Integrating Materials and Manufacturing Innovation最新文献

Additive manufacturing (AM) technologies offer unprecedented design flexibility but are limited by a lack of understanding of the material microstructure formed under their extreme and transient processing conditions and its subsequent transformation during post-build processing. As part of the 2022 AM Bench Challenge, sponsored by the National Institute of Standards and Technology, this study focuses on the phase composition and phase evolution of AM nickel alloy 718, a nickel-based superalloy, to provide benchmark data essential for the validation of computational models for microstructural predictions. We employed high-energy synchrotron X-ray diffraction, in situ synchrotron X-ray scattering, as well as high-resolution transmission electron microscopy for our analyses. The study uncovers critical aspects of the microstructure in its as-built state, its transformation during homogenization, and its phase evolution during subsequent aging heat treatment. Specifically, we identified secondary phases, monitored the dissolution and coarsening of microstructural elements, and observed the formation and stability of γ’ and γ” phases. The results provide the rigorous benchmark data required to understand the atomic and microstructural transformations of AM nickel alloy 718, thereby enhancing the reliability and applicability of AM models for predicting phase evolution and mechanical properties.

The 2022 Asynchronous AM-Bench challenge was designed to test the ability of simulations to accurately predict laser power absorption as well as various melt pool behaviors (width, depth, and solidification) during laser melting of solid metal during stationary and scanned laser illumination. In this challenge, participants were asked to predict a series of experimental outcomes. Experimental data were obtained from a series of experiments performed at the Advanced Photon Source at Argonne National Laboratories in 2019. These experiments combined integrating sphere radiometry with high-speed X-ray imaging, allowing for the simultaneous recording of absolute laser power absorption and two-dimensional, projected images of the melt pool. All challenge problems were based on experiments using bare aluminum solid metal. Participants were provided with pertinent experimental information like laser power, scan speed, laser spot size, and material composition. Additionally, participants were given absorptance and X-ray imaging data from stationary and scanned laser experiments on solid Ti–6Al–4V that could be used for testing their models before attempting challenge problems. In total, this challenge received 56 submissions from eight different research groups for eight individual challenge problems. The data for this challenge, and associated information, are available for download from the NIST Public Data Repository. This paper summarizes the results from the 2022 Asynchronous AM-Bench challenge as well as discusses the lessons learned to help inform future challenges.

To understand the correlation between process, structures, and properties in laser powder bed fusion (L-PBF), it is essential to use numerical analysis as well as experimental approaches. A finite element thermal analysis uses a moving heat source model represented as a volumetric heat flux to simulate heat input by laser. Because of its computational efficiency, finite element thermal analysis is suitable for iterative procedures such as parametric study and process optimization. However, to obtain valid simulated results, the heat source model must be calibrated by comparison with experimental results for each laser scanning condition. The need for re-calibration limits the applicable window of laser scanning conditions in the thermal analysis. Thus, the current study developed a novel heat source model that is valid and precise under any laser scanning condition within a wide process window. As a secondary objective in the development, we quantitatively evaluated and compared the four heat source models proposed to date. It was found that the most suitable heat source model for the L-PBF is conical one among them. Then, a multiple linear regression analysis was performed to represent the heat source model as a function of laser power and scanning velocity. Consequently, the thermal analysis with the novel model is valid and precise within the wide process window of L-PBF.

Abstract

High-strength aluminum alloys used in aerospace and automotive applications obtain their strength through precipitation hardening. Achieving the desired mechanical properties requires precise control over the nanometer-sized precipitates. However, the microstructure of these alloys changes over time due to aging, leading to a deterioration in strength. Typically, the size, number, and distribution of precipitates for a quantitative assessment of microstructural changes are determined by manual analysis, which is subjective and time-consuming. In our work, we introduce a progressive and automatable approach that enables a more efficient, objective, and reproducible analysis of precipitates. The method involves several sequential steps using an image repository containing dark-field transmission electron microscopy (DF-TEM) images depicting various aging states of an aluminum alloy. During the process, precipitation contours are generated and quantitatively evaluated, and the results are comprehensibly transferred into semantic data structures. The use and deployment of Jupyter Notebooks, along with the beneficial implementation of Semantic Web technologies, significantly enhances the reproducibility and comparability of the findings. This work serves as an exemplar of FAIR image and research data management.

Abstract

The problem of generating microstructures of complex materials in silico has been approached from various directions including simulation, Markov, deep learning and descriptor-based approaches. This work presents a hybrid method that is inspired by all four categories and has interesting scalability properties. A neural cellular automaton is trained to evolve microstructures based on local information. Unlike most machine learning-based approaches, it does not directly require a data set of reference micrographs, but is trained from statistical microstructure descriptors that can stem from a single reference. This means that the training cost scales only with the complexity of the structure and associated descriptors. Since the size of the reconstructed structures can be set during inference, even extremely large structures can be efficiently generated. Similarly, the method is very efficient if many structures are to be reconstructed from the same descriptor for statistical evaluations. The method is formulated and discussed in detail by means of various numerical experiments, demonstrating its utility and scalability.

Abstract

Indentation testing has played a major role for many materials design processes as a convenient and relatively cheap experiment. However, extracting the data from indentation tests requires complex post-processing or an integrated simulation and experiment framework. Accordingly, the simulation of indentation has become a post-processing routine for indentation tests. Providing a highly efficient, computationally scalable, and open-source platform for indentation simulation provides invaluable machinery for materials design process. An open-source PRISMS-Indentation module is presented here as a multi-scale elasto-plastic virtual indentation framework. The module is implemented as a part of PRISMS-Plasticity software which covers length scales of macroscopic plasticity and crystal plasticity. The contact problem is handled using a primal–dual active set method. The framework is first tested against analytical solution of Hertzian theory for contact using an isotropic elasticity model. The robustness of the framework is then investigated in simulations of indentation of annealed Cu microstructures. Unstructured meshes with hexahedral elements and variable mesh density are used to demonstrate potential for speedup in indentation simulations.

X-ray diffraction patterns contain information about the atomistic structure and microstructure (defect population) of materials, extracting detailed information from diffraction patterns is complex, demanding and relies on prior knowledge. We hypothesize that deep-learning techniques can help to perform an effective and accurate analysis with high throughput rates. To demonstrate this concept, we applied a novel deep learning framework to determine the evolution of the (upbeta )-phase volume fraction in a Ti–6Al–4V alloy during heat-treatment from video sequences of 2D diffraction patterns recorded in transmission and with highly monochromatic radiation in a synchrotron beamline. In particular, we studied the impact of network design on prediction reliability and computational performance. Networks of different architectures were trained using 3008 experimental 2D patterns. A well-tuned model was found to reproduce the phase fractions of another experimental data set, consisting of 1100 diffraction patterns, with a mean-square error as small as (2.6 times 10^{-4}). The average prediction error of (upbeta )-phase volume fraction was within (1.6 times 10^{-2}) (in each diffraction pattern) of the values obtained by conventional methods. Our work demonstrates that convolutional neural networks can evaluate high energy X-ray diffraction patterns with a remarkable level of reliability. Furthermore, it demonstrates the significance of network design on the reliability of predictions and computational performance. The most complex models do not necessarily result in highest accuracy and may even fail to learn from the data.

The additive manufacturing benchmarking challenge described in this work was aimed at the prediction of average stress–strain properties for tensile specimens that were excised from blocks of non-heat-treated IN625 manufactured by laser powder bed fusion. Two different laser scan strategies were considered: an X-only raster and an XY raster, which involved a 90(^circ ) rotation in the scan direction between subsequent layers. To measure anisotropy, multiple tensile orientations with respect to the build direction were investigated (e.g., parallel, perpendicular, and intervals in between). Benchmark participants were provided grain structure information via electron backscatter diffraction measurements, as well as the stress–strain response for tensile specimens manufactured parallel to the build direction and produced by the XY scan strategy. Then, participants were asked to predict tensile properties, like the ultimate tensile strength, for the remaining specimens and orientations. Interestingly, the measured mechanical properties did not vary linearly as a function of tensile orientation. Moreover, specimens manufactured with the XY scan strategy exhibited greater yield strength than those corresponding to the X-only scan strategy, regardless of orientation. The benchmark data have been made publicly available for anyone that is interested [1]. For the modeling aspect of the challenge, five teams participated in this benchmark. While most of the models incorporated a crystal plasticity framework, one team chose to use a more semiempirical approach and to great success. However, no team excelled at all the predictions, and all teams were seemingly challenged with the predictions associated with the X-only scan strategy.

Abstract

Advancements in high-throughput data generation and physics-informed artificial intelligence and machine-learning algorithms are rapidly challenging the status quo for how materials data is collected, analyzed, and communicated with the world. Machine-learning algorithms can be executed in just a few lines of code by researchers with minimal data science expertise. This perspective addresses the reality that the ecosystems which have been constructed to nurture new materials discovery and development are not yet well equipped to take advantage of the radically more powerful and accessible computational and algorithmic tools which have the immediate potential to enhance the pace of scientific advancement in this field. A novel architecture for managing materials data is proposed and discussed from the standpoint of how historical and emerging subfields of materials science could have been or might still significantly improve the impact of materials discoveries to the many human societal needs for new materials.

Austenitic 347H stainless steel offers superior mechanical properties and corrosion resistance required for extreme operating conditions such as high temperature. The change in microstructure due to composition and process variations is expected to impact material properties. Identifying microstructural features such as grain boundaries thus becomes an important task in the process-microstructure-properties loop. Applying convolutional neural network (CNN)-based deep learning models is a powerful technique to detect features from material micrographs in an automated manner. In contrast to microstructural classification, supervised CNN models for segmentation tasks require pixel-wise annotation labels. However, manual labeling of the images for the segmentation task poses a major bottleneck for generating training data and labels in a reliable and reproducible way within a reasonable timeframe. Microstructural characterization especially needs to be expedited for faster material discovery by changing alloy compositions. In this study, we attempt to overcome such limitations by utilizing multimodal microscopy to generate labels directly instead of manual labeling. We combine scanning electron microscopy images of 347H stainless steel as training data and electron backscatter diffraction micrographs as pixel-wise labels for grain boundary detection as a semantic segmentation task. The viability of our method is evaluated by considering a set of deep CNN architectures. We demonstrate that despite producing instrumentation drift during data collection between two modes of microscopy, this method performs comparably to similar segmentation tasks that used manual labeling. Additionally, we find that naïve pixel-wise segmentation results in small gaps and missing boundaries in the predicted grain boundary map. By incorporating topological information during model training, the connectivity of the grain boundary network and segmentation performance is improved. Finally, our approach is validated by accurate computation on downstream tasks of predicting the underlying grain morphology distributions which are the ultimate quantities of interest for microstructural characterization.