SUMMARY

Metal additive manufacturing has provided a viable processing pathway for materials considered difficult for traditional manufacturing methods. However, one challenge with extending this manufacturing technique to new materials is developing and optimizing processing parameters that can be used to fabricate parts. This work presents a statistics-based methodology that can be used to determine optimized process parameters for the directed energy deposition-laser beam-powder blown process. The Julia programming language is used to implement the presented methodology. The method can be used to scale from single laser tracks without deposition to building parts with simple geometries. The methodology is applied in a case study on the Niobium alloy C-103. Using data from single laser track experiments to set bounds for the parameters, the work details using an active learning approach to optimize process parameters for the fabrication of thin walls as well as exploring the process space. This approach results in the determination of multiple process parameter sets that can be used to be fully dense parts (>99% relative density), with a maximum relative density of 99.87%. The resulting statistical model is then analyzed using Shapley values to determine the relative importance of different process parameters and the relationships between the values of the process parameters and the resulting densities. These optima are discovered with the fabrication of just over 50 specimens, suggesting a reduction in the number of experiments needed to map the process space of other materials.

Teams link: bit.ly/3DK6EUt