Mechanistic modeling of heat transfer and fluid flow during fusion-based additive manufacturing of structural alloys
Open Access
- Author:
- Knapp, Gerald
- Graduate Program:
- Materials Science and Engineering
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- May 19, 2021
- Committee Members:
- Edward Reutzel, Outside Unit & Field Member
Jingjing Li, Outside Field Member
Allison Beese, Major Field Member
John Mauro, Program Head/Chair
Tarasankar Debroy, Co-Chair & Dissertation Advisor
Todd Palmer, Co-Chair & Dissertation Advisor - Keywords:
- numerical modeling
heat transfer
mass transfer
additive manufacturing
alloys - Abstract:
- During fusion-based additive manufacturing (AM) of metals and alloys, a part is built up layer-by-layer by melting the feedstock to add material only where it is needed selectively. The small volumes of material that solidify at any given time lead to significantly different material processing routes than traditional manufacturing, such as casting, forging, and welding. Additionally, differences in AM processes from more conventional metallurgical processes arise from the heat sources’ size and intensity, feedstock characteristics, complex scanning strategies, and repeated heating and cooling due to layer-by-layer deposition. These differences culminate in novel and largely unquantified heat and mass transfer phenomena in the molten pool underneath the focused heat source. The molten pool’s characteristics directly relate to the part microstructure and defect formation, so understanding the heat and mass transfer within the pool is key to controlling process-property relationships. While AM processes have numerous input parameters, such as heat source power, scanning speed, and hatch spacing, these do not directly correlate with the output variables of interest to the process-property relationship. Instead, heat and mass transfer within the molten pool determines the resultant solidification microstructure, fusion zone, and defects like lack-of-fusion porosity. However, test matrix or statistical approaches for investigating process parameters’ effects on process outcomes obfuscates the physical mechanisms that operate inside the molten pool. It is, therefore, necessary to gain direct insight into the mechanisms at play within the molten pool under the variety of conditions that occur during the different types of AM processing. This dissertation investigates the overarching role of heat and mass transfer mechanisms in the molten pool by examining a broad range of alloys and processing conditions that fall under the umbrella of AM. Four prominent AM processes are studied; namely, laser directed energy deposition (DED-L), wire and arc additive manufacturing (WAAM), and laser and electron beam powder bed fusion (PBF-L and PBF-EB). Mechanistic models were developed to conduct virtual experiments that provided insight into the molten pool for various AM alloys in collaboration with physical experiments for model validation. Specifically, the roles of alloy composition, alloy properties, and process-specific parameters were investigated with a focus on the molten pool geometry, solidification, and solidification microstructure. Whereas physical experiments are limited in their ability to interrogate the molten pool due to the molten pool’s extreme conditions and optical opacity, virtual experiments can thoroughly interrogate the spatially and temporally varying conditions within the molten pool. Furthermore, the developed mechanistic models allowed the roles of unique heat and mass transfer mechanisms in each AM process to be quantified. The role of alloy composition on deposit geometry was first investigated for the DED-L process. During DED-L, powder feeding allows for spatial variation in composition and enables the creation of functionally graded materials. However, variations in deposit geometry as a function of material composition resulted in difficulties for fabrication of dense parts that meet geometric tolerances. Many factors that influence heat transfer and fluid flow within the molten pool change as a function of composition, resulting in spatial variations in the heat transfer and fluid flow. Here a mechanistic numerical heat transfer and fluid flow model was coupled with thermodynamic data to identify changes in fluid flow as a function of composition in a functionally graded ferritic-to-austenitic steel part made by DED-L. Surface-active elements and their changing thermodynamic activity in the molten pool alloy as a function of liquid composition were found to be the main cause for differences in deposit geometry. Due to interactions of surface-active oxygen with minor alloying elements in the austenitic alloy, e.g., Al and Ti, surface-active elements in the austenitic compositions had much less activity than in the ferritic compositions. Calculated deposit geometries showed that this difference in temperature-dependence of surface tension was the main factor in differences in the molten pool geometry rather than the alloys’ thermophysical properties. These results highlight the importance of considering the interactions of the liquid alloy’s thermodynamics and heat transfer and fluid flow during DED-L of functionally graded materials. The role of composition and process-specific phenomena was then investigated for WAAM processing. Because of the relatively large deposits compared to other processes, even small changes in molten pool size can lead to sizeable lack-of-fusion defects between passes and layers that impact bulk part properties. Previous molten pool-scale models for WAAM focused on droplet transfer from the wire. However, there is a lack of mechanistic models that can simulate multiple passes and layers due to the computational complexity of interface calculations. A transient implementation of the free surface energy minimization technique was implemented to enable the study of molten pool effects on the multi-layer, multi-pass deposit geometry during WAAM. The model was validated against several single-pass experimental results from literature and multi-pass results obtained from collaborators’ experiments. Analysis of the underlying mechanisms of heat and mass transfer during WAAM of H13 tool steel and Ti-6Al-4V alloy showed that the arc pressure and fluid flow significantly impacted the deposit geometry. Fluid flow was shown to play a more significant role in H13 than in Ti-6Al-4V. Additionally, hatch spacing and heat accumulation were shown to impact the pool geometry and solidification characteristics during solidification. Compared to WAAM and DED-L, the small scale of PBF processes poses unique challenges for microstructural control. Depending on the alloy and process parameters, the solidification microstructure can undergo a columnar-to-equiaxed transition (CET). In particular, the processing window for Inconel 718 alloy by PBF-EB has been shown to produce either columnar or equiaxed grains. However, while experimentally demonstrated, the exact mechanisms contributing to the CET within the pool were unclear. A validated transient heat transfer and fluid flow model was used with the Kurz-Giovanola-Trivedi CET model to predict transient microstructural morphology under multiple processing conditions. Spatial and transient variations in the solidification morphology were shown to be sensitive to Marangoni convection and the temperature/preheat of the substrate during melting. Despite the relatively small size of the molten pool compared to DED-L and WAAM processes, heat and mass transfer driven by fluid flow still played a role in the outcomes of the PBF-EB process. In cases where columnar solidification morphology occurs, the orientation of the solidification microstructure preferentially grows along the temperature gradients during solidification. Because temperature gradients during solidification are perpendicular to the molten pool’s boundary, accurate prediction of fusion zone geometry is key to predicting the overall part microstructure and properties. While heat and mass transfer due to fluid flow is understood to be important for heat transfer in fusion-based processing, theory and results in welding literature indicate that fluid flow would be less influential in the smaller molten pool of PBF processes compared to typical laser welding conditions. A validated mechanistic model of PBF-L was used to show the differences between predicted molten pools with and without fluid flow during conduction-limited melting in alloys with negative surface tension temperature coefficient (i.e., no surface-active elements). Prominent AM alloys were used for the simulation dataset, namely Stainless Steel 316-L, Ti-6Al-4V, Inconel 718, Inconel 625, and Duplex Stainless Steel 2205. Significant geometric differences in the molten pool geometries are characterized and discussed concerning the impact on microstructure. Coupling the molten pool model to microstructural calculations done by collaborators indicated that the orientation of the solidification microstructure was affected by fluid flow-driven heat and mass transfer. Despite the relatively small molten pool compared to other AM processes, the fluid flow-driven heat and mass transfer during PBF-L simulations were shown to impact part-scale microstructure. By quantifying the mechanisms of heat and mass transfer in the molten pool, this dissertation provides a new understanding of molten pool phenomena and where they deviate from conventional welding and joining processes. In particular, the driving forces and impact of fluid flow in the molten pool were quantified for PBF, DED, and WAAM processes, and the results highlight the importance of capturing accurate model pool geometries and temperature distributions during modeling. This understanding further enables using mechanistic modeling as an exploratory and predictive tool for AM and provides insights into the molten pool that could not have been observed experimentally. The crucial roles of surface-active elements, substrate heating, and surface deformation on the resultant molten pool geometry and solidification metallurgy indicate several areas of future research by including these variables in machine learning and statistical control models for AM. Furthermore, as numerical models are integrated into AM process, and part design workflows, this physical understanding of the impact of the underlying physics on part characteristics will allow for the informed balancing of accuracy, cost, and speed of mechanistic models.