A digital twin is a digital representation of a real-world entity that contains both the structure and the dynamics of the real-world entity. It is connected to its real-world counterpart, so it continuously updates itself representing a near real-time status of its real counterpart. The Digital Twin also contains the history of the existence of the real entity. Digital twins of real-world entities, like manufacturing facilities, distribution systems, transportation systems, public utilities, like power and water supply networks, and even entire cities are enabling a wide range of innovative services.
The structure, dynamics, and data enable simulation of the system. Simulating “what-if” scenarios helps designers, planners, and decision-makers to understand the behavior of the system. It allows them to virtually “see” the future of the real entity. Incorporating artificial intelligence, machine learning, and software analytics, the Digital Twin provides the ability to learn from its own operational data, from human experts, from other similar entities, and from the environment. Thus, the Digital Twin makes it possible to predict the behavior of the real entity. GE has reported that the Digital Twin has been used to predict the remaining life of a turbine blade on a specific aircraft engine with great accuracy.  The Digital Twin also provides a safe environment for experimentation and training.
Through the Digital Twin, decision-makers at various levels, like operators, planners, designers, and even process control software have access to the latest process plans and operating rules and conditions. The Digital Twin, therefore, makes it possible to transcend information silos, providing the right information at the right place and time. According to Dr. Michael Grieves, Digital Twin supports conceptualization, comparison, and collaboration.
The primary objective of a Digital Twin is to optimize its real-world counterpart over its entire lifecycle.
Throughout its lifecycle, an entity goes through multiple and overlapping cycles of Predict, Plan and Execute. Prediction identifies the need, specifically, the need for change. Planning draws up the arrangement' of the system and processes for implementing the change. Execution essentially consists of Monitoring and Control. While Monitoring tracks the system performance with respect to the plan, Control makes the necessary correction to bring the process back on track when it deviates from the plan. The monitored data is an input for prediction. When the need for a change is predicted, the cycle begins once again, or it may trigger another process in a connected system.
The cycle of predict, plan, execute described above is a rather simplified representation, the actual working is very complex. Some plans are required as input for prediction. For example, investment plans are used to predict production output, which in turn are an input to predict maintenance requirements and the operation of other systems in the supply chain. Production plans also influence the prediction of the need for investment. For example, with a sub-optimal production plan, there a need for larger capacity is predicted – in terms of machines, workforce, and time. Whereas with an optimal plan the same output can be achieved with a lower capacity.
Planning is a Key Component
A system is designed and operated based on plans. The future state of the system or the system's output at a future time is dictated to a large extent by the current plans. Therefore, for Prediction, the availability of various plans is extremely vital. For example, to predict the remaining life of a tool or a machine component at the end of a given time period, the Digital Twin not only needs historical data, but also the operation plan until that period.
By leveraging data, coupled with analytics and artificial intelligence, the Digital Twin is a learning system. In addition to the simulation capability, the Digital Twin provides a look-ahead capability to predict undesirable condition and favorable opportunities. This capability can identify the need to change the current plan. For example, the plan will need to change because of the unavailability of a specific resource or a change in the demand pattern will require accumulating certain inventory.
Monitoring, a core functionality of a Digital Twin, is the process which observes the performance of the system and compares it to a standard or a target. The plan provides the necessary performance standards for comparison. For example, customers or transporters use the Digital Twin to monitor the progress of their deliveries against the planned delivery dates. The data obtained by monitoring is an important input to Prediction.
If there is a deviation from the target, corrective action is taken by the Digital Twin or the Digital Twin provides the necessary input to decision makers. The Digital Twin connected to the actuators of the physical systems controls the system . In case of manually operated systems, the result of monitoring provides an input for analysis and decision making.
A plan must be tested and validated in the light of the system's inherent uncertainty and variability. This is achieved by the Digital Twin's simulation capability. Simulating the system improves the understanding of the system's behavior. This leads to better decisions and control over the system. The Planner needs to be in constant and close interaction with, Predict, Monitor and Control, the other core functions of the Digital Twin. Therefore, for efficient working of the Digital Twin, the design of the Digital Twin must incorporate the Planner.
Planning for Optimization
For optimal operation of a system, the activities and processes of various resources must be organized and coordinated. This is achieved by the planning process. Sub-optimal plans cannot deliver optimal system performance. Therefore, the Digital Twin requires a suitable Planning & Scheduling System to generate plans considering all requirements and constraints of system, across the entire planning horizon. Ignoring some constraints and requirements will lead to plans which are far from optimal. The Digital Twin cannot deliver optimal performance with such plans. Such plans when executed will require frequent changes which be which will adversely impact optimality. Many a times, such plans will be inexecutable.
The operation of a manufacturing system is optimized by maximizing resource utilization, by balancing workload across resources, and minimizing inventory levels, with just-in-time deliveries. In addition, the related supply and distribution systems require that the production be smooth and on-time, to minimize variability and uncertainty. Besides, the optimal plan must also consider maintenance requirements, so that the system continues to deliver optimal performance throughout its lifecycle. Besides, the plan needs to be optimized over the entire planning horizon and entire supply chain. For instance, minimizing inventory in the short term may lead to shortages in the supply chain over the longer term, while maximizing resource utilization may lead to wasteful blocking of storage and financial resources.
When a plan is in operation, the Digital Twin monitors the performance. Variability in the system, due to process delays and quality issues, will require suitable adjustments to maintain optimality. When the Digital Twin predicts a change in conditions that require a change in the plan, the Planning system will create a new plan or adjust the current plan for the new conditions. Thus, the Digital Twin working closely with a suitable Planning and Scheduling system will deliver optimal system performance.
 Parris, Collin J.;Laflen, Brandon J.;Grabb, Mark L.;Kalitan Danielle M., "The Future for Industrial Services: The Digital Twin," Infosys, [Online]. Available. [Accessed 23 November 2018].
 M. W. Grieves, "Digital Twin: Manufacturing Excellence through Virtual Factory Replication," 2014. [Online]. Available.
 Deloitte, "Industry 4.0 and the digital twin," 2017. [Online]. Available. [Accessed 30 December 2018].
 M. Batty, "Digital twins," Environment and Planning B: Urban Analytics and City Science, vol. 45, no. 5, p. 817–820, 2018.