AI-based diagnostics: A look at the surface is enough
There are good diagnostic options to see what’s going on underneath the material just by looking at the surface. During material tests on aircraft, for example, the deformation of the overall structure can be calculated from images of the wings. Up until now, however, a real look inside has required much more complex measures such as X-rays or simply cutting open. This is exactly where a team comes up with a new idea: an AI that provides detailed information about the interior based on the properties of the material surface.
Identifying internal defects from the outside: An application example from the researchers
A look at the surface says it all
As the two researchers, doctoral student Zhenze Yang and her professor for civil and environmental technology Markus Buehler, wrote in the journal Advanced Materials (via WITH), they use the strengths of AI systems to recognize patterns in large data sets for their approach. In this case, the model was fed with vast amounts of data to link the surface measurements to the associated internal properties. The next step was the important phase of revising, refining and supplementing the model. To do this, the AI makes predictions that are then compared with actual data. Deviations are recognized and other results are checked again in the same process, taking into account the corrections.
As Yang writes, this method once again proves to be extremely effective. The system can very reliably identify uniform as well as combined materials and thus delivers good results for both materials and biological samples. The scientists are amazed, however, that the AI also achieves a lot in completely different areas. “Not only is it limited to solid mechanics problems, but it can also be applied to other engineering disciplines such as fluid dynamics and other fields,” Yang said loudly Phys.
Buehler adds that the technology could be used in everyday life in the short term, for example to supplement more expensive technologies in the maintenance of aircraft. The AI provides clues as to where structural problems are to be expected, investigations can be carried out in a more targeted manner – at least that’s the idea. Initially, however, he expects it to be used primarily in laboratories, which can use it to develop and test completely new materials. The results of the two researchers are freely available to everyone on Github.
- AI system draws conclusions about the inner structure from the surface
- Database linked to surface measurements and internal properties
- System recognizes uniform and combined materials very reliably
- AI can also be used in other engineering disciplines such as fluid dynamics
- Mainly used in the laboratory to develop and test materials
- AI can complement more expensive techniques in aircraft maintenance
- Freely available on Github
See also:
