a StuxNet attack, it is no surprise Ruh is concerned about the security impact associated with M2M communications and intelligence in machines. His biggest concern is the insider threat. He said the machine needs to understand about the user, the human who is accessing its controls. This suggests it requires authentication and sophisticated role-based security, the kind of technology associated with commercial operating systems. Ruh points to Apple as an example of how security can be managed. He say: "Apple defined and locked down iOS ahead of time. We can apply such a sandbox approach in our system."
On the user interface side, he said GE has standardised on HTML5, but at lower layers of the software stack, the company works with Android, Windows and Linux. Ruh predicts that machines will have apps, rather like iPhone and Android apps to extend their functionality, although the programming model will be very different, and clearly there will be a need for a robust security framework before an app will be allowed to run on a machine.
The role of data scientists in the future
GE has established a 200-person site in Silicon Valley specialising in developing services, powered by software and sensor networks to support its vision of zero downtime and real time controls for machine efficiency. While it is a software centre, and programming is key to the development of these new services. Ruh is seeking new skills.
He said: "We need data scientists. It is the most in-demand skillset and we're looking for a certain kind of person." According to McKinsey, 1.5 million data scientists are needed globally.
"The technology providers gave us a solution to business intelligence but analytics is a lot harder than people think. Ten to 15 years ago people were told BI gives greater insight, but the BI tools have been used just for better reporting." This is where the data scientists role fits. Ruh is looking not merely for great mathematicians or engineers, but for people who can understand the meaning of the data sets when applied to a machine operating in the real world.
Two years ago IBM demonstrated how its Watson supercomputer could outsmart a human by winning the US game show, Jeopardy. No matter how sophisticated such a machine or how it is programmed, Ruh does not think it can answer the kind of questions Ruh' team of technology researchers are trying to crack. He said: "Watson cannot tell me when this machine part will break." He said predictive analytics builds upon Watson-like machine-learning capabilities.
Given GE's footprint in the UK, Ruh is hoping to establish a similar facility in the UK. Cambridge seems to be the preferred location at the moment.