The models developed were applied to the demonstrator system and used for systematically studying the influence of process parameters. Thus, process conditions which are attractive in terms of minimum stress, distortion and surface roughness can be identified. E.g. it was found that due to the very high conductivity of Cu, the meltpool is circular and highly sensitive to the peak intensity, which in turn may lead to a very rapid change from a shallow meltpool in conduction mode to a deep melt pool in keyhole mode.
As input to the model validation process, accurate stress data is required, which requires a methodology to consider coarse grains and texture as a typical characteristic of LPBF-generated microstructure. Applying the method, it was found that the residual stress inside Cu components is low due to the limited yield strength, while the residual stress in the ceramic substrates is significantly higher. Also, the influence of different beam shapes on the microstructure could be revealed. With a standard Gaussian beam, relatively small grains but a strong texture were observed, while a top hat beam profile produced large grains with negligible texture for the Cu specimen.
One key result in terms of process development is the fact that IR laser sources are able to generate fully dense Cu species with a density of ≥ 99.7% at optimum process conditions. However, the process window is small, leading either to insufficient densities or the destruction of the sensitive substrate if exceeded. Also, it was found that differences in functional properties are rather small for samples built with a Gaussian vs. samples built with a top hat beam profile.
For secure positioning and handling, a simple yet efficient fixture system was developed. It proved to handle the delicate substrates without introducing any defects. An electromagnet-based mechanism allows a fast exchange of base plates in the LPBF system.
Furthermore, a camera-based positioning system is developed, which determines the position of individual substrates with an accuracy of ≤ 30 µm in x- and y-direction and ≤ 0.1° concerning rotation. Also, a structure-borne sensor system was evaluated with regard to the detection of substrate cracking. It was found that transforming the time frequency domain signal into an envelope description and reducing the latter to a 2D representation by PCA allows detecting larger cracks with an F1-score of > 99.9% applying the SVM algorithm. Additionally, the ability of an OCT to resolve the surface height was evaluated and a resolution in the range of ≤ 30 µm in z-direction was found.
On the material side, a robust process to regenerate oxidised Cu powder was developed. It reduces the oxygen content of used powder by reduction in H2-atmosphere without impacting flowability or process parameters. Preliminary tests with recycled powder showed a comparable dependence of porosity and surface roughness on laser power and scan speed as virgin powder. Also, the effects of surface modification by etching were evaluated. However, absorption improved little, and the process window remained limited. Efforts to produce test samples of metal matrix composites from SiC and Cu failed, though, due to the large density differences and a lack of adequate homogenization.
