Fraunhofer ISC, Center for High Temperature Materials and Design HTL, Bayreuth
Project Cluster DiMaWert
Digitisation of material development along the value chains
Completed funded project
Motivation
Methods of artificial intelligence
Artificial intelligence (AI) is developing rapidly and is being used in more and more areas of technology and development. The use of special neural networks has already led to major successes, particularly in the field of adaptive algorithms. Pattern recognition in complex data sets is an established field of AI. Due to the further development of sensor technology, such data is increasingly being generated in material development and production. Numerous other AI methods are available for various tasks.
Objective
Top-down material design for multiphase ceramics
In this sub-project, the benefits of AI methods were to be tested at all stages of the value chain. Different AI methods were to be used as examples in other sub-projects, compared with conventional techniques and, if successful, established. Among other things, AI techniques should enable top-down modeling of material and process properties in accordance with specified application requirements. The expertise of ISC employees with regard to the use of AI methods should be significantly expanded.
Results
AI-supported defect evaluation in ceramic components
- AI methods have been successfully implemented in the field of material design and material testing. A probabilistic machine learning method was developed for the top-down material design of ceramic composite materials. This makes it possible to adapt existing material systems quickly and specifically to the required specifications.
- In the field of material testing, neural networks were trained to detect defects in ceramic components and refractory bricks on the basis of computer tomography images. Using simulation-based and statistical methods, the effects of individual defects on the service life can be reliably estimated.
- Finally, various digital tools for software development were introduced, thanks to which the implementation of new methods and customer-specific applications will be more efficient in future.