This investigation explores fatigue performance of laser powder bed fusion (LPBF)-fabricated AlSi10Mg alloy subjected to various post-processing treatments: stress relief (SR), Hot Isostatic Pressing (HIP), and HIP + T6, all followed by prolonged thermal aging at 177 °C for up to 1000 h. Microstructural evolution, hardness, density, and uniaxial fatigue performance, including SEM fractography, were examined to recommend practical service life thresholds. Microstructural analysis confirmed that HIP and HIP + T6 treatments effectively eliminated process induced lack of fusion porosity, which was identified as the direct cause of premature elastic regime failure in SR-treated samples. While HIP + T6 achieved the highest initial hardness, both HIP-treated groups showed susceptibility to thermal degradation. The HIP-only samples, in particular, suffered a fatigue life reduction of up to 38% after just 10 h of aging. Fracture analysis showed that unaged SR samples failed in a comparatively more brittle manner, characterized by tear ridges, whereas HIP and HIP + T6 samples displayed an initially ductile fracture with prominent dimple features. Prolonged aging caused the HIP + T6 samples to replace ductile dimple features with a mix of dimple features along with uneven patterns, which may correlate with their decreased fatigue performance. The results of this paper conclusively demonstrate that while HIP post- processing provides robust initial fatigue strength, its benefits are severely compromised by thermal exposure. And for applications limited to the elastic regime, SR post-processing may be suitable, as long as printing parameters are optimized to prevent lack of fusion defects.

AlSi7Mg (F357) alloy specimens were fabricated in two different laser powder bed fusion (LPBF) systems: EOS M290 and SLM 280HL. Vertical (Z) and horizontal (XY) orientations were fabricated, and five different thermal post processes were applied to samples, individually. According to ASTM F3318-18, the considered thermal conditions were as-built, stress relieved (SR1), HIP, T6 and HIP+T6. Subsequently, the individual samples were aged at 140 °C and 177 °C for 100 h and for up to 1000 h. Tensile specimens were machined down from the aged samples and tested as per ASTM E8/E8M-21. While the yield stress (YS), elongation (%), and Vickers microindentation hardness (HV) were somewhat different for the as-built components, the general trends for the different heat treatments were essentially the same. As-built and SR1 treated microstructures were dominated by microdendritic cells, while the HIP, T6 and T6 + HIP component microstructures consisted of recrystallized grains containing eutectic Si particles of various sizes and shapes within the grain interiors and the grain boundaries; which gave rise to wide-ranging mechanical properties. As an example of these widely-ranging mechanical properties, it was observed that components fabricated in the Z or build direction in the EOS system exhibited a YS, elongation, and HV of 225 MPa, 13%, and HV120, while when HIPed and unaged exhibited values of 87 MPa, 25%, and HV51, respectively. These same HIPed components when aged at 177 °C for 1000 h exhibited values of 81 MPa, 41% and HV44. The mechanical properties of the unaged, HIPed and aged fabricated in Z direction in SLM system were 85 MPa, 31%, HV51, and 80 MPa, 42%, and HV47, respectively, providing support for LPBF system fabrication compatibility. These measured mechanical property values represent a small fraction of the more than 1600 mechanical property measurements (YS, UTS, elongation, and hardness (HV)) in this study.

A critical aspect of generating mechanical property data is the accurate estimation of the cross-sectional area of the specimen being tested for computing stress. This ordinarily trivial matter assumes greater significance in the context of thin wall structures (defined here as less than 2 mm thick) fabricated using the laser powder bed fusion process commonly used in metal additive manufacturing. This is primarily on account of the significant surface roughness associated with additively manufactured parts as well as due to variations in cross-sectional area across the gauge. In this work, the thin-wall measurement capabilities of four different metrologies are examined – standard micrometer, point micrometer, blue light 3D scanning, and x-ray computed tomography (XCT) scanning. Seven specimens covering a thickness range of 0.3 to 2 mm are measured by each of the metrologies and tested mechanically, with stress-strain curves derived using the different section area estimates from these metrologies. XCT scanning is chosen as the baseline on account of having the highest resolution of the metrologies under study. Deviations from XCT scanning are established, showing that the standard micrometer is a poor method of choice, particularly at low thicknesses, but that point micrometer measurements are within 5% of XCT measurements across all thicknesses. Blue light scanning is also shown to be a reliable measurement method at all but the lowest thickness, and is also used to estimate the coefficient of variation (CV) in section area within each specimen, showing that the CV rises sharply as specimens get thinner. This paper thus demonstrates, for the first time, the validity of using a point-micrometer for section area estimation, while also showing how variations in section area for thin specimens can contribute to initiation of failure at lower strengths, and thus at least in part explain mechanical debits observed with reductions in thickness.

The Archimedes method is a commonly used approach for measuring the density of specimens fabricated with metal additive manufacturing. Most of the specimens fabricated for this purpose are cubes with 10 mm sides. This work examines the appropriateness of using the Archimedes method for estimating the density of thin-wall specimens (0.3-2 mm thick). The significantly higher surface area to volume ratio of these specimens, as well as their cumbersome size, can introduce measurement challenges not seen for the standard cubic specimens. The method is assessed for thin-walls in three ways: first, a Measurement Capability Analysis is conducted, quantifying repeatability and reproducibility and comparing them to those obtained for cubic specimens. Then, densities are assessed for their sensitivity for detecting differences due to changes in thickness, post-processing conditions (Hot Isostatic Pressing, or HIP) and build location. Finally, density values are correlated to mechanical tensile test metrics to establish trends and R-2 values. The Archimedes method is shown to be capable of measuring thin wall densities for specimens as thin as 0.3 mm, though standard deviation is shown to increase as thickness reduces. The method is also sensitive enough to detect differences due to thickness and HIP, and to a lesser extent, build location. Finally, densities obtained from this method yield interesting insights for modulus, ultimate tensile strength and elongation, with R-2 values in some cases exceeding 0.6. This work focuses on Inconel 718 tensile test specimens fabricated using the laser powder bed fusion process and demonstrates that the Archimedes method is a capable and critical tool in developing thin wall structures.

In this investigation, the optimization of mechanical properties with thermal post-processing treatments was analyzed across a wide range of variants. A major aspect of additive manufacturing is the correlation between heat treatments and the effects on the mechanical properties and microstructure of the printed materials. Therefore, the present paper describes a comprehensive overview of post-process heat treatments for Laser Powder Bed Fusion fabricated AlSi10Mg alloy consisting of stress relief anneals at 190 ̊C and 285 ̊C for 2 h, hot isostatic pressing at 515 ̊C for 3 h, hot isostatic pressing + T6 treatment for 6 h, and final aging of each of these conditions at 177 ̊C for up to 1000 h. This has resulted in 40 experimental variants: 20 in the vertical and 20 in the horizontal tensile direction. After tensile testing, the resulting mechanical properties (ultimate tensile strength, yield strength, and elongation) and stress–strain curves are analyzed for comparison between all variants. Ultra-fine cellular, micro dendritic structures (0.6–1.2 μm) along with melt-band structures dominated the asbuilt and stress relief anneal conditions. In contrast, hot isostatic pressing and hot isostatic pressing + T6 conditions were dominated by ~10 μm, equiaxed, recrystallized grain structures and pseudo-eutectic silicon particles with varying sizes and size distributions. Microhardness and fractography results also corresponded to their specific heat treatment and microstructure. The comparison and correlation of the heat treatments are presented to help advance the selection of design strategies for high performance applications.

Creating an environment to enable the seamless integration of experiment, computation, and data within a laboratory environment is essential to enabling the practice of Integrated Computational Materials Engineering. Such an environment depends on the connection of experimental equipment and high performance computing resources to a collaborative software environment that supports research teams through simulation tool sharing and archival data management in a secure manner. Key functions of such a system include project management, workflow management, tool staging, data provenance tracking, and user authentication. An overview will be provided on efforts to establish such an integrated collaborative environment in a research laboratory involved in material and process discovery and development in both structural and functional materials.

Electron beam melting (EBM) additive manufacturing (AM) technology has allowed the layerwise fabrication of parts from metal powder precursor materials that are selectively melted using an electron beam. An advantage of EBM technology over conventional manufacturing processes has been the capability to change processing variables (e.g., beam current, beam speed, and beam focus) throughout part fabrication, enabling the processing of a wide variety of materials. In this research, additional scans were implemented in an attempt to promote grain coarsening through the added thermal energy. It is hypothesized that the additional energy caused coarsening of Ti-6Al-4V microstructure that has been shown to increase mechanical properties of as-fabricated parts as well as improve surface characteristics (e.g., reduced porosity). Fatigue testing was performed on an L-bracket using a loading configuration designed to cause failure at the corner (i.e., intersection of the two members) of the bracket. Results showed 22% fatigue life improvement from L-brackets with as-fabricated conditions to L-brackets with a graded microstructure resulting from the selective addition of thermal energy in the expected failure region. Three L-brackets were fabricated and exposed to a triple melt cycle (compared to the standard single melt cycle) during fabrication, machined to specific dimensions, and tested. Results for fatigue performance were within ∼1% of wrought L-brackets. The work from this research shows that new design procedures can be implemented for AM technologies that involve evaluation of stress concentration sites using finite element analysis and implementation of scanning strategies during fabrication that help improve performance by spatially adjusting thermal energy at potential failure sites or high stress regions.

We illustrate how emerging methods in artificial intelligence (AI) may be useful in materials science. Historically, these methods were developed in the area of materials process control and, more recently, in the nascent field of materials discovery. However, machine intelligence is of much broader import and our primary objective here is to illustrate how such methods may be used to circumvent some serious roadblocks in the computer simulation of a significant class of computationally hard problems in materials science. This is illustrated by a new approach to solving the dynamics of the N-body problem for large numbers of objects of essentially arbitrarily complex geometry or interaction potential. The approach, based on a particulate artificial neural net dynamics algorithm (PANNDA) is more than two orders of magnitude faster than existing methods when applied to large systems and is only marginally slower (∼10%) than the theoretical lower limiting case of hard spheres. In this method an artificial neural net is trained to predict accurately the time to next collision for binary encounters spanning the Hilbert space of relative positions, orientations and momenta (linear and angular). This approach, which can be extended to soft complex systems, enables construction of exact, albeit numerical, models for the thermodynamic, transport and non-equilibrium properties of very large ensembles of hard or soft objects of arbitrarily complex shape or interaction potential. Our results open up the possibility of immediate application to an usually wide spectrum of contemporary computationally intractable “hard” problems ranging from granular materials with asperities through inclusion of complex many-body terms in the intermolecular interaction in molecular dynamics calculations of complex fluids and polymers.

We describe a novel physical application of the OctTree data structure [P. Meagher, Comput. Graphics Image Process 19(2) (1982) 129–147] in a dynamically tessellating algorithm, in conjunction with an object-oriented, constructive solid geometry library (DOC), to efficiently determine pore size distributions in large multi-particle systems. We apply the DOC library to investigate the evolving dynamics of pore formation in multi-particle systems, such as a mixture of smooth hard cubes and spheres and a collection of frictional soft spheres. We demonstrate that the algorithm is able to provide insight into the effect of structural changes on the porosity network; for example, during the uniaxial compaction of soft spheres, we find the number density of pores increases while the mean volume of the pores decreases. This trend is responsible for a shift in the distribution of the pore volumes to favour smaller volumes. We anticipate that the DOC method will have wider applications in the area of granular materials for studying the changes in pore structure in both experimental and numerical systems as a complement to the analysis of particle packing.

The molecular dynamics package DL_POLY has at its heart a number of versatile and efficient dynamics algorithms that can readily be adapted to extend the application of this code well beyond the time and length scales typically associated with atomistic simulations. In order to achieve this, it is necessary to substitute the appropriate interparticle potentials and forces in place of the default functional forms in DL_POLY, which are mainly suitable for molecular systems. To facilitate this, it may be required to incorporate additional factors, into the simulation, such as velocity-dependent dissipation effects (friction), rotational degrees of freedom and non-spherosymmetric forces. In this paper, we will discuss some of the practical details of implementing these changes to DL_POLY (version 2) together with applications of discrete particle dynamics methods, such as dissipative particle dynamics (DPD) and granular dynamics (GD) (also known as the discrete or distinct element method, DEM) to particle packing in composite systems and pharmaceutical powders. We also consider how well the approach of simulating particles of arbitrary shape using rigid assemblies of fused soft spheres (each individually interacting via pairwise continuous potentials) compares to true hard-body simulations of polygonal particles.

We report a method of conducting molecular dynamics (MD) simulations that uses an artificial neural net (ANN) to significantly increase computational speed. The technique enables dynamical simulation of hard objects with essentially arbitrarily complex geometry and is well suited to the simulation of granular matter over a wide range of densities. In hard systems, binary collisions are well defined and the ANN approach enables an efficient algorithm to determine the time to next collision with high accuracy. The method has been used to enable an MD study of an ensemble of 1800 hard, smooth, impenetrable equilateral triangles in a two-dimensional periodic space. At high packing fraction (0.6<rho<0.9), the hard-triangle system exists as a liquid-crystalline-like phase (LCP) in which there is no long-range translational order but in which there is nearly perfect long-range orientational order. As the packing fraction decreases, the LCP undergoes a transition to a fluid state in which the long-range orientational correlation vanishes but short-range order is retained. Long-lived clusters, notably hexamers, are clearly apparent in the liquid phase and appear to be stabilized by a sort of internal "orientational" osmotic pressure. Insofar as can be inferred from our machine calculations, the transition between the LCP and the liquid occurs around rhosimilar to0.57 and appears to be second order. At low density, the hard-triangle system undergoes "chattering" collisions in which pairs of triangles collide and become associated, undergoing multiple collisions with each other before colliding with a third particle. The radial distribution function obtained from both molecular dynamics and Monte Carlo calculations shows a weak peak at low packing fraction.

Discrete modeling of processes at the atomic-scale affords practical approaches to complex materials of interest commercially and to the US Air Force. Reductions in computation times can be large, suggesting the possibility of real-time modeling of thin film growth and the consequent development of processing routes to achieve specific physical and chemical properties. Formulation of the model to be used is critical in achieving such computational gains. Frameworks for these models such as Monte Carlo and molecular dynamics can be used conceptually, but they cannot be applied in practice because of the high number of required computations per time step. The simplest discrete model involves the Potts model to simulate energies, then to create a partition function of probabilities for various states and configurations, followed by a decision algorithm that determines the state of surface atoms. Although the inclusion of defects, dopants, atom complexes, surface reconstruction and crystal orientations can be included directly in this modeling approach, the resulting collection of behaviors is very entangled with logical and mathematical functions. Hence, the time to exercise the model increases noticeably. It is shown that this problem can be reduced dramatically by employing neuro-computing methods.