Now is the Time to Put Your Digital Twin to Work!
Guest Blogger: Ian McGregor, Director, Emulate 3D
A Digital Twin is most simply defined as the virtual representation of a physical asset, and the term was originally used by Dr Michael Grieves, in a course delivered at the University of Michigan in 2003. The concept consists of data being fed from the real world to the virtual representation, and information flowing back, suggesting that the data has undergone a process which transforms it into useful real world information. While this process will sound familiar to those who are involved with industrial simulation, the Digital Twin concept has broader aims than classic project-based simulation. The physical asset being represented virtually can be anything to which the concept usefully applies - a bolt, a machine, a system, or a network of systems.
The Digital Twin was conceived as a virtual indicator of the state of a production process or a manufacturing line, for example – like a SCADA system where near real-time data keeps the virtual representation current, and feeds useful information to users. In 2003 even this was a considerable challenge, as few systems outside high-value industries like semiconductor manufacturing were equipped with real-time data acquisition capability. The subsequent multiplication of inexpensive sensors and interconnected MES has brought us to the age of the Internet of Things, and a wealth of data of all types. Couple this with the cloud to store Big Data economically, and enough computing power to process it to generate useful information, and it becomes clearer why the Digital Twin concept is only now coming of age.
Simulation - the Beating Heart of Every Digital Twin
Narrowing the scope of the concept to warehousing operations allows us to focus on the more familiar realities of discrete event industrial simulation, and to consider where it fits within the Digital Twin concept. Start from the traditional position of a simulation model as a means of analysing and improving a system, then adapt the resulting final iteration of that model to be an online representation of the system as it operates, fed by real-time or near real-time data from the actual system. Add the ability to click on model objects and drill down to a variety of data from operational results to maintenance requirements, and you have implemented a SCADA system.
From Analysis to Operational Use
The original simulation model that was used to reduce the risk associated with each decision was based on a mix of predictions, assumptions, and experience, but as the Digital Twin has now moved from the analysis phase into operation, the situation has changed. The model can now be improved to reflect not only the new reality in which it exists, complete with real data, but also the changed objectives. It is no longer primarily an exploratory model, but is now more useful as a predictive tool. The model structure is defined by the actual system, and data from real performance can be used to calibrate the model. Differences in performance between simulated and real operations can be used to detect operational drift, or perhaps deviance from best practices. Actual data and a more accurate, calibrated model combine to create a powerful predictive tool for evaluating short term order throughput, resource use, and maintenance schedules, for example.
Do I Need a Digital Twin?
As the amount of data collected in manufacturing and distribution increases, and the ability to store and process it grows, the opportunities offered by the Digital Twin concept grow with it. However, the word “twin” may be misleading – while the notion of one all-encompassing virtual copy of a production facility, warehouse, or fulfilment center may be appealing, it isn’t realistic. Even with unlimited computing power, storage capacity, and processing speed, it isn’t optimal to have the same model for product throughput analysis and the maintenance requirements of one machine within the system, for example. In reality, the SCADA-style drill-down will imperceptibly access different models, each using different data subsets relevant to their objective of efficiently creating useful outputs.
Conclusion – Digital Twins Need Simulation
A Digital Twin is a virtual representation of a real system, and can be calibrated to respond in the same way the real system would to external influences such as order data, resource availability, and breakdowns. As such, it is a powerful tool for improving throughput, reliability, product mixes, maintenance schedules and so on. Digital Twins can be tested to breaking point in ways which are unthinkable, uneconomic, and disruptive in the real world, and that’s where their real value lies. As the operating conditions under which automated systems are used continue to change, the Digital Twin is the trusted sandbox where solutions are developed and refined to take the right decisions, based on rigorous analysis and experimentation, with low risk and cost.