Researchers at the University of Wisconsin–Madison are advancing digital-twin technology from a passive replica to an active decision-support tool that helps engineers optimize systems and predict future behavior. Harsh Apurva Sharma, an assistant professor of mechanical engineering focused on computational science and dynamic systems, explains that a digital twin begins as a computer model of a physical asset, such as a turbine, aircraft wing, energy grid, or factory system, and evolves as sensors feed real-time operating data into it. With frequent updates, the digital representation reflects the physical asset’s current condition, enabling engineers to analyze performance and make informed choices about maintenance, efficiency improvements, and system changes.
Digital twins differ from traditional simulations because they incorporate dynamic data from sensors on the physical object and update continuously. This bi-directional interaction lets engineers run “what-if” scenarios using the digital twin to see how systems might respond to changes before making adjustments in reality. For example, sensor data from an aircraft wing accumulated across many flights can reveal subtle structural degradation. That evolving profile helps maintenance planners decide what interventions are needed and when.
Today’s digital twins are already deployed in aerospace and manufacturing. Companies such as GE and Rolls-Royce build digital twins of jet engines to schedule predictive maintenance based on in-flight data, and automakers use twins of production lines to test layout or process changes before disrupting operations. The technology’s reach extends beyond industry: researchers are exploring digital twins for climate modeling and disaster planning.
Sharma notes that digital twins still face challenges, especially when data are sparse or incomplete. Highly predictive models require data that may not be directly measurable, and indirect data must be interpreted carefully. A major research frontier is making digital twins not just descriptive but truly predictive so they can forecast performance under unseen conditions.
In sum, modern digital twins act as dynamic decision engines, blending physical sensor input with computational models to guide choices about performance, maintenance, and future scenarios. Engineers are moving beyond static digital replicas to systems that add real-world context and predictive insight.