Study on Algorithms for the Virtual Assembly and Best Combinations of In-line Measured Injection-Molded Parts

Geometrical deviations of manufactured parts affect the quality of the resulting product. In an assembly, deviations accumulate so that the functional requirements may no longer be met. Therefore, the compliance of geometrical interfaces is of great importance in order to achieve a high-quality product. Modern 3D sensors capture dense point clouds of manufactured parts and thus enable creating a part’s geometrical digital twin. However, there are no suitable registration techniques that allow computing the assembly of the measured parts. With the knowledge of the assembly configuration, the geometrical accumulation can be determined, ultimately allowing the prediction of the quality of the final product. In the authors preceding work, the virtual assembly registration technique was presented methodologically. This paper focusses on the implementation of a dedicated use case. As real-world problem, the assembly of injection-molded covers and housings is examined. The data acquisition is fully integrated into an in-line measurement process. Therefore, a fringe projection system is used for capturing 3D data. Specific segmentation approaches are presented in order to detect the part and to segment assembly-relevant geometries in the measurement data. Due to possible artefacts in the point clouds, strategies for increasing the robustness are discussed. Finally, the developed quality prediction is embedded in the production context. In order to achieve time, cost, and quality benefits, a selective assembly strategy is applied that reduces the number of scrap parts by a factor of two and more.