It’s all about speed. Predix, a cloud-based platform for analytics, management, and app development, went live to the public in February. GE hopes to spur interest by creating a fast, plug-and-play way to bring sensors and machines on board.
The kits are being announced today in beta at Predix Transform, the company’s first developer conference for the platform. As of last week, the conference had nearly 1,700 attendees signed up for the event, being held in Las Vegas.
More than half of those could be GE employees, estimates Greg Petroff, GE’s chief experience officer — but the company is OK with that. Predix went live inside GE just two years ago, and not all of the company’s 14,000 software developers have even heard of the platform, he says.
In fact, that’s another reason for the kits: to help spread the Predix gospel internally.
The platform operates in the cloud, doing calculations on a computer model of the item being analyzed, be it a wind turbine, a factory machine, or the soil at a vineyard. In GE parlance, it works on a “digital twin” of the “asset.” Predix can also apply GE algorithms, such as models for the laws of physics.
That’s great for running what-if scenarios or figuring out how quickly an engine will wear down. But the real utility comes from adding information about how the real-world asset is doing. Configuring that edge device — tweaking the Predix edge software and getting the device to connect back to Predix in the cloud — can take weeks, GE claims.
The company wants badly for this step to not be a stumbling block in the adoption of Predix.
In less heavy industries — winemaking would be one example — the platform could prove its worth quickly if GE could “make it fast and really straightforward to get connected,” Petroff says.
The kits are designed to be simple and inexpensive. They include a hardware board with a small microprocessor; so far, kits come in Rasberry Pi, Intel Edison, or GE Field Agent configurations. When connected to a laptop, the board’s software steps you through installation, which includes scripts to provision the connection to the cloud. It’s a no-brainer process that can take about 15 minutes, GE claims.
The future of the kits is partly in the hands of developers, as GE would like to see what its customers do with the concept. The hope is that developers from different industries will build specialty kits, using GE’s software. In fact, the whole of Predix is meant to be open, so that developers could even build their own digital-twin models if they wanted to.
“It’s likely that we’ll have people who aren’t in GE building algorithms as good as, or better than, the ones we have in GE. We’re completely OK with that,” Petroff says.
Items on GE’s to-do list for the kits include a certification program to approve hardware vendors to run the Predix stack and ruggedized versions suitable for heavy industrial uses.