A digital twin is an exact copy of a process, product or service which is used to simulate the introduction of an innovation. This is the meaning of Digital Twin, theorized in 2002 at the University of Michigan and originally applied to Product Lifecycle Management, which provides for the two “twins”, one physical and the other digital, a “life as a couple” and an evolution which can be carried out together. The Digital Twin, applicable both to existing realities and during the design phase, therefore has innumerable advantages referable not only to the possibility of predicting, by anticipating these, problems during production, but also to improve product development and reduce the costs of manufacturing prototypes.
After talking about the challenges and advantages of the Digital Twin, you can ask yourself which are the technologies which contribute to its realization.
In order to introduce a Digital Twin into the company, according to the White Paper recently published by Engineering, it is necessary to be able to count on various “conditions”, i.e. to have the right connectivity available, the necessary level of digitalization and a “quantum satis” of Artificial Intelligence.
It almost goes without saying that a Digital Twin needs an adequate connectivity for collecting the data to be stored in Cloud and made available to others. When building the twin, there is a part which acquires operational and environmental data through sensors which capture the signals and one which analyzes the Big Data collected, which are aggregated and combined with other company data.
The emergence of the MEMS (Micro-Electro-Mechanical-Systems) technology which drastically reduced the price of many standard sensors, has certainly helped with this collection process.
When we speak of digitalization for Digital Twin, we refer to the need for applications useful for simulation. Simulation modeling is an application which combines connectivity, Big Data and IoT in a digital model, aligned almost in real time with the systems it reproduces.
The technologies for creating digital models depend on the nature of the systems themselves: while a mechanical component can be simulated using techniques based on physics and mathematics (CFD and FEM), a more complex system, like a production line, can be modeled using agent-based-modeling or discrete event techniques.
As regards data, reference is instead made to Data-Driven Modeling (DDM) which, unlike simulation models which require knowledge of the operation of the real system, borrows advanced mathematical and statistical techniques to analyze the data which characterize a system and to find relationships between input and output.
When imagining a digital model, one cannot forget there is a part of Artificial Intelligence, useful for independent learning, based on the machines’ experience, capable of perceiving the environment, analyzing situations and identifying the best decisions in order to reach the default goal.
AI is applicable in Analytics, which analyze and visualize the information collected by the sensors in order to discover, interpret and transmit significant data trends and make decisions based on these, as well as to manage the action requests from the real system, to reproduce it through “actuators” subjected to human control and that trigger the physical process.