PHYSICS-BASED HIGH FIDELITY MODELING OF HEAT AND MASS TRANSFER IN LASER ADDITIVE MANUFACTURING PROCESSES WITH APPLICATIONS TO PROCESS QUANTIFICATION AND OPTIMIZATION
In the advent of laser additive manufacturing (LAM), extensive efforts have been taken to optimize the properties of resulting manufactured products. Since optimizing these processes experimentally is expensive from both an equipment and materials perspective, modeling of the processes is critical to gain insight into the key parameters necessary to produce a high-quality manufactured component. Physics-based high fidelity modeling of additive manufacturing processes can provide information to predict material properties via track geometry and temperature field; however, previous models require tuning factors that prevent prediction of deposition processes over a wide range of materials or operating conditions.
The overall objective of this research was to develop a methodology to systematically describe each aspect of the LAM process (laser-powder interaction, powder-surface interaction, and heat transfer mechanics) and use the relevant information to feed into various models to predict microstructures, phases, and properties of the resulting deposition. The methodology was demonstrated on a variety of deposition systems, including blown-powder systems and powder bed systems to show the robustness of the method and the predictive capabilities of simulating each of the aforementioned aspects of the process to obtain the track geometry and temperature field, the key factors necessary to determine material properties of as-built components. Although the interactions of the powder, laser irradiation, and substrate are different in nature and must be modeled with due-diligence, these were found to be boundary conditions for a common-core deposition model applicable for any LAM process.
For blown-powder systems, computational fluid dynamics (CFD) was used to calculate the average spatial distribution of powder as the powder is ejected from a gas-assisted nozzle. This was then coupled to the molten pool dynamics model, which involves melting, fluid flow and subsequent heat transfer to the surrounding areas, which are solved by a set of coupled momentum, continuity and energy equations with proper source terms and boundary conditions with the free surface tracked using the levelset method. This model was subsequently applied to H13 and Ti-6Al-4V powder being deposited on their respective substrates in a single-track configuration to understand the temperature field and track geometry throughout the LAM process. These studies enabled the prediction of the phases, microstructure, residual stress and hardness of as-built components produced with blown-powder LAM for these two materials. More importantly, predictions of capture efficiency were obtained, as opposed to using capture efficiency as an input, which previous researchers relied on as a model tuning parameter. The study of Ti-6Al-4V was taken further by simulating a multi-track deposition with the same LAM parameters and was shown to predict the molten pool region, heat affected zone, and track geometry after three tracks were simulated without the need for any model tuning. Since powder concentration could be calculated throughout the computational domain, the effect of standoff distance on the deposition process was studied to optimize the best cladding condition for Stellite-6 cladding of a mild steel substrate, wherein the cladding is often performed with the laser focal point above the substrate surface to minimize dilution.
The AM model has been extended to powder bed additive manufacturing by modeling particle-particle and particle-surface interactions via a discrete element model coupled with the molten pool dynamics model developed for the blown powder model. Particles of Ti-6Al-4V were modeled with aerodynamic drag and cohesive forces to demonstrate the effect of evaporative-driven gas flow during powder bed deposition, a phenomenon which has been observed experimentally, but had yet to be modeled with reasonable accuracy when coupled to a LAM process model. Simulation of the powder bed formation was included to consider particles of multiple sizes and multiple spreading passes, which is necessary for obtaining a physically representative powder bed. Finally, the model was updated with a robust dual-mesh algorithm that allows for the simulation of high scanning speed processes for large manufactured components without excessive computational effort associated with large-scale simulations.
With these modeling processes being used to predict the geometry and temperature field of a deposition, regardless of the powder feed mechanism, the results could be used to verify optimal LAM parameters from experimentation. Unfortunately, computational effort and cost for modeling of these processes for a large domain is prohibitively expensive, which is needed to determine the resultant microstructure and mechanical properties of industrial large-scale parts. Though having a high-fidelity model of the deposition process enables accurate prediction of the track geometry and temperature fields, methods to increase model throughput are necessary to obtain accurate process predictions without excessive computational effort. A combination of an Arbitrary Lagrangian-Eulerian mesh formulation and volumetric powder-bed heating methods decrease computational effort compared to analogous models in literature by up to 95%.