Search NASA⌕ Search

SEARCH · Search NASA

Results for “Discrete Element Method”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 487 records · Page 27
All sourcesNASA NTRSDOE OSTI
DOE OSTI2019

Preconditioning a Newton-Krylov solver for all-speed melt pool flow physics

n this paper, we introduce a multigrid block-based preconditioner for solving linear systems arising from a Discontinuous Galerkin discretization of the all-speed Navier-Stokes equations with phase change. The equations are discretized in conservative form with a reconstructed Discontinuous Galerkin (rDG) method and integrated with fully-implicit time discretization schemes. To robustly converge the numerically stiff systems, we use the Newton-Krylov framework with a primitive-variable formulation (pressure, velocity, and temperature), which is better conditioned than the conservative-variable form at low-Mach number. In the limit of large acoustic CFL number and viscous Fourier number, there is a strong coupling between the velocity-pressure system and the linear systems become non-diagonally dominant. To effectively solve these ill-conditioned systems, an approximate block factorization preconditioner is developed, which uses the Schur complement to reduce a 3 x 3 block system into a sequence of two 2 x 2 block systems: velocity-pressure,v P , and velocity-temperature, v T . We compare the performance of the v P -v T Schur complement preconditioner to classic preconditioning strategies: monolithic algebraic multigrid (AMG), element-block SOR, and primitive variable block Gauss-Seidel. The performance of the preconditioned solver is investigated in the limit of large CFL and Fourier numbers for low-Mach lid-driven cavity flow, Rayleigh-Bénard melt convection, compressible internally heated convection, and 3D laser-induced melt pool flow. Here, numerical results demonstrate that the v P -v T Schur complement preconditioned solver scales well both algorithmically and in parallel, and is robust for highly ill-conditioned systems, for all tested rDG discretization schemes (up to 4th-order).

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Exploring constituent redistribution in irradiated U-19Pu-14Zr fuel via electron probe microanalysis

Here, the phenomena of constituent redistribution, wherein a previously homogeneous metallic fuel forms discrete, radially concentric compositional zones upon irradiation was investigated by examining an irradiated U-19Pu-14Zr fuel (where numbers represent wt. %) with a burnup of 11.5 at.% with electron probe microanalysis (EPMA) and quadruple inductively coupled plasma mass spectroscopy (Q-ICP-MS). EPMA-generated U, Pu, and Zr compositional data obtained from a diameter traverse of the sample was converted to mass and was used to: 1) compare the overall fuel element analysis results between the two methods, 2) determine the number of compositionally distinct zones forming as a result of constituent redistribution; and 3) quantify the post-irradiation loss or gain of U, Pu, and Zr atoms in each distinct compositional zone. Weight percent concentrations of U, Pu, and Zr for the overall cross section compare favorably between the two analytical methods, suggesting that the spatially resolved EPMA analysis complements bulk chemical analysis. Among the four identified compositional zones, post-irradiation quantification of U, Pu, and Zr elemental atom content changes shows that the quantity of U atoms lost from the innermost zone is slightly less than the quantity of U atoms gained by the middle two zones, and the quantity of Zr atoms lost from the high-U third zone is slightly less than is gained by the two innermost zones. Pu is lost from all four zones, although the innermost zone and the high-U third zone lose a significantly higher percentage (> 22 %) of their initial Pu atoms than the other two zones. For all three elements, EPMA cannot distinguish between atoms lost due to transport to a different zone from atoms lost due to nuclear processes; however, the insight gained from using this process can be used to experiment with new modeling techniques to predict constituent redistribution in U-Pu-Zr fuels.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗
DOE OSTI2022

A mixed, unified forward/inverse framework for earthquake problems: fault implementation and coseismic slip estimate

SUMMARY We introduce a new finite-element (FE) based computational framework to solve forward and inverse elastic deformation problems for earthquake faulting via the adjoint method. Based on two advanced computational libraries, FEniCS and hIPPYlib for the forward and inverse problems, respectively, this framework is flexible, transparent and easily extensible. We represent a fault discontinuity through a mixed FE elasticity formulation, which approximates the stress with higher order accuracy and exposes the prescribed slip explicitly in the variational form without using conventional split node and decomposition discrete approaches. This also allows the first order optimality condition, that is the vanishing of the gradient, to be expressed in continuous form, which leads to consistent discretizations of all field variables, including the slip. We show comparisons with the standard, pure displacement formulation and a model containing an in-plane mode II crack, whose slip is prescribed via the split node technique. We demonstrate the potential of this new computational framework by performing a linear coseismic slip inversion through adjoint-based optimization methods, without requiring computation of elastic Green’s functions. Specifically, we consider a penalized least squares formulation, which in a Bayesian setting—under the assumption of Gaussian noise and prior—reflects the negative log of the posterior distribution. The comparison of the inversion results with a standard, linear inverse theory approach based on Okada’s solutions shows analogous results. Preliminary uncertainties are estimated via eigenvalue analysis of the Hessian of the penalized least squares objective function. Our implementation is fully open-source and Jupyter notebooks to reproduce our results are provided. The extension to a fully Bayesian framework for detailed uncertainty quantification and non-linear inversions, including for heterogeneous media earthquake problems, will be analysed in a forthcoming paper.

58 GEOSCIENCES↗
DOE OSTI2024

Leveraging a Neural Network-Enhanced Reproducing Kernel Particle Method for Multiphysics Degradation Modeling of Energy Storage Materials

Energy storage materials exhibit strong electro-chemo-mechanical coupling and highly anisotropic material properties, contributing to the formation and propagation of micro-cracking during charge/discharge cycling and resulting in reduced performance and service life. A coupled electro-chemo-mechanical reproducing kernel particle method (RKPM) formulation has been developed to analyze this system. With microstructural images supplied by the National Renewable Energy Laboratory (NREL), pixel-based model construction by RKPM is used to represent the complex material microstructures that dictate the coupled physics of these systems. Traditional electro-chemo-mechanical models rely on mesh-based finite element methods, which can lead to difficulties in meshing such complex geometries and capturing crack propagation due to mesh dependency. Here, a neural network-enhanced reproducing kernel particle method (NN-RKPM) [1, 2] is introduced to effectively model damage and crack propagation in the material microstructures; the location, orientation, and solution transition near a localization are automatically captured by superimposed block-level NN optimizations. This NN enrichment approach allows for effective modeling of localizations via a fixed background discretization, relieving tedious efforts for adaptive refinement in traditional mesh-based methods. Applications to the heterogeneous microstructures of Li-ion battery cathodes will be presented to demonstrate the effectiveness of the proposed methods. NN-RKPM is additionally used to inform how crack opening and closure in turn affect the coupled chemical equations and material microstructure. Reference: [1] Baek, J., Chen, J. S., Susuki, K., "Neural Network enhanced Reproducing Kernel Particle Method for Modeling Localizations," International Journal for Numerical Methods in Engineering, Vol. 123, pp 4422-4454, https://doi.org/10.1002/nme.7040, 2022. [2] Baek, J., Chen, J. S., "A Neural Network-Based Enrichment of Reproducing Kernel Approximation for Modeling Brittle Fracture", Computer Methods in Applied Mechanics and Engineering Vol. 410, 116590, 2024.

degradation↗
DOE OSTI2023

Continuous multimodal technologies in industrial microbiology: potential for achieving high process performance and agility

The field of industrial microbiology has great potential to tackle the most pressing challenges facing society, but today, this entails much more than classical fermentation technology followed by product upgrading. Methods are needed to confer unit-operation-level functionality and whole-system-level robustness, as well as flexibility. As examples of this concept, we focus here on methodologies often deployed at industrial scale as discrete processes—such as biological, chemical, and electrochemical unit operations—that are also capable of substantial complementarity when combined sequentially and operated continuously. A key and relatively new type of conversion step is made possible by catalyst immobilization methods at both the enzymatic and cellular level. Cases drawn from electro-fermentation for acetoin production and syngas-mediated microbial electrosynthesis of ethanol exemplify the synergy of combined operational approaches. Because modular processes can be treated as a series of reactor elements, considerable flexibility is possible. As discussed, however, not all processes are amenable to this intensification.

09 BIOMASS FUELS↗
DOE OSTI2024

Progressive Hedging Decomposition for Solutions of Large-Scale Process Family Design Problems

Rapid, wide-scale deployment of green process systems, such as carbon capture or water desalination systems, is essential for combatting climate change. Methods relying on traditional design or modularity fail to capture the benefits of both economies of numbers and economies of scale. We have proposed process family design, which designs a family of processes simultaneously exploiting opportunities for common elements. In previous work, we explored different optimization formulations to solve this problem. In this work, we develop a decomposition approach to tackle larger problems efficiently. We solve a water desalination case study, which is too large to solve within a reasonable timeframe with the discretization formulation. We exploit the block angular structure of the discretization problem to decompose and solve using Progressive Hedging (PH). We use the open-source Python package mpi-sppy to execute PH which allows us to leverage parallelization and a HPC cluster to further improve solution time.

Stinchfield, Georgia↗
DOE OSTI2018

A Weighted Least-Squares Transport Equation Compatible with Source Iteration and Voids

We present that second-order forms of the transport equation allow the use of continuous finite elements (CFEMs). This can be desired in multiphysics calculations where other physics require CFEM discretizations. Second-order transport operators are generally self-adjoint, yielding symmetric positive-definite (SPD) matrices, which allow the use of efficient linear algebra solvers with an enormous advantage in memory usage. Least-squares (LS) forms of the transport equation can circumvent the void problems of other second-order forms but are almost always nonconservative. Additionally, the standard LS form is not compatible with discrete ordinates method (SN) iterative solution techniques such as source iteration. A new form of the LS transport equation has recently been developed that is compatible with voids and standard SN iterative solution techniques. Performing nonlinear diffusion acceleration (NDA) using an independently differenced low-order equation enforces conservation for the whole system and makes this equation suitable for reactor physics calculations. In this context, “independent” means that both the transport and low-order solutions converge to the same scalar flux and current as the spatial mesh is refined, but for a given mesh, the solutions are not necessarily equal. In this paper we show that introducing a weight function into this LS equation improves issues with causality and can render our equation equal to the self-adjoint angular flux (SAAF) equation. Causality is a principle of the transport equation that states that information travels only downstream along characteristics. This principle can be violated numerically. We show how to limit the weight function in voids and demonstrate the effect of this limit on accuracy. Using the C5G7 benchmark, we compare our method to the SAAF formulation with a void treatment (SAAFτ) that is not self-adjoint and has a nonsymmetric coefficient matrix. Finallyl, we show that the weighted LS equation with NDA gives acceptable accuracy relative to the SAAFτ equation while maintaining a SPD system matrix.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Multilevel well modeling in aggregation-based nonlinear multigrid for multiphase flow in porous media

A full approximation scheme (FAS) nonlinear multigrid solver for two-phase flow and transport problems driven by wells with multiple perforations is developed here. It is an extension to our previous work on FAS solvers for diffusion and transport problems. The solver is applicable to discrete problems defined on unstructured grids as the coarsening algorithm is aggregation-based and algebraic. To construct coarse basis that can better capture the radial flow near wells, coarse grids in which perforated well cells are not near the coarse-element interface are desired. This is achieved by an aggregation algorithm proposed in this paper that makes use of the location of well cells in the cell-connectivity graph. Numerical examples in which the FAS solver is compared against Newton's method on benchmark problems are given. In particular, for a refined version of the SAIGUP model, the FAS solver is at least 35% faster than Newton's method for time steps with a CFL number greater than 10.

58 GEOSCIENCES↗
DOE OSTI2022

MECHANISTIC MULTIPHYSICS MODELING OF CLADDING RUPTURE IN NUCLEAR FUEL RODS DURING LOSS-OF-COOLANT ACCIDENT CONDITIONS

The Loss of Coolant Accident (LOCA) is a design basis accident that is included as part of the safety analysis of nuclear power plants. As the nuclear industry desires to increase the cycle length and discharge burnup of existing nuclear power plants they must demonstrate safe operation during a LOCA on high burnup fuel. During a LOCA transient on high burnup fuel rods, the rods may undergo a process known as fuel fragmentation, relocation, and dispersal (FFRD). To permit dispersal, the cladding encapsulating the fuel must rupture with an opening size large enough to allow the fragmented fuel particles to release. Current licensing tools used by industry and the United States Nuclear Regulatory Commission are limited in geometric fidelity and materials that can be analyzed. These simulation tools generally employ a quasi-two-dimensional (1.5D or Layered1D) or 2D-RZ axisymmetric geometric representations exclusively. While a valid approach under some instances, there are times when important physics have an asymmetric behavior in the fuel rod. Examples include fuel fragmentation, thermal-hydraulic boundary conditions, and cladding rupture, all of which are important for LOCA analysis. As industry pursues burnup extensions it must be demonstrated that fuel dispersal can be mitigated or eliminated. To do this, an understanding of the rupture opening after cladding failure is required. This work presents the development of a model for predicting the size and location of the rupture opening in failed fuel rods during LOCA conditions using advanced modeling and simulation tools. In order to supply the rupture model with appropriate boundary conditions, improvements to fuel fragmentation, axial relocation and oxidation modeling were required. First, the eXtended Finite Element Method (XFEM) is used to mechanistically predict the number of fuel fragments that form due to material strength randomization, criteria for strength randomization, mesh density, power ramping rates and irradiation effects. These predictions with associated uncertainty were compared to empirical correlations developed for UO2 verifying that they can be used with increased confidence in subsequent axial relocation analyses. Secondly, a new first-of-its-kind Layered2D computational framework was developed that provides the ability to apply azimuthally varying boundary conditions while providing discrete layers to track fuel movement during the LOCA. An existing fuel axial relocation model developed for Layered1D was extended to work within the Layered2D framework. A large sensitivity study was performed on the initial version of the model to identify modeling parameters of particular importance, with the emissivity used for radiation after blowdown being the primary source of uncertainty. Then, a simplistic approach to incorporate mechanical degradation of the cladding due to oxidation was also developed to investigate the impact of reduced cladding thickness on predictions of the time to failure of cladding tubes. It was found that the cladding will typically rupture prior to a reduction in thickness that is sufficient to impact the rupture behavior. A model was then developed for predicting cladding rupture that transfers the cladding surface temperatures, rod internal and external pressures, fast neutron fluence, and fast neutron flux from a more detailed Layered1D, Layered2D, or 2D-RZ analysis to a 3D cladding only analysis. Comparisons of the rupture model to a few experiments indicate reasonable predictions. The rupture model was then applied to two accident tolerant fuel concepts (FeCrAl and Cr-coated Zircaloy) where it predicted that both ATF concepts would have smaller rupture openings and delayed rupture times than the standard Zircaloy-4 cladding material under identical loading conditions.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗
DOE OSTI2019

A communication-avoiding 3D sparse triangular solver

We present a novel distributed memory algorithm to improve the strong scalability of the solution of a sparse triangular system. This operation appears in the solve phase of direct methods for solving general sparse linear systems, Ax = b. Our 3D sparse triangular solver employs several techniques, including a 3D MPI process grid, elimination tree parallelism, and data replication, all of which reduce the per-process communication when combined. We present analytical models to understand the communication cost of our algorithm and show that our 3D sparse triangular solver can reduce the per-process communication volume asymptotically by a factor of O(n1/4) and O(n1/6) for problems arising from the finite element discretizations of 2D "planar" and 3D "non-planar" PDEs, respectively. We implement our algorithm for use in SuperLU_DIST3D, using a hybrid MPI+OpenMP programming model. Our 3D triangular solve algorithm, when run on 12k cores of Cray XC30, outperforms the current state-of-the-art 2D algorithm by 7.2x for planar and 2.7x for the non-planar sparse matrices, respectively.

Sao, Piyush↗
DOE OSTI2018

A non-uniform corrosion model and meso-scale fracture modelling of concrete

Corrosion-induced concrete cracking is a significant durability problem for reinforced concrete structures. Considerable research has been carried out in the last few decades to understand and model the expansion mechanism of the corrosion products around the reinforcing bar and simulate the cracking behaviour of the concrete cover. In this paper, a new corrosion model based on non-uniform corrosion expansion is formulated and validated against experimental data. A meso-scale fracture model, consisting of aggregates, cement paste/mortar and ITZ, is established for the cases of both middle and side reinforcing bars. Under the developed corrosion and concrete fracture model, the cracking phenomena of the concrete cover are accurately simulated. It has been found that the non-uniform corrosion model can be used to express the realistic corrosion rust progression around the reinforcing bar, with the best accuracy. It has also been found that some microcracks occur before they are connected to form the dominating discrete crack which usually appears on the concrete surface. Moreover, the effects of the corrosion variables, as well as other key material and geometric parameters, on surface cracking of concrete are investigated.

36 MATERIALS SCIENCE↗
DOE OSTI2020

Finite Volume Discretization of the Euler Equations in Pronghorn

Modeling flow and heat transfer in high temperature gas reactors (HTGR) requires the ability to model a wide range of flow speeds from slow (natural convection), to intermediate (forced-flow conditions), to supersonic regimes (depressurization) for a wide range of geometries including the pebble bed, upper and lower plenum, and risers. In previous work, Pronghorn has effectively modeled low-to-medium speed flows in scenarios such as the one described in the two-dimensional PBMR-400 benchmark, using its finite-element-based streamline-upwind Petrov-Galerkin (SUPG) stabilized implementation of the Euler equations. However, limitations of this method become apparent when dealing with more complicated geometries (e.g. imposing slip boundary conditions at nodes belonging to two different boundaries) and when gas speeds are fast enough for shocks and supersonic flow to occur. For these problems, the finite-element-based solver lacks robustness and is plagued by slow iterative convergence or even divergence. In order to address these challenges, the Pronghorn code at INL has been updated with new, modified versions of its original equations. The new Pronghorn models are built on the finite volume method with a Harten-Lax-van Leer-Contact (HLLC) Riemann solver based numerical flux method, which (1) allows imposing slip boundary conditions much more robustly and (2) performs well for a wide range of flow speeds. The finite-volume-based flow solver will form the basis for a robust coarse-mesh thermal-hydraulics capability in Pronghorn.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2020

Remotely powered line monitor

Improved devices, apparatuses, systems, and methods involving the use of field probes with voltage-controlled variable impedance elements and surface-launched acoustic wave devices, including surface acoustic wave (SAW) devices and SAW sensor-tag wireless interface devices to measure voltage and current in current carrying conductors (CCCs) via measurement of the electric fields and magnetic fields around said CCCs are disclosed. The devices, apparatuses, systems, and methods taught in the present invention include a remotely powered line-mounted measurement unit that requires no connection to electrical ground for operation, and that has no internal energy source, no energy harvesting or power conditioning circuitry, no discrete energy storage components, and no radio. The devices, apparatuses, systems, and methods taught in the present invention also provide for determining the phase (leading or lagging) of the current relative to the voltage in AC power systems, and also provide for measurement of temperature of the conductor.

24 POWER TRANSMISSION AND DISTRIBUTION↗
DOE OSTI2023

A Refinement-by-Superposition -Method for (curl)- and (div)-Conforming Discretizations

Here, we present refinement-by-superposition (RBS) hp-refinement infrastructure for computational electromagnetics (CEMs), which permits exponential rates of convergence. In contrast to dominant approaches to hp-refinement for continuous Galerkin methods, which rely on explicit constraint equations, the multilevel strategy presented drastically reduces the implementation complexity. Through the RBS methodology, enforcement of continuity occurs by construction, enabling arbitrary levels of refinement with ease, and without the practical (but not theoretical) limitations of constrained-node refinement. We outline the construction of the RBS hp-method for refinement with H (curl)- and H (div)-conforming finite cells. Numerical simulations for the 2-D finite element method (FEM) solution of the Maxwell eigenvalue problem demonstrate the effectiveness of RBS hp-refinement. As an additional goal of this work, we aim to promote the use of mixed-order (low- and high-order) elements in practical CEM applications.

42 ENGINEERING↗
DOE OSTI2018

Model Order Reduction of Nonviscously Damped Structural Dynamic Models

A reduced order modeling capability has been developed to reduce the computational burden associated with time-domain solutions of structural dynamic models with linear viscoelastic materials. The discretized equations-of-motion produce convolution integrals resulting in a linear system with nonviscous damping forces. The challenge associated with the reduction of nonviscously damped, linear systems is the selection and computation of the appropriate modal basis to perform modal projection. The system produces a nonlinear eigenvalue problem that is challenging to solve and requires use of specialized algorithms not readily available in commercial finite element packages. This SAND report summarizes the LDRD discoveries of a reduction scheme developed for monolithic finite element models and provides preliminary investigations to extensions of the method using component mode synthesis. In addition, this report provides a background overview of structural dynamic modeling of structures with linear viscoelastic materials, and provides an overview of a new code capability in Sierra Structural Dynamics to output the system level matrices computed on multiple processors.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2018

A new method for analyzing and visualizing plasma simulations using a phase-space tessellation

We apply a novel phase-space interpolation technique referred to as the simplex-in-cell (SIC) method to analyze two- and three-dimensional particle-in-cell (PIC) simulations of electromagnetic plasmas. SIC relies on a discretization of the initial phase-space distribution function into simplices, which allows an approximation to the full, continuously defined distribution function to be constructed at any later time in the simulation. This allows densities, currents, and even full momentum distribution functions to be measured at any point in the simulation domain without averaging over control volumes. The SIC approach applies to any PIC simulation for which a tessellation of the initial particle distribution can be constructed. In this study, we use outputs from standard PIC simulations of the Weibel instability and compare physical quantities such as charge and current densities calculated in postprocessing using SIC and standard particle deposits. Using 2D simulations with 1–65 536 particles-per-cell, we find that SIC eliminates discrete particle noise and in some cases can reach a given noise level using ~1000 times fewer simulation particles than with standard particle deposition schemes. In regions of low density, such as between current filaments, SIC is able to capture small amplitude features even with fewer particles than gridpoints due to the deformable nature of the SIC volume elements. Here, by calculating momentum distributions, we show how SIC can capture low density tails in the spectrum using far fewer particles than are necessary for standard particle deposits. We calculate the charge density on spatial grids of increasing resolution to demonstrate the ability of SIC to reveal fine-scale details that are not accessible with standard particle deposits. Finally, we show how SIC can be extended to 3D and give an example of its use to calculate the charge density from 3D PIC simulations of the Weibel instability. These results motivate the future implementation of SIC directly in the simulation force calculation for a novel low-noise electromagnetic plasma simulation method.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2021

A review of rare earth elements and yttrium in coal ash: Content, modes of occurrences, combustion behavior, and extraction methods

Rare earth elements and yttrium (REY) have attracted considerable attention over the last decade because of their vital roles in clean energy, consumer product, national defense and security applications, among other uses. Due to the retention of REY during coal burning, coal combustion ash is considered as potential alternative sources for REY. Understanding the content, speciation, retention and/or transformation behavior of REY during coal combustion not only expands our knowledge of the combustion behavior of the trace elements in coal, but also provides basis for modeling REY partitioning during coal combustion and for developing economically viable REY recovery technologies. This review makes a critical summary of recent progress in the study of REY in coal ash. The contents and the extraction potentials of REY in coal ash derived from 15 major coal-producing countries worldwide were summarized and evaluated. Various analytical methods for determining REY bulk contents and speciation, together with the solid sample pretreatment, analytical accuracy and precision, advantages and disadvantages were summarized and compared. Modern analytical approaches combined indirect methods (e.g., sequential extraction) shed light on the physical distribution, mineralogy, and the chemical state of REY in coal ash. Three types of REY occurrences in coal ash, including Si-Al glassy association, discrete minerals or compounds, and organic association (bound with unburned carbon) were defined in the review. The glassy association can be further divided into REY minerals closely bound to glass phases and dispersed throughout the glassy structure. REY partitioning in various emission streams, the size distribution, and their enrichment behavior in coal ash were discussed. Additionally, thermal behavior and transformation of various REY forms in coal during combustion process, including organic-associated REY, REY phosphates, REY carbonates, clay-bound REY and among others were summarized. Two possible retention mechanisms of REY by aluminosilicate glass at boiler temperature were proposed: the incorporation of the individual REY phases into the glass as inclusions and the diffusion of REY phases throughout Si-Al glass structures in the melting process. Feed coal mineral types, mineral-mineral associations, boiler conditions, and other factors control the retention process. After coal combustion, the speciation of REY in fly ash may be modified by the reactions of REY phases with flue gas components. Further, an overview of REY transformation mechanisms during coal combustion was deeply discussed. Finally, current extraction techniques for REY recovery from coal combustion ash were introduced. Future outlooks and research problems were also identified.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2022

Learning in continuous action space for developing high dimensional potential energy models

Reinforcement learning (RL) approaches that combine a tree search with deep learning have found remarkable success in searching exorbitantly large, albeit discrete action spaces, as in chess, Shogi and Go. Many real-world materials discovery and design applications, however, involve multi-dimensional search problems and learning domains that have continuous action spaces. Exploring high-dimensional potential energy models of materials is an example. Traditionally, these searches are time consuming (often several years for a single bulk system) and driven by human intuition and/or expertise and more recently by global/local optimization searches that have issues with convergence and/or do not scale well with the search dimensionality. Here, in a departure from discrete action and other gradient-based approaches, we introduce a RL strategy based on decision trees that incorporates modified rewards for improved exploration, efficient sampling during playouts and a “window scaling scheme" for enhanced exploitation, to enable efficient and scalable search for continuous action space problems. Using high-dimensional artificial landscapes and control RL problems, we successfully benchmark our approach against popular global optimization schemes and state of the art policy gradient methods, respectively. We demonstrate its efficacy to parameterize potential models (physics based and high-dimensional neural networks) for 54 different elemental systems across the periodic table as well as alloys. We analyze error trends across different elements in the latent space and trace their origin to elemental structural diversity and the smoothness of the element energy surface. Broadly, our RL strategy will be applicable to many other physical science problems involving search over continuous action spaces.

36 MATERIALS SCIENCE↗