Twinning for and by Systems Engineering
The Cyber-Physical Systems (of Systems) we design, maintain and above all, evolve, over increasingly long periods of time, are ever increasing in complexity. In the mean time, demands on quality, maintainability, sustainability, etc. become more and more stringent. A host of desirable extra-functional system features such condition monitoring, fault diagnosis, predictive maintenance, and optimization may be realized when adopting a “twinning” paradigm.
In this paradigm, a twinning architecture (a system in its own right) is used whereby a virtual instance of a System under Study (SuS) or “asset” is continually updated with the asset’s health, performance, maintenance, etc. status information, and this throughout the asset’s life-cycle (requirements analysis, design, production, assembly, operation and optimization, maintenance, re-purposing, disposal).
We take a product family/line approach to structure desirable system features, the different conceptual twinning architectures needed to realize these, and finally, the different means to deploy/realize these architectures.
This builds on many existing techniques, architectures and standards from real-time simulation, co-simulation, systems and control theory, IoT, knowledge management, machine learning, surrogate modelling, etc.
The notion of “twinning experiment” will be introduced. As there is one experiment per Property of Interest (PoI), multiple experiments need to be orchestrated. This leads to an ecosystem of interacting twins.
The data/information/knowledge obtained from twinning experiments needs to be recorded for later use. Such use may enable optimization of future versions of asset design, or even online optimization of the current asset. To make the recording of multi-formalism, multi-abstraction, multi-fidelity, … models and data scalable, a virtual, federated knowledge repository is needed.
With this repository, a-synchronous “inferencing” for the discovery of new knowledge becomes possible. This inferencing may in its own right spawn new twinning experiments. As multiple inferencing and experiment processes may exist concurrently, the repository acts as a blackboard. At this level, the term “cognitive twin” (of an asset) is used.
Hans Vangheluwe is a professor in the Computer Science department of the University of Antwerp. His Modelling, Simulation and Design Lab (MSDL), part of the Antwerp Systems and Software Modelling (AnSyMo) group is a core research laboratory in the Design and Optimization cluster of Flanders Make, the strategic research centre for the Flemish manufacturing industry.
In his research on multi-paradigm modelling, he studies the foundations and applications of modelling language engineering. This covers the entire spectrum, from acausal modelling languages such as Modelica for lumped parameter modelling of physical systems, to discrete-event simulation languages such as DEVS and GPSS to model software and production systems. He investigates modular combinations of these formalisms, of views and of abstractions. He develops scalable (meta-)modelling and (co-)simulation tools to help engineers design, build, optimise and maintain Cyber-Physical Systems.
He was the co-founder and coordinator of the EU ESPRIT Basic Research Working Group 8467 “Simulation in Europe”, a founding member of the Modelica language Design Team, and the chair of the EU COST Action IC1404 on Multi-Paradigm Modelling for Cyber-Physical Systems (MPM4CPS).
Slides available here: https://msdl.uantwerpen.be/cloud/public/893466