Search NASA⌕ Search

SEARCH · Search NASA

Results for “computer performance”

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 1,765 records · Page 98
All sourcesNASA NTRSDOE OSTI
NASA NTRS2003

Aviation Safety Modeling and Simulation (ASMM) Propulsion Fleet Modeling: A Tool for Semi-Automatic Construction of CORBA-based Applications from Legacy Fortran Programs

Within NASA's Aviation Safety Program, NASA GRC participates in the Modeling and Simulation Project called ASMM. NASA GRC s focus is to characterize the propulsion systems performance from a fleet management and maintenance perspective by modeling and through simulation predict the characteristics of two classes of commercial engines (CFM56 and GE90). In prior years, the High Performance Computing and Communication (HPCC) program funded, NASA Glenn in developing a large scale, detailed simulations for the analysis and design of aircraft engines called the Numerical Propulsion System Simulation (NPSS). Three major aspects of this modeling included the integration of different engine components, coupling of multiple disciplines, and engine component zooming at appropriate level fidelity, require relatively tight coupling of different analysis codes. Most of these codes in aerodynamics and solid mechanics are written in Fortran. Refitting these legacy Fortran codes with distributed objects can increase these codes reusability. Aviation Safety s modeling and simulation use in characterizing fleet management has similar needs. The modeling and simulation of these propulsion systems use existing Fortran and C codes that are instrumental in determining the performance of the fleet. The research centers on building a CORBA-based development environment for programmers to easily wrap and couple legacy Fortran codes. This environment consists of a C++ wrapper library to hide the details of CORBA and an efficient remote variable scheme to facilitate data exchange between the client and the server model. Additionally, a Web Service model should also be constructed for evaluation of this technology s use over the next two- three years.

Sang, Janche↗
DOE OSTI2022

Weighted relaxation for multigrid reduction in time

Current trends in computer architectures now mean that faster computation speed must come primarily from increased concurrency, not faster clock speeds, which are stagnating. Thus, this situation creates bottlenecks for serial algorithms, including the well-known bottleneck for sequential time-integration, where each individual time-value (i.e., time-step) is computed sequentially. One approach to alleviate this and achieve parallelism in time is with multigrid. Here, in this work, we consider multigrid-reduction-in-time (MGRIT), a multilevel method applied to the time dimension that computes multiple time-steps in parallel. Like all multigrid methods, MGRIT relies on the complementary relationship between relaxation on a fine-grid and a correction from the coarse grid to solve the problem. All current MGRIT implementations are based on unweighted-Jacobi relaxation; here we introduce the concept of weighted relaxation to MGRIT. We derive new convergence bounds for weighted relaxation, and use this analysis to guide the selection of relaxation weights. Numerical results then demonstrate that by choosing appropriate non-unitary relaxation weights, one can achieve faster convergence rates and lower iteration counts for MGRIT when compared with unweighted relaxation. In most cases, weighted relaxation yields a 10%–20% saving in iterations, which is significant when using large high-performance computers. For A-stable integration schemes, results also illustrate that under-relaxation can restore convergence in some cases where unweighted relaxation is not convergent.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

OmniFed: A Modular Framework for Configurable Federated Learning from Edge to HPC

Federated Learning (FL) is critical for edge and High Performance Computing (HPC) where data is not centralized and privacy is crucial. We present OmniFed, a modular framework designed around decoupling and clear separation of concerns for configuration, orchestration, communication, and training logic. Its architecture supports configuration-driven prototyping and code-level override-what-you-need customization. We also support different topologies, mixed communication protocols within a single deployment, and popular training algorithms. It also offers optional privacy mechanisms including Differential Privacy (DP), Homomorphic Encryption (HE), and Secure Aggregation (SA), as well as compression strategies. These capabilities are exposed through well-defined extension points, allowing users to customize topology and orchestration, learning logic, and privacy/compression plugins, all while preserving the integrity of the core system. We evaluate multiple models and algorithms to measure various performance metrics. By unifying topology configuration, mixed-protocol communication, and pluggable modules in one stack, OmniFed streamlines FL deployment across heterogeneous environments. Github repository is available at https://github.com/at-aaims/OmniFed.

Tyagi, Sahil [ORNL] (ORCID:0009000783144745)↗
DOE OSTI2024

Distributed-Memory Sparse Deep Neural Network Inference Using Global Arrays

Partitioned Global Address Space (PGAS) models exhibit tremendous promise in developing efficient and productive distributed-memory parallel applications. They have been used extensively in scientific computations due to conveniently offering a ``shared-memory''-like model and convenient interfaces that separate communication with synchronization. Traditionally, PGAS communication models have been applied to dense/contiguously distributed data, but most modern applications depict varied levels of sparsity. Existing PGAS models require certain adaptations to support distributed sparse computations, since associated computations often require matrix arithmetic, in addition to data movement. The Global Arrays toolkit from Pacific Northwest National Laboratory (PNNL) is one of the earliest PGAS models to combine one-sided data communication and distributed matrix operations and is still used in the popular NWChem quantum chemistry suite. Recently, we have expanded the Global Arrays toolkit to support common sparse operations, like sparse matrix-dense matrix multiplies (SpMM), sparse matrix-sparse matrix multiplication (SpGEMM) and Sampled Dense-Dense Matrix Multiplication (SDDMM). As it turns out, these operations are the bedrock of sparse Deep Learning (DL); sparse deep neural networks and Graph Neural Networks (GNNs) have gained increasing attention recently in achieving speedups on training and inference with reduced memory footprints. Unlike scientific applications in High Performance Computing (HPC), modern (distributed-memory capable) DL toolkits often rely on non-standardized and closed-source vendor software optimizations, creating challenges in software-hardware co-design at scale. Our goal is to support a variety of distributed-memory sparse matrix operations and helper functions in the newly created Sparse Global Arrays (SGA), such that it is possible to build portable and productive Machine Learning scenarios for algorithm/software and hardware codesign purposes. Contemporary data-parallel schemes for training/inference are undergoing a major overhaul since model replication limits scalability and causes resource inefficiencies. As such, we have adopted tensor parallelism in decomposing the model and inputs, to mitigate memory issues. Current implementation is built on top of MPI and uses CPUs to maximize the portability across the platforms.

Distributed computing, machine learning↗
DOE OSTI2024

Addressing GPU memory limitations for Graph Neural Networks in High-Energy Physics applications

Introduction Reconstructing low-level particle tracks in neutrino physics can address some of the most fundamental questions about the universe. However, processing petabytes of raw data using deep learning techniques poses a challenging problem in the field of High Energy Physics (HEP). In the Exa.TrkX Project, an illustrative HEP application, preprocessed simulation data is fed into a state-of-art Graph Neural Network (GNN) model, accelerated by GPUs. However, limited GPU memory often leads to Out-of-Memory (OOM) exceptions during training, due to the large size of models and datasets. This problem is exacerbated when deploying models on High-Performance Computing (HPC) systems designed for large-scale applications. Methods We observe a high workload imbalance issue during GNN model training caused by the irregular sizes of input graph samples in HEP datasets, contributing to OOM exceptions. We aim to scale GNNs on HPC systems, by prioritizing workload balance in graph inputs while maintaining model accuracy. Our paper introduces diverse balancing strategies aimed at decreasing the maximum GPU memory footprint and avoiding the OOM exception, across various datasets. Results Our experiments showcase memory reduction of up to 32.14% compared to the baseline. We also demonstrate the proposed strategies can avoid OOM in application. Additionally, we create a distributed multi-GPU implementation using these samplers to demonstrate the scalability of these techniques on the HEP dataset. Discussion By assessing the performance of these strategies as data loading samplers across multiple datasets, we can gauge their effectiveness in both single-GPU and distributed environments. Our experiments, conducted on datasets of varying sizes and across multiple GPUs, broaden the applicability of our work to various GNN applications that handle input datasets with irregular graph sizes.

Lee, Claire Songhyun↗
NASA NTRS1991

Knowledge-based decision support for Space Station assembly sequence planning

A complete Personal Analysis Assistant (PAA) for Space Station Freedom (SSF) assembly sequence planning consists of three software components: the system infrastructure, intra-flight value added, and inter-flight value added. The system infrastructure is the substrate on which software elements providing inter-flight and intra-flight value-added functionality are built. It provides the capability for building representations of assembly sequence plans and specification of constraints and analysis options. Intra-flight value-added provides functionality that will, given the manifest for each flight, define cargo elements, place them in the National Space Transportation System (NSTS) cargo bay, compute performance measure values, and identify violated constraints. Inter-flight value-added provides functionality that will, given major milestone dates and capability requirements, determine the number and dates of required flights and develop a manifest for each flight. The current project is Phase 1 of a projected two phase program and delivers the system infrastructure. Intra- and inter-flight value-added were to be developed in Phase 2, which has not been funded. Based on experience derived from hundreds of projects conducted over the past seven years, ISX developed an Intelligent Systems Engineering (ISE) methodology that combines the methods of systems engineering and knowledge engineering to meet the special systems development requirements posed by intelligent systems, systems that blend artificial intelligence and other advanced technologies with more conventional computing technologies. The ISE methodology defines a phased program process that begins with an application assessment designed to provide a preliminary determination of the relative technical risks and payoffs associated with a potential application, and then moves through requirements analysis, system design, and development.

Source record↗
DOE OSTI2019

Generation and characterization of carbon fiber microstructure in atomistic simulations

A clear understanding of carbon fiber (CF) microstructure is necessary for the development of high strength CFs. Here, we present an atomistic approach for generating and characterizing realistic mi- crostructures of CFs. Large-scale reactive molecular dynamics simulations are used to generate a set of distinct CF microstructures. Comprehensive characterization of the simulated microstructures is enabled by the development of a suite of computational structural analysis tools capable of evaluation of hy- bridization states of carbon atoms, populations and orientations of individual carbon rings, degree of graphitization, and pore size distribution. The calculation of X-ray diffraction profiles provides a direct link between the structural features of simulated samples and experimental data available for CFs. The CF generation algorithm is shown to produce microstructures with experimental densities and with structural characteristics matching those of PAN-based CFs. The key structural features affecting the properties of CFs, such as the relative fractions of graphitic, turbostratic and amorphous micro- constituents, degree of alignment, pore size distributions, and chemical cross-linking can be effectively controlled in simulations, thus enabling efficient exploration of structureeproperties relationships in CFs. The capabilities of the developed approach are illustrated by performing computational analysis of the mechanical deformation and fracture of CFs under axial tensile loading.

36 MATERIALS SCIENCE↗
DOE OSTI2022

Inference-Engine v0.1.0

Given a pre-trained neural network, Inference-Engine performs maps network inputs to outputs by executing the forward pass through the provided network. Although the predominant programming language for machine-learning is Python, most high-performance computing (HPC) applications are written in Fortran, C, or C++. Inference-Engine aims to support HPC programs and is written in Fortran, a language with a large feature set supporting interoperability with C. This software exposes concurrency in a portable way by using standard language features that some modern Fortran compilers can exploit with various optimizations, including offloading computation to a Graphics Processing Unit (GPU). In particular, this software makes extensive use of Fortran's "do concurrent" parallel loop construct, implicitly parallel array statements, and pure procedures that can be invoked inside "do concurrent" blocks. Inference-Engine also supports dynamic choice of inference methods at runtime. Two current options include one method that uses Fortran's "dot_product" intrinsic function inside "do concurrent" blocks and another method that instead uses Fortran' "matmul" array intrinsic function. We plan to investigate automatic compiler offloading of "do concurrent" calculations to GPUs and compile-time substitution of optimized libraries such as the Basic Linear Algebra Library (BLAS) for "matmul" invocations. We also envision the potential for the choice of which method to use could happen at program launch based on in situ performance measurements on any given platform.

Rouson, Damian↗
DOE OSTI2020

Development and Validation of a Two-Phase Thermal-Hydraulic CFD Code NEK-2P

A project is underway to develop, verify and validate an advanced two-phase flow modeling capability for the highly-scalable, high-performance Computational Fluid Dynamics (CFD) code NEK5000. The goal of this work is to verify and validate the two-phase version of the NEK5000 code, named NEK-2P, to simulate the two-phase flow and heat transfer phenomena that occur in a Boiling Water Reactor (BWR) fuel bundle under various operating conditions. The NEK-2P two-phase flow models follow the approach used for the Extended Boiling Framework (EBF) previously developed at Argonne but include more fundamental physical models of boiling phenomena and advanced numerical algorithms for improved computational accuracy, robustness, and computational speed. The development of the NEK-2P two-phase solver and the implementation of the Extended Boiling Framework two-phase models were initially supported by Argonne National Laboratory (Argonne) through a Laboratory Directed Research and Development (LDRD) project during FY2014-2016. The development and validation of the two-phase models through analyses of selected two-phase boiling flow experiments was supported by the Nuclear Energy Advanced Modeling and Simulation (NEAMS) program in FY2017-2020. This report focuses on verification and validation of the water-steam boiling model NEK-2P Two-Phase, CFD code. The NEK-2P was validated with Nuclear Power Engineering Corporation (NUPEC) Pressurized Water Reactor (PWR) Sub-channel and Bundle Test (PSBT) void distribution benchmark. Three different simulations were performed and analyzed for various operating conditions such as wall-heat flux and sub-cooled inlet temperatures. Reasonably good agreement with measured data was obtained in predicting the measured void distributions. Simulations were performed for Virginia Tech. (VT) 3x3 rod bundle geometry with and without spacers. The preliminary results were presented for Simplified Spacer Grid (SSG). In addition, the implementation of interface reconstruction model was tested with one of the Becker benchmark Critical Heat Flux (CHF) experiments.

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗
NASA NTRS2019

Testbed Results for Scalar and Vector Radiative Transfer Computations of Light in Atmosphere-Ocean Systems

We generate and tabulate reflectance values of the Stokes parameters I, Q, and U of upwelling radiance just above a rough ocean surface and at the top of the atmosphere (TOA) for 100 scattering geometries, four atmosphere-ocean systems, and four wavelengths. The atmosphere-ocean systems increase in complexity from (a) a molecular atmosphere above a rough ocean surface (AOS-I model); to (b) a pure water body below a rough ocean surface (AOS-II model); to (c) a fully-coupled simple atmosphere-ocean system (AOS-III model) containing a molecular atmosphere, rough ocean surface, and pure water; to (d) a fully-coupled complex atmosphere-ocean system (AOS-IV model) that includes scattering by molecules, rough ocean surface, pure water, and hydrosols. Our wavelengths (350, 450, 550, and 650 nm) capture the ultraviolet-visible range. Our tables provide radiative transfer (RT) testbed results for atmosphere-ocean systems with an accuracy that surpasses the measurement accuracy of state-of-the-art polarimeters. To validate the accuracy of these tables we performed computations using three independent RT codes that provide deterministic numerical solutions for the RT equation. The agreement is 10(exp –5) for AOS-IV model, and 10(exp –6) for the other models. The degree of linear polarization computed by these RT codes differs by ≤0.2% for 15 isolated cases of tabulated reflectance values, and by ≤0.1% for all remaining cases. We also provide comparisons with results obtained by a stochastic RT code for AOS-I model. The agreement between the deterministic and stochastic results for this model is 10(exp –5) at TOA, and 10(exp –6) above the ocean surface.

Polarimeters↗
DOE OSTI2022

DenKv: Addressing Design Trade-offs of Key-value Stores for Scientific Applications

High-performance computing (HPC) facilities have employed flash-based storage tier near to compute nodes to absorb high I/O demand by HPC applications during periodic system-level checkpoints. To accelerate these checkpoints, proxy-based distributed key-value stores (PD-KVS) gained particular attention for their flexibility to support multiple backends and different network configurations. PD-KVS rely internally on monolithic KVS, such as LevelDB or RocksDB, to exploit the KV interface and query support. However, PD-KVS are unaware of the high redundancy factor in checkpoint data, which can be up to GBs to TBs, and therefore, tend to generate high write and space amplification on these storage layers. In this paper, we propose DenKv which is deduplication-extended node-local LSM-tree-based KVS. DenKv employs asynchronous partially inline dedup (APID) and aims to maintain the performance characteristics of LSM-tree-based KVS while reducing the write and space amplification problems. We implemented DenKv atop BlobDB and showed that our proposed solution maintains performance while reducing write amplification up to 2× and space amplification by 4× on average.

Khan, Awais↗
DOE OSTI2021

Distributionally Robust Partially Observable Markov Decision Process with Moment-Based Ambiguity

In this paper, we consider a distributionally robust partially observable Markov decision process (DR-POMDP), where the distribution of the transition-observation probabilities is unknown at the beginning of each decision period, but their realizations can be inferred using side information at the end of each period after an action being taken. We build an ambiguity set of the joint distribution using bounded moments via conic constraints and seek an optimal policy to maximize the worst-case (minimum) reward for any distribution in the set. We show that the value function of DR-POMDP is piecewise linear convex with respect to the belief state and propose a heuristic search value iteration method for obtaining lower and upper bounds of the value function. We conduct numerical studies and demonstrate the computational performance of our approach via testing instances of a dynamic epidemic control problem. Our results show that DR-POMDP can produce more robust policies under misspecified distributions of transition-observation probabilities as compared to POMDP but has less costly solutions than robust POMDP. The DR-POMDP policies are also insensitive to varying parameter in the ambiguity set and to noise added to the true transition-observation probability values obtained at the end of each decision period.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

A Case Study of Multimodal, Multi-institutional Data Management for the Combinatorial Materials Science Community

Although the convergence of high-performance computing, automation, and machine learning has significantly altered the materials design timeline, transformative advances in functional materials and acceleration of their design will require addressing the deficiencies that currently exist in materials informatics, particularly a lack of standardized experimental data management. The challenges associated with experimental data management are especially true for combinatorial materials science, where advancements in automation of experimental workflows have produced datasets that are often too large and too complex for human reasoning. The data management challenge is further compounded by the multimodal and multi-institutional nature of these datasets, as they tend to be distributed across multiple institutions and can vary substantially in format, size, and content. Furthermore, modern materials engineering requires the tuning of not only composition but also of phase and microstructure to elucidate processing–structure–property–performance relationships. To adequately map a materials design space from such datasets, an ideal materials data infrastructure would contain data and metadata describing (i) synthesis and processing conditions, (ii) characterization results, and (iii) property and performance measurements. In this work, we present a case study for the low-barrier development of such a dashboard that enables standardized organization, analysis, and visualization of a large data lake consisting of combinatorial datasets of synthesis and processing conditions, X-ray diffraction patterns, and materials property measurements generated at several different institutions. While this dashboard was developed specifically for data-driven thermoelectric materials discovery, we envision the adaptation of this prototype to other materials applications, and, more ambitiously, future integration into an all-encompassing materials data management infrastructure.

36 MATERIALS SCIENCE↗
DOE OSTI2023

Thermal Spectrum Molten Salt-Fueled Reactor Reference Plant Model

This report details a fully coupled neutronics thermal hydraulics reference plant model for a thermal spectrum molten salt-fueled reactor. The Multiphysics model is developed on the Nuclear Regulatory Commission's Comprehensive Reactor Analysis Bundle available on the Idaho National Laboratory high-performance computer, which natively and seamlessly couples Griffin and Sam codes of the Multiphysics Object-Oriented Simulation Environment (MOOSE) based applications. Griffin provides the reactor physics capabilities, k-eigenvalue, delayed neutron precursor concentrations, and transient solutions for stationary and flowing fuels. Sam provides the solution of mass, momentum, and energy conservation equations of the whole system with a set of system components for one-dimensional single-phase flow. The neutronics feedback model relies primarily on fuel and moderator temperatures, fuel salt density, and delayed neutron precursors distributions. This report presents our results for the coupled steady state core and an unprotected loss of flow event. Although this model is prototypical regarding capabilities in the Comprehensive Reactor Analysis Bundle, its results are consistent with experimental data of the pump transient tests that were performed at zero power of the Molten Salt Reactor Experiment (MSRE).

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
NASA NTRS2019

A Mixed Integer Efficient Global Optimization Algorithm with Multiple Infill Strategy - Applied to a Wing Topology Optimization Problem

With the advancement in high performance computing and numerical optimization techniques,engineering design optimization problems are becoming more complex, larger scale,higher fidelity, and computationally more demanding, requiring longer run times than ever before. There exists methodologies and techniques that can address some of these challenges but very few can address all, and most are limited in the extent that these concerns can be addressed. With the goal of addressing such challenging engineering problems, we developed anew optimization framework, named AMIEGO, that combines concepts from surrogate-based optimization approaches, gradient-based numerical methods, Partial Least Squares, evolutionary algorithms, and Branch-and-Bound, providing newer capabilities that were not previouslyperceived. However, the original version of this framework, in the process of adaptive samplingto explore and exploit the design space, finds only a single sample point per iteration. The efforthere builds upon this previously developed optimization framework to include multiple infillsampling capability that combines the concept of generalized expected improvement function,unsupervised learning, and multi-objective evolutionary technique. To demonstrate, AMIEGOwith the multiple infill capability (called AMIEGO-MIMOS) solves a series of increasingly difficultengineering design optimization problems. The results reveal the performance of the newapproach is problem dependent. When applied to a ten-bar truss problem, the newly proposedmultiple infill strategy consistently leads to a better design solutions when compared to theexisting CPTV method (implemented with the context of the AMIEGO framework). On theother hand, when applied to a mixed-integer high fidelity wing topology optimization problem- MIMOS, despite showing a steeper convergence at the start, eventually leads to an inferiorsolution as compared to CPTV approach. These results also reveal that a small number ofstarting points, in general, are sufficient to lead to a good overall solution.

Mixed-integer optimization↗
DOE OSTI2021

Reaction Mechanism Generator v3.0: Advances in Automatic Mechanism Generation

In chemical kinetics research, kinetic models containing hundreds of species and tens of thousands of elementary reactions are commonly used to understand and predict the behavior of reactive chemical systems. Reaction Mechanism Generator (RMG) is a software suite developed to automatically generate such models by incorporating and extrapolating from a database of known thermochemical and kinetic parameters. Here, we present the recent version 3 release of RMG and highlight improvements since the previously published description of RMG v1.0. Most notably, RMG can now generate heterogeneous catalysis models in addition to the previously available gas- and liquid-phase capabilities. For model analysis, new methods for local and global uncertainty analysis have been implemented to supplement first-order sensitivity analysis. The RMG database of thermochemical and kinetic parameters has been significantly expanded to cover more types of chemistry. The present release includes parallelization for faster model generation and a new molecule isomorphism approach to improve computational performance. RMG has also been updated to use Python 3, ensuring compatibility with the latest cheminformatics and machine learning packages. Overall, RMG v3.0 includes many changes which improve the accuracy of the generated chemical mechanisms and allow for exploration of a wider range of chemical systems.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2024

MARVEL Reactor Fuel Performance Report (Rev.2)

The Microreactor Applications Research Validation and EvaLuation (MARVEL) project is producing a high temperature liquid metal-cooled nuclear test bed at Idaho National Laboratory (INL) to ultimately improve the integration of microreactors to end-user applications. This ambitious effort seeks to design, authorize, construct, test, and operate the reactor within five years. In order to construct and operate the MARVEL reactor in a timely manner, the system will utilize materials and component designs which have already been used, qualified, or licensed from previous reactors. The MARVEL reactor will be located at the INL Transient Reactor Test (TREAT) facility in the north high-bay equipment pit and will use the existing 304 stainless steel-clad U-ZrH1.6 pin-type fuel system developed by General Atomics and purchased from TRIGA International. This fuel has been previously qualified under the United States Department of Energy’s (US DOE) Reduced Enrichment for Research and Test Reactors (RERTR) Program. Even though the regulator of the MARVEL reactor is the US DOE, the standards and overall approach recommended by the Nuclear Regulatory Commission is well-defined and utilized here. Following NUREG-1537 regulatory guidance, this report documents the authorization case for the MARVEL fuel system’s application to MARVEL and establishes stable and predictable fuel performance during the most thermophysically unfavorable conditions achievable in the MARVEL reactor. To that end, this report provides a comprehensive survey of the known thermophysical properties, performance, and quantitative relationships associated with the MARVEL reactor fuel element and uses this information to determine its mechanical integrity and risk of reaching unacceptable conditions during the most extreme accident scenarios predicted for the reactor using the most conservative assumptions available. The information contained herein is compiled from a combination of historical reports and peer reviewed scientific publication manuscripts. Known mechanisms under which the fuel is susceptible to failure are highlighted and compared to conditions that could exist in the MARVEL reactor during an unanticipated transient or accident scenario. The two scenarios considered for analysis in this report are (1) an unprotected loss of flow accident and (2) a hypothetical unprotected loss of coolant accident during the loss of flow accident. Preliminary 2D steady-state analyses herein indicate that both fuel-cladding chemical interactions and fuel-cladding mechanical interactions are negligible throughout the fuel’s operational cycle under both normal and high temperature accident scenario conditions. Although higher fidelity 3D time-dependent modeling and simulations are planned, the following may be concluded presently. The MARVEL fuel element maintains its geometric stability and structural integrity during the most extreme accident scenarios predicted for the MARVEL reactor. The hoop stress during the unprotected loss of flow accident reaches about -1.27 MPa; this negative stress indicates that it is compressive rather than tensile. The compressive stress is a result of the NaK pressure on the outside of the fuel element, caused by the restricted thermal expansion of the NaK coolant, exceeding the internal pressures generated inside of the fuel element. The hoop stress generated in the cladding during the unprotected loss of coolant accident reaches a maximum of about approximately 10 MPa, which is nearly an order of magnitude less than the predicted yield strength of the cladding under high-temperature accident scenario conditions. Calculations were compared with results from high performance computational simulations using BISON and are in very close agreement. A conservative MARVEL fuel meat peak temperature limit of 900 °C is recommended presently, which is about 180 °C higher than the peak fuel temperature predicted to occur during the most extreme accident. Based on the known properties and behavior of the MARVEL fuel element, the fuel successfully meets its design and safety requirements under normal and most extreme accident conditions with a large safety margin.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗
DOE OSTI2019

A Quantitative Study of Deep Learning Training on Heterogeneous Supercomputers

Deep learning (DL) has become a key technique for solving complex problems in scientific research and discovery. DL training for science is substantially challenging because it has to deal with massive quantities of multi-dimensional data. High- performance computing (HPC) supercomputers are increasingly being employed for meeting the exponentially growing demand for DL. Multiple GPUs and high-speed interconnect network are needed for supporting DL on HPC systems. However, the excessive use of GPUs without considering effective benefits leads to inefficient resource utilization of these expensive setups. In this paper, we conduct a quantitative analysis to gauge the efficacy of DL workloads on the latest HPC system and identify viability of next-generation DL-optimized heterogeneous supercomputers for enabling researchers to develop more efficient resource management and distributed DL middleware. We evaluate well- known DL models with large-scale datasets using the popular TensorFlow framework, and provide a thorough evaluation in- cluding scalability, accuracy, variability, storage resource, GPU- GPU/GPU-CPU data transfer, and GPU utilization. Our analysis reveals that the latest heterogeneous supercomputing cluster shows varying performance trend as compared to the existing literature for single- and multi-node training. To the best of our knowledge, this is the first work to conduct such a quantitative and comprehensive study of DL training on a supercomputing system with multiple GPUs.

Han, Jingoo↗