Search NASA⌕ Search

SEARCH · Search NASA

Results for “Performance Gap”

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 55 records · Page 3
All sourcesNASA NTRSDOE OSTI
DOE OSTI2023

Evaluating performance and portability of high-level programming models: Julia, Python/Numba, and Kokkos on exascale nodes

We explore the performance and portability of the high-level programming models: the LLVM-based Julia and Python/Numba, and Kokkos on high-performance computing (HPC) nodes: AMD Epyc CPUs and MI250X graphical processing units (GPUs) on Frontier’s test bed Crusher system and Ampere’s Arm-based CPUs and NVIDIA’s A100 GPUs on the Wombat system at the Oak Ridge Leadership Computing Facilities. We compare the default performance of a hand-rolled dense matrix multiplication algorithm on CPUs against vendor-compiled C/OpenMP implementations, and on each GPU against CUDA and HIP. Rather than focusing on the kernel optimization per-se, we select this naive approach to resemble exploratory work in science and as a lower-bound for performance to isolate the effect of each programming model. Julia and Kokkos perform comparably with C/OpenMP on CPUs, while Julia implementations are competitive with CUDA and HIP on GPUs. Performance gaps are identified on NVIDIA A100 GPUs for Julia’s single precision and Kokkos, and for Python/Numba in all scenarios. We also comment on half-precision support, productivity, performance portability metrics, and platform readiness. We expect to contribute to the understanding and direction for high-level, high-productivity languages in HPC as the first-generation exascale systems are deployed.

Godoy, William↗
DOE OSTI2024

Effects of Rolling Reduction on Critical Current Density and Microstructure of Bi-2212 Wires

Bi-2212 Rutherford cables have been fabricated into flat racetrack coils and canted-cosine-theta dipole magnets. The performance gap between the magnets made with Rutherford cables and the “short-sample-limit” is about 30%. To better understand the influence of Rutherford cable processing on the strand performance, we studied three Bi-2212 wires with filament architectures of 37 × 18 and 55 × 18 and diameters of 0.8 and 1.0 mm. To simulate the deformation caused by cabling process, the three wires were rolled with thickness reductions ranging from 10% to 30%. The aspect ratios of rolled strands are between 1.29 and 2.05. The low aspect-ratio wire is also an interesting form for fabricating solenoid coils with higher packing density. The round and rolled strands were heat-treated under 50 bar and with maximum heat treatment temperatures of 885.5 °C and 890.5 °C. The rolling deformation reduced filament size uniformity, resulting in filament merging in fully heat-treated wires. It was found that rolling reduction reduced wire critical current density (JE) by 16 to 18%, but the JE decrease saturated at 15 to 20% of the thickness reduction. It is believed that the reduced JE results from the filament merging caused by rolling and non-uniform shrinking during overpressure heat treatment.

Bi-2212 wire↗
NASA NTRS1992

Processor architecture and data buffering

A set of architectures from three major architecture families: stack, register, and memory-to-memory is discussed. It is shown that scalable architectures are not applicable for low-density technologies because they require at least 32 words of local memory. Software support is shown to be capable of bridging the performance gap between scalable and nonscalable architectures. A register architecture with 32 words of local memory allocated interprocedurally outperforms scalable architectures with equal sizes local memories and even some with larger size local memories. The performance advantage of unscalable architectures becomes significant when in addition to quality compile-time support, a small cache is added to an unscalable architecture. A 32-register architecture with 512 byte cache executes 20 percent less cycles when compared with an 8-set multiple overlapping set organization.

Mulder, Hans↗
DOE OSTI2022

Surface engineering with oxidized Ti 3 C 2 T x MXene enables efficient and stable p-i-n-structured CsPbI 3 perovskite solar cells

All-inorganic CsPbI 3 perovskite has a near-ideal band gap, high thermal stability, and simple material composition, thus presenting a promising option for developing perovskite/Si tandem solar cells. However, CsPbI 3 undergoes a rapid phase transition under exposure to moisture and exhibits a significant performance gap relative to other perovskite compounds, particularly in the p-i-n structure favored for perovskite/Si tandems. In this work, we demonstrate highly efficient and stable p-i-n-structured CsPbI 3 perovskite solar cells by surface engineering the CsPbI 3 layer with oxidized Ti 3 C 2 T x MXene (OMXene) nanoplates via spray coatings. OMXene provides a physical barrier against moisture and improves charge separation at the perovskite-electron transporting layer interface via an enhanced electric field. Consequently, we demonstrated CsPbI 3 /OMXene-based p-i-n devices with efficiencies of 19.69% for 0.096-cm 2 cells and 14.64% for 25-cm 2 minimodules. The encapsulated minimodule showed good stability, retaining ~85% of the initial efficiency under simultaneous damp heat (85°C/85% relative humidity) and 1-sun light soaking for over 1,000 h.

14 SOLAR ENERGY↗
DOE OSTI2019

Computer-aided detection using non-convolutional neural network Gaussian processes

Deep convolutional neural networks (CNNs) have in recent years achieved record-breaking performance on many image classification tasks and are therefore well-suited for computer aided detection (CAD). The need for uncertainty quantification for CAD motivates the need for a probabilistic framework for deep learning. The most well-known probabilistic neural network model is the Bayesian neural network (BNN), but BNNs are notoriously difficult to sample for large complex network architectures, and as such their use is restricted to small problems. It is known that the limit of BNNs as their widths increase toward infinity is a Gaussian process (GP), and there has been considerable research interest in these infinitely wide BNNs. Recently, this classic result has been extended to deep architectures in what is termed the neural network Gaussian process (NNGP) model. In this work, we implement an NNGP model and apply it to the ChestXRay14 dataset at the full resolution of 1024x1024 pixels. Even without any convolutional aspects to the network architecture and without any data augmentation, our five layer deep NNGP model outperforms other non-convolutional models and therefore helps to narrow the performance gap between non-convolutional and convolutional models. Our NNGP model is fully Bayesian and therefore offers uncertainty information through its predictive variance that can be used to formulate a predictive confidence measure. We show that the performance of the NNGP model is significantly boosted after low-confidence predictions are rejected, suggesting that convolution is most beneficial only for these low-confidence examples. Finally, our results indicate that an extremely large fully-connected neural network with appropriate regularization could perform as well as the NNGP if not for the computational bottleneck resulting from the large number of model parameters.

Agrawal, Devanshu↗
DOE OSTI2024

High protonic resistance of hydrocarbon-based cathodes in PEM fuel cells under low humidity conditions: Origin, implication, and mitigation

Hydrocarbon-based electrodes for proton-exchange membrane fuel cells face challenges in closing the performance gap with electrodes based on perfluorosulfonic acid ionomers, particularly under low humidity conditions. Alongside increased oxygen transport resistance and higher kinetic-induced overpotentials, the protonic resistance of these fluorine-free electrodes is the primary hurdle to improved performance. This study systematically investigates the origin and impact of the cathode protonic resistance on fuel cell performance, utilizing sulfonated phenylated polyphenylenes as hydrocarbon ionomers. Electrochemical characterization at low relative humidity (≤50 %) reveal a high protonic resistance arising from both lower conductivity of the hydrocarbon thin film compared to the bulk membrane and increased cathode tortuosity at a gas transport-optimized ionomer to carbon (I/C) ratio of 0.2. The poor protonic resistance at low relative humidities leads to a non-homogeneous current distribution across the thickness of the cathode electrode, resulting in lower catalyst utilization. To address this issue, reducing the thickness of the cathode CL while maintaining a constant Pt loading (i.e., increasing the Pt on carbon ratio) significantly reduces protonic resistance. This improvement compensates for the kinetic disadvantages of highly loaded carbon particles and results in a considerable performance increase by 40 % at 0.75 V under low relative humidities.

25 ENERGY STORAGE↗
DOE OSTI2022

Surface reaction for efficient and stable inverted perovskite solar cells

We report perovskite solar cells (PSCs) with an inverted structure (often referred to as the p-i-n architecture) are attractive for future commercialization due to their easily scalable fabrication, reliable operation, and compatibility with a wide range of perovskite-based tandem device architectures. However, the power conversion efficiency (PCE) of p-i-n PSCs falls behind n-i-p (or normal) structure counterparts. This large performance gap could undermine efforts to adopt p-i-n architectures, despite their other advantages. Given the remarkable advances in perovskite bulk materials optimization over the past decade, interface engineering has become the most important strategy to push PSC performance to its limit. Here, we report a reactive surface engineering approach based on a simple post-growth treatment of 3-(Aminomethyl)pyridine (3-APy) on top of a perovskite thin film. First, the 3-APy molecule selectively reacts with surface FA+, reducing perovskite surface roughness and surface potential fluctuations associated with surface steps/terraces. Second, the reaction product on the perovskite surface decreases the formation energy of charged iodine-vacancies, leading to effective n-type doping with a reduced work function in the surface region. With this reactive surface engineering, the resulting p-i-n PSCs obtained a PCE over 25%, along with retaining 87% of the initial PCE after over 2400 h of one-sun operation at about 55 degrees C in air.

14 SOLAR ENERGY↗
DOE OSTI2023

Sparse Symmetric Format for Tucker Decomposition

Tensor-based methods are receiving renewed attention in recent years due to their prevalence in diverse real-world applications. There is considerable literature on tensor representations and algorithms for tensor decompositions, both for dense and sparse tensors. Many applications in hypergraph analytics, machine learning, psychometry, and signal processing result in tensors that are both sparse and symmetric, making them an important class for further study. Similar to the critical Tensor Times Matrix chain operation (TTM c ) in general sparse tensors, the $\underline{S}$ parse $\underline{S}$ ymmetric $\underline{T}$ ensor $\underline{T}$ imes $\underline{S}$ ame $\underline{M}$ atrix $\underline{c}$ hain (S 3 TTM c ) operation is compute and memory intensive due to high tensor order and the associated factorial explosion in the number of non-zeros. We present the novel Compressed Sparse Symmetric (CSS) format for sparse symmetric tensors, along with an efficient parallel algorithm for the S 3 TTM c operation. We theoretically establish that S 3 TTM c on CSS achieves a better memory versus run-time trade-off compared to state-of-the-art implementations, and visualize the variation of the performance gap over the parameter space. We demonstrate experimental findings that confirm these results and achieve up to 2.72× speedup on synthetic and real datasets. The scaling of the algorithm on different test architectures is also showcased to highlight the effect of machine characteristics on algorithm performance.

42 ENGINEERING↗
DOE OSTI2025

Electric Drive Technologies Research: ELT223 Component Modeling, Co-Optimization, and Trade-Space Evaluation Annual Report

This project is intended to support the development of new traction drive systems that meet the targets of 100 kW/L for power electronics and 50 kW/L for electric machines with reliable operation to 300,000 miles. To meet these goals, new designs must be identified that make use of state-of-the-art and next-generation electronic materials and design methods. Designs must exploit synergies between components, for example converters designed for high-frequency switching using wide band gap (WBG) devices and ceramic capacitors. This project included: (1) a survey of available technologies; (2) investigating new technologies, that for example, reduce volume of thermal management or magnetic components; (3) the development of computer aided design tools that consider the converter volume, reliability, and electrical performance; (4) exercising the design software to evaluate performance gaps and predict the impact of certain technologies and design approaches, i.e. GaN semiconductors, ceramic capacitors, ceramic thermal management components, and select topologies; (5) building and testing hardware prototypes to validate models and concepts. The design tools enable co-optimization of the power module and passive elements and provide some design guidance. At the end of the project, new advanced computing methods, such as machine learning approaches, were considered.

33 ADVANCED PROPULSION SYSTEMS↗
NASA NTRSDate not supplied

X-57 Traction Power and Command Systems Development

This paper describes the final, as-built design of the X‑57 Maxwell aircraft power and command system architecture that implements the electrified propulsion capability. The development of the traction power, command, and avionics power subsystems proceeded as planned at the initial project critical design review, but improvements to the design were identified following the development of detailed operations concepts and integrated subsystem and system testing. The redundant architecture with A‑side and B‑side buses provided a robust framework for a developmental system that would turn out to have a lower technology readiness than had been assumed upon project formulation. As the project team identified reliability or performance gaps in the electrified propulsion powertrain components and their interactions with the other vehicle systems, the traction power, command, and avionics architectures were modified to accommodate the modified systems. Following the publication of the planned design approach at the critical design review milestone [1], the project team developed the flight hardware and software, integrated the systems, and adjusted the design and qualification activities to address gaps in the components, system architecture, and requirements as the gaps were realized. In addition to inherent challenges in development of these subsystems, the integration of this new technology with adjacent critical systems in an aircraft configuration posed additional challenges that drove design considerations across the subsystem development and the other vehicle systems. Integration complications included electromagnetic compatibility, thermal performance, and tolerance of single-point failures internal and external to the powerplant. Development of a qualification program was required for the new motors, inverters, and batteries, and hardware performance during component qualification fed back into the development process and led lessons learned and redesign of key elements.

distributed propulsion↗
DOE OSTI2022

Fusion pilot plant performance and the role of a sustained high power density tokamak

Here, recent U.S. fusion development strategy reports all recommend that the U.S. should pursue innovative science and technology to enable construction of a fusion pilot plant (FPP) that produces net electricity from fusion at low capital cost. Compact tokamaks have been proposed as a means of potentially reducing the capital cost of a FPP. However, compact steady-state tokamak FPPs face the challenge of integrating a high fraction of self-driven current with high core confinement, plasma pressure, and high divertor parallel heat flux. This integration is sufficiently challenging that a dedicated sustained-high-power-density (SHPD) tokamak facility is proposed by the U.S. community as the optimal way to close this integration gap. Performance projections for the steady-state tokamak FPP regime are presented and a preliminary SHPD device with substantial flexibility in lower aspect ratio (A = 2–2.5), shaping, and divertor configuration to narrow gaps to an FPP is described.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2022

Fusion Pilot Plant performance and the role of a Sustained High Power Density tokamak

Recent U.S. fusion development strategy reports all recommend that the U.S. should pursue innovative science and technology to enable construction of a Fusion Pilot Plant (FPP) that produces net electricity from fusion at low capital cost. Compact tokamaks have been proposed as a means of potentially reducing the capital cost of a fusion pilot plant. However, compact steady-state tokamak FPPs face the challenge of integrating a high fraction of self-driven current with high core confinement, plasma pressure, and high divertor parallel heat flux. This integration is sufficiently challenging that a dedicated sustained-high-power-density (SHPD) tokamak facility is proposed by the U.S. community as the optimal way to close this integration gap. Performance projections for the steady-state tokamak FPP regime are presented and a preliminary SHPD device with substantial flexibility in lower aspect ratio (A=2-2.5), shaping, and divertor configuration to narrow gaps to a FPP is described.

core-edge integration↗
DOE OSTI2025

Benchmarking Operators in Deep Neural Networks for Improving Performance Portability of SYCL

SYCL is a portable programming model for heterogeneous computing, so it is important to obtain reasonable performance portability of SYCL. Towards the goal of better understanding and improving performance portability of SYCL for machine learning workloads, we have been developing benchmarks for basic operators in deep neural networks (DNNs). These operators could be offloaded to heterogeneous computing devices such as graphics processing units (GPUs) to speed up computation. In this paper, we introduce the benchmarks, evaluate the performance of the operators on GPU-based systems, and describe the causes of the performance gap between the SYCL and Compute Unified Device Architecture (CUDA) kernels. We find that the causes are related to the utilization of the texture cache for read-only data, optimization of the memory accesses with strength reduction, use of local memory, and register usage per thread. We hope that the efforts of developing benchmarks for studying performance portability will stimulate discussion and interactions within the community.

Jin, Zheming [ORNL] (ORCID:000000027197780X)↗
DOE OSTI2017

Performance evaluation of an agent-based occupancy simulation model

Occupancy is an important factor driving building performance. Static and homogeneous occupant schedules, commonly used in building performance simulation, contribute to issues such as performance gaps between simulated and measured energy use in buildings. Stochastic occupancy models have been recently developed and applied to better represent spatial and temporal diversity of occupants in buildings. However, there is very limited evaluation of the usability and accuracy of these models. This study used measured occupancy data from a real office building to evaluate the performance of an agent-based occupancy simulation model: the Occupancy Simulator. The occupancy patterns of various occupant types were first derived from the measured occupant schedule data using statistical analysis. Then the performance of the simulation model was evaluated and verified based on (1) whether the distribution of observed occupancy behavior patterns follows the theoretical ones included in the Occupancy Simulator, and (2) whether the simulator can reproduce a variety of occupancy patterns accurately. Results demonstrated the feasibility of applying the Occupancy Simulator to simulate a range of occupancy presence and movement behaviors for regular types of occupants in office buildings, and to generate stochastic occupant schedules at the room and individual occupant levels for building performance simulation. For future work, model validation is recommended, which includes collecting and using detailed interval occupancy data of all spaces in an office building to validate the simulated occupant schedules from the Occupancy Simulator.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗
DOE OSTI2024

Evaluating Operators in Deep Neural Networks for Improving Performance Portability of SYCL

SYCL is a portable programming model for heterogeneous computing, so it is important to obtain reasonable performance portability of SYCL. Towards the goal of better understanding and improving performance portability of SYCL for machine learning workloads, we have been developing benchmarks for basic operators in deep neural networks (DNNs). These operators could be offloaded to heterogeneous computing devices such as graphics processing units (GPUs) to speed up computation. In this work, we introduce the benchmarks, evaluate the performance of the operators on GPU-based systems, and describe the causes of the performance gap between the SYCL and Compute Unified Device Architecture (CUDA) kernels. We find that the causes are related to the utilization of the texture cache for read-only data, optimization of the memory accesses with strength reduction, shared local memory accesses, and register usage per thread. We hope that the efforts of developing benchmarks for studying performance portability will stimulate discussion and interactions within the community.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2016

Can High-Performance Equipment Lead to a Low-Performance Building?

The performance-based compliance alternative available in most energy codes, intended to provide energy efficiency equivalent to that of prescriptive compliance while allowing innovation and design flexibility, can instead result in sub-standard energy performance in both the short and the long term. The potential deficiencies in modeled buildings originate with subtleties in the energy modeling rules, allowing building systems that consume more energy than their real-world, prescriptively-designed counterparts. This performance gap is exacerbated over subsequent decades as less efficient permanent features of the building remain while elements with shorter lives are regularly upgraded in most buildings. This paper summarizes an investigation into the topic for Pacific Northwest National Laboratory and the City of Seattle, including identification of the principal deficiencies exploited in the modeling path, and several potential code amendments that could resolve these deficiencies and establish better equivalency between prescriptive and performance compliance paths. The study, focusing on Seattle and Washington State energy codes, offers lessons and implications for other jurisdictions and energy codes.

Building Energy Codes↗
DOE OSTI2021

Synthesis and Structure–Activity Characterization of a Single-Site MoO 2 Catalytic Center Anchored on Reduced Graphene Oxide

Molecularly derived single-site heterogeneous catalysts can bridge the understanding and performance gaps between conventional homogeneous and heterogeneous catalysis, guiding the rational design of next-generation catalysts. While impressive advances have been made with well-defined oxide supports, the structural complexity of other supports and the nature of the grafted surface species present an intriguing challenge. In this study, single-site Mo(=O) 2 species grafted onto reduced graphene oxide (rGO/MoO 2 ) are characterized by XPS, DRIFTS, powder XRD, N 2 physisorption, NH 3 -TPD, aqueous contact angle, active site poisoning assay, Mo EXAFS, model compound single-crystal XRD, DFT, and catalytic performance. NH 3 -TPD reveals that the anchored MoO 2 moiety is not strongly acidic, while Mo 3d 5/2 XPS assigns the oxidation state as Mo(VI), and XRD shows little rGO periodicity change on MoO 2 grafting. Contact angle analysis shows that MoO 2 grafting consumes rGO surface polar groups, yielding a more hydrophobic surface. The rGO/MoO 2 DRIFTS assigns features at 959 and 927 cm –1 to the symmetric and antisymmetric Mo=O stretching modes, respectively, of an isolated cis-(O=Mo=O) moiety, in agreement with DFT computation. Moreover, the Mo EXAFS rGO/MoO 2 structural data are consistent with isolated (C–O) 2– Mo(=O) 2 species having two Mo=O bonds and two Mo–O bonds at distances of 1.69(3) and 1.90(3) Å, respectively. rGO/MoO 2 is also more active than the previously reported AC/MoO 2 catalyst, with reductive carbonyl coupling TOFs approaching 1.81 × 10 3 h –1 . rGO/MoO 2 is environmentally robust and multiply recyclable with 69 ± 2% of the Mo sites catalytically significant. Altogether, rGO/MoO 2 is a structurally well-defined and versatile single-site Mo(VI) dioxo heterogeneous catalytic system.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2023

Efficacy of vibro-impact energy harvesting absorbers on controlling dynamical systems under vortex-induced vibrations and base excitation

Here this study investigates improving the efficacy of an energy harvesting absorber's ability to control a structure under vortex-induced vibrations, base excitation, and a combination of the two by including mechanical amplitude stoppers. The nonlinear reduced-order model is developed through modifying trilinear spring models to represent the impact forces, a modified van der Pol oscillator to represent the forcing due to the vortex-induced vibrations and using the Euler-Lagrange principle to express the equations of motion. It is seen that a soft stopper stiffness and a 5mm gap performs the most effectively of increasing the power generated from the absorber while still greatly reducing the primary structure's amplitude. By changing the stopper's location towards the middle of the energy harvesting absorber, the large effects of the impact forces are reduced and improves the efficacy of medium and hard stopper stiffnesses to generate near the amount of power the soft stopper does, while greatly improving the control of the primary structure. When the system is under combined loadings, the large oscillations of the synchronization region cause the effective configuration to be that of a 27.5 mm gap with soft stiffnesses. The results shows that medium stiffness stoppers with small gaps generate large aperiodic regions due to the high impact force. When the oscillations are close to the stoppers, the beating phenomenon is observed and is not overpowered by the vibro-impact force.

16 TIDAL AND WAVE POWER↗