Defining the Digital Twin and its Applications

This post first appeared in ARC's blog, Viewpoints.

In recent years, the term digital twins has been in the limelight, creating a buzz across all industries.  At the ARC Industry Forum in Orlando, ARC’s Greg Gorbach, an industry leader in digitalization and IoT, and Chad Stoecker, Global Vice President of Industrial Managed Services for GE Digital, discussed defining the digital twin.  This blog captures some of the salient points and quotes of their conversation.  The interview can be viewed in entirety in the video below.  

“Every software company has some version or other of the digital twin, and probably some other companies as well.  How does GE Digital define digital twin and what technologies do you bring to the party?” was Greg’s opening question. 

“I really think a digital twin is defined by the outcomes it's trying to achieve.  Using ARC's framework on a digital twin, it is composed of either engineering technology, which tends to be process simulation or 3D models, or operational technology, or information technology.  We really define our digital twin as focusing on enabling the lifecycle of decision-making for field service - the integration of operational technology and information technology,” responded Chad.

Substantiating the value of digital twins, Chad said that GE Digital helped customers manage more than 8,000 assets in their center last year, and this helped generate $187 million of customer documented savings in 2019. The company has assisted customers with a range of digital twin applications - for jet engines in flight; submersible pumps in oil wells; turbines on power plants; and packaging palletizing machines in manufacturing. The application and content configuration is different for every customer and use case.  “Ultimately we're helping customers move corrective maintenance to predictive maintenance.  We're helping customers take their strategies from being defined once every five or 10 years to being defined and optimized in real time,” explained Chad.

Greg asked about those who are in the early stages of their digital transformation journey.  For them perhaps the digital twin would seem far-fetched, almost like science fiction.

“Today, you are seeing the industry move from digital twin as a pilot or as a use case or case study to digital twin at scale, transforming whole organizations.  I think the key is to make sure you go with an experienced vendor who can size the technology for your particular application,” responded Chad.  One of the most important things for customers to understand is the holistic risk ranking of their equipment.  For some types of assets, the most economical decision is a run to failure model.  But other assets are more critical, and that's where the more critical technologies can be applied.  The right technology must be used for the right application to make sure you get fast ROI.

Unplanned downtime is really about improving reliability - digital reliability, explained Chad.  The same techniques and standards that have always been in play, but new technology really allows it to be applied at scale, allows you to digitize the work process to ensure that you're doing it.  The value that you get by leveraging these machine learning models is early warning.  

“Today, customers might react to hard alarm limits.  When something hits an alarm, then they take action.  But by that time, the failure might have already exhibited secondary and tertiary effects that are really costly.  So what the machine learning can allow you to do is dive inside the hard alarm limits, without being inundated with false alerts.  You can start to take action when the model and the actual signals deviate from the model, as opposed to waiting for the hard alarm.  That difference might be hours, it might be days, it might be weeks, it might be months, but it allows maintenance organizations to go into an economic planning cycle,” said Chad.

The company’s core machine learning technology was spun out of Argonne National Labs 20 years ago.  It has been creating digital twins for more than 15 years, and that digital twin blueprint (over 300 types) now has millions of asset-run hours gone through it.  “So, we bring that blueprint to the table.  That really allows the customers to get ROI on their solution much faster.  Reliability is a multi-vertical discipline, so a lot of the blueprints will apply no matter what vertical we approach,” said Chad.

“It's a key component, but not the entire suite,” said Chad.  A digital twin encompasses the configuration required to do a complete plan and action for an asset’s lifecycle, and a key component of it is machine learning.  Today, customers have more data than ever before.  Sensors are getting cheaper and you can take video feeds with drones etc.  Chad concluded, “Where machine learning works really well is to pre-filter through all that data, identify causes early, suggest diagnoses, and really help serve up the information to make your reliability engineers more effective.”