Additive Manufacturing

Advanced manufacturing techniques like selective laser melting (SLM) and directed energy deposition (DED) have the potential to allow efficient fabrication of complex metal parts through layer-by-layer addition of material. However, because each deposition requires the melting and re-solidification of metal, the range and scales of cyclic temperature excursions experienced by the material are more extreme than in traditional manufacturing processes like casting. The resulting microstructures, including grain size, porosity, and compositional segregation, depend heavily on process parameters like input laser power and scan path, as well as part geometry. The lack of predictive models for this dependence has hindered the adoption of additive manufacturing technologies. Our group’s key accomplishment is the development of an integrated set of multiscale simulation methods to predict the relationship between process parameters, material microstructure, and the resulting part behavior.

We have implemented HPC models for microstructure development using a cellular automaton (CA) method, coupling solidification and grain growth to the local thermal history in a material. The temperature field, in turn, can be predicted using detailed, multiphase thermo-fluid simulations of the molten material during solidification . This approach has been used to predict the microstructure resulting from single-track laser melting in a metal plate (Figures 1 and 2). Our simulation results for this problem have been validated against experiments, and recently won three 1^st Place awards in the AM-Bench challenge, a double-blind modeling challenge sponsored by NIST soliciting predictions for a series of controlled benchmark experiments. This model has also been applied to multilayer DED of nickel-based superalloys, giving intricate microstructures that match experimentally built geometries (Figure 3); these are the first 3D simulations of DED-built material to capture these microstructural details over multiple layers.

The thermal history at any location in the part, in turn, depends on geometry and process conditions at the part scale. To capture this, we have developed simulation software based on a finite element solution of the energy equation with phase change [7, 8]. This software tool, GAMMA, takes as input the part geometry, process parameters, laser path, and boundary conditions, and computes the temperature as a function of time throughout the part. Current work focuses both on improving the efficiency of this code using novel numerical methods and HPC implementations, and increasing modeling accuracy by incorporating fine-scale melt-pool information into the macroscale simulation.

GO-MELT represents one of our most recent developments in multiscale simulations. In GO-MELT, we combine three separate thermal solvers and couple them together using variational multiscale terms. As a result, we can get efficient and accurate simulations that can resolve the melt pool in the fine scales (micrometer scale) while modeling a part-scale build (centimeter scale). The code is open-sourced and can be found in the following GitHub repository: GO-MELT GitHub.