Materials science in the artificial intelligence age: high-throughput library generation, machine learning, and a pathway from correlations to the underpinning physics

Vasudevan, Rama K.; Choudhary, Kamal; Mehta, Apurva; Smith, Ryan P.; Kusne, Gilad; Tavazza, Francesca; Vlček, Lukáš; Ziatdinov, Maxim; Kalinin, Sergei V.; Hattrick‐Simpers, Jason

doi:10.1557/mrc.2019.95

Cited by 149 publications

(109 citation statements)

References 174 publications

(191 reference statements)

Supporting

Mentioning

109

Contrasting

Order By: Relevance

“…The presented modeling chain will be used together with high-throughput screening of feasible alloy compositions, quick-and-dirty based on CALPHAD-thermodynamic databases, and completely new compositions can be explored with density functional theory (DFT)-based approaches. This can be complemented with high-throughput experimentation, including rapid combinatorial materials synthesis and characterization schemes [38,45].…”

Section: Discussionmentioning

confidence: 99%

Process-Structure-Properties-Performance Modeling for Selective Laser Melting

Pinomaa

Yashchuk

Lindroos

et al. 2019

Metals

View full text Add to dashboard Cite

Selective laser melting (SLM) is a promising manufacturing technique where the part design, from performance and properties process control and alloying, can be accelerated with integrated computational materials engineering (ICME). This paper demonstrates a process-structure-properties-performance modeling framework for SLM. For powder-bed scale melt pool modeling, we present a diffuse-interface multiphase computational fluid dynamics model which couples Navier-Stokes, Cahn-Hilliard, and heat-transfer equations. A computationally efficient large-scale heat-transfer model is used to describe the temperature evolution in larger volumes. Phase field modeling is used to demonstrate how epitaxial growth of Ti-6-4 can be interrupted with inoculants to obtain an equiaxed polycrystalline structure. These structures are enriched with a synthetic lath martensite substructure, and their micromechanical response are investigated with a crystal plasticity model. The fatigue performance of these structures are analyzed, with spherical porelike defects and high-aspect-ratio cracklike defects incorporated, and a cycle-amplitude fatigue graph is produced to quantify the fatigue behavior of the structures. The simulated fatigue life presents trends consistent with the literature in terms of high cycle and low cycle fatigue, and the role of defects in dominating the respective performance of the produced SLM structures. The proposed ICME workflow emphasizes the possibilities arising from the vast design space exploitable with respect to manufacturing systems, powders, respective alloy chemistries, and microstructures. By digitalizing the whole workflow and enabling a thorough and detailed virtual evaluation of the causal relationships, the promise of product-targeted materials and solutions for metal additive manufacturing becomes closer to practical engineering application.to be modified to maintain similar properties achieved with traditional techniques such as casting or wrought parts. Compared to traditional manufacturing methods, due to local powder fusion in rapid solidification conditions and solid-state heat cycling, the resulting microstructures present unique features such as metastable phases, smaller scale microsegregation, formation of unexpected secondary phases, spherical or high-aspect-ratio pores, oxide inclusions, microstructural residual stresses, melt pool boundaries, grains with complex morphology, and strong texture.The aforementioned structural features in SLM lead to properties and performance that are unexpected [1]. For example, stainless steel 316L can achieve remarkably high hardness and ductility [2] but has a susceptibility to pitting corrosion [3]; Inconel 625 nickel superalloy contains detrimental intermetallic and carbide phases [4]; Ti-6-4 tends to produce hard needlelike α martensite which leads to high hardness but low ductility; tool steels are extremely difficult to fabricate with SLM due to their complex metallurgy, including a tendency to form brittle martensite and high volume fractions of ...

show abstract

Section: Discussionmentioning

confidence: 99%

Process-Structure-Properties-Performance Modeling for Selective Laser Melting

Pinomaa

Yashchuk

Lindroos

et al. 2019

Metals

View full text Add to dashboard Cite

show abstract

“…There is now, however, a small, but steadily growing, subset of SPM groups who are adopting machine 1 Inverse imaging nonetheless still only provides limited information on the geometric structure of the apex and is often difficult to interpret [22,23]. 2 As Vasudevan et al [29] point out, however, excitement about AI-driven data analysis and experimental design has ebbed and flowed for many decades, with periods of intense interest followed by 'AI winters' . [31]; (B)Using a somewhat similar methodology to that employed for (A), Aldritt and co-workers [34] have trained a convolutional neural net to recognise specific molecular geometries in ultrahigh resolution AFM images; (C) Ziatdinov et al determined the orientation/rotation of single molecules on a solid surface via an artificial neural network.…”

Section: More Human Than Human: Beyond the Single Molecule Limitmentioning

confidence: 99%

“…' Shortly before Zhang et al's work was published in July this year, a related approach to the automated extraction of 'buried' information from AFM, rather than STM, data was described in what is very likely to be a highly influential paper, by Benjamin Aldritt and his colleagues (and recently accepted in the journal Science Advances) [34]. Building on previous machine learning protocols developed by Sergei Kalinin and co-workers at Oak Ridge National Laboratories (among others) [29,36,37], Aldritt et al have developed what they describe as automated structure discovery AFM (ASD-AFM), a deep learning framework based on a similar methodology to that of Zhang et al, whereby a large set of simulated images is generated-in this case via a combination of density functional theory (DFT) optimisation of molecular structures and the probe-particle model of tip-sample interactions developed by Hapala et al [38,39] -and a convolutional neural net is used to determine the best match between experimental data and a molecular geometry. Figure 2 illustrates the general CNN methodology adopted by Aldritt et al, alongside Ziatdinov, Maksov, and Kalinin's earlier work on determining the rotational state of adsorbed molecules from STM data [35].…”

Section: More Human Than Human: Beyond the Single Molecule Limitmentioning

confidence: 99%

Machine learning at the (sub)atomic scale: next generation scanning probe microscopy

Gordon

Moriarty

2020

Mach. Learn.: Sci. Technol.

View full text Add to dashboard Cite

We discuss the exciting prospects for a step change in our ability to map and modify matter at the atomic/molecular level by embedding machine learning algorithms in scanning probe microscopy (with a particular focus on scanning tunnelling microscopy, STM). This nano-AI hybrid approach has the far-reaching potential to realise a technology capable of the automated analysis, actuation, and assembly of matter with a precision down to the single chemical bond limit.

show abstract

“…Data-driven methods for materials development have become increasingly prevalent over the past decade. [1][2][3][4][5] One widespread machine learning approach for materials development is screening. [2,[6][7][8] In materials screening, a machine learning model is trained to predict materials properties given the chemical formula and processing information and then is applied to a set of candidate materials to predict their properties.…”

Section: Introductionmentioning

confidence: 99%