Multi-scale computational modeling of selective laser melting for process improvements
Selective Laser Melting (SLM) is a powder-based additive manufacturing technology that is already in use in a number of industries and applications. However, it suffers from large variations in build outcomes and difficulties in setting process parameters. This drives up costs and makes certification procedures difficult. Accurate models of the SLM additive manufacturing process have the potential to reduce testing and experimentation required when using new materials or part designs. By predicting part quality directly from the process inputs, problems such as incomplete melting or thermal stresses introduced by large temperature gradients can be identified and corrected. Simulating the complete build of an SLM part is challenging due to the difference in scale between the size of particles being melted and the size of the part being produced. Because of this, bed-scale continuum models, in which the powder bed is represented as a continuous medium, are typically used as the computational cost of resolving individual particles in a full-scale bed is prohibitive. However, continuum models require as inputs volume-averaged, effective properties of the powder which are often unknown and difficult to measure or obtain by experimental means. This work develops a multi-scale modeling approach in which particle-scale models, where individual particles are resolved in a small, representative, domain, are used to predict the effective optical properties, thermal conductivity, and melt behavior of a powder which are then used in bed-scale continuum models to predict temperature history, melt percentage, and thermal stresses in an SLM build. At every step in the process, uncertainties due to processing parameters, material properties, and bed configurations are quantified, allowing uncertainties estimated at the particle-scale to be propagated up through the bed-scale models to the final predictions.