23. Internationales Stuttgarter Symposium

Automobil- und Motorentechnik

4. - 5. Juli 2023

Session: Poster |

Intelligent Analysis of Components with regard to significant Features for subsequent Classification

Alexander Nüßgen, Technische Hochschule Köln

The knowledge explosion in all areas, including the automotive industry, is both an opportunity and a risk. In order to shape the effects as positively as possible and create sustainable added value, the amount of information must be processed in a targeted manner and important content must be separated from ignorable content. In the automotive industry and especially in the development of new components, it is possible to look back on many past projects and thus knowledge. However, the decisive factor is whether this information is available in a suitably processed form or the knowledge is even held by just a single expert. A new and intelligent method is therefore required to analyze existing data appropriately and at the same time prepare it ideally for further applications, such as use within forecast models based on artificial intelligence. To achieve the aforementioned goal, several steps need to be taken. First, the voxelization of the component takes place, which results in the three-dimensional component or CAD file being mathematically readable and thus a kind of translation takes place. This is done by rasterizing the component based on a previously selected resolution and other upcoming steps. Subsequently, it is possible to perform a suitable segmentation of the component. The aim is to detect areas in a component where features and form elements can be found. These could be, for example, areas that contain recesses or have other previously defined properties. Other regions, on the other hand, can be ignored after inspection by voxelization and segmentation: There is no sustainable valuable knowledge here. In a final step, the segmented areas including the features present there can be analyzed accordingly. For this purpose, the shape of the features is examined by means of ray tracing. In summary, these individual steps provide a possible method for coping with the knowledge explosions described at the beginning. As a result, not only can knowledge that is unimportant for the application be deliberately ignored, but it can also be suitably analyzed and further processed.