|
Title
|
|
|
|
A digital twin approach to support the evolution of cyber-physical systems
| |
|
Author
|
|
|
|
| |
|
Abstract
|
|
|
|
| Computer modeling and simulation is found in many industrial applications, for example there exist CAD models, finite element simulations of mechanical parts, simulations of chemical plants, models of remaining life prediction and many more. Those models and simulations are used both in the design of systems, and during their operations. Furthermore, during a system's operation, data is collected and stored at a large scale, this is a result of what is commonly called the fourth industrial revolution, or industry 4.0. This revolution is that of computer automation and large scale data exchange, seen in the manufacturing industry of the 21st century. These computer models, and the widespread collection of data gave rise to the concept of a digital twin. A digital twin is a virtual representation or a mirroring of a real-world system. Data is continually collected from that real-world system, and fed back to the digital twin such that it contains the most up-to-date model of the system it mirrors. In exchange, a digital twin offers its users a variety of helpful services based on its models and large collection of data, examples of such services are historical data replay, virtual testing, virtual commissioning and optimization. Despite stemming from the manufacturing field, digital twins find applications in many more domains such as healthcare, geophysical processes and construction engineering. A key aspect in all digital twins is that of evolution, and this in two different ways. The first way deals with the mirroring of the real-world system. When that real-world system evolves, for example, because components are replaced or because it ages and shows signs of wear and tear, the models in the digital twin should reflect that evolution. The second way deals with the evolution of the services in the twin itself. Like any other software system, the purpose and requirements of the digital twin can evolve over time. In this thesis the focus is on a subset of issues encountered in these two types of digital twin evolution. It provides reusable techniques that aid digital twin developers with handling the evolution of their digital twin. The first part of this thesis tackles the digital twin evolution driven by requirements/purpose changes of the digital twin itself. Here, the challenge is interweaving new software components of the digital twin with the existing components, and solving their conflicting interactions. To this end, we developed a notation and a set of common transformation templates that occur during a twin's evolution. This notation allows twin developers to visualize and reason about the impact the evolution has on the digital twin. The second part tackles the digital twin evolution driven by changes in the real-world system. The main issue we identify is how to handle divergences between the real-world system and the twin as the real-world system ages over its lifetime. It is important, since the accuracy of the twin's services depends on the agreement between the model and the real-world system's behavior. To this end, we drew inspiration from model verification and validation techniques to provide a method that detects deviations between a digital twin's model and a real-world system. We additionally applied state-of-the-art classification methods to further pinpoint the cause of the divergence as to ease the model updating process. Combining those two techniques we demonstrate a workflow that incorporates them for the continuous detection of system variations during a software release cycle to provide accurate testing results when the digital twin is used for virtual testing purposes. Supporting these two parts, we developed a domain specific language for defining model validation experiments and a gantry crane twin system. The domain specific language allows non-domain experts to define and execute model validation experiments without requiring knowledge of model simulation. The gantry crane twin system consists of a physical gantry crane and an accompanying digital twin, it is used throughout this thesis to demonstrate or evaluate the proposed techniques. |
| |
|
Language
|
|
|
|
English
| |
|
Publication
|
|
|
|
Antwerpen
:
University of Antwerp, Faculty of Applied Engineering
,
2024
| |
|
DOI
|
|
|
|
10.63028/10067/2105580151162165141
| |
|
Volume/pages
|
|
|
|
186 p.
| |
|
Note
|
|
|
|
:
Denil, Joachim [Supervisor]
| |
|
Full text (open access)
|
|
|
|
| |
|