Listen. The machines are talking. And as they communicate with each other (given we’re the ones who’ve designed the Industrial Internet for useful eavesdropping), we tune in and discover innumerable opportunities to improve they way they can work for us. David Parkinson, Regional IT Leader at GE Australia & New Zealand Oil & Gas, is something of a Dr Dolittle with these talking machines—he knows how to learn from their conversations. He explains how we we can engage in this data dialogue, and why we must.
GEreports: What’s the machine conversation starter?
David: You’ve got the machine sending a whole bunch of data which is essentially zeros and ones—binary code—and you’ve got them connected to a back end which is where the analytics happen, or you’ve got a processor running to say, “Hey, this is the right data that I’m getting. It’s functioning right.” Or “This is the wrong data. I need to change something.” And then sending the result of that data analysis back to the machine to tell it: “Slow down” or “Speed up” or “Turn left, or right, or off, or on”. That’s where we start talking about the Industrial Internet.
GEreports: How quickly has the Industrial Internet exploded?
David: GE has been selling remote monitoring and diagnostics technology for a long time, 20 plus years. You’d have a piece of equipment, put a sensor on it, work out how that sensor will transport or network its information back to somewhere—whether it’s into a big processing HQ or just to a terminal on that machine. This is something that we’ve been doing for ages.
GEreports: What’s different about the way that technology is being used now?
David: The next level of remote monitoring and diagnostics is based on expanding your system to be not just that one machine with its one sensor, but thousands of machines, either in that system—the machine next to that machine—or with lots of those same or similar sorts of machines and applying analytics. In aviation, for example, GE has sensors on all of our jet engines. We make thousands of jet engines. We have the sensors on all of those engines—hundreds of sensors per aircraft. We can pull all of that information back and look at that data.
The point is to run software that connects machines, analyse the data and then make better decisions using that data.
It’s easy when you’re the original equipment manufacturer [OEM]. When GE makes the equipment, we know the physics, the chemistry, the design and engineering. We know it inside out. So when the sensors on it send the data back, that data can be analysed in context of the piece of machinery, and really good decisions can be made around what needs to be done to make the machine work better. It’s a GE-built machine, so when that 1 and 0 binary code comes through, we know if that reading is related to vibrations, or heat, or oil heat, or wind velocity, and whether that’s a good signal or a bad signal.
GEreports: For the Industrial Internet to be truly transformational, doesn’t this data sharing need to open up across rival manufacturers?
David: Yes, that’s it. The next stage in the Industrial Internet is when we get lots of machines that are kind of the same, from different manufacturers, allowing their data to be collected and compared. So then if there are six machines and they are all giving us a 1, but the seventh machine is giving us a 0, we can assume that the one that’s giving us a 0 is the odd one out; there might something wrong with it. You can take someone else’s equipment, put a model over the data and start understanding what those data points are telling you about its behaviour.
David Parkinson instructing a GE employee’s son during a session of the company’s Coding for Kids initiative.
GEreports: Then it’s about asking questions?
David: Yes. GE sees the Industrial Internet as monitoring sets of data and having sets of models across all types of industries, and then contextualising it depending on who’s asking for that data. So if I’m a pilot, the questions I ask are: “How much fuel am I using at the moment? Have I got enough fuel to get to this next destination? Do I have to adjust my speed to slow down so that I’m burning less fuel, so that I’ve got less risk of running out of fuel on my flight?” They’re the types of questions that I would ask of that dataset.
Whereas if I’m GE, or another manufacturer, I might be asking of that dataset, “How can I make this plan fly further? How is that part performing over time? Do I need to go to a new supplier or material because we’re finding failures on that same part?”
Same data, same dataset, different question. But because we’re gathering all of that data and putting it in one place, we can start asking different questions depending on your use case, who you are and what context you want to put it in.
GEreports: So the machines talk, and the collected conversations of those individual ‘chats’ build into something much more powerful?
David: Absolutely—that’s what’s changed. We’ve invested in Predix, a software platform, to do it. We’ve built all these tools to support our customers with our own equipment, why can’t we apply that same intellectual property across other manufacturer’s equipment?’ The point is to run software that connects machines, analyses the data in the machines, and then helps the customers who run those machines make better decisions, using that data.
GEreports: And this is already happening—other companies are hooking up other people’s machines to GE’s software platforms?
David: Yes, it’s happening everywhere. Predix has been built with other manufacturer’s equipment in mind. We’re putting our knowledge as an OEM into these software products to monitor our own equipment, and then wrapping up a system for our customers, because they won’t have only GE equipment.
GEreports: Can you sketch out an example of how it can work?
David: Think of a mine site where you’ve got all the machines getting, say, iron ore out of the ground. A truck comes and picks it up and drives it to get processed onto the back of a train. That train then travels across Australia to the nearest port. When it arrives at the port, the ore has to get loaded off the train and put onto a boat, and then that boat heads for China, or whereever.
There’s only a small benefit in us just giving a mining company some advanced analytics on how that truck going around the mine could be better utilised and run longer and tip more coal if the train’s not there when it needs to be for that truck to tip the coal into and so on. You’ve got to create a pit-to-port solution, and that’s what the Industrial Internet will bring. It will schedule the boat so that the boat is arriving as the train is pulling up. The boat’s not going to be sitting in the harbour for two days waiting for the train to come.
GEreports: And these principles apply for other industries, too?
David: Absolutely. One of the things we’re tackling in healthcare, for example, is hospital equipment. Where is that mobile equipment in the hospital when you need it? We have a system that tracks it, and enables maximum utilisation of these valuable assets, saving huge amounts of time and money.
It’s the same in every industry. It doesn’t matter whether you’re running a coal mine, or a hospital, or an oil rig, or a taxi service. Are my assets working as hard as they possibly can, being as productive as they possibly can be right now? Where are they? What are they doing? Then you’ve got asset management, which is the next level. When do I need to fix that asset? How is it performing at the moment? Can I see that it’s not going to work properly, and what do I need to do next to make sure that it continues to be as productive as possible? How can I plan or schedule that downtime?
You can put that concept over any of the businesses that we work in or work with. That’s machines “talking”: giving us the data to enable us to work out how to use them better—and that’s the heart of the Industrial Internet.