LASER CLADDING OF ALUMINUM ALLOYS AND HIGH-FIDELITY MODELING OF THE MOLTEN POOL DYNAMICS IN LASER MELTING OF METALS

This research focuses on understanding and improving various metal additive manufacturing processes. The first half is dedicated to experimental investigations and methods for improving the laser cladding of aluminum alloys. The second half is dedicated to high-fidelity modeling of the laser melting process and methods for reducing the computational burden.

First, laser cladding is a surface enhancement and repair process in which a high-powered laser beam is used to deposit a thin (0.05 mm to 2 mm) layer of material onto a metal substrate with no cracking, minimal porosity, and satisfactory mechanical properties. In this work, a 4 kW High Power Diode Laser (HPDL) is used with off-axis powder injection to deposit single-tracks of aluminum alloy 6061 powder on a 6061-T6511 substrate. The process parameters were varied to identify the possible processing window in which a successful clad is achieved. Geometrical characteristics were correlated to the processing parameters and the trends were discussed. Microhardness testing was employed to examine the mechanical properties of the clad in the as-deposited and precipitation heat-treated conditions. Transmission electron microscopy (TEM) was used to investigate the precipitate structures in the clad and substrate as an explanation for the hardness variations. Experiments were completed on two substrate widths to understand the effect of domain size on the process map, layer size, and hardness.

Second, a method to deposit quench-sensitive age-hardening aluminum alloy clads is presented, which produces a hardness similar to the T6 temper without the requirement of solution heat treatment. A high-powered diode laser is scanned across the workpiece surface and material feedstock is delivered and melted via off-axis powder injection. The cladding process is immediately followed by quenching with liquid nitrogen, which improves the cooling rate of the quench-sensitive material and increases the hardness response to subsequent precipitation heat treatment. The method was demonstrated on the laser cladding of aluminum alloy 6061 powder on 6061-T6511 extruded bar substrates of 12.7 mm thickness. Single-track single-layer clads were deposited at a laser power of 3746 W, scan speed of 5 mm/s, and powder feed rate of 18 g/min. The in-situ liquid nitrogen quenching improved the clad hardness by 15.7% from 73.1 HV to 84.6 HV and the heat-affected zone hardness by 19.3% from 87.1 HV to 103.9 HV. Extending the process to multi-track multi-layer cladding further increased the clad hardness to 89.3 HV, close to the T6 temper hardness of 90 HV. Transmission electron microscopy revealed the increased precipitate density in the liquid nitrogen quenched clads was responsible for the higher hardness.

Third, a high-fidelity model of the molten pool dynamics during the laser melting of metals is presented for accurate prediction of the molten pool size and morphology at operating conditions relevant to laser powder bed fusion. The goal of this research is to improve the accuracy of previous models, present a thorough experimental validation, and quantify the model's sensitivity to various properties and parameters. The model is based on an OpenFOAM compressible Volume-of-Fluid (VOF) solver that is modified to include the physics relevant to laser melting. Improvements to previous works include the utilization of a compressible solver to incorporate temperature-dependent density, implementation of temperature-dependent surface tension and viscosity, utilization of the geometric isoAdvector VOF method, selection of a least squares method for the gradient calculations, and careful selection of physically accurate material properties. These model improvements resulted in accurate prediction of the molten pool depth and width (mean absolute error of 7% and 5%, respectively) across eleven operating conditions spanning the conduction and keyhole regimes with laser powers ranging from 100 W to 325 W and scan speeds from 250 mm/s to 1,200 mm/s. The validation included in-house experiments on 304 L stainless steel and experiments from the National Institute of Standards and Technology on Inconel 718. Incorporating the large density change from the ambient temperature to vaporization temperature and utilizing a least squares scheme for the gradient calculation were identified as important factors for the predictive accuracy of the model. The model sensitivity to the wide range of literature values for laser absorptivity, liquid thermal conductivity, and vaporization temperature was quantified. Literature sources were analyzed to identify the most physically accurate property values and reduce the impact of their variability on model predictions.

Finally, an original surrogate model is presented for the accurate and computationally efficient prediction of molten pool size in multi-track laser melting over a large domain at operating conditions relevant to laser powder bed fusion. The thermal models available for the laser melting process range from heat conduction models to high-fidelity computational fluid dynamics (CFD) models. High-fidelity models provide a comprehensive treatment of the relevant physics of heat conduction, fluid flow, solidification, vaporization, laser propagation, etc. A carefully implemented high-fidelity model is capable of accurately predicting the molten pool dynamics in a broad range of operating conditions. However, the high computational expense limits their application to a few short tracks on small domains. Conduction models, on the other hand, are orders of magnitude cheaper to evaluate but lack the necessary physics for accurate predictions. This research presents a surrogate model that combines the computational efficiency of the conduction model with the accuracy of the high-fidelity model. A conduction model and high-fidelity model are simulated over a small scan pattern to generate training data of the highly transient molten pool depth and width. A surrogate model, consisting of a fuzzy basis function network, is trained with the aforementioned data. The conduction model is then simulated over a larger scan pattern, the results are input into the trained surrogate model, thereby outputting high-fidelity predictions of the molten pool size over a larger scan pattern. Comparison with experimental results shows this surrogate modeling framework provides reasonably accurate predictions of the molten pool size and is a valid way to extend computationally intensive high-fidelity models to larger and more industrially relevant scan patterns.