In a show-stopping moment at Minds + Machines 2016, GE Vice President of Software Research Colin Parris donned an augmented reality headset and used voice recognition on the conference main stage to show how digital twin can overlay the digital capabilities for servicing a steam turbine. The technology alerted him to wear detected in the turbine, showed him the trouble spot, and offered several options on how to address the situation based on the asset’s history.
It was an arresting inside look of digital twin at work -- offering a view of the past, present, and future of a critical industrial asset. GE has 551,000 digital twins out there today, with more created every day. “We see the digital twin as a key technology to fully digitize the physical world,” said Dimitri Volkmann, a digital twin thought leader for GE Digital.
The technology goes beyond looking at data; it creates virtual 3D models to understand how equipment will perform and then presents options to extend an asset’s life for better business outcomes.
Digital twin eliminates guesswork from determining the best course of action to service critical physical assets, from engines to power turbines. Moving forward, easy access to this unique combination of deep knowledge and intelligence about your assets paves the road to optimization and business transformation.
In order to create a complete digital model, GE technology digs into any data related to the industrial asset, following the digital thread that defines its lifecycle. The thread starts with the development of a new piece of equipment, say a power generation system or a new jet engine, from the design through the build phase. This thread continues into the operation of the asset and its service history—all to predict what will come next, and suggest improvements and optimization throughout the cycle.
The digital twin is built on Predix, GE’s platform for the Industrial Internet, which enables it to unravel that data and discern what may happen next with an asset, while continuously learning and improving the models. This technology is a natural fit for the power infrastructure and the aviation industry, where unexpected equipment failure is not an option. In aviation, for example, information collected from jet engines and flight recorders can reveal commonalities in asset performance. “We realized that if we analyze this data and crunch it, we start finding patterns,” Volkmann said.
Through digital twins, the need to service jet engines can be determined in advance, and furthermore helps plan ways to extend the use of the asset. After a plane spends much of its operational life in the dry, sandy air of the Middle East, our technology could suggest redeploying the plane in a different clime such as the Pacific Northwest, offering cooler, moist air to reduce risk of engine failure. Another option might be to reroute the plane closer to maintenance facilities that can regularly provide service.
The idea behind the digital twin is to go further than working with models; the costs of maintenance versus replacing an entire asset are also considered. If a company orders jet engines, the revenue projections become part of the digital twin, along with the designs for the engines, specific materials used in construction, and information on the factory where they were built. When the engine starts up, and when it is serviced, sensors feed that data into the twin.
Yet the digital twin technology spans across all industries where the value is in assets and more generally complex systems. Its ability to deliver early warnings, predictions, and optimization is fairly universal. We foresee the digital twin to be applied to humans, playing a significant role in healthcare.
What the digital twin produces, when bundling this data with intelligence, is a view of the each asset’s history and its potential future performance. This continuum of information leads to early warnings, predictions, ideas for optimization, and most importantly a plan of action to keep assets in service longer. The future of digital twin, Volkmann said, will be about sending commands to machines in response to those forecasts. “If you close the loop, with data and predictions, you can act directly on the asset itself,” he said.