Data-based stochastic 3D structure modelling for automatic learning of mechanical properties

German-French research project on the topic:

A large number of materials have a polycrystalline structure. This means that they consist of many individual crystals that are separated from each other by grain boundaries. This also includes titanium aluminides, which are used, for example, in the construction of engines or gas turbines. The 3D microstructure of such materials, i.e., the spatial-geometrical arrangement of the individual crystals, has a decisive influence on the mechanical properties, such as the fracture behaviour of the materials.

Stochastik, Uni Ulm, Project Smile, Figure: 2D visualisation of (differently coloured) single crystals in a titanium aluminide.
Figure: 2D visualisation of (differently coloured) single crystals in a titanium aluminide.

The tomographic reconstruction of the 3D microstructure of real materials is very time-consuming and cost-intensive. Therefore, mathematical models are needed to systematically investigate the influence of the microstructure on mechanical material properties. They allow the simulation of virtual, but realistic 3D microstructures on the computer. The mechanical properties of these structures are then determined by numerical simulations. In particular, the use of AI methods to accelerate the numerical simulation of the fracture behaviour of titanium aluminides is a promising approach. The method development required to establish this approach is the subject of the SMILE project, in which Prof. Dr. Evgeny Spodarev, Prof. Dr. Volker Schmidt and Dr. Matthias Neumann at the Institute of Stochastics at Ulm University, together with their French partners Dr. François Willot, Prof. Dr. Henry Proudhon and Dr. Samy Blusseau from Mines ParisTech, are making a contribution to Franco-German cooperation in the field of AI.

In combination with stochastic 3D structure modelling, the efficient numerical simulation of mechanical properties developed in the SMILE project will be able to generate a large database to quantify the microstructure influence on the fracture behaviour of titanium aluminides. Here, AI methods will be used to automatically learn mechanical properties of the structures based on numerical simulations on virtual microstructures. In particular, AI is used for an efficient  selection of training data in order to keep the computational effort for the numerical simulations, which are required for learning, as low as possible. In addition, the SMILE project also includes a data-driven component to ensure that the methodology, which is developed in SMILE on the basis of virtual structures, can also be applied to real microstructures. This means that the obtained results will be validated using tomographic image data measured during the course of the project. More precisely, the spatial arrangement (3D) of the microstructure of titanium aluminides as well as their temporal evolution under mechanical loading (4D) will be resolved. The SMILE project combines complementary expertises of the German-French partners at Ulm University and Mines Paris Tech to enable an efficient elucidation of microstructure-property relationships for titanium aluminides using AI. The project thus makes an important contribution to the further digitisation of materials science.

 

SMILE is funded by the Federal Ministry of Education and Research (BMBF) and the Ministère de l' enseignement supériereur, de la recherche et de l' innovation (MESRI).

Contact

Prof. Dr. Evgeny Spodarev

Director of the Institute
Institute of Stochastics

Helmholtzstr. 18, Room No. 165
D-89069 Ulm

Tel.: +49 (0) 7 31 - 50 23530
Fax: +49 (0) 7 31 - 50 23649

E-Mail: evgeny.spodarev(at)uni-ulm.de
web: Prof. Dr. Evgeny Spodarev

Other Topics from the faculty of Mathematics and Economics

American mathematicians are calling on their colleagues to refrain from collaborating with the police. They say it lends a scientific appearance to a racist system.

Mathematics is clarity. It can describe aspects of our ambiguous world, but it is itself unambiguous. In mathematics, only that which can be proven is true. It seems detached from the agitated reality we are currently seeing on the streets of the United States, where George Floyd’s death first triggered demonstrations and then eventually the toppling of monuments. But now a group of mathematicians has brought their sedate science into this tempest.

IThe ever-increasing presence of cars in cities leads to more and more congestion as well as air and noise pollution, among other things. It also takes up a lot of space. It is therefore vital to strengthen local public transport as a means for more sustainable urban mobility. The Institute of Sustainable Corporate Management at Ulm University is thus investigating whether temporarily free local public transport can be an effective measure in this regard. read more

Experience shows that the expectations people have of the standard of living they will able to achieve in old age often deviate from reality. There are numerous arguments in favour of an early, comprehensible and realistic presentation of the retirement income that can be expected. This requires an easily accessible source of relevant information on the status of one’s own individual pension coverage – summarized across all three pillars (statutory, company and private pensions). This information should be as complete, as comprehensible, as reliable and as comparable as possible. In addition, it is crucial that cross-pillar pension information also be efficiently usable and cost effective, to ensure it is both accepted by the people and supported by pension institutions.

The study is available as a complete document and as a summary.

Developing alternatives for sustainable textile consumption among young people

A joint project of Ulm University and the Technical University of Berlin, BNTextillabor aims to conduct research on more sustainable fashion consumption among young people. The consumption behaviour of young people was investigated and concepts for a more sustainable textile consumption were developed. Can the changed behaviour of young people subsequently be transferred to other areas of consumption? Based on the results, teaching and learning formats for different types of schools will be developed with the objective of influencing the behaviour of young people in a sustainable manner.

As part of an advanced CSE project, the use of a leap motion controller (LMC) in the field of biomechanics was to be reviewed. The LMC is a computer hardware sensor device, which, like a mouse, supports hand and finger movements as input, but does not require hand contact or touch. This sensor is designed to help record a hand movement that is then transferred to the AnyBody Modeling System™ (AMS). The AMS can be used to determine muscle and joint loads for a given body motion in an inverse dynamic simulation that will help optimise the healing processes of a fracture.

The analysis of fracture healing processes and the optimisation of patient treatment is an ongoing area of research. In addition to experimental methods, computational modelling also plays an important role in understanding the process of bone growth. Mechanical stimulation is an essential factor for simulating the healing process.

How social and mobile technologies can help

The risk of long-term unemployment is particularly high for unemployed persons over the age of 50. Researchers at the Institute for Business Analytics at Ulm University have developed a new, innovative approach to support this target group in their job search using social media: digital peer group counselling (DIGIPEG). Unemployed individuals over 50 support each other digitally, anonymously and voluntarily. DIGIPEG was introduced throughout Baden-Württemberg as part of a two-year application-oriented research project. Results show that digital peer groups significantly improve the job search – as measured by application activity, application skills and return to employment.

Those who lose their jobs at 50+ generally have a harder time getting back into the workforce than younger unemployed individuals. In addition to the job search, the psychological burden is also a challenge for those affected. Along with the risk of long-term unemployment, the threat of poverty in old age increases. Particularly in these times of demographic change, solutions to this problem are needed more than ever.

Discrimination by algorithms is increasingly perceived as a societal and legal problem. In response, a number of criteria for implementing algorithmic fairness in machine learning have been developed in the literature. However, some of them are known to contradict each other, both philosophically and/or mathematically. In recently published work, BCCP Senior Fellow Philipp Hacker, together with co-authors Meike Zehlike and Emil Wiedemann, propose the continuous fairness algorithm (CFAθ), which enables a continuous interpolation between two contradictory fairness definitions, namely individual and group fairness. Individual fairness is commonly understood as "similar individuals should be treated similarly." Group fairness posits that the chance of receiving a positive outcome should be equal across all protected demographic groups. read more