In this week’s Helmholtz AI project showcase, scientists from the German Aerospace Center (DLR) and the Karlsruhe Institute of Technology (KIT) are working on AI-based digital methods to reduce the time, cost and characterization effort in the development process of components and structures of materials.

Could you introduce yourself, giving your affiliation, area of work, and of course, the project title?

The title of our funded Project is ‘FoAIm - Coupling of complex multiscale structural foam characteristics with AI based methods’. We are two project partners: The Institute of Vehicle Concepts and the Institute for Applied Materials – Microstructure Modelling and Simulation (IAM-MMS).

The Institute of Vehicle Concepts is a part of the German Aerospace Center (DLR). We research, develop and evaluate new vehicle concepts and technologies in the light of future demands of transportation systems. Based on a comprehensive understanding of road and rail vehicles, the department “Materials and Process Applications for Road and Rail Vehicles” is specialised in transferring new and complex material and process solutions to innovative applications. The IAM-MMS is located at the Karlsruhe Institute of Technology (KIT) and comprises expertise in the development of scalable simulation software for high performance material simulations as well as in curating huge datasets by automated workflows within a developed research data infrastructure.

In simple words, what specifically is your project about? And, how and why do you think it is a high risk, high gain endeavour?

In our project the correlation of microstructural properties, such as porosity, pore size and wall thickness, and the resulting macroscopic mechanical material behaviour of foams is to be analysed by using artificial intelligence (AI) based digital methods. The objective and the high gain-potentials are that intelligent algorithms for automated modelling of microstructures combined with machine learning approaches can identify and predict relevant material parameters allowing for significantly reducing the necessary time, costs and characterization effort within the development process of components and structures. In foam materials there is a risk of not capturing all irregularities e.g. undetected micropores and shape variations, which are necessary for describing the materials behaviour. Furthermore, depending on the complexity of the problem, there is a risk of providing not enough data to train CNNs properly. As a result, predictions have a high chance to fail. From the perspective of a data scientist, the critical aspect of dimensionality in the database could be insufficient which in turn will very likely lead to an overfitting of the data and ultimately the materials design will not meet the requirements.

How important has the Helmholtz AI funding and platform been to carry out this project?

The Helmholtz AI platform offers the opportunity to cooperate with other Helmholtz centers and combine the knowledge to accelerate the research and development of suitable AI-methods for different use cases. For us the funding and the platform is a huge driver for distributing applied AI across the Helmholtz community.

What else would you like to share?

We are excited to see how FoAIm and the other funded projects can contribute to solve complex problems with AI and we would like to warmly invite all interested experts to get in contact with us.

Figure: FoAIm project structure. The individual work steps are illustrated including the dependency graph of the generated data.