For today’s systems, we take the first steps in this direction. Fusing IoT with system analytics and data science in an unprecedented digitalization push, we reduce total costs of ownership of assets with data-driven control and operations. In this digital fusion of IoT, virtualization and data science, we see the rise of digital twins.

Using real-time data to understand, learn and reason, digital twins mirror the physical system. Often used for adaptive control, they enable live optimizations where, e.g., fleet owners and plant operators successfully reduced the total cost of ownership (TCO) of assets, minimizing resource consumption as well as system downtime with predictive health management and condition based maintenance. They link into business intelligence, e.g., as they provide the numbers for after-sale logistics. For engineers, they allow trials and inspections that do not compromise operations, safeguarding continuous integration of updates, upgrades, or re-configurations, which are all necessary to cope with the heightened dynamics modern high-tech systems face.

Their usefulness to tune and improve system operations led to their rise in all major listings of key technologies for Industry 4.0 and smart cities, buildings, or mobility, as they are widely seen as key to the digitalization of assets but also of system engineering.

You are most welcome to join us on this journey.

How to do this best and realize digital twins on industrial scale is an area of active development. We set individual roadmaps according to a company’s starting points, i.e., the availability of assets with live data or of machine-readable design or simulation models. Some of this is shown in the figure above, but please contact us for details.

Knowledge & data meet in digital twins

Furthermore, we use generative techniques to realize digital twins efficiently and approach this from the data and the engineering knowledge angles at the same time.

Given the direction of the digital twin, i.e., if it aims at virtualization, just-in-time aspects of operations, or at prognosis, we start with a major decision: whether the digital twin is realized best with a single computational model or if several, partially linked models are better suited to execute its functions. The latter approach is often easier to build but harder to manage, while the former will typically require a multi-layered approach that spans over many stakeholder concerns.

Our design decision is guided by a value of information analysis that also supports us to decide the internal structure of the digital twin, i.e., the selection and placement of its reasoning layers and data items. As this is where the engineering knowledge couples to the data-centric information flows, we advocate meet-in-the-middle procedures based on model-to-model transformations that bridge various levels of details here for consistency and efficiency.