IntroductionWe are going to talk about the Digital Twins course that is being offered by the University of Oxford Department of Continuing Education and is based on MATLAB and Unity. Learners will study model-based design under the framework of the digital twin and its advanced modeling techniques like AR, VR and others.
BackgroundA Digital Twin is a model of a physical entity that is developed and maintained throughout its lifecycle, from concept, to development, production, operation and decommissioning. The model is continuously updated to reflect the current state of the entity, and as a result, it offers many opportunities to analyse and improve business decision-making. In modern Digital Twin development, simulation software integrates a suite of technologies, such as photo-realistic simulators and artificial intelligence, to generate predictive models that can be deployed on a broad operational system. The development workflow resembles that of a model-based design framework due to the need for system integration, V&V and deployment. Hence, when we designed the course Digital Twins: Enhancing Model-based Design with AR, VR, and MR (Online) at the University of Oxford, we considered the disruptive nature of emerging technologies for the digital twin and their impact on industrial model-based design. This blog post explains our rationale for this course, implications for industry, and techniques to implement these ideas in MATLAB.
The evolution of model-based designThe ideas behind Model-based design are already mature. The Model-based design approach places virtual models at the centre of the development process to improve the delivery of engineered products. Model-based design (MBD) combines the concepts of modelling and automation to
- accelerate the exploration of new designs and perform fast, repeatable tests with modelling and simulation;
- eliminate manual steps and reduce human error by automating steps such as reporting, coding, and verification;
- create a digital thread with traceability from requirements and system architecture to component design and tests;
- perform predictive maintenance, detect faults, and optimize the system in operation using models as digital twins.
- Digital Twins are becoming a crucial part of MBD as companies seek to continuously improve product versions by analysing complete lifecycles;
- Simulations & Simulators with physics-based engines accelerate modelling workflows. Developers less frequently specify model behaviour, but rather assemble systems as a network of parametric components with prescribed physical behaviour - as in a Simscape model, in a CAD package or in a gaming engine;
- Workflows span many domains and verticals, where technologies are shared across industries, e.g. LiDAR-based mapping in automotive and in construction.
- Model granularity in each vertical is increasing, and components need to be represented at different levels of fidelity to support system-level performance optimisation;
- AI strategies complement modelling, design and analysis (explained below)
- Augmented, Virtual and Mixed Reality enhance capabilities to design engineering products, to test them against virtual scenarios, and to deliver new product experiences;
- Computational Additive & Destructive manufacturing techniques, such as 3D printing and CNC milling, become part of the design process.
Implications for industry – Digital twins and Model-based designThe evolution of MBD has several implications. Here we will outline some of the impact on industry
1) Impact of AI and algorithms on digital twinsA range of modelling algorithms could be used for Digital Twin development. For example, algorithms that implement
- Physics-based modelling, where a developer implements the governing equations of a physical system and solves them via a combination of analytical and numerical techniques, g. Runge-Kutta solvers for ordinary differential equations, or Finite-Element methods for partial differential equations. Fidelity of the models can be improved with parameter-based estimation from hardware data. The models can also be used to support observers in a system, e.g. Kalman estimators.
- Data-driven modelling, where models are deduced from input-output component data, g. using regressions, look-up tables, or system identification techniques;
- AI-based modelling, in which learning techniques are used to improve the fidelity and dynamic performance of data-driven surrogate models.
2) Digital twins – unifying the model-based design flowDigital twins help better inform design and operation stages:
- System Requirements, Functionality and Architectures, are improved on from previous product iterations.
- Subcomponent model development is accelerated by data collected from the field. Using historical data can also reduce risk of plant-model mismatch.
- Integration & Testing – model-based scenarios and scenes allow for broader testing and more robust designs, especially in cases that are expensive to reproduce, g. faults,
- Operations and Sustainment – mismatches between the Digital Twin and the physical asset inform decisions for operation and maintenance.
3) Use of Augmented and Virtual Reality in MBDDigital twins can support traditional development workflows, but in a broader sense, they have also evolved the way developers
- design with CAD and CAM
- collaborate in design
- deploy a component in an ecosystem and visualize of the ecosystem
- model an ecosystem over time
- automatically generate components, e.g. using optimisation and AI,
4) New Applications and services based on digital twinsDigital twins could also lay the foundations for new applications and services – for example, for
- Design Optimization
- Verification and Validation
- Mixing Virtual and Real
- Virtual Commissioning
- Fault Detection, Diagnostics, and Prognosis
- Predictive Maintenance
- Operation Optimization
5) System complexityThese emerging technologies are impacting many industries, especially large enterprises where products are composed of many technologies. Components are deployed in often unknown and dynamic environments where diverse factors must be understood and visualised.
Strategies for implementing in MATLABIn the course, we explore how MATLAB & Simulink support:
- development, analysis and optimisation of designs;
- physics-based model development and synthetic data generation;
- data-driven model development using AI;
- interoperability with 3D simulators;
- deployment to cloud & edge.
ConclusionWhen we designed the course Digital Twins: Enhancing Model-based Design with AR, VR, and MR (Online), we were aware that this is a pioneering course – in an area that is dynamic and rapidly evolving – but has significant potential in many areas. Have a comment for us? Please, leave a comment below!
To leave a comment, please click here to sign in to your MathWorks Account or create a new one.