Conclusion

The work presented in this article shows that artificial intelligence models enable fully automated processing of tomographic data. This processing integrates the localization of discontinuities (segmentation), as well as their classification to retain only those discontinuities considered as defects. To date, the proposed software has been evaluated and validated on aluminum alloys produced by gravity casting and die-casting.

Ultimately, the aim is to implement the models developed in industrial software, in order to make tomography-based production inspections of metal products more reliable, and to provide industrial producers and users of these tomography-controllable products with new inspection automation solutions, based on qualified solutions. The ultimate aim is to encourage the choice of cast solutions, since casting allows for more complex, streamlined...