Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Numerical techniques”

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 1,387 records · Page 77
All sourcesNASA NTRSDOE OSTI
DOE OSTI2024

Algorithms for Non-Negative Matrix Factorization on Noisy Data With Negative Values

Non-negative matrix factorization (NMF) is a dimensionality reduction technique that has shown promise for analyzing noisy data, especially astronomical data. For these datasets, the observed data may contain negative values due to noise even when the true underlying physical signal is strictly positive. Prior NMF work has not treated negative data in a statistically consistent manner, which becomes problematic for low signal-to-noise data with many negative values. In this paper we present two algorithms, Shift-NMF and Nearly-NMF, that can handle both the noisiness of the input data and also any introduced negativity. Both of these algorithms use the negative data space without clipping or masking and recover non-negative signals without any introduced positive offset that occurs when clipping or masking negative data. We demonstrate this numerically on both simple and more realistic examples, and prove that both algorithms have monotonically decreasing update rules.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2022

Enriched immersed boundary method (EIBM) for interface-coupled multi-physics and applications to convective conjugate heat transfer

It is challenging to develop numerical methods that simultaneously maintain accuracy of resolving boundary conditions and mesh flexibility to handle the interface in interface-coupled multi-physics problems involving large property discontinuity. On the one hand, boundary-fitted methods possess high accuracy in capturing the interfacial phenomenon but involve complicated volumetric mesh generation, mesh-motion, and even re-meshing procedures. On the other hand, immersed boundary methods (IBM) provide mesh flexibility but sometimes suffer from inferior interface representations, leading to poor enforcement of boundary conditions. This paper presents an enriched immersed boundary method (EIBM) to overcome this challenge and demonstrates its efficacy in conjugate heat transfer, a representative example in interface-coupled multi-physics systems that have implications for many industrial processes. Further, the core technique of the method is to enhance IBM’s accuracy of the fluid–solid interface by enriching the degrees of freedom of the cut elements to enforce temperature and flux compatibilities and resolve all the physical unknowns on the background mesh to simplify the volumetric mesh generation. We implement the EIBM under the framework of a variational multiscale formulation for coupled Navier–Stokes and thermodynamics equations. The enriched DoFs enable better enforcement of temperature and flux compatibilities with large conductivity ratios across the fluid–solid interface. At the same time, the immersed nature of the proposed method still attains mesh flexibility. The EIBM’s accuracy is thoroughly evaluated through a set of examples, ranging from benchmark problems with analytical solutions to real-world cooling processes of a moving metallic structure with complex geometry.

42 ENGINEERING↗
DOE OSTI2021

Dynamic Rupture Modeling of the M7.1, 2019 Ridgecrest, California, Earthquake

An important part of the simulation technique implemented in the strong motion prediction recipe developed in Japan, originally proposed by Irikura and Miyake (2011) is the earthquake rupture characterization that is based on established empirical scaling relationships between source parameters and seismic moment. These scaling relationships and rupture characterization are derived from recorded data and are tested by simulations. Recent analysis of results obtained with the Recipe for the 2016 M7.1 Kumamoto earthquake performed by several researchers, indicate that in order to reproduce near-fault long-period ground motions with permanent displacements, typical for earthquakes with surface rupturing, the recipe needs modifications. In particular, improvements are required in characterizing the rupture kinematics at shallow depths, above the seismogenic zone (Irikura et al., 2019). Modeling of rupture dynamics for recorded earthquakes, and numerical simulations for different rupture scenarios, provides physical constrains on slip evolution in time and space, which can be used in adjustments of empirical source scaling relationships for large crustal earthquakes. In this study we mainly focused on rupture dynamics modeling of the 2019 M7 Ridgecrest, California earthquake. The characterization of the kinematic rupture model derived from the dynamic rupture models of this earthquake can inform further refinements of rupture modeling proposed by the Recipe. Through different rupture scenarios we investigated potential relationships between characteristics of Long Period Ground Motion Generation Area (LMGA), used in the Recipe, and asperity depth and rupture initiation.

58 GEOSCIENCES↗
DOE OSTI2021

RAVEN Theory Manual

RAVEN is a software framework able to perform parametric and stochastic analysis based on the response of complex system codes. The initial development was aimed at providing dynamic risk analysis capabilities to the thermohydraulic code RELAP-7, currently under development at Idaho National Laboratory (INL). Although the initial goal has been fully accomplished, RAVEN is now a multi-purpose stochastic and uncertainty quantification platform, capable of communicating with any system code. In fact, the provided Application Programming Interfaces (APIs) allow RAVEN to interact with any code as long as all the parameters that need to be perturbed are accessible by input files or via python interfaces. RAVEN is capable of investigating system response and explore input space using various sampling schemes such as Monte Carlo, grid, or Latin hypercube. However, RAVEN strength lies in its system feature discovery capabilities such as: constructing limit surfaces, separating regions of the input space leading to system failure, and using dynamic supervised learning techniques. The development of RAVEN started in 2012 when, within the Nuclear Energy Advanced Modeling and Simulation (NEAMS) program, the need to provide a modern risk evaluation framework arose. RAVEN’s principal assignment is to provide the necessary software and algorithms in order to employ the concepts developed by the Risk Informed Safety Margin Characterization (RISMC) program. RISMC is one of the pathways defined within the Light Water Reactor Sustainability (LWRS) program. In the RISMC approach, the goal is not just to identify the frequency of an event potentially leading to a system failure, but the proximity (or lack thereof) to key safety-related events. Hence, the approach is interested in identifying and increasing the safety margins related to those events. A safety margin is a numerical value quantifying the probability that a safety metric (e.g. peak pressure in a pipe) is exceeded under certain conditions. Most of the capabilities, implemented having RELAP-7 as a principal focus, are easily deployable to other system codes. For this reason, several side activates have been employed (e.g. RELAP5-3D, any MOOSE-based App, etc.) or are currently ongoing for coupling RAVEN with several different software. The aim of this document is to provide a set of commented examples that can help the user to become familiar with the RAVEN code usage.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

Reduced order models for thermal radiative transfer problems based on moment equations and data-driven approximations of the Eddington tensor

Here a new group of structure and asymptotic preserving reduced-order models (ROMs) for multidimensional nonlinear thermal radiative transfer (TRT) problems is presented. They are formulated by means of the nonlinear projective approach and data compression techniques. The nonlinear projection is applied to the Boltzmann transport equation (BTE) to derive a hierarchy of low-order moment equations. Approximation of the Eddington tensor that provides exact closure for the system of moment equations is found with projection-based data-driven methodologies. These include the (i) proper orthogonal decomposition (POD), (ii) dynamic mode decomposition (DMD) and (iii) a variant of the DMD. A parameterization is derived for this ROM for the temperature of radiation incoming to the problem domain (the radiation drive temperature). This parameterization is informed from results of a dimensionless study of the TRT problem. Analysis of the ROMs is performed on the classical Fleck-Cummings TRT multigroup test problem in 2D geometry with a radiation-driven Marshak wave. Numerical results are presented to demonstrate the performance of these ROMs for the simulation of evolving radiation and heat waves. Results show these models to be sufficiently accurate for practical computations with rather low-rank representations of the Eddington tensor. As the rank of the approximation is increased, the errors of solutions generated by the ROMs gradually decreases.

42 ENGINEERING↗
DOE OSTI2020

An investigation of Newton-Sketch and subsampled Newton methods

Sketching, a dimensionality reduction technique, has received much attention in the statistics community. In this paper, we study sketching in the context of Newton's method for solving finite-sum optimization problems in which the number of variables and data points are both large. In this work, we study two forms of sketching that perform dimensionality reduction in data space: Hessian subsampling and randomized Hadamard transformations. Each has its own advantages, and their relative tradeoffs have not been investigated in the optimization literature. Additionally, our study focuses on practical versions of the two methods in which the resulting linear systems of equations are solved approximately, at every iteration, using an iterative solver. The advantages of using the conjugate gradient method vs. a stochastic gradient iteration are revealed through a set of numerical experiments, and a complexity analysis of the Hessian subsampling method is presented.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2021

RAVEN User Manual

RAVEN is a generic software framework to perform parametric and probabilistic analysis based on the response of complex system codes. The initial development was aimed to provide dynamic risk analysis capabilities to the Thermo-Hydraulic code RELAP-7, currently under development at the Idaho National Laboratory (INL). Although the initial goal has been fully accomplished, RAVEN is now a multi-purpose probabilistic and uncertainty quantification platform, capable to agnostically communicate with any system code. This agnosticism includes providing Application Programming Interfaces (APIs). These APIs are used to allow RAVEN to interact with any code as long as all the parameters that need to be perturbed are accessible by inputs files or via python interfaces. RAVEN is capable of investigating the system response, and investigating the input space using Monte Carlo, Grid, or Latin Hyper Cube sampling schemes, but its strength is focused to- ward system feature discovery, such as limit surfaces, separating regions of the input space leading to system failure, using dynamic supervised learning techniques. The development of RAVEN has started in 2012, when, within the Nuclear Energy Advanced Modeling and Simulation (NEAMS) program, the need to provide a modern risk evaluation framework became stronger. RAVEN principal assignment is to provide the necessary software and algorithms in order to employ the concept developed by the Risk Informed Safety Margin Characterization (RISMC) program. RISMC is one of the pathways defined within the Light Water Reactor Sustainability (LWRS) program. In the RISMC approach, the goal is not just the individuation of the frequency of an event potentially leading to a system failure, but the closeness (or not) to key safety-related events. Hence, the approach is interested in identifying and increasing the safety margins related to those events. A safety margin is a numerical value quantifying the probability that a safety metric (e.g. for an important process such as peak pressure in a pipe) is exceeded under certain conditions. The initial development of RAVEN has been focused on providing dynamic risk assessment capability to RELAP-7, currently under development at the INL and, likely, future replacement of the RELAP5-3D code. Most the capabilities that have been implemented having RELAP-7 as principal focus are easily deployable for other system codes. For this reason, several side activates are currently ongoing for coupling RAVEN with soft- ware such as RELAP5-3D, etc. The aim of this document is the explanation of the input requirements, focalizing on the input structure.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2026

Embedding Fiber Optic Sensors in Stainless Steel using Spark Plasma Sintering for Structural Health Monitoring in Harsh Environments

Embedded fiber optic sensors such as fiber Bragg gratings (FBGs) offer a unique route for distributed real-time in-situ imaging of various engineering parameters for numerous purposes. This study advanced the current sensor embedding approaches by exploring a spark plasma sintering (SPS)-assisted technology to embed FBGs in high-temperature structural materials and demonstrated the capability of temperature measurement. In this approach, single-mode FBGs were integrated into stainless steel (SS) 316L components using SPS, followed by the evaluation of the bonding quality between the FBGs and matrix, the optical attenuation of the fibers induced by embedding, and the sensing characters of the FBGs under temperature stimuli. The results demonstrated that superior bonding was achieved between the FBGs and highly-densified SS316L. Examination of the behavior of Bragg gratings validated signal fidelity after embedding. Real-time thermal imaging under temperature cycling using the FBGs demonstrated the effectiveness of the technique for smart materials manufacturing.

36 - MATERIALS SCIENCE↗
DOE OSTI2020

The Adjoint Petrov–Galerkin method for non-linear model reduction

Here, we formulate a new projection-based reduced-order modeling technique for non-linear dynamical systems. The proposed technique, which we refer to as the Adjoint Petrov–Galerkin (APG) method, is derived by decomposing the generalized coordinates of a dynamical system into a resolved coarse-scale set and an unresolved fine-scale set. A Markovian finite memory assumption within the Mori–Zwanzig formalism is then used to develop a reduced-order representation of the coarse scales. This procedure leads to a closed reduced-order model that displays commonalities with the adjoint stabilization method used in finite elements. The formulation is shown to be equivalent to a Petrov–Galerkin method with a non-linear, time-varying test basis, thus sharing some similarities with the Least-Squares Petrov–Galerkin method. Theoretical analysis examining a priori error bounds and computational cost is presented. Numerical experiments on the compressible Navier–Stokes equations demonstrate that the proposed method can lead to improvements in numerical accuracy, robustness, and computational efficiency over the Galerkin method on problems of practical interest. Improvements in numerical accuracy and computational efficiency over the Least-Squares Petrov–Galerkin method are observed in most cases.

42 ENGINEERING↗
DOE OSTI2020

Interactions Between the Amazonian Rainforest and Cumuli Clouds: A Large‐Eddy Simulation, High‐Resolution ECMWF, and Observational Intercomparison Study

Abstract The explicit coupling at meter and second scales of vegetation's responses to the atmospheric‐boundary layer dynamics drives a dynamic heterogeneity that influences canopy‐top fluxes and cloud formation. Focusing on a representative day during the Amazonian dry season, we investigate the diurnal cycle of energy, moisture and carbon dioxide at the canopy top, and the transition from clear to cloudy conditions. To this end, we compare results from a large‐eddy simulation technique, a high‐resolution global weather model, and a complete observational data set collected during the GoAmazon14/15 campaign. The overall model‐observation comparisons of radiation and canopy‐top fluxes, turbulence, and cloud dynamics are very satisfactory, with all the modeled variables lying within the standard deviation of the monthly aggregated observations. Our analysis indicates that the timing of the change in the daylight carbon exchange, from a sink to a source, remains uncertain and is probably related to the stomata closure caused by the increase in vapor pressure deficit during the afternoon. We demonstrate quantitatively that heat and moisture transport from the subcloud layer into the cloud layer are misrepresented by the global model, yielding low values of specific humidity and thermal instability above the cloud base. Finally, the numerical simulations and observational data are adequate settings for benchmarking more comprehensive studies of plant responses, microphysics, and radiation.

Vilà‐Guerau de Arellano, J.↗
DOE OSTI2021

Cherenkov and Transition Radiation as Low-Energy Background Sources in SuperCDMS Detectors

The Weakly Interacting Massive Particle (WIMP) has historically been a prime candidate for dark matter due to its elegant compatibility with the Minimally Supersymmetric Standard Model (MSSM). Recent dark matter experiments have ruled out much of the GeV-TeV mass range predicted by the MSSM, however, and the newest generation of direct detection experiments, such as SuperCDMS SNOLAB, have begunto explore sub-GeV dark matter. As experiments push towards lower masses, sensitivity to eV-scale electron recoil events has become increasingly important. A variety of unexplained excesses at energy deposits of 1-100 eV have been found inmany such low-mass experiments, across different detection techniques and at different excess rates. These low-energy events are not thought to be dark matter, and they must be understood and effectively removed to further improve detector energy resolution. This thesis will outline the simulation of one possible source of low-energy excess at SuperCDMS: optical photons produced from charges in uniform motion, specifically from Cherenkov radiation (CR), transition radiation (TR), and the intermediate hybrid transition-Cherenkov radiation (HR). In the latter case, the standard formulae for HR contain divergences at certain angles in transparent media, which are an artifact of their derivation and become problematic to simulate. To rsolve this, we present a novel approach to normalize the divergent HR peaks, which allows HR to transition smoothly between TR and CR in a numerical simulation. We also add TR and CR as a physics process to the SuperCDMS Monte Carlo simulation package SuperSim, based on the Geant4 simulation toolkit. To verify the physics of our addition, we show some test simulations in comparison to theoretical predictions of optical radiation intensity. We also use the simulation to make some preliminary predictions on the contribution of TR and CR to the low energy background, with the expectation that more thorough analyses will be conducted in the future.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2023

PETSc/TAO Users Manual (Rev. 3.19)

This manual describes the use of the Portable, Extensible Toolkit for Scientific Computation (PETSc) and the Toolkit for Advanced Optimization (TAO) for the numerical solution of partial differential equations and related problems on high-performance computers. PETSc/TAO is a suite of data structures and routines that provide the building blocks for the implementation of large-scale application codes on parallel (and serial) computers. PETSc uses the MPI standard for all distributed memory communication. PETSc/TAO includes a large suite of parallel linear solvers, nonlinear solvers, time integrators, and opti mization that may be used in application codes written in Fortran, C, C++, and Python (via petsc4py; see Getting Started). PETSc provides many of the mechanisms needed within parallel application codes, such as parallel matrix and vector assembly routines. The library is organized hierarchically, enabling users to employ the level of abstraction that is most appropriate for a particular problem. By using techniques of object-oriented programming, PETSc provides enormous flexibility for users. PETSc is a sophisticated set of software tools; as such, for some users it initially has a much steeper learning curve than packages such as MATLAB or a simple subroutine library. In particular, for individuals without some computer science background, experience programming in C, C++, python, or Fortran and experience using a debugger such as gdb or lldb, it may require a significant amount of time to take full advantage of the features that enable efficient software use. However, the power of the PETSc design and the algorithms it incorporates may make the efficient implementation of many application codes simpler than “rolling them” yourself. For many tasks a package such as MATLAB is often the best tool; PETSc is not intended for the classes of problems for which effective MATLAB code can be written. There are several packages, built on PETSc, that may satisfy your needs without requiring directly using PETSc. We recommend reviewing these packages functionality before starting to code directly with PETSc. PETSc can be used to provide a “MPI parallel linear solver” in an otherwise sequential, or OpenMP parallel code. This approach cannot provide extremely large improvements in the application time by utilizing large numbers of MPI processes but can still improve the performance. Certainly all parts of a previously sequential code need not be parallelized but the matrix generation portion must be parallelized to expect true scalability to large numbers of MPI processes. See PCMPI for details on how to utilize the PETSc MPI linear solver server. Since PETSc is under continued development, small changes in usage and calling sequences of routines will occur. PETSc has been supported for twenty-five years; see mailing list information on our website for information on contacting support.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

GEN3D Experimental and Numerical Development of GEN3 Durability Models

Understanding of the high temperature durability of particles and the materials that contain them is critical to next generation of concentrating solar power (CSP) technology. Here we studied the durability of particles and their containment materials under extreme UV cycling, thermal cycling, and in low-speed high temperature mechanical wear situations. The optical stability of seven candidate particles has been determined following exposure to the high temperature conditions present in a generation 3 particle-based CSP technology. Particle solar weighted absorptance and emittance measured periodically during 10,000 high solar irradiance exposure cycles and up to 400 hours of isothermal aging has been documented. The particle aging due to repeated exposure to concentrated solar flux represented the 30-year lifetime of a power plant. Models were fit to the absorptivity and emissivity data following the isothermal aging provides the projected optical degradation of the particles as a function of temperature. Mechanical wear was studied through the use of custom developed wear testing facilities for measuring high temperature impact wear, abrasion wear, and particle attrition. Additionally, a novel technique for measuring the high temperature mechanical properties of single particles was developed. Through these tests it was observed that high nickel alloys generally showed lower wear than comparable iron based steels, particularly at elevated temperatures of 800°C. Mechanical wear at these temperatures is a highly complex phenomenon combining both mechanical wear and oxidation. Additionally, the containment materials wear rates are influenced by the particles (both hardness and roundness), making the wear mechanisms complex. The initial wear test conducted in the abrasion test rig revealed a substantial amount of oxide materials in the particle bed after testing (in relative to later tests), and substantially more wear, likely indicating a need for concern in startup operation of particle facilities to not incur high wear from the presence of oxides. Particle attrition experiments have only been conducted for a single material but increase size distribution, reduction in circularity, and particle diameter is observed. Efforts to develop predictive models was limited due to a testing campaign that prioritized testing materials for particle pathway developers over building a comprehensive design of experiments. The results discussed in this report inform future CSP developers and researchers further de risking the technology and assisting in its future development. ParticleBased CSP development provides a path to dispatchable solar power generation with storage at a price competitive in the current energy market. Lowering the cost of CSP technology provides a carbon free power generation solution that can assist in the transition from fossil fuels to renewable sources of electricity.

14 SOLAR ENERGY↗
DOE OSTI2020

PETSc Users Manual (Rev. 3.13)

This manual describes the use of PETSc for the numerical solution of partial differential equations and related problems on high-performance computers. The Portable, Extensible Toolkit for Scientific Computation (PETSc) is a suite of data structures and routines that provide the building blocks for the implementation of large-scale application codes on parallel (and serial) computers. PETSc uses the MPI standard for all message-passing communication. PETSc includes an expanding suite of parallel linear solvers, nonlinear solvers, and time integrators that may be used in application codes written in Fortran, C, C++, and Python. PETSc provides many of the mechanisms needed within parallel application codes, such as parallel matrix and vector assembly routines. The library is organized hierarchically, enabling users to employ the level of abstraction that is most appropriate for a particular problem. By using techniques of object-oriented programming, PETSc provides enormous flexibility for users. PETSc is a sophisticated set of software tools; as such, for some users it initially has a much steeper learning curve than a simple subroutine library. In particular, for individuals without some computer science background, experience programming in C, C++, python, or Fortran and experience using a debugger such as gdb or dbx, it may require a significant amount of time to take full advantage of the features that enable efficient software use. However, the power of the PETSc design and the algorithms it incorporates may make the efficient implementation of many application codes simpler than \rolling them" yourself.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Pore-scale visualization of natural hydrate-bearing sediments

Accurate modeling of gas hydrate reservoir productivity and geomechanical risks associated with subsurface dissociation of natural gas hydrates (NGH) requires the determination of model parameters through physical testing on natural hydrate-bearing sediments (HBS). This involves investigating the hydro-mechanical behavior of undisturbed hydrate samples from nature under in situ conditions using pressure core characterization and analysis, which provides a unique opportunity for research. By employing state-of-the-art micro computed tomography imagery on cryogenically preserved, hydrate-bearing sediment samples, we can determine hydrate saturation as well as permeability with and without the presence of hydrates in the sediment. Furthermore, utilizing a machine learning based image segmentation technique, it is possible to extract pore space and grain information. Subsections of the entire image volume were used to determine anisotropic permeabilities using a finite-difference method Stokes solver (FDMSS). Additionally, permeability measurements on whole pressure and temperature preserved hydrate-bearing core were analyzed by utilizing the National Energy Technology Laboratory’s (NETL) Pressure Core Characterization and X-ray CT Visualization Tool (PCXT) to manipulate, cut, and analyze pressure preserved sediment. Permeabilities were measured under a broad range of vertical stress states to simulate expected pressure changes during production scenarios, and the results show that permeabilities derived from images are in agreement with those from traditional core derived experiments. The collected stress-dependent permeability, permeability anisotropy, and corresponding gas hydrate saturations provide valuable input into numerical simulations of reservoir productivity. These properties have been proven to be key parameters determining a long-term reservoir response under depressurization.

Liu, Mengwei [Oak Ridge Institute for Science and ↗
DOE OSTI2020

Residual Stress Modeling with Phase Transformation for Wire Arc Additive Manufacturing of B91 Steel

Wire arc additive manufacturing (WAAM) is an energy-efficient manufacturing technique used for near-net-shape production of functional industrial components. However, heat accumulation during deposition and the associated mechanical and metallurgical changes result in complex residual stress profiles across the cross section of the fabricated components. These residual stresses are detrimental to the service life of the components. In this study, sequentially coupled thermomechanical analysis of WAAM B91 steel is conducted to quantify the residual stress variation across the component. The thermomechanical analysis includes a transient heat transfer model and a static stress model that incorporates the transformation-induced plasticity due to martensitic phase transformation. Here, the experimentally calibrated heat transfer model mirrors the temperature variation of the system during the deposition. The results from the stress model are validated via x-ray diffraction measurements, and the numerical results are in good agreement with the experimental data.

20 FOSSIL-FUELED POWER PLANTS↗
DOE OSTI2026

Negative fluxes and cell-miss errors in the random ray method

The random ray method is a recently developed stochastic method for solving neutral particle transport problems based on the method of characteristics. Perhaps surprisingly for a characteristics-based method using flat sources, we note that the random ray method can produce negative fluxes which may be numerically troublesome in several situations. These occur most severely in fixed source problems where the source is in a region with a small cross section. Additionally, we briefly discuss another source of bias which can occur in similar situations, namely a ray missing a mesh with a strong source and small cross section, resulting in the entirety of the source being unphysically deposited locally. This paper describes the mechanism by which negative fluxes may occur and several different methods to mitigate their effects. These fixes are tested on an eigenvalue problem, a ‘fusion-like’ shielding problem, and a shielding problem featuring an adjoint calculation. Even when extremely coarse random ray quadratures are used such that 20%–30% of cells are missed during a given iteration, use of the preferred fix technique ensures local flux tally errors remain trivial (below 1%). The preferred fix is now the default option in SCONE and OpenMC.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2023

Improving deep learning model performance under parametric constraints for materials informatics applications

Abstract Modern machine learning (ML) and deep learning (DL) techniques using high-dimensional data representations have helped accelerate the materials discovery process by efficiently detecting hidden patterns in existing datasets and linking input representations to output properties for a better understanding of the scientific phenomenon. While a deep neural network comprised of fully connected layers has been widely used for materials property prediction, simply creating a deeper model with a large number of layers often faces with vanishing gradient problem, causing a degradation in the performance, thereby limiting usage. In this paper, we study and propose architectural principles to address the question of improving the performance of model training and inference under fixed parametric constraints. Here, we present a general deep-learning framework based on branched residual learning (BRNet) with fully connected layers that can work with any numerical vector-based representation as input to build accurate models to predict materials properties. We perform model training for materials properties using numerical vectors representing different composition-based attributes of the respective materials and compare the performance of the proposed models against traditional ML and existing DL architectures. We find that the proposed models are significantly more accurate than the ML/DL models for all data sizes by using different composition-based attributes as input. Further, branched learning requires fewer parameters and results in faster model training due to better convergence during the training phase than existing neural networks, thereby efficiently building accurate models for predicting materials properties.

36 MATERIALS SCIENCE↗