Paving the Path for Digital Twins in HPC

The concept of digital twins (DT) was established more than twenty years ago. But, over the last five years, the topic has been receiving increasing amounts of attention, both in academia and industry. This can, for example, be seen in a tremendous increase in the annual number of publications [2]. For a digital twin, one creates and operates one or more virtual representations of a physical entity. The virtual representations contain relevant information about the physical entity, which is ideally much cheaper to obtain compared to using the latter. When operating the digital twin, the physical entity and its virtual representations are connected, which means that state information is exchanged in both directions.

The initial focus of research on digital twins was product life cycle management, which covers the creation, production and operation of products [1]. In industry, this is still of significant interest, for instance, in the development of autonomous driving solutions or for the realisation of preventive maintenance concepts. The use of digital twins for large ecosystems is a more recent development. The most outstanding initiative in this direction is Destination Earth (see digital-strategy.ec.europa.eu/en/policies/destination-earth ), which aims to develop a highly accurate digital model of the earth on a global scale. One goal is to facilitate informed policy decision-making by means of digital twins, as such a twin of the earth would allow us to explore the impact of human activities on natural phenomena and vice versa. One important feature of Destination Earth is the need for high-performance computing (HPC) to model the weather and climate, and the behaviour of the oceans, as well as surface and subsurface water flow, amongst other things.

One effort that is working towards realising the vision of Destination Earth is the BioDT project ( biodt.eu ), where PDC is one of more than twenty partners throughout Europe. The goal of this project is to design and demonstrate a digital twin platform that will improve our understanding of biodiversity dynamics by developing means that provide advanced modelling, simulation and prediction capabilities. The work will be guided by various real-life use cases. One class of use cases is concerned with the question of how the evolution of species is impacted by environmental changes. A more specific question is, for instance, how bird populations will change in the northern part of Europe depending on forest management strategies. Addressing this question requires the integration of forest and biodiversity models, as well as all the necessary environmental data. Another important use case relates to food security. It is being observed that environmental changes are accelerating the decline of pollinators like honey bees. Multiple factors are assumed to impact honey bee dynamics. BioDT plans to port agent-based models to HPC systems and to add support for dynamic model updating based on updated environmental data. 

The integration and upscaling of various models, as well as the integration of various data sources from already existing research infrastructures in this field, will be one of the key challenges for BioDT. Once this is achieved, researchers will be able to exploit Europe’s most powerful supercomputer system, namely LUMI. At the same time, the project will have paved the path for getting future supercomputing systems ready to deploy and operate digital twins.

References

1. M. Grieves, J. Vickers, Digital Twin: Mitigating Unpredictable, Undesirable Emergent Behaviour in Complex Systems. Springer International Publishing, 2017.
2. M. Sjavor et al., The Digital Twin Concept in Industry – A Review and Systematization, proceedings of the ETFA 2020 conference.