Accelerating Phase Field Simulations Through Convolution-Only Deep Learning and Diffusion Models
Explore the source record for details and available documents.
Safta, Cosmin [Sandia National Laboratories (SNL-C↗
SEARCH · Engineering Papers
Results for “accelerator modeling”
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.
A comparison of the results of dynamic wind-tunnel tests with theoretical predictions for an aeromechanical gust-alleviation system for light airplanes
Dynamic wind tunnel tests have been conducted on a 1/6-scale model of a general aviation airplane equipped with an all-mechanical gust alleviation system which uses auxiliary aerodynamic surfaces to drive the flaps. The longitudinal short period motions were studied under simulated gust conditions in order to verify the mathematical model used in a previous study to predict the performance of the full scale system and determine the amount of normal acceleration alleviation which could be attained. The model responses were measured for different configurations with the system active and without the system active for comparison. The tests confirmed the general relationships between the experimental variables and the model responses predicted by the mathematical model, but there were significant differences in the magnitudes of the responses. The experimental results for the model were used to estimate a reduction of 30 percent in the rms normal acceleration response of a similar full scale airplane in atmospheric turbulence.
Stewart, E. C.↗
Is there a limit on solar flare proton fluxes
The possibility that solar flare proton fluxes are limited in magnitude by saturation effects inherent to the acceleration mechanism is explored. If cyclotron damping of Alfven waves acts to accelerate protons, the criterion that the damping time is comparable to the acceleration time provides a fast particle number density at which protons load the wave spectrum. The limiting flux at 1 AU is obtained by a volume integration over the acceleration region and redistribution into an interplanetary emission cone. A simplified explosion model permits a delineation between volume and surface acceleration mechanisms in terms of a temporal parameter, the effective duration of acceleration. It is concluded that the limiting flux may be a realistic concept, but that further investigation is warranted to sharpen the criterion.
Barbosa, D. D.↗
MEPHESTO: Modeling Energy-Performance in Heterogeneous SoCs and Their Trade-Offs
Integrated shared memory heterogeneous architectures are pervasive because they satisfy the diverse needs of mobile, autonomous, and edge computing platforms. Although specialized processing units (PUs) that share a unified system memory improve performance and energy efficiency by reducing data movement, they also increase contention for this memory since the PUs interact with each other. Prior work has investigated performance degradation due to memory contention, but few have studied the relationship of power and energy to memory contention. Moreover, a comprehensive solution that models memory contention for kernel placement on contemporary heterogeneous systems on chip (SoCs) in response to energy and performance has been largely unaddressed.This paper presents MEPHESTO, a novel and holistic approach for managing this balance. The authors characterize applications and PUs in terms of two memory contention factors - time factors and power factors - to achieve the desired trade-off between energy and performance for collocated kernel execution on heterogeneous systems. The authors believe that this investigation is the first to combine all of these factors and present a simple knob-based approach that expresses the target trade-off. The approach is evaluated on a diverse integrated shared memory heterogeneous system with a CPU, GPU, and programmable vision accelerator. By using an empirical model for memory contention that provides up to 92% accuracy, the kernel collocation approach can provide a near-optimal ordering and placement based on the user-defined, energy-performance trade-off parameter. Moreover, the dynamic programming-based heuristics provide up to 30% better energy or 20% performance benefits when compared with the greedy approaches commonly employed by previous studies.
Alaul haque monil, Mohammad↗
High-Level Synthesis of Irregular Applications: A Case Study on Influence Maximization
The Influence Maximization problem is the problem of identifying a small cohort of actors from a broader population that, when initially activated in a diffusion process, are expected to result in a large number of activations in the population. While the problem is known to be NP-hard, several approximation algorithms have been devised by leveraging its submodular structure. While these algorithms are theoretically efficient, they are computationally very expensive in practice. This work advances the current state-of-the-art parallelization scheme for the IMM algorithm by devising the adoption of custom hardware accelerators implemented on FPGAs by leveraging High Level Synthesis from OpenCL. We study the performance of our proposed approach by exploring optimizations tailored at improving the parallel efficiency of the accelerators and highlight their effects and limitations in accelerating complex graph analytic applications. Our experimental evaluation shows that FPGA acceleration can improve the performance of the LT diffusion model up to 1.72x for the entire application and up to 2.90x for its most important kernel with respect to a CPU only parallel execution. The FPGA acceleration of the LT model shows also a 1.54x reduction in energy consumption when compared to a parallel CPU only run.
Neff, Reece W.↗
wa-hls4ml: A Benchmark and Surrogate Models for hls4ml Resource and Latency Estimation
As machine learning (ML) is increasingly implemented in hardware to address real-time challenges in scientific applications, the development of advanced toolchains has significantly reduced the time required to iterate on various designs. These advancements have solved major obstacles, but also exposed new challenges. For example, processes that were not previously considered bottlenecks, such as hardware synthesis, are becoming limiting factors in the rapid iteration of designs. To mitigate these emerging constraints, multipleefforts have been undertaken to develop an ML-based surrogate model that estimates resource usage of synthesized ML accelerator architectures. We introduce wa-hls4ml, a benchmark for ML accelerator resource and latency estimation, and its corresponding initial dataset of over 680 000 fully connected and convolutional neural networks, all synthesized using hls4ml and targeting Xilinx FPGAs. The benchmark evaluates the performance of resource and latency predictors against several common ML model architectures, primarily originating from scientific domains, as exemplar models, and the average performance across a subset of the dataset. Additionally, we introduce GNN- and transformer-based surrogate models that predict latency and resources for ML accelerators. We present the architecture and performance of the models and find that the models generally predict latency and resources for the 75% percentile within several percent of the synthesized resources on the synthetic test dataset.
Hawks, Benjamin G. [Fermilab]↗
Cosmic ray drift, shock wave acceleration and the anomalous component of cosmic rays
A model of the anomalous component of the quiet-time cosmic ray flux is presented in which ex-interstellar neutral particles are accelerated continuously in the polar regions of the solar-wind termination shock, and then drift into the equatorial regions of the inner heliosphere. The observed solar-cycle variations, radial gradient, and apparent latitude gradient of the anomalous component are a natural consequence of this model.
Pesses, M. E.↗
Cosmic ray drift, shock wave acceleration, and the anomalous component of cosmic rays
A model of the anomalous component of the quiet-time cosmic ray flux is presented in which ex-interstellar neutral particles are accelerated continuously in the polar regions of the solar-wind termination shock and then drift into the equatorial regions of the inner heliosphere. The observed solar-cycle variations, radial gradient, and apparent latitude gradient of the anomalous component are a natural consequence of this model.
Pesses, M. E.↗
Resistive diffusion of force-free magnetic fields in a passive medium. III - Acceleration of flare particles
A one-dimensional model is considered in which an increasingly large electric field is induced by a rapidly evolving magnetic field. In the case of solar flares, energies are estimated to which protons and electrons may be directly accelerated by such an induced electric field.
Low, B. C.↗
APOLLO: a facility-scale differentiable virtual accelerator for Fermilab
As the design complexity of modern accelerators grows, there is more interest in using advanced simulations that have fast execution time or yield additional insights like gradients. The FAST/IOTA facility has been working on implementing and experimentally validating an end-to-end digital twin that is both fast and gradient-aware, allowing for rapid prototyping of new software and experiments with minimal beam time costs. Our framework integrates physics and ML codes for linac and ring simulation through a set of generic interfaces between surrogate and physics-based sections. To reproduce device inputs and outputs, system state is exposed as a deterministic discrete event simulator. Because Fermilab is undergoing control system transition, both EPICS and ACNET frontends are supported. Recently, we have begun transitioning to a new community lattice standard, PALS, as well as developing standardized infrastructure for data ingest and normalization to prepare for model calibration during FAST proton injector commissioning. We discuss implementation details as well as challenges, and future plans to extend modelling to main complex proton accelerators like PIPII and Booster.
Kuklev, Nikita [Fermilab]↗
The preferential acceleration of heavy nuclei in solar flares.
A two-stage model is proposed for the mechanism of heavy nuclei (Z greater than 8) accelerated in solar flares. In the first stage, fully stripped ions are first accelerated to suprathermal energies. In the second stage, a fraction of these ions are Fermi-accelerated to higher energies. Transport to the Fermi-acceleration region involves electron pickup, while injection into Fermi acceleration is rigidity dependent. The degree of enhancement of the abundance of heavy nuclei may be expected to vary for each individual flare.
Cartwright, B. G.↗
Microwave and hard X-ray imaging of a solar flare on 1980 November 5
VLA and SMM hard X ray data on the solar flares of November 5, 1980 are analyzed and compared with data from other sources. The VLA provided measurements at 15 GHz at 10 sec intervals, using left and right circular polarizations with a 0.6 arcsec resolution. The hard X ray imaging spectrometer on the SMM obtained data in six bands from 3.5-30 keV, with 8 x 8 arcsec resolution and 1.5 sec separation. The data were examined for a possible nonthermal source for the microwave component of the emissions detected, the origin of 16-30 keV excess fluxes, the relation between the X ray and microwave sources, the magnetic connection between observed loops, and the physical characteristics of the radiating loop. The data were consistent with a model that assumes fast electrons are accelerated to a single power-law energy distribution and freely stream along the magnetic field. The data also agreed with a thick-target model for solar flare X ray emission.
Hoyng, P.↗
Simulations of lunar gravity field determination for Lunar Observer
The current plan for the Lunar Observer (LO) mission is to launch in the late 1990s and insert LO into a 100 km polar circular mapping orbit. However, prior to the mapping orbit, LO will be placed in a gravity calibration orbit (GCO) at a higher altitude to determine the gravity field of the moon. This paper examines the abilities of two GCO orbits (at 200 and 500 km altitudes) to recover a high degree and order gravity truth model that includes spherical harmonics and mascons by estimating different degree and order gravity fields with spherical harmonics only. This is achieved by comparing radial accelerations from the true and estimated models at the mapping altitude and by comparing trajectory propagations. For the gravity fields estimated (up to 30th degree and order), the 500 km GCO was just as successful as the 200 km GCO in determining the gravity field.
Konopliv, Alex↗
Towards a New Generation of Agricultural System Data, Models and Knowledge Products: Design and Improvement
This paper presents ideas for a new generation of agricultural system models that could meet the needs of a growing community of end-users exemplified by a set of Use Cases. We envision new data, models and knowledge products that could accelerate the innovation process that is needed to achieve the goal of achieving sustainable local, regional and global food security. We identify desirable features for models, and describe some of the potential advances that we envisage for model components and their integration. We propose an implementation strategy that would link a "pre-competitive" space for model development to a "competitive space" for knowledge product development and through private-public partnerships for new data infrastructure. Specific model improvements would be based on further testing and evaluation of existing models, the development and testing of modular model components and integration, and linkages of model integration platforms to new data management and visualization tools.
Agricultural systems↗
Real-Time Drilling Optimization System for Improved Overall Rate of Penetration and Reduced Cost Per Foot in Geothermal Drilling
The key to success in geothermal drilling is economic feasibility, and a major cost in the development of geothermal resources is the actual drilling of the wells. In this project, a real-time drilling optimization system for geothermal drilling was developed. The system couples three individual components while drilling. The first component is a drill stem vibration analysis model, the second is Mechanical Specific Energy (MSE) analyses, and the third is a detailed PDC Rate of Penetration (ROP) drill bit model for optimum RPM and WOB combinations. The benefit of the coupled system is that the range of WOB and RPM could be selected to avoid drill stem vibrations. Secondly, MSE is used as an efficiency measure and the detailed PDC drill bit model ensures the drill bit does not endure temperatures that exceed the temperature at which the PDC cutters experience accelerated wear. The new detailed PDC bit model is based on rock/bit interaction that physically tracks the PDC cutter wear flats as the bit drills ahead giving the capability to calculate the temperature being generated underneath the worn cutters to better advise on operational parameters to avoid accelerated cutter wear and failure and to ensure that operational parameters are applied so that overall ROP is maximized. By combining the drill stem vibrations and the detailed PDC bit cutter wear and “safe” non-accelerated cutter wear temperature and optimum ranges of operating parameters, it results in higher ROP and lower cost drilling. Single cutter PDC testing performed in different lithologies at Sandia was utilized to verify the PDC cutter forces and depth of cut for new and worn cutters. Based on single cutter PDC temperature modeling, verification using single cutter data from the testing done by National Oilwell Varco (NOV) was performed. Sandia’s Hard-Rock Drilling Facility (HRDF) was utilized to test different drill bit configurations with different cutter designs and wear status with different induced modes of vibration to obtain the critical bit RPM/WOB ranges resulting in ineffective drilling and low ROP. The collected test data were further used to verify and calibrate the full hole PDC ROP model that was developed based on single cutter interaction data. A full coupled drill stem vibration model was formulated and verified with geothermal field data from the Chocolate Mountain Aerial Gunnery Range (CMAGR). A graphical user interface (GUI) was developed using Tkinter library in the computer programming language Python, which integrates all the developed models in one system. The developed system consists mainly of the PDC ROP model, PDC bit wear model, PDC cutter temperature model, Mechanical Specific Energy (MSE) model, and drillstring vibration model integrated into one system. The developed system can be used for both, post well analysis and real-time optimization using different criteria such as ROP maximization or MSE minimization. The software uses Differential Evolution Algorithm (DEA) to find optimum values for operational parameters based on last foot drilled while avoiding the drillstring vibration and cutter temperature critical operating parameters.
15 GEOTHERMAL ENERGY↗
Fast Fermi acceleration in the plasma sheet boundary layer
A longstanding question in the field of magnetospheric physics is the source of the energetic particles which are commonly observed along the plasma-sheet boundary layer (PSBL). Several models have been suggested for the acceleration of these particles. Here, a means is suggested by which the fast Fermi acceleration mechanism (Wu, 1984) can accelerate electrons at the plasma sheet and perhaps account for some of the observations. In this scheme, a localized hydromagnetic disturbance propagating through the tail lobe region impinges upon the PSBL, deforming it and displacing it in toward the central plasma sheet. The boundary layer can then act like a moving magnetic mirror. If the disturbance is propagating nearly perpendicular to the layer, then its velocity projected parallel to the layer (and the magnetic field) can be very large, resulting in significant acceleration of reflected particles.
Wu, C. S.↗
Stochastic Particle Acceleration in Impulsive Solar Flares
The acceleration of a huge number of electrons and ions to relativistic energies over timescales ranging from several seconds to several tens of seconds is the fundamental problem in high-energy solar physics. The cascading turbulence model we have developed has been shown previously (e.g., Miller 2000; Miller & Roberts 1995; Miner, LaRosa, & Moore 1996) to account for all the bulk features (such as acceleration timescales, fluxes, total number of energetic particles, and maximum energies) of electron and proton acceleration in impulsive solar flares. While the simulation of this acceleration process is involved, the essential idea of the model is quite simple, and consists of just a few parts: 1. During the primary flare energy release phase, we assume that low-amplitude MHD Alfven and fast mode waves are excited at long wavelengths, say comparable to the size of the event (although the results are actually insensitive to this initial wavelength). While an assumption, this appears reasonable in light of the likely highly turbulent nature of the flare. 2. These waves then cascade in a Kolmogorov-like fashion to smaller wavelengths (e.g., Verma et al. 1996), forming a power-law spectral density in wavenumber space through the inertial range. 3. When the mean wavenumber of the fast mode waves has increased sufficiently, the transit-time acceleration rate (Miller 1997) for superAlfvenic electrons can overcome Coulomb energy losses, and these electrons are accelerated out of the thermal distribution and to relativistic energies (Miller et al. 1996). As the Alfven waves cascade to higher wavenumbers, they can cyclotron resonate with progressively lower energy protons. Eventually, they will resonate with protons in the tail of the thermal distribution, which will then be accelerated to relativistic energies as well (Miller & Roberts 1995). Hence, both ions and electrons are stochastically accelerated, albeit by different mechanisms and different waves. 4. When the protons become superAlfvenic (above about 1 MeV/nucleon), they too can suffer transit-time acceleration by the fast mode waves and will receive an extra acceleration "kick." The basic overall objective of this 1 year effort was to construct a spatially-dependent version of this acceleration model and this has been realized.
Miller, James A.↗
An Optimization Approach to Support Science Decision Making for Lunar Surface Exploration
Introduction: Scientific exploration is one of the three pillars of NASA’s Moon2Mars architecture, with crew surface extra vehicular activities (EVA) serving a critical enabling function. Development of surface EVA operational planning and execution, specifically integrating science and flight control teams (FCT), is currently being explored through analog scenarios. This integration, exercised, for example, through the Joint EVA and Hu-man Surface Mobility Test Team (JETT), allows for science input on EVA activities in near real-time through a Science Evaluation Room (SER), or Arte-mis science backroom, which integrates with the broader FCT through the Science Officer. The SER works within the FCT to support dynamic EVA planning in response to changes in operational constraints as well as science opportunities and re-prioritization, increasing the mission science return and accelerating the accomplishment of the Moon2Mars science objectives. The SER works within the FCT to provide recommendations to traverse execution in near real-time. One challenge is the requirement to deliver SER inputs to the FCT on operationally relevant timelines. Failure to do so may result in suboptimal execution of science exploration EVAs or even loss of key science objectives. To close this gap, we present a network optimization tool to allow the SER to provide rapid input to the FCT in response to changes in operational constraints or science opportunities. Inputs are predicated on approved science objectives, and clear rationale must be provided to the FCT for any requested change. Accordingly, this tool incorporates the Science Traceability Matrix (STM), SER prioritization scheme, and station characterization and action planning with operational constraints such as duration, traverse speed, and distance to maximize science objectives based on SER priorities, consistent with FCT operational requirements. Method: As a proof of concept, we used an existing linear programing software package used to simulate optimal routes through cellular metabolism. We built a Demonstrative Model with three STM objectives and four stations on a region of the Moon. The objectives were given an arbitrary prioritization and mapped to the stations through four possible crew actions. (Figs. 1 and 2). This station to STM mapping is consistent with the method used by the JETT5 Science Team to develop analog surface EVA science planning. We used a grid system with the landing site at the origin and the four stations placed across the positive x,y quadrant. Actions were assigned to each station and the accomplishment of those actions resulted in a numerical “reward” based on the ability of that action to achieve science objectives. The aggregate reward from each individual STM objective contributes to a global score (Science Yield), weighted by its priority. Operational constraints included a requirement to start and end at the landing site, 5 minutes each for initial station characterization and “clean up,” and variable total EVA time, traverse rate (fixed to 0.5 meters per second in our example), and time to perform each action (10, 5, 7, and 15 min for actions 1, 2, 3, and 4, respectively). Additional constraints and variables will be added in the future (e.g., sample mass, number of stations, traverse route constraints, illumination). Optimization. We converted the connections (arcs) between these stations (nodes) into a mixed integer linear programming optimization problem (arcs = constraints, nodes = variables) with the objective to maximize Science Yield. For any action, the Science Yield is equal to the relevance of that action to an STM objective [3, 2, and 1 point(s) for High, Med., and Low relevance, respectively], multiplied by the STM Objective Priority [3, 2, and 1 point(s) for High, Med., and Low priority, respectively]. This resulted in a model that computes the optimal station and action combination to maximize the Science Yield. These weightings can be adjusted by the SER as desired. Results: We explored three test cases for the Demonstrative Model. First, we set the maximum EVA duration to 120 minutes and computed the optimal route (Fig. 3A). The model suggested per-forming Actions 1 and 2 at Station P01, followed by Actions 1 and 2 at Station P02, and finally Actions 1 and 3 at Station P04 before returning to the Landing Site. Second, we adjusted the STM Objective Priori-ty order and computed the new optimal route (Fig. 3B). Under this situation, the model suggested per-forming all Actions at Station P02 followed by all Actions at Station P03. The previous test cases were relevant to SER planning activities. Next, we explored providing mid-EVA replanning input to the FCT. Scenario: While executing the Route in Fig. 3A the crew finishes at Station P01 and FCT decides that the EVA needs to finish in 45 minutes back at the Landing Site. FCT asks SER to recommend changes to the plan to accommodate this operation-al change. Using the model and incorporating these new constraints (start at Station P01, max. time of 45 min), the model suggested performing Actions 2 and 4 at Station P03 (Fig. 4), requiring 41 minutes to complete and return to the Landing Site. Interestingly, Station 3 was not part of the original route. Using the model, we determined the EVA would need 66 minutes, instead of 45, in order for the original Station P04 to yield a larger Science Yield than Station P03. The parametrization and simulation was per-formed in less than a minute, demonstrating the operational relevance of the approach. Future Efforts: The results from the Demonstrative Model suggest this tool can accelerate SER decision making on operationally relevant timelines. Use in analog activities, such as JETT5 or follow-ons, which have over a dozen stations for a crew to explore and over a dozen actions per station, will provide needed validation of the utility of this tool for planning EVAs, replanning mid-EVA, or planning follow-on EVAs based on previous results. Further integration with FCT execution monitoring tools may provide additional efficiency gains, al-lowing rapid and iterative exploration of operation-al and science decision space by the FCT and SER.
Science Operations↗