General Electric (GE) was co-founded in 1897 by Thomas Edison. Today, 120 years later, GE is the single company with the longest continual presence in the Dow Jones Industrial Average, and is undergoing one of the most dramatic transformation initiatives of any major company. Mainstream legacy businesses should take note. In a matter of only a few years, GE has migrated from being an industrial and consumer products and financial services firm to a “digital industrial” company with a strong focus on the “Industrial Internet” and $7 billion in software sales in 2016.
This is the story of how GE has accomplished this digital transformation by leveraging AI and Machine Learning fueled by the power of Big Data.
The GE transformation is an effort that is still in progress, but one which is increasingly looking like a success story, as chronicled in the 2016 MIT Sloan Management Review story GE’s Big Bet on Data and Analytics. GE’s software offering, Predix, has become well-established. Less well-understood is GE’s focus on analytics and AI to make sense of the massive volumes of Big Data that are being captured by its industrial devices. Bill Ruh, the CEO of GE Digital and the company’s Chief Digital Officer, emphasizes the role and importance of data and analytics in the company’s transformation. In a recent blog post on the GE site, Ruh wrote about Waking up as a Software and Analytics Company. Ruh observes, “It’s not enough to connect machines. You have to make your machines smarter. You need to figure out the best ways for embedding intelligence into machines and devices. Then you need to develop the best techniques for collecting the data generated by those machines and devices, analyzing that data and generating usable insights that will enable you to run your equipment more efficiently and optimize your operations and supply chains.” This is how companies become data-driven organizations.
In a recent interview that we conducted with Ruh, he emphasized the importance of Machine Learning as one approach that has been particularly beneficial in helping GE leverage the power of Big Data and the Internet of Things (IoT). Machine learning technology, according to Ruh, is critical to making the “digital twin” concept successful. A digital twin is a digital replica, or data-based representation of an industrial machine. When sensors in those machines — for example, a jet engine, a gas turbine or a windmill — gather data on the machine’s attributes (heat, vibration, noise and the like), the data is collected in the “cloud” and organized into a model “twin” that allows analysis that replicates the machine’s performance. The digital twin model can then be used to diagnose faults and predict the need for maintenance, ultimately reducing or eliminating unplanned downtime in that machine. The digital twin concept can be extended to aggregations of machines — a plant or fleet can be digitally twinned as well.
The data never stops flowing into these digital twin models, which can be populated by many unique variables. Because there may also be changes over time relative to which variables and models best predict the need for required maintenance, machine learning represents the best technology approach to addressing these requirements. Machine learning approaches make it possible to learn from new data and to modify predictive models over time. Ruh points out that machine learning makes it possible to identify anomalies, signatures and trends in machine performance and develop understanding of patterns of behavior. In addition, Machine Learning can be applied to help identify efficiencies within a machine and use this as a best practice for other machines.
