Automated 2D-3D-Mapping and Assessment of Defects Obtained from 2D Image Detection on a 3D Model for Efficient Repair of Industrial Turbine Blades

Manufacturing processes suited to repair worn-out machine parts are increasingly gaining traction in industrial practice. This evolution is driven by the desire to contribute to sustainable production by extending the lifetime of a part, and methods and machine tools have evolved in the last years, such that the refurbishment reaches equivalent quality compared to new parts. Yet, repair processes often remain too tedious and expensive to be profitable, and are therefore not put in practice. This could be prevented by increased efficiency and automation. In order to repair parts, the defects need to be detected and assessed. This task lies in the field of Reverse Engineering (RE). In the scope of repair and overhaul, RE aims to obtain data and models that can be fed to subsequent applications, such as path planning for additive manufacturing. This paper presents a highly automated process for defect inspection by the example of industrial turbine blades. The current process requires many analogue work steps and human intervention. The developed software processes defects obtained by AI-based image classification. It mainly consists of camera scene calibration, 3D pose estimation, and 2D-3D-mapping. The gained value is the new composition of existing technologies and their customization for turbine blades. Accuracy and computational duration are assessed. The presented method is able to enhance the reparation process and can be deployed to different applications and industries.