Dynamic analogies establish analogies between electrical, mechanical, acoustic, magnetic and electronic systems. Many systems can be used to create an analogue model. And when a model is run on an analogue or digital computer, this is known as the simulation process.

Around 50 years ago, during the Apollo 13 mission, the command module suffered an explosion in one of the oxygen tanks, generating a series of warning alarms for the astronauts. How to diagnose and resolve the problem of a failing physical asset 330,000 km from Earth and without direct human intervention (the three astronauts were trapped inside and could not even see the damage)?

NASA had 15 simulators that were used to train astronauts and mission controllers on all aspects of the mission, including multiple failure scenarios. Gene Kranz, NASA's flight director for Apollo 13, says simulators were one of the most complex technologies in the entire space program. The only material reality in the simulation training program was just the crew, cockpit and mission control consoles. Everything else was made up of a few computers, lots of formulas and highly qualified technicians.

A simulator is not a "digital twin", as [1] states, but how NASA mission controllers were able to quickly adapt and modify simulations to match the conditions of a real-life damaged spacecraft, to be able to investigate, reject and improve the strategies needed to bring the astronauts back, was probably the first use of the "digital twin".

• They are most useful when referring to physical assets beyond direct human reach. Modern “digital twins” typically use 3D models and augmented/virtual reality to achieve this goal;
• They require constant “feedback” of data from the physical asset that can be used to update its state and support engineering decisions. Modern “digital twins” typically use IoT to achieve this goal;
• They must be flexible enough to react to changes in the physical asset. Modern "digital twins" typically use simulations to provide critical information, emulating the physical and dynamic behaviour of assets, requiring data analysis to make decisions about deterministic or non-deterministic behaviour, using linear or precision models and discrete approaches based on events with variables or fixed sampling time;
• There was no single "digital twin" for the Apollo program; NASA used 15 different simulators to control various aspects of the mission. Contemporary “digital twins” consist of multiple interaction models that can be combined to account for different aspects of performance. For example, predictive “machine learning” algorithms can be used to carry out simulations and complement other simulations carried out with discrete or continuous, deterministic or stochastic models;
• The events of Apollo 13 took place over three and a half days, during which many adaptations and reengineering took place. How long would it take to install new “digital twins” models if there were changes to the corresponding physical asset? Perhaps the use of 3D virtual models and augmented/virtual reality could be sufficient. However, it remains difficult to work on both digital and physical models of an asset simultaneously.

 Building a “digital twin” follows a simple workflow. First, mathematical models of a process or product are built by adding geometry and physical characteristics such as data tables, formulas and boundary conditions. Secondly, the behavior of the application is specified, for example, what information should be presented, how the simulations and predictions will appear and how the sensors will be connected. In the end, the application is run as a stand-alone application or installed in the “cloud” to be viewed in a web browser.

Figure 2 – Example of a digital twin of the International Space Station, based on Ontologies and Cloud Computing. Image reference: Microsoft.