An app store for machines
Delegates to the idea-sharing conference staged on consecutive days in Brisbane, Melbourne and Sydney watched in real time as Predix ingested the data flow from every airline, worldwide, running on GE engines, and showed their performance in an easily navigated dashboard. One engine alone, says Mark Sheppard, GM and chief commercial officer for GE Digital in the Asia Pacific region, “produces 500 gigabytes of data per flight and for each engine I can see its performance in all the variables we measure, against the performance of the fleet.” Performance anomalies, when they occur, are shown simply by a graph of engine data output against the fleet’s statistical average.
“While this is mainly a visualisation tool to make sense of all that data, we have airlines starting to use it for performance and productivity gains,” says Sheppard.
1% savings in multibillion dollar industries add up
In industries from aviation to oil and gas, GE is hunting down “the power of 1%”, said Geoff Culbert CEO of GE Australia and NZ, speaking at the University of Technology, Sydney. “If you can reduce downtime in the oil and gas industry by 1%, through predictive analytics and the use of data, that’s worth $5-7 billion a year to the global oil and gas industry. If you can reduce fuel burn in the aviation industry by 1%, you potentially save $2 billion to $3 billion a year.”
GE has been using software analytics to reveal aviation fuel-efficiency insights for decades, but it’s the computing power of Predix operating in the cloud that is allowing the company to draw in huge seemingly disparate data sets—engine-performance indicators, weather conditions, atmospheric particulate readings, aircraft loads—and analyse their cumulative influence on machine performance.
“There are many things you don’t want in industrial strength, like coffee …” GE Predix advertising campaign
GE began seeing locomotives, hospital equipment and electricity substations in cloud formations, in around 2009, says its global chairman and CEO Jeffrey Immelt. That is, it started seeing the potential for digitally enabled and connected industrial machinery to be run significantly more efficiently and realised what that could mean for industrial productivity. With its massive installed base of industrial assets, GE wanted to ensure that it would be the company to offer customers those benefits, and to create the software that would become essential to driving machine learning. “The digital initiative ... is the biggest transformation in the history of the company,” says Immelt.
GM and chief commercial officer Asia Pacific for GE Digital, Mark Sheppard, demonstrates Predix at the GE CSIRO Digital Industrial Series conference in Brisbane. Image: Natalie FilatoffIn developing Predix over the past few years, GE has imbued its cloud-based platform with five vital characteristics:
The 5 principles of Predix
- It ingests real-time data, even from dirty, remote sources. “Data from a lot of our customers comes from dirty places. If you’ve got a locomotive hauling ore across the Pilbara, the sensors are dirty, they’re grindy and there are going to be holes in the data. Physically getting that data is tough,” says Sheppard.
- It runs GE’s proven analytics. Sheppard: “We needed a platform that could take in all the analytical routines we’ve built already—and we’ve been making software for a long time. These routines, for condition monitoring for, say, our aviation business and our healthcare space, are now embedded in Predix.”
- It has built-in data security. Compared to cyber security for information technology (IT), cyber security for operational technology (OT) is in its infancy. “There are a lot of vulnerabilities,” says Sheppard, “and as we got into this area of connecting machines, we made sure our platform had the security built in, so that our customers could connect securely.”
- It allows agile building of apps, by incorporating an array of pre-coded functions. Sheppard uses Apple’s iOS as another example of an operating system developers love to work with: “Apple doesn’t make apps itself. It created an ecosystem where others can quickly develop on the platform. iOS does a lot of work for you, handling all of the screen processing for example. That’s what Predix does—a lot of the grunt work is done and lives in the platform, so if you’re an application developer, you can now build an Industrial Internet application very, very quickly.”
- It’s agnostic in its acceptance of data. “We’ve built this platform to connect to pretty much anything; we’ve got ISA 95 (the standard automated interface) connectivity, we’ve got ERP [Enterprise Resource Planning software) connectivity, we can hook into a variety of different machines and sensors, and we can work with our competitors to take data in,” says Sheppard.
“That’s the philosophy of Predix,” he adds, “And this is what it looks like …”
Keeping freight trains right side UP!
One screen of the decision-support tool called “UP” that GE has built on Predix for its freight-rail customers in the US shows bubbles hovering like various sized balloons over the United States: “The size of each bubble represents the size of the cost of a derailment incident, and the colour indicates what caused it,” says Sheppard. The biggest bubble on the screen is red, indicating that rain caused the accident, and it represents a derailment of 53 carriages that cost the rail company $5 million. UP combines eight years’ of accumulated data, integrating weather conditions and train movements, which allows GE to predict future “blowovers”. Also built into the app is the cost of holding a train back in threatening conditions. “So we can calculate the relative cost of a likely derailing, as opposed to the cost of delaying it, and we can model whether it’s better to hold it back,” says Sheppard.
GE’s UP application on Predix helps managers of freight-rail companies in the US understand the risks of running transport in poor weather conditions, and the costs of holding back freight trains at risk of blowover.Many companies, say GE’s experts, make the assumption that they need to collect new data in order to gain valuable insights into their operations. In fact most utilities and companies large and small already have enough useful data to get started in incrementally improving their operations. In factories where equipment has never been connected, even being able to visualise data generated by their operation to date can lead to huge cost-saving improvements to processes and maintenance schedules.
“People collect a lot of treasure, and they don’t know it, they don’t use it!” Fang Chen, group leader and senior principal researcher in Analytics, Data61
Sheppard and his team look for what he calls “money box projects”. He says, “My advice to most of the companies we work with is to work with the data you have today. Let’s get some wins out of that, and then use the savings and the benefits from that to look at what else we need to add data-collecting sensors to.”
He says Industrial Internet hype has led to overspending by many companies in a bid to achieve the promised big gains. “But we are encouraging companies to find small gold-nugget projects to start with. It might be static data they have. We can take an FTP file and put it through the analytics tools, visualise it and do some proof of concepts—that’s very easy to do. And from there we can usually flesh out where the benefits and opportunities might lie before we do any work to hook the machines up in real time and get the whole process working. You’d be surprised what you can find in terms of initial benefits.”