QuaDSensS

Initial situation

Powder bed fusion of metals using a laser beam (PBF-LB/M) is established as an industrial manufacturing process for prototypes, individual components, as well as small to medium batch sizes. However, process stability is limited by a multitude of coupled influencing factors whose effects can only be controlled to a limited extent through process parameters. As a result, statistically unavoidable component irregularities and defects such as pores, lack-of-fusion defects, or cracks occur, which can act as crack initiators and significantly reduce mechanical properties.
For safety-critical applications, reliable quality assurance is therefore mandatory. For this purpose, among other measures, witness specimens are used for component qualification. However, these provide only a global assessment of the build job and do not allow conclusions to be drawn regarding the location, type, or severity of defects in the actual component. Complementary local inspection using non-destructive testing methods, such as CT scanning of each individual part, is in principle possible, but due to the high time and cost requirements, it significantly limits the economic viability and scalability of additive manufacturing.

Objective

The objective of the project is the process-integrated detection, classification, and quantification of defects in the PBF-LB/M process, with explicit consideration of the material-specific self-healing effect. Instead of relying on purely post-process component inspection, a knowledge-based quality assessment is to be derived directly from process monitoring data. Defect characterization thereby includes both the evaluation of the probability of occurrence and the quantification of relevant defect attributes such as size, shape, and location. By fusing multiple process monitoring signals, information gaps of individual sensors are to be closed and the overall informative value of the monitoring system significantly enhanced. Defect classification is achieved through the combination of domain knowledge on defect formation mechanisms with characteristic data patterns extracted from the sensor signals. In the long term, this approach aims to reduce the need for time-consuming and cost-intensive post-process quality assurance.

Approach

For defect quantification, systematic experimental studies are conducted at different structural levels of the PBF-LB/M process. These include investigations on single melt tracks (1D), area-based experiments with overlapping melt tracks (2D), and volumetric specimens (3D), in which the relevant process parameters are deliberately varied and the manufacturing process is monitored using different sensor systems.
The defects induced in this manner are subsequently detected and quantitatively characterized by means of sequential micro-CT scans. By spatially and temporally correlating the defect information with the process monitoring data, a direct assignment between local process states and the formation of specific defect types is enabled.
From the individual sensor signals, statistically as independent as possible features are systematically extracted, allowing not only the identification of the defect type but also a robust assessment of defect size and probability of occurrence. After an individual evaluation of the respective sensor signals, they are combined in an appropriate manner to further increase the informative value of the process assessment, for example through the targeted selection of different spatial or temporal levels of analysis.
A particular focus is placed on the influence of defect self-healing resulting from (partial) remelting, and its effects on both the resulting defect morphology and its representation in the process monitoring signals are investigated.

Acknowledgments
The research project is funded by the German Research Foundation (Deutsche Forschungsgemeinschaft DFG). We thank for the opportunity to work on this project.

Funding source

Consortium partners