Assessing the printability of alloys in fusion-based additive manufacturing: towards criteria for alloy selection

Abstract

Fusion-based Additive Manufacturing (AM) has recently gained much attention due to the process's flexibility and the possibility of fabricating complex metal components. The technology offers tremendous advantages in production time, geometrical flexibility and material savings. However, the process is quite unique and distinctive from traditional manufacturing methods, which brings a lot of unsolved metallurgical issues during processing and post-processing. Moreover, the limited materials portfolio restricts the broader adoption of the technology. The purpose of this study was to assess the printability of alloys in fusion-based AM, where the “printability” is defined as “the ability of a material to be consolidated layer-upon-layer to form a designed part with desired microstructure and properties that ensure the performance of the printed material over its intended service life”. This PhD thesis gains a deeper understanding of the interplay between the process characteristics, material physical properties and the printability aspects of the alloys during fusion-based AM. Out of the fusion-based technologies, laser powder bed fusion (LPBF) is arguably the most common. It has quite similar process characteristics to electron beam powder bed fusion (EB-PBF) and direct energy deposition (DED) methods, so LPBF was chosen as a representative technology for fusion-based AM. It is, however, recognised that slightly different processing conditions in EB-PBF and DED might require some modifications of the developed theories. Liquid-state cracks are one of several defect types commonly observed in fusion-based AM, limiting the reliability of components. Cracks tend to create high stress concentration locations that can significantly limit mechanical performance, particularly in fatigue. It is of paramount importance to prevent liquid-state cracking in LPBF to enable the reliable long-term performance of alloys. The tendency of alloys to form solidification cracks was correlated with the material solidification behaviour using fusion welding concepts, particularly solidification gradient. Solidification gradient served as a good assessment indicator to explain why 316L steel is immune to solidification cracking, while IN718, HEA and Hastelloy-X contain some degree of cracks. Although the solidification gradient alone could not rank different alloys according to the crack density, it could explain the difference in the number density of cracks between Hastelloy X variants with slightly different chemical compositions. It was found that minor compositional changes in Si and Mo content in Hastelloy X can drastically influence the solidification behaviour and cracking tendency, hence the printability. The solidification cracking assessment criteria was then used to develop a design methodology against liquid-state cracking by combining the Scheil–Gulliver solidification simulations and Machine Learning analysis to design alloys for Fusion Additive Manufacturing. Applying this design approach resulted in a Fe–20Cr–7Al–4Mo–3Ni. The alloy was successfully printed with relative densities of over 99 %. The microstructure of printed material was extensively characterised through scanning and transmission electron microscopy, energy dispersive spectroscopy and X-ray diffraction, confirming a single-phase material with low texture and negligible chemical segregation. Neither solidification nor liquation cracks were detected, supporting the validity of the methodology; however, the alloy suffered from solid-state cracking, hindering the ductility and reliability in structural applications. Another key aspect of printability is the ability to achieve desired microstructure and mechanical properties. Process parameters, in particular, scanning strategy, can significantly influence the consolidation, solidification microstructure, and mechanical properties of the alloy. A range of print parameters was used for a comprehensive assessment of printability of CoCrFeMnNi high entropy alloy, providing a basis to establish the relationship between process, microstructure, and mechanical properties. The study demonstrates a high relative density of the alloy fabricated with energy density in the range 62.7–109.8 J/mm3. It is shown that the scan strategy plays an important role in consolidation. For the same energy density, the rotation of 67° between two consecutive layers tends to yield higher consolidation than other considered strategies. Moreover, the scan strategy is found to be most influential in microstructure development. The scan strategy rotation angle controls the extent to which epitaxial growth can occur, and hence the crystallographic texture and the grain morphology. Amongst four considered strategies, the 0°- and 90°-rotation meander led to the strongest preferred texture while the 67°-rotation resulted in a weaker texture. The 67°-rotation strategies led to broadened grains with lower aspect ratios. The understanding of texture and grain size explains the observed mechanical properties (such as flow stress and plastic anisotropy) of the alloy. Mechanical properties of the printed alloy can be further tailored via post-processing, particularly heat treatment. The microstructural development and strengthening mechanisms of IN718 manufactured by fusion-based AM was investigated and compared to a wrought variant. Two different heat treatments were performed, a standard solution heat treatment (SHT) commonly used for conventional IN718 and a modified SHT developed for LPBF IN718 from prior literature. CALPAHD tools were used to gain insights into the dynamics of precipitate formation, while characterisation techniques were used to compare the morphology. It was found that LPBF processed IN718 had superior mechanical properties compared to the wrought variant, both heat-treated with the conventional SHT. By deconvoluting the strengthening contributions of various strengthening mechanisms, it was identified that primarily the strengthening in heat-treated IN718 comes from order and coherence strengthening form γ’’. However, despite the underaged size of the precipitates in LPBF IN718 after conventional heat-treatment, the contributions from prior solidification cells and increased dislocation density had a high enough impact to raise the strength of the alloy above the strength of conventional IN718 with optimal precipitate size. Additionally, it was found that the oxide scale on the surface of LPBF processed IN718 after the heat treatment had a much lower impact on the mechanical properties of the alloy in comparison to near-surface porosity. Hence, to enhance printability, a reliable method to optimise process parameters has to be developed. To accelerate the search for an optimal printability window during the fusion-based AM, an approach incorporating the analytical Rosenthal model and machine learning based on prior literature results was developed. Over 2000 observations of process parameters and resulting sample properties were collected to train the machine learning models. First, a Gaussian regression model was trained to predict the laser absorptivity in Fe-based, Ni-based and Al-based alloys depending on the process parameters. This process-dependent absorptivity was incorporated into the analytical Rosenthal model to accurately calculate the melt-pool depth depending on the alloy, laser power, scan speed and laser spot size. A neural network ensemble was developed to predict the melt-pool depth as a function of the aforementioned process parameters. These models were used to assess the printability of 316L, Hastelloy X, IN718 and Ti6Al4V, in particular, the consolidation, based on the ratios between melt-pool depth and layer thickness; and melt-pool width and hatch spacing. Such consolidation plots identify regions of optimal printability based on the process parameters without the need for lengthy experimental trials.