To gain the organizational insights and greater agility the mirrored world promises, you first need a comprehensive and robust data foundation for your twins. When intelligent twins are connected in mirror environments, they are a powerful way to turn data into actionable, big-picture insights. But incomplete or incorrect data will lead to false conclusions.
High-quality historical data is critical for intelligent twins — it’s what the twin uses to monitor real-time machine performance, build models of business processes and high-value assets, and more. But COVID-19 has made historic data increasingly unreliable. Everything from traffic and shopping patterns to energy consumption and international travel changed abruptly due to the pandemic. These anomalous changes in behavior and activity patterns have sent many machine learning models that have been trained on “normal” behavior off course, impacting supply chains, inventory management, marketing, and more. Going forward, enterprises cannot rely on historic data blindly — they need to check and correct their models as the world changes.
On top of historical data, federal agencies need a strategy for real-time data collection, or they'll miss out on the real-time analytics intelligent twins can provide. There are two sides to this: investment in sensors and IoT devices to collect data and the tools to prepare, analyze, and visualize the massive amounts of information gathered. Today, many agencies are already investing in IoT devices and sensors, but some struggle to fully utilize the data these devices generate. New cloud-based services and platforms are being developed to bridge this gap and help enterprises achieve real-time insights. Snowflake, for instance, which Barron’s recently described as a “growth juggernaut,” offers clients data warehousing as a service, which can load continuously generated real-time data, requires no manual effort, and can even digest semi-structured data.
From there, intelligent twins can make real-time data actionable in the moment, as many of the examples above illustrate. Going even further, some enterprises are starting to explore how multiple intelligent twins, connected in mirror environments, can use real-time data to safely increase autonomy. GEMINA (Generating Electricity Managed by Intelligent Nuclear Assets) is a U.S. Department of Energy program funding research projects that use AI and digital twin technology to increase the flexibility and autonomy of nuclear reactor systems and reduce operation and maintenance costs. Two of the projects to receive funding are tied to GE Hitachi’s BWRX-300 boiling water reactor design. GE Research intends to move from time-based to condition-based predictive maintenance, which will lead to significant savings. To make this possible, they will develop an array of digital twins for continuous monitoring, diagnostics, prognostics, and early warnings for the reactors. They will also develop a "Humble AI" framework that defaults to a safe operation mode when confronted with situations the algorithm does not recognize. In doing so, the system ensures the secure handling of uncertainties and increases the feasibility of more autonomous operations.
As they continue building out their mirrored worlds, agencies will also need to think about data integration across multiple twins or multiple sub-components that feed into a single twin. API connections can help achieve that data synchronization, enabling different twins or components to connect and interact.