Abstract

The digital twin has become an intrinsic part of every product creation process. Basically, it is a virtual copy of a real asset, integrating all data, models and other structured digital information of a product, a plant, an infrastructure system, or a production process. The data constituting the digital twin can be generated during design, engineering, manufacturing, commissioning, operation, and/or service. While the digital twin can have multiple appearances, the objective is always to have a digital representation suited to the purpose in terms of level of detail, completeness, accuracy, and execution speed. Consistent, traceable integration of all information is key to leverage existing and create new business opportunities. The true power of the digital twin lies in its relationship with its physical counterpart. The design engineering models should hence be as accurate as possible to represent the as-manufactured and as-built physical asset. Data acquired on the physical asset can validate, update and enrich the digital twin and provide information to improve the design. Powerful IoT solutions are instrumental to this. Inversely, the knowledge contained in the digital representation can be transformed into value for the physical asset itself. To this purpose, specific encapsulations can be made starting from the digital twin to model a specific set of behaviors against a specific product context, thereby delivering a stand-alone executable representation. Such instantiated, self-contained and encapsulated, model can be referred to as the Executable Digital Twin. The key element is that such Executable Digital Twin can be used outside its authoring environment and can be leveraged by anyone at any point of a products lifecycle on any certified device, from edge to cloud, without the need for heavy simulation software. While the integration of model-based functions in operational environments has been done since long, the derivation of such models from the digital twin, keeping full consistency and traceability, is novel and provides a huge leveraging potential. Examples of such usage are as embedded models for virtual sensing, model-based control, performance monitoring or X-in-the-loop hybrid testing applications. But also the use of the Executable Digital Twin as a companion model to accompany the physical asset through its lifecycle for performance assessment, system integration or decision support applications offers new value propositions. Key enabling technologies to create the Executable Digital Twin are fast simulation methods, Model Order Reduction, state estimation and standard model delivery (e.g. as FMU). Machine Learning methods are instrumental to deliver compact models of complex non-linear systems over wide range of operational conditions. In the execution phase, open platforms accepting external models with their execution engines and linking these where needed in a co-simulation environment are key. As the digital twin hereby starts to lead an own life across supply chains and in open -cloud- environments, IP protection is a key challenge. The presentation will address the approach behind the Executable Digital Twin, report the specific developments made on the level of the enabling technologies and illustrate the power of the approach in various examples related to test-based product engineering, the integration with manufacturing control platforms as well as operational system monitoring. Several case studies will be treated including virtual sensing and hybrid (XiL) testing in automotive and aerospace applications and decision support and process monitoring applications in the manufacturing and process industry. The related value propositions include data augmentation, reducing sensor cost, speeding up system validation and optimizing operational performance and asset availability.