Digital Twins Introduction: Concepts and Engineering Perspectives

Digital twins are an emerging, enabling technology for industry to transition to the next level of digitisation.

They are meant to understand and control assets in nature, in industry and in society at large.

They were originally conceived at NASA for the space program, they have emerged as an engineering discipline, based on best practices.

What is a digital twin

NASA’s definition of a DT

“an integrated multi-physics, multi-scale, probabilistic simulation of a vehicle or system that uses the best available physical models, sensor updates, fleet history, etc., to mirror the life of its flying twin. It is ultra-realistic and may consider one or more important and interdependent vehicle systems”

A Digital Twin (DT) is a live replica of a Physical System (Physical Twin, PT). DT is connected to PT in near real-time via data-streams (e.g. by sensors). DT takes information from the PT, uses them to do observations and take decisions/actions to be performed on the PT.

A Digital Twin integrates aspects of models and control systems.

Lifecycle Management

Digital Thread: The digital twin is meant to adapt, as the underlying assets evolve with time. It evolves in tandem with the asset.

This allows to:

  1. Connect the designs, requirements and software that go into the system represented by the DT

  2. Connect the different phases of the system to the DT: design, development, operation, decommissioning, …

    • The DT accompanies and adapts to the evolution of a system in all its phases

Thus proposing a new Software Engineering (SE) paradigm:

  • Models are an integrated part of the system, as they are also used in SE phases following the system design phase.

  • The purpose of the system is building software to maintain models, instead of building models to maintain software.

  • Changes in the assets trigger model evolutions.

  • CPS in-the-large: distributed, heterogeneous

Conceptual Layers of a Digital Twin

Conceptual Layers

There are four conceptual layers that constituite a Digital Twin:

  • Descriptive: Insight into the past (“what happened?” scenarios)

  • Predictive: Understanding the future (“what may happen?” scenarios)

  • Prescriptive: Advise on possible outcomes (“what if?” scenarios)

  • Reactive: Automated decision making

Digital Twins and Formal Methods

Role of Formal Methods in DT

  • Conceptual clearness, semantics, compositionality

  • Ensure correctness

  • Better tool support

  • Beyond simulation: evaluate or predict worst-cases, what-if scenarios, etc…

Role of Knowledge Representation in DT

  • A DT can be seen as composed by a Structural Twin which uniformly represent knowledge about PT and DT

  • Reasoning support that can exploit this knowledge

  • Allows correctness properties to be expressed as relations between PT and DT

The Semantically Reflected Digital Twin

Now we see an example of a digital twin architecture description using formal methods based on three technologies techniques. This is based on the use of the following tools:

SWT: Semantic Web Technologies for uniform knowledge representation and integration of domain knowledge (discussed later).

FMI: The Functional Mock-Up Interface standard for interfaces between PT and DT, as well as simulations (discussed later).

SMOL: Semantic Reflection to reason about PT and DT through the integration of SWT and FMI into a programming language (discussed later). The system is implemented in SMOL, a unique language designed specifically for integration of SWT and programming.

Example: House heating

An example for a digital twin architecture can be made for an house with a heating system.

A DT can be divided into:

  • Structural Twin: formed by
    • Asset model:

      • Domain knowledge: a connection between all the “parts” of the house (rooms, heaters, walls) to represent the concept of “House” as long as the simulators for each part.

      • Instance: a specific instance of the domain knowledge for a particular house.

    • Twin model:

      • Domain & Instance: instance of the domain knowledge for the behavioral twin

  • Behavioral Twin: formed by

    • Digital twin infrastructure: which coordinates the simulation of all the units (parts of the house).

    • Twin configuration: which control the coupled simulation of all the units (parts of the house), hence the relations between different simulations.

Tool Installation

Here is a list of the softwares and assets that will be used in this tutorial.

Once installed Docker on your device:

  • Run docker pull ghcr.io/smolang/smol:latest

  • Run docker pull openmodelica/openmodelica:v1.19.2-minimal