An Open Source Digital Twin Framework
Aho, Panu (2020)
Aho, Panu
2020
Johtamisen ja tietotekniikan DI-tutkinto-ohjelma
Tekniikan ja luonnontieteiden tiedekunta - Faculty of Engineering and Natural Sciences
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Hyväksymispäivämäärä
2020-01-08
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tuni-201912096712
https://urn.fi/URN:NBN:fi:tuni-201912096712
Tiivistelmä
In this thesis, the utility and ideal composition of high-level programming frameworks to facilitate digital twin experiments were studied. Digital twins are a specific class of simulation artefacts that exist in the cyber domain parallel to their physical counterparts, reflecting their lives in a particularly detailed manner. As such, digital twins are conceived as one of the key enabling technologies in the context of intelligent life cycle management of industrial equipment. Hence, open source solutions with which digital twins can be built, executed and evaluated will likely see an increase in demand in the coming years.
A theoretical framework for the digital twin is first established by reviewing the concepts of simulation, co-simulation and tool integration. Based on the findings, the digital twin is formulated as a specific co-simulation class consisting of software agents that interact with one of two possible types of external actors, i.e., sensory measurement streams originating from physical assets or simulation models that make use of the mentioned streams as inputs.
The empirical part of the thesis consists of describing ModelConductor, an original Python library that supports the development of digital twin co-simulation experiments in presence of online input data. Along with describing the main features, a selection of illustrative use cases are presented. From a software engineering point of view, a high-level programmatic syntax is demonstrated through the examples that facilitates rapid prototyping and experimentation with various types of digital twin setups.
As a major contribution of the thesis, object-oriented software engineering approach has been demonstrated to be a plausible means to construct and execute digital twins. Such an approach could potentially have consequences on digital twin related tasks being increasingly performed by software engineers in addition to domain experts in various engineering disciplines. In particular, the development of intelligent life cycle services such as predictive maintenance, for example, could benefit from workflow harmonization between the communities of digital twins and artificial intelligence, wherein high-level open source solutions are today used almost exclusively.
A theoretical framework for the digital twin is first established by reviewing the concepts of simulation, co-simulation and tool integration. Based on the findings, the digital twin is formulated as a specific co-simulation class consisting of software agents that interact with one of two possible types of external actors, i.e., sensory measurement streams originating from physical assets or simulation models that make use of the mentioned streams as inputs.
The empirical part of the thesis consists of describing ModelConductor, an original Python library that supports the development of digital twin co-simulation experiments in presence of online input data. Along with describing the main features, a selection of illustrative use cases are presented. From a software engineering point of view, a high-level programmatic syntax is demonstrated through the examples that facilitates rapid prototyping and experimentation with various types of digital twin setups.
As a major contribution of the thesis, object-oriented software engineering approach has been demonstrated to be a plausible means to construct and execute digital twins. Such an approach could potentially have consequences on digital twin related tasks being increasingly performed by software engineers in addition to domain experts in various engineering disciplines. In particular, the development of intelligent life cycle services such as predictive maintenance, for example, could benefit from workflow harmonization between the communities of digital twins and artificial intelligence, wherein high-level open source solutions are today used almost exclusively.