Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “network acceleration”

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 937 records · Page 52
All sourcesNASA NTRSDOE OSTI
DOE OSTI2026

Linking microstructure to creep behavior in vertically and horizontally built LPBF Haynes 282 compared with wrought material via θ -projection

Laser Powder Bed Fusion (LPBF) has emerged as a promising route for fabricating intricate geometries in high-performance alloys. Haynes 282 (H282) is a strong candidate for applications such as heat exchangers or engines due to its excellent creep strength and thermal stability; however, the long-term creep behavior of LPBF-processed H282 remains poorly understood. In this study, the θ -projection method is used to analyze and extrapolate the creep behavior of vertically built LPBF, horizontally built LPBF, compared to wrought H282 tested at 816 °C. Vertically built LPBF H282 exhibits the lowest minimum creep rate (MCR), while the horizontally built condition shows a higher MCR comparable to that of wrought H282. Despite these differences, both LPBF conditions exhibit significantly shorter rupture life and reduced rupture strain than the wrought material, with the most severe degradation observed in the horizontal builds, consistent with an earlier onset of tertiary creep and accelerated strain-rate evolution. Microstructural characterization reveals that both LPBF and wrought H282 exhibit abundant twin-related boundary character; however, their grain boundary topologies differ markedly. The wrought alloy contains a higher fraction of low-angle grain boundaries and continuous twin lamellae, whereas the LPBF microstructure is characterized by a suppressed low-angle boundary population and fragmented twin-related boundaries embedded within irregular high-angle grain boundary networks. Fractographic analysis further reveals predominantly intergranular cracking in LPBF H282, accompanied by grain-boundary-decorated carbides, Al 2 O 3 inclusions, and high-aspect-ratio pores. These results demonstrate that grain boundary topology, rather than minimum creep rate alone, plays a critical role in governing creep damage accumulation and rupture behavior in LPBF and wrought H282.

Creep↗
DOE OSTI2024

Structural evolution of liquid silicates under conditions in Super-Earth interiors

Molten silicates at depth are crucial for planetary evolution, yet their local structure and physical properties under extreme conditions remain elusive due to experimental challenges. In this study, we utilize in situ X-ray diffraction (XRD) at the Matter in Extreme Conditions (MEC) end-station of the Linear Coherent Linac Source (LCLS) at SLAC National Accelerator Laboratory to investigate liquid silicates. Using an ultrabright X-ray source and a high-power optical laser, we probed the local atomic arrangement of shock-compressed liquid (Mg,Fe)SiO 3 with varying Fe content, at pressures from 81(9) to 385(40) GPa. We compared these findings to ab initio molecular dynamics simulations under similar conditions. Results indicate continuous densification of theO-O and Mg-Si networks beyond Earth’s interior pressure range, potentially altering melt properties at extreme conditions. This could have significant implications for early planetary evolution, leading to notable differences in differentiation processes between smaller rocky planets, such as Earth and Venus, and super-Earths, which are exoplanets withmasses nearly three times that of Earth.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2022

A Length Adaptive Algorithm-Hardware Co-design of Transformer on FPGA Through Sparse Attention and Dynamic Pipelining

Transformers are considered one of the most important deep learning models since 2018, in part because it establishes state-of-the-art (SOTA) records and could potentially replace existing Deep Neural Networks (DNNs). Despite the remarkable triumphs, the prolonged turnaround time of Transformer models is a widely recognized roadblock. The variety of sequence lengths imposes additional computing overhead where inputs need to be zero-padded to the maximum sentence length in the batch to accommodate the parallel computing platforms. This paper targets the field-programmable gate array (FPGA) and proposes a coherent sequence length adaptive algorithm–hardware co-design for Transformer acceleration. Particularly, we develop a hardware-friendly sparse attention operator and a length-aware hardware resource scheduling algorithm. The proposed sparse attention operator brings the complexity of attention-based models down to linear complexity and alleviates the off-chip memory traffic. The proposed length-aware resource hardware scheduling algorithm dynamically allocates the hardware resources to fill up the pipeline slots and eliminates bubbles for NLP tasks. Experiments show that our design has very small accuracy loss and has 80.2 × and 2.6 × speedup compared to CPU and GPU implementation, and 4 × higher energy efficiency than state-of-the-art GPU accelerator optimized via CUBLAS GEMM.

Peng, Hongwu↗
DOE OSTI2022

Hypothesis testing via AI: Generating physically interpretable models of scientific data with machine learning (Full Technical Report)

Deep learning has demonstrated an exceptional ability to solve complex tasks (an engineering success); however, it has done so at the expense of the ability to generate new knowledge (a scientific failure). We propose an alternative framework—entitled Deep Symbolic Regression (DSR)—in which artificial neural networks (NNs) rapidly generate hypotheses about physical relationships among inputs. This framework bypasses the need to interpret an NN altogether, while still leveraging the representational power of deep learning. The resulting models are tractable mathematical expressions, which are inherently and readily human interpretable and can provide insights into underlying physical phenomena. Further, we fold this methodology into the scientific process by allowing the scientist to directly integrate a priori knowledge and beliefs to accelerate learning. We demonstrate this methodology on symbolic regression—the problem of rediscovering underlying expressions describing a dataset—and achieve state-of-the-art performance across a wide variety of symbolic regression problems. Further, we generalize our DSR framework to apply to the more general class of symbolic optimization problems, in which one seeks to optimize a sequence of symbols or “tokens” under a black-box reward function. Examples of other symbolic optimization problems include neural architecture search and computational antibody design. Our generalized tool, Deep Symbolic Optimization (DSO), has been demonstrated on the task of learning symbolic control policies for reinforcement learning environments, and has been adopted as an enabling capability for computational antibody design.

97 MATHEMATICS AND COMPUTING↗
NASA NTRS1993

A broad-band microseismometer for planetary operations

There has recently been renewed interest in the development of instrumentation for making measurements on the surface of Mars. This is due to the Mars Environmental Survey (MESUR) Mission, for which approximately 16 small, long-lived (2-10 years), relatively inexpensive surface stations will be deployed in a planet-wide network. This will allow the investigation of processes (such as seismology and meteorology) which require the simultaneous measurement of phenomena at many widely spaced locations on the surface over a considerable length of time. Due to the large number of vehicles involved, the mass, power, and cost of the payload will be severely constrained. A seismometer has been identified as one of the highest priority instruments in the MESUR straw-man payload. The requirements for an effective seismic experiment on Mars place a number of constraints on any viable sensor design. First, a large number of sensors must be deployed in a long-lived global network in order to be able to locate many events reliably, provide good spatial sampling of the interior, and increase the probability of seismic detection in the event of localized seismicity and/or high attenuation. From a practical standpoint, this means that individual surface stations will necessarily be constrained in terms of cost, mass, and power. Landing and thermal control systems will probably be simple, in order to minimize cost, resulting in large impact accelerations and wide daily and seasonal thermal swings. The level of seismic noise will determine the maximum usable sensitivity for seismometer. Unfortunately, the ambient seismic noise level for Mars is not well known. However lunar seismic noise levels are several orders of magnitude below that of the Earth. Sensitivities on the order of 10(exp -11)g over a bandwidth of .04 to 20 Hz are thought to be necessary to fulfill the science objectives for a seimometer placed on the Martian surface. Silicon micromachined sensor technology offers techniques for the fabrication of monolithic, robust, compact, lower power and mass accelerometers. Conventional micro-machined accelerometers have been developed and are commercially available for high frequency and large acceleration measurements. The new seismometer we are developing incorporates certain principles of conventional silicon micromachined accelerometer technology. However, currently available silicon micromachined sensors offer inadequate sensitivity and bandwidth for the Mars seismometer application. Our implementation of an advanced silicon micromachined seismometer is based on principles recently developed at JPL for high-sensitivity position sensor technology.

Banerdt, W. B.↗
DOE OSTI2024

wa-hls4ml: A GNN Surrogate Model for hls4ml

Recent advancements in use of machine learning techniques on field-programmable gate arrays (FPGAs) have allowed for implementation of embedded neural networks with extremely low latency. This is invaluable for particle detectors at the Large Hadron Collider, where latency and used area must be strictly bounded. The hls4ml framework is a procedure for converting from trained machine learning model software, to a synthesis result that can be used on an FPGA. However, running the pipeline is a time-consuming procedure, and there is a strong risk of failure. In particular, it is possible that the model is unable to be converted into a synthesis result, or that the resource consumption of the model will exceed the resources of the target FPGA. To aid with this development, we introduce wa-hls4ml, a surrogate model which uses a graph neural network to emulate the structure of the source models. The goal is to estimate the chance of success and resource consumption of an arbitrary model when passed through the hls4ml procedure, without the time consumption of actually running the pipeline.

43 PARTICLE ACCELERATORS↗
DOE OSTI2024

Electromagnetic shower separation in LArTPCs: study of the calorimetric method at LArIAT

This thesis reports a study of the prospects of using calorimetry for separating photon- from electron-initiated electromagnetic (EM) showers in liquid argon time projection chambers (LArTPCs), a prevalent need for neutrino oscillation experiments based on the LArTPC technology as $\gamma$-initiated showers are a significant background to electron neutrino ($\nu_e$) events, whose signature is electron showers. From events simulated in the LArIAT (Liquid Argon In A Testbeam) experiment's detector, we quantitatively evaluated the dependence of the calorimetric separation on the initial momentum and position (along the beamline) of the primary particle. We could conclude that the separation is more efficient if the primary particle is created inside the chamber, whereas the impact of the momentum is not as significant. Furthermore, a first step was taken towards applying the calorimetric method to the showers measured in LArIAT. A convolutional neural network (CNN) trained on simulation was deployed to find EM showers among the events collected by the experiment. The charge deposition at the beginning of the showers found is centered at a plausible value, although the width of the distribution demands further inquiry.

43 PARTICLE ACCELERATORS↗
DOE OSTI2025

A percolating path to green iron

About 1.9 gigatons of steel is produced every year, emitting 8% (3.6 gigatons) of global CO 2 in the process. More than 50% of the CO 2 emissions come from a single step of steel production, known as ironmaking. Hydrogen- based direct reduction (HyDR) of iron oxide to iron has emerged as an emission-free ironmaking alternative. However, multiple physical and chemical phenomena ranging from nanometers to meters inside HyDR reactors alter the microstructure and pore networks in iron oxide pellets, in ways that resist gaseous transport of H 2 /H 2 O, slow reaction rates, and disrupt continuous reactor operation. Using synchrotron nano X-ray computed tomography and percolation theory, we quantify the evolution of pores in iron oxide pellets and demonstrate how nanoscale pore connectivity influences micro- and macroscale flow properties such as permeability, diffusivity, and tortuosity. Our modeling framework connects disparate scales and offers opportunities to accelerate HyDR.

hydrogen↗
DOE OSTI2022

Machine learning for design principles for single atom catalysts towards electrochemical reactions

Machine learning (ML) integrated density functional theory (DFT) calculations have recently been used to accelerate the design and discovery of heterogeneous catalysts such as single atom catalysts (SACs) through the establishment of deep structure–activity relationships. Here, this review provides recent progress in the ML-aided rational design of heterogeneous catalysts with the focus on SACs in terms of structure–activity relationships, feature importance analysis, high-throughput screening, stability, and metal–support interactions for electrochemistry. Support vector machine (SVM), random forest regression (RFR), and deep neural networks (DNN) along with atomic properties are mainly used for the design of SACs. The ML results have shown that the number of electrons in the d orbital, oxide formation enthalpy, ionization energy, Bader charge, d-band center, and enthalpy of vaporization are mainly the most important parameters for the defining of the structure–activity relationships for electrochemistry. However, the black-box nature of ML techniques occasionally makes a physical interpretation of descriptors, such as the Bader charge, d-band center, and enthalpy of vaporization, non-trivial. At the current stage, ML application is limited by the lack of a large and high-quality database. Future prospects for the development of a large database and a generalized ML algorithm for SAC design are discussed to give insights for further studies in this field.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
NASA NTRS2001

Monitoring the Microgravity Environment Quality On-Board the International Space Station Using Soft Computing Techniques

This paper presents an artificial intelligence monitoring system developed by the NASA Glenn Principal Investigator Microgravity Services project to help the principal investigator teams identify the primary vibratory disturbance sources that are active, at any moment in time, on-board the International Space Station, which might impact the microgravity environment their experiments are exposed to. From the Principal Investigator Microgravity Services' web site, the principal investigator teams can monitor via a graphical display, in near real time, which event(s) is/are on, such as crew activities, pumps, fans, centrifuges, compressor, crew exercise, platform structural modes, etc., and decide whether or not to run their experiments based on the acceleration environment associated with a specific event. This monitoring system is focused primarily on detecting the vibratory disturbance sources, but could be used as well to detect some of the transient disturbance sources, depending on the events duration. The system has built-in capability to detect both known and unknown vibratory disturbance sources. Several soft computing techniques such as Kohonen's Self-Organizing Feature Map, Learning Vector Quantization, Back-Propagation Neural Networks, and Fuzzy Logic were used to design the system.

Jules, Kenol↗
DOE OSTI2026

Enhanced Ionization Charge Identification in the Short-Baseline Neutrino Program Neutrino Detectors with Deep Neural Networks

We present a deep neural net-based region of interest detection method (DNN ROI) for signal processing in the liquid argon time projection chambers of the Short-Baseline Neutrino (SBN) Program, SBND and ICARUS. DNN ROI addresses limitations of the traditional wire-by-wire thresholding algorithm by leveraging the full two-dimensional detector readout and cross-plane matching information. To account for detector performance variations, we explore training with augmented samples. We find that DNN ROI outperforms the traditional method in both low-level ROI identification performance and high-level reconstruction metrics for high-energy cosmic and accelerator neutrino interaction products, while also being more robust against detector variations, with or without sample augmentation.

Abratenko, P. [Tufts U.]↗
NASA NTRS1990

A GNM mission and system design proposal

Here, the author takes an advocacy position for the proposed Mars Global Network Mission (GNM); it is not intended to be an objective review, although both pros and cons are presented in summary. The mission consists of launches from earth in the '96, '98, and '01 opportunities on Delta-class launch vehicles (approx. 1000 kg injected to Mars in 8 to 10 ft diameter shroud). The trans Mars boost stage injects a stack of small independent, aeroshelled spacecraft. The stack separates from the boost stage and each rigid (as opposed to deployable) aeroshell flies to Mars on its own, performing midcourse maneuvers as necessary. Each spacecraft flies a unique trajectory which is targeted to achieve approach atmospheric interface at the desired latitude and lighting conditions; arrival times may vary by a month or more. A direct entry is performed, there is no propulsive orbit capture. The aeroshelled rough-landers are targeted to achieve a desired attitude and entry flight path angle, and then follow a passive ballistic trajectory until terminal descent. Based on sensed acceleration (integrated to deduce altitude), the aft aeroshell skirt is jettisoned; a short later a supersonic parachute is deployed. The ballistic coefficient of the parachute is sized to achieve terminal velocity at about 8 km. However the parachute is not deployed until a few Km above the surface to minimize wind-induced drift. The nose cap descent imaging begins, a laser altimeter also measures true altitude. Based on range and range rate to the surface, the parachute is jettisoned and the lander uses descent engines to achieve touchdown velocity. A contact sensor shuts down the motors to avoid cratering, and the lander rough-lands at less than 5 m/sec. The remaining aeroshell and a deployable bladder attenuate landing loads and minimize the possibility of tip over. Science instruments are deployed and activated, and the network is established.

Bailey, Stephen↗
DOE OSTI2022

Multi-variance replica exchange SGMCMC for inverse and forward problems via Bayesian PINN

Physics-informed neural network (PINN) has been successfully applied in solving a variety of nonlinear non-convex forward and inverse problems. However, the training is challenging because of the non-convex loss functions and the multiple optima in the Bayesian inverse problem. In this work, we propose a multi-variance replica exchange stochastic gradient Langevin dynamics method to tackle the challenge of the multiple local optima in the optimization and the challenge of the multiple modal posterior distribution in the inverse problem. Replica exchange methods are capable of escaping from the local traps and accelerating the convergence; two chains with different temperatures are designed where the low temperature chain aims for the local convergence, and the target of the high temperature chain is to travel globally and explore the whole loss function entropy landscape. However, it may not be efficient to solve mathematical inversion problems by using the vanilla replica method directly since the method doubles the computational cost in evaluating the forward solvers (likelihood functions) in the two chains. To address this issue, we propose to make different assumptions on the energy function estimation and this facilities one to use solvers of different fidelities in the likelihood function evaluation. More precisely, one can use a solver with low fidelity in the high temperature chain while using a solver with high fidelity in the low temperature chain. Our proposed method significantly lowers the computational cost in the high temperature chain, meanwhile preserving the accuracy and converging very fast. Here we give an unbiased estimate of the swapping rate and give an estimation of the discretization error of the scheme. To verify our idea, we design and solve four inverse problems which have multiple modes. The proposed method is also employed to train the Bayesian PINN to solve the forward and inverse problems; faster and more accurate convergence has been observed when compared to the stochastic gradient Langevin dynamics (SGLD) method and vanilla replica exchange methods.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

Hierarchical Optimal Power Flow with Improved Gradient Evaluation

Existing algorithms to solve alternating-current optimal power flow (AC-OPF) often exploit linear approximations to simplify system models and accelerate computations. In this paper, we improve a recent hierarchical OPF algorithm, which rested on primal-dual gradients evaluated in a linearized distribution power flow model. Specifically, we identify a risk of voltage violation arising from the model linearization, and propose a more accurate gradient evaluation method to eliminate that risk. We further develop a hierarchical primal-dual algorithm to solve OPF based on the proposed gradient evaluation method. Numerical results on IEEE networks show that our algorithm can enhance voltage safety with satisfactory computational efficiency.

distributed algorithm↗
NASA NTRS1981

Future prospects for gamma-ray astronomy

As gamma-ray astronomy moves from the discovery to the exploratory phase, the promise of gamma-ray astrophysics noted by theorists in the late 1940s and 1950s is beginning to be realized. In the future, satellites should carry instruments that will have over an order of magnitude greater sensitivity than those flown thus far, and, for at least some portions of the gamma-ray energy range, these detectors will also have substantially improved energy and angular resolution. The information to be obtained from these experiments should greatly enhance our knowledge of several astrophysical phenomena including the very energetic and nuclear processes associated with compact objects, astrophysical nucleosynthesis, solar particle acceleration, the chemical composition of the planets and other bodies of the solar system, the structure of our galaxy, the origin and dynamic pressure effects of the cosmic rays, high energy particles and energetic processes in other galaxies especially active ones, and the degree of matter-antimatter symmetry of the universe. The gamma-ray results of the forthcoming programs such as Gamma-I, the Gamma Ray Observatory, the gamma-ray burst network, Solar Polar, and very high energy gamma-ray telescopes on the ground will almost certainly provide justification for more sophisticated telescopes. These advanced instruments might be placed on the Space Platform currently under study by N.A.S.A.

Fichtel, C.↗
NASA NTRS2001

Amyloplast sedimentation dynamics in maize columella cells support a new model for the gravity-sensing apparatus of roots

Quantitative analysis of statolith sedimentation behavior was accomplished using videomicroscopy of living columella cells of corn (Zea mays) roots, which displayed no systematic cytoplasmic streaming. Following 90 degrees rotation of the root, the statoliths moved downward along the distal wall and then spread out along the bottom with an average velocity of 1.7 microm min(-1). When statolith trajectories traversed the complete width or length of the cell, they initially moved horizontally toward channel-initiation sites and then moved vertically through the channels to the lower side of the reoriented cell where they again dispersed. These statoliths exhibited a significantly lower average velocity than those sedimenting on distal-to-side trajectories. In addition, although statoliths undergoing distal-to-side sedimentation began at their highest velocity and slowed monotonically as they approached the lower cell membrane, statoliths crossing the cell's central region remained slow initially and accelerated to maximum speed once they reached a channel. The statoliths accelerated sooner, and the channeling effect was less pronounced in roots treated with cytochalasin D. Parallel ultrastructural studies of high-pressure frozen-freeze-substituted columella cells suggest that the low-resistance statolith pathway in the cell periphery corresponds to the sharp interface between the endoplasmic reticulum (ER)-rich cortical and the ER-devoid central region of these cells. The central region is also shown to contain an actin-based cytoskeletal network in which the individual, straight, actin-like filaments are randomly distributed. To explain these findings as well as the results of physical simulation experiments, we have formulated a new, tensegrity-based model of gravity sensing in columella cells. This model envisages the cytoplasm as pervaded by an actin-based cytoskeletal network that is denser in the ER-devoid central region than in the ER-rich cell cortex and is linked to stretch receptors in the plasma membrane. Sedimenting statoliths are postulated to produce a directional signal by locally disrupting the network and thereby altering the balance of forces acting on the receptors in different plasma membrane regions.

NASA Program Fundamental Space Biology↗
DOE OSTI2025

Hong-Ou-Mandel interference with a coexisting clock using transceivers for synchronization over deployed fiber

Interference between independently generated photons is a key step towards distributing entanglement over long distances, but it requires synchronization between the distantly-located photon sources. Synchronizing the clocks of such photon sources using coexisting two-way classical optical communications over the same fiber that transports the quantum photonic signals is a promising approach for achieving photon-photon interference over long distances, enabling entanglement distribution for quantum networking using the deployed fiber infrastructure. Here, we demonstrate photon-photon interference by observing the Hong-Ou-Mandel dip between two distantly-located sources: a weak coherent-state source obtained by attenuating the output of a laser and a heralded single-photon source. We achieve a maximum dip visibility of 0.58 +/- 0.04 when the two sources are connected via 4.3 km of deployed fiber. Dip visibilities > 0.5 are nonclassical and a first step towards achieving teleportation over the deployed fiber infrastructure. In our experiment, the classical optical communication is achieved with - 21 dBm of optical signal launch power, which is used to synchronize the clocks in the two independent, distantly-located photon sources. The impact of spontaneous Raman scattering from the classical optical signals is mitigated by appropriate choice of the quantum- and classical-channel wavelengths. All equipment used in our experiment (the photon sources and the synchronization setup) is commercially available. Finally, our experiment represents a scalable approach to enabling practical quantum networking with commercial equipment and coexistence with classical communications in optical fiber.

47 OTHER INSTRUMENTATION↗
DOE OSTI2024

Sensitivity of Simulations of Double-detonation Type Ia Supernovae to Integration Methodology

Abstract We study the coupling of hydrodynamics and reactions in simulations of the double-detonation model for Type Ia supernovae. When assessing the convergence of simulations, the focus is usually on spatial resolution; however, the method of coupling the physics together as well as the tolerances used in integrating a reaction network also play an important role. In this paper, we explore how the choices made in both coupling and integrating the reaction portion of a simulation (operator/Strang splitting versus the simplified spectral deferred corrections method we introduced previously) influences the accuracy, efficiency, and nucleosynthesis of simulations of double detonations. We find no need to limit reaction rates or reduce the simulation time step to the reaction timescale. The entire simulation methodology used here is GPU-accelerated and made freely available as part of the Castro simulation code.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗