Digital Twin

The word digital twin is used within many different contexts and can include parts, components, assets or systems. At Systems Navigator, we develop digital twins for systems, where we use simulation models of factories, terminals, ports or even entire supply chains, and turn them into a digital twin of the system.

​

For us, a digital twin is an operational decision support system that supports operators, planners, analysts and managers to make decisions based on an accurate prediction of the near future, depending on the measures and actions that are possible.

How does a Digital Twin generate value?

Make better decisions. What we do with data ultimately effects our ability to make better decisions. Better decision making saves money, increases service and/or quality and maximizes the use of your assets & materials. Digital twins promise better decision making by operators and analysts as they can predict what is likely to happen in the near future. Additional to this prediction, digital twins can also predict how the system will perform after making change, in other words it can automatically perform what-if analysis.

​

Generating value for your organization. Discrete event simulation (DES) model results have a unique ability in being able to determine the likelihood of a certain event or situation to happen. In the context of a digital twin system, we like to call these system alarms. The way digital twins can improve the scheduling for complex systems lies in the fact that they can incorporate these schedules, run them through the DES model while also using the actual system status, and quickly calculate what the near future looks like. The key here is speed of calculation, including of actuals and automatic alternatives.

Turning a Digital Model into a Digital Twin

Turning a simulation model into a digital twin involves a 3-step process that starts with the creation of a simulation model, and ends with a completely integrated operational decision support system for both operators & system planners as well as analysts.

New or existing model

In order to use discrete event simulation models as part of a digital twin, they typically require more detail. The digital twin needs to be able to predict the system for the next hours/shift/day or maybe week as a maximum timeline. Model results therefore need to be written accordingly, in a new data structure. To incorporate both more detail and a shorter timeline is usually fairly easy to do, avoiding the need to build a brand-new model for the digital twin of a system.

Data requirements

To be accurate at predicting the near future, the model needs to be able to use data from operational systems related to orders, materials, recipes, bill of materials and actual system status. Furthermore, the model now also needs to know the exact schedule of labor & resource availability, as well as the upcoming workload such as orders, ship arrivals & other external scheduled events.

Model population

Building a model is not the same as populating a model from data. The latter approach is more complex, but preferred when used for a digital twin. The complexity does not lie in the number of modelling constructs, but mainly in how they are being used over time. With the capability of model population, the digital twin becomes easier to maintain over time, as its make-up can be tied to data from legacy systems such as MES and ERP.

Execution

Model execution within the context of a digital twin needs to be done automatically and at fixed intervals, to deliver accurate predictions to operators. Operators and analysts must also have the capability to create and execute experiments at any moment in order to determine what the consequences are of applying a certain measure. To assist operators even further, Digital Twins may even execute multiple scenarios at the same time in order to give the operator an overview of the options available to him, and predict the outcome of all these options ranked by preferred outcome.

Analysis

As simulation engines have limited capabilities in this area, dashboards for analysis are required that automatically collect the latest model results, highlight alarms in terms of predicted performance, and clearly show options ranked by outcome. We've developed Scenario Navigator to support detailed analysis of model results. This way operators are better in control of the system, make better decisions using the digital twin and can act on time to avoid unwanted situations such as no more space in warehouse, stopping a production line, delaying a ship etc.

Deployment

The digital twin needs to be hosted on a server, preferably on a virtual machine somewhere in the cloud or on premise. To ensure performance, the system needs to have sufficient computation power to give new predictions quickly to operators and analysts who try to find the best way forward. Being able to scale up capacity when needed, and spread execution over the available hardware (VMs) results in a digital twin that is up to the task at hand.