CD Laboratory for Data-assisted simulations of complex flows

This CD Laboratory explores ways in which physical phenomena of interest can be simulated for better understanding using completely virtual experiments, and how this can be done as quickly as possible despite their complexity.

Conventional experiments can be carried out to gain a better understanding of physical phenomena. However, when it comes to investigating high-temperature industrial processes under extreme conditions, for example, the factors of danger and cost should not be underestimated. However, if the physical laws relevant to the phenomenon under investigation are translated into program code and solved numerically by a computer, the phenomenon in question can be simulated in a virtual experiment. This means that not only can empirical values be collected for a real plant, but processes that are yet to be developed can also be tested in advance for their feasibility – without any risk.

The disadvantage here is that while experiments in the real world take place in real time, the computing effort for virtual experiments can be very high – the phenomena of interest are often so complex that it can take several days to simulate them for just a few seconds. A prime example of this are particulate problems, which are of great importance from both a scientific and practical perspective: these involve systems of solid particles through which one or more gas or liquid phases flow, resulting in mass, momentum and heat exchange between particles and fluid phases.

For science, understanding such processes is a particularly exciting challenge, as many small particles generate the long-term behaviour of a large whole through numerous short interactions. From a practical point of view, over 75% of all raw materials are granular and often exhibit unexpected behaviour in industrial plants as a result of these complex processes: the more knowledge that can be gathered in this regard, the better production problems and inefficiencies can be prevented.

However, the problem of the potentially enormous amount of time required for virtual experiments remains – and this is where the CD Laboratory comes in: by combining classic simulation techniques with modern, data-driven AI methods, it should be possible to combine physical equations to be solved with information on system behaviour on a larger scale over longer periods of time – which in turn allows for much faster calculations. This would allow data-driven simulations to run in real time, just like experiments in the real world – or even faster: in critical moments, a computer model could predict within seconds how a reactor will develop over the next hour, and this information could be used to quickly decide whether an emergency shutdown is necessary, for example.

The application-oriented basic research conducted by the CD Laboratory will thus make an important contribution to greater safety and production efficiency in industry and energy supply, but it also shows great potential in terms of reducing greenhouse gas emissions and improving the achievability of climate targets: on the one hand, real-time simulations of complex flows can also be used to reduce the energy requirements of industrial plants already in operation. On the other hand, these simulations can also be used to search for new hydrogen-based processes for steel production instead of carbon-based ones. Simulation methods developed at the CD Laboratory are therefore applied in particular to problems related to environmental and climate protection in order to identify promising new approaches or optimise existing processes in the computer model, which will benefit society as a whole.