Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “continuous models”

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 1,369 records · Page 76

Experiments with a four-bladed cyclic pitch stirring model rotor, part 3

The experimental work with the 2-bladed 16-inch diameter model rotor has been continued with a 4-bladed 16.5 inch diameter rotor capable of progressing and regressing cyclic pitch excitation (cyclic pitch stirring). Advance ratios of 0, .19 and .38 were tested at rotor speeds corresponding to non-dimensional blade natural frequencies of 1:14 and 1.19. The results are presented in the form of the first 5 Fourier components of the periodic response modulating function which for a periodic system takes the place of the complex response amplitude ratio of a constant system. In addition, the first and second harmonics of the trim response are presented. The test data are compared to analytical data without rotor wake and to the test data obtained earlier with the two-bladed rotor model of half the blade solidity ratio.

Hohenemser, K. H.↗

Magnetohydrodynamic MACH Code Used to Simulate Magnetoplasmadynamic Thrusters

The On-Board Propulsion program at the NASA Glenn Research Center is utilizing a state of-the-art numerical simulation to model the performance of high-power electromagnetic plasma thrusters. Such thrusters are envisioned for use in lunar and Mars cargo transport, piloted interplanetary expeditions, and deep-space robotic exploration of the solar system. The experimental portion of this program is described in reference 1. This article describes the numerical modeling program used to guide the experimental research. The synergistic use of numerical simulations and experimental research has spurred the rapid advancement of high-power thruster technologies for a variety of bold new NASA missions. From its inception as a U.S. Department of Defense code in the mid-1980's, the Multiblock Arbitrary Coordinate Hydromagnetic (MACH) simulation tool has been used by the plasma physics community to model a diverse range of plasma problems--including plasma opening switches, inertial confinement fusion concepts, compact toroid formation and acceleration, z-pinch implosion physics, laser-target interactions, and a variety of plasma thrusters. The MACH2 code used at Glenn is a time-dependent, two-dimensional, axisymmetric, multimaterial code with a multiblock structure. MACH3, a more recent three-dimensional version of the code, is currently undergoing beta tests. The MACH computational mesh moves in an arbitrary Lagrangian-Eulerian (ALE) fashion that allows the simulation of diffusive-dominated and dispersive-dominated problems, and the mesh can be refined via a variety of adaptive schemes to capture regions of varying characteristic scale. The mass continuity and momentum equations model a compressible viscous fluid, and three energy equations are used to simulate nonthermal equilibrium between electrons, ions, and the radiation field. Magnetic fields are modeled by an induction equation that includes resistive diffusion, the Hall effect, and a thermal source for magnetic fields. Various models of plasma resistivity are included, along with ablation models and multiport circuit solvers. The set of equations is closed using either an ideal gas or real equation of state.

Mikellides, Pavlos G.↗

The 3-D world modeling with updating capability based on combinatorial geometry

A 3-D world modeling technique using range data is discribed. Range data quantify the distances from the sensor focal plane to the object surface, i.e., the 3-D coordinates of discrete points on the object surface are known. The approach proposed herein for 3-D world modeling is based on the Combinatorial Geometry (CG) method which is widely used in Monte Carlo particle transport calculations. First, each measured point on the object surface is surrounded by a small sphere with a radius determined by the range to that point. Then, the 3-D shapes of the visible surfaces are obtained by taking the (Boolean) union of all the spheres. The result is an unambiguous representation of the object's boundary surfaces. The pre-learned partial knowledge of the environment can be also represented using the CG Method with a relatively small amount of data. Using the CG type of representation, distances in desired directions to boundary surfaces of various objects are efficiently calculated. This feature is particularly useful for continuously verifying the world model against the data provided by a range finder, and for integrating range data from successive locations of the robot during motion. The efficiency of the proposed approach is illustrated by simulations of a spherical robot in a 3-D room in the presence of moving obstacles and inadequate prelearned partial knowledge of the environment.

Goldstein, M.↗

System and Propagation Availability Analysis for NASA's Advanced Air Transportation Technologies

This report summarizes the research on the System and Propagation Availability Analysis for NASA's project on Advanced Air Transportation Technologies (AATT). The objectives of the project were to determine the communication systems requirements and architecture, and to investigate the effect of propagation on the transmission of space information. In this report, results from the first year investigation are presented and limitations are highlighted. To study the propagation links, an understanding of the total system architecture is necessary since the links form the major component of the overall architecture. This study was conducted by way of analysis, modeling and simulation on the system communication links. The overall goals was to develop an understanding of the space communication requirements relevant to the AATT project, and then analyze the links taking into consideration system availability under adverse atmospheric weather conditions. This project began with a preliminary study of the end-to-end system architecture by modeling a representative communication system in MATLAB SIMULINK. Based on the defining concepts, the possibility of computer modeling was determined. The investigations continue with the parametric studies of the communication system architecture. These studies were also carried out with SIMULINK modeling and simulation. After a series of modifications, two end-to-end communication links were identified as the most probable models for the communication architecture. Link budget calculations were then performed in MATHCAD and MATLAB for the identified communication scenarios. A remarkable outcome of this project is the development of a graphic user interface (GUI) program for the computation of the link budget parameters in real time. Using this program, one can interactively compute the link budget requirements after supplying a few necessary parameters. It provides a framework for the eventual automation of several computations required in many experimental NASA missions. For the first year of this project, most of the stated objectives were accomplished. We were able to identify probable communication systems architectures, model and analyze several communication links, perform numerous simulation on different system models, and then develop a program for the link budget analysis. However, most of the work is still unfinished. The effect of propagation on the transmission of information in the identified communication channels has not been performed. Propagation effects cannot be studied until the system under consideration is identified and characterized. To study the propagation links, an understanding of the total communications architecture is necessary. It is important to mention that the original project was intended for two years and the results presented here are only for the first year of research. It is prudent therefore that these efforts be continued in order to obtain a complete picture of the system and propagation availability requirements.

Ugweje, Okechukwu C.↗

Unpacking model inadequacy: The quantification of silver release from TRISO fuel by considering empirical and mechanistic approaches

Increasing adoption of the proposed tristructural isotropic (TRISO) particle fuel for both advanced and existing reactors makes it critical to assess and address any uncertainties and inadequacies of TRISO fission product release models. Model inadequacy stems from simplifications made to the computational model when compared to the experiments. The modeling and simulation efforts conducted using the BISON fuel performance code, along with the experimental campaigns carried out under the Advanced Gas Reactor Fuel Development and Qualification Program, afford a unique opportunity to conduct a rigorous modeling inadequacy assessment within the Bayesian uncertainty quantification (UQ) framework. Here, this study compares the standard Bayesian framework against the Kennedy-O'Hagan (KOH) framework, which explicitly represents modeling inadequacy, in regard to UQ for TRISO silver release models. For this purpose, both the traditional Arrhenius equation fitted to experimental data and the more advanced lower-length-scale (LLS)-informed model, which considers microstructure information, are independently considered. Applying the inverse UQ process on the AGR-2 and -3/4 datasets revealed modeling inadequacy to be the most dominant source of uncertainty. Experimental noise uncertainty is also significant; however, model parameter uncertainty can be considered negligible. Interestingly, both the Arrhenius equation and the LLS-informed model demonstrated similar levels of modeling inadequacy. For the forward predictive UQ, the KOH framework improved both the accuracy and quality of quantified uncertainties in comparison to the standard Bayesian framework. This is true for both the Arrhenius equation and the LLS-informed model. In comparing these modeling approaches, both demonstrated similar performance at the engineering scale, while the LLS-informed model expectedly outperformed the Arrhenius equation at the mesoscale. These conclusions highlight the importance of explicitly accounting for modeling inadequacy in the UQ process, and reinforce the need for continuous refinement of physics-based models in order to address the modeling inadequacy.

Advanced reactors↗

A brief overview of compartmental modeling for intake of plutonium via wounds

Here, the aim of this study is to present several approaches that have been used to model the behavior of radioactive materials (specifically Pu) in contaminated wounds. We also review some attempts by the health physics community to validate and revise the National Council on Radiation Protection and Measurements (NCRP) 156 biokinetic model for wounds, and present some general recommendations based on the review. Modeling of intake via the wound pathway is complicated because of a large array of wound characteristics (e.g. solubility and chemistry of the material, type and depth of the tissue injury, anatomical location of injury). Moreover, because a majority of the documented wound cases in humans are medically treated (excised or treated with chelation), the data to develop biokinetic models for unperturbed wound exposures are limited. Since the NCRP wound model was largely developed from animal data, it is important to continue to validate and improve the model using human data whenever plausible.

61 RADIATION PROTECTION AND DOSIMETRY↗

Arctic tropospheric ozone: assessment of current knowledge and model performance

As the third most important greenhouse gas (GHG) after carbon dioxide (CO 2 ) and methane (CH 4 ), tropospheric ozone (O 3 ) is also an air pollutant causing damage to human health and ecosystems. This study brings together recent research on observations and modeling of tropospheric O 3 in the Arctic, a rapidly warming and sensitive environment. At different locations in the Arctic, the observed surface O 3 seasonal cycles are quite different. Coastal Arctic locations, for example, have a minimum in the springtime due to O 3 depletion events resulting from surface bromine chemistry. In contrast, other Arctic locations have a maximum in the spring. The 12 state-of-the-art models used in this study lack the surface halogen chemistry needed to simulate coastal Arctic surface O 3 depletion in the springtime; however, the multi-model median (MMM) has accurate seasonal cycles at non-coastal Arctic locations. There is a large amount of variability among models, which has been previously reported, and we show that there continues to be no convergence among models or improved accuracy in simulating tropospheric O 3 and its precursor species. The MMM underestimates Arctic surface O 3 by 5% to 15% depending on the location. The vertical distribution of tropospheric O 3 is studied from recent ozonesonde measurements and the models. The models are highly variable, simulating free-tropospheric O 3 within a range of ±50% depending on the model and the altitude. The MMM performs best, within ±8% for most locations and seasons. However, nearly all models overestimate O 3 near the tropopause (~300 hPa or ~8 km), likely due to ongoing issues with underestimating the altitude of the tropopause and excessive downward transport of stratospheric O 3 at high latitudes. For example, the MMM is biased high by about 20% at Eureka. Observed and simulated O 3 precursors (CO, NO x , and reservoir PAN) are evaluated throughout the troposphere. Models underestimate wintertime CO everywhere, likely due to a combination of underestimating CO emissions and possibly overestimating OH. Throughout the vertical profile (compared to aircraft measurements), the MMM underestimates both CO and NO x but overestimates PAN. Perhaps as a result of competing deficiencies, the MMM O 3 matches the observed O 3 reasonably well. Our findings suggest that despite model updates over the last decade, model results are as highly variable as ever and have not increased in accuracy for representing Arctic tropospheric O 3 .

54 ENVIRONMENTAL SCIENCES↗

Multivalued-Velocity-Field Model of Turbulence

Report presents multi-valued-velocity model used in calculations of turbulence formation. Developed as part of continuing study of turbulent fluid motion, model expected to evolve into comprehensive mathematical tool to explain origin and effects of turbulence. Model has great theoretical and practical value in such fields as aerodynamics, meteorology, and combustion.

Zak, M. A.↗

Arctic Tropospheric Ozone: Assessment of Current Knowledge and Model Performance

As the third most important greenhouse gas (GHG) after carbon dioxide (CO2) and methane (CH4), tropospheric ozone (O3) is also an air pollutant causing damage to human health and ecosystems. This study brings together recent research on observations and modeling of tropospheric O3 in the Arctic, a rapidly warming and sensitive environment. At different locations in the Arctic, the observed surface O3 seasonal cycles are quite different. Coastal Arctic locations, for example, have a minimum in the springtime due to O3 depletion events resulting from surface bromine chemistry. In contrast, other Arctic locations have a maximum in the spring. The 12 state-of-the-art models used in this study lack the surface halogen chemistry needed to simulate coastal Arctic surface O3 depletion in the springtime; however, the multi-model median (MMM) has accurate seasonal cycles at non-coastal Arctic locations. There is a large amount of variability among models, which has been previously reported, and we show that there continues to be no convergence among models or improved accuracy in simulating tropospheric O3 and its precursor species. The MMM underestimates Arctic surface O3 by 5 % to 15 % depending on the location. The vertical distribution of tropospheric O3 is studied from recent ozonesonde measurements and the models. The models are highly variable, simulating free-tropospheric O3 within a range of ±50 % depending on the model and the altitude. The MMM performs best, within ±8 % for most locations and seasons. However, nearly all models overestimate O3 near the tropopause (∼300 hPa or ∼8 km), likely due to ongoing issues with underestimating the altitude of the tropopause and excessive downward transport of stratospheric O3 at high latitudes. For example, the MMM is biased high by about 20 % at Eureka. Observed and simulated O3 precursors (CO, NOx, and reservoir PAN) are evaluated throughout the troposphere. Models underestimate wintertime CO everywhere, likely due to a combination of underestimating CO emissions and possibly overestimating OH. Throughout the vertical profile (compared to aircraft measurements), the MMM underestimates both CO and NOx but overestimates PAN. Perhaps as a result of competing deficiencies, the MMM O3 matches the observed O3 reasonably well. Our findings suggest that despite model updates over the last decade, model results are as highly variable as ever and have not increased in accuracy for representing Arctic tropospheric O3.

Arctic↗

Compaction localization and constitutive behavior of weak porous sandstone.

A combined experimental and constitutive modeling program for weak porous sandstone deformation is described. A series of axisymmetric compression tests were performed over a range of mean stresses to study dilatational, compactional and transitional regimes. Experimental results were used both to derive constitutive parameters for testing localization theory and to parameterize a poroelastic-plastic model. Observed strain localization, imaged syn-deformationally using acoustic emissions, includes high- and low-angle shear and low angle compactional features or 'bands'. Isotropic elastic moduli measured via unloading loops show a progressive degradation pre-failure as decreasing functions of work-conjugate plastic strains and increasing functions of stress magnitude. The degradation pathway is unique for samples which underwent localization versus those that underwent spatially pervasive pore collapse. Total shear and volume strains are partitioned into elastic and plastic portions including the ''coupling'' strain associated with modulus degradation. Plastic strain calculated with and without the coupling term is compared with regard to localization predictions. Both coupled and uncoupled cases predict high angle shear bands for uniaxial and low mean stress conditions on the dilatational side of the yield surface. Uncoupled predictions show progressively lower angle shear bands approaching the transitional regime (stress conditions approaching the 'cap' surface). When elastic-plastic coupling is accounted for, compaction bands are predicted for the transitional regime, as are observed in the experiments. Finite element modeling efforts are described using a 3-invariant, mixed-hardening, continuous yield surface, elasto-plasticity model that includes several features important for porous sandstone constitutive behavior and observed experimentally, including non-associativity, nonlinear elasticity, elastic-plastic coupling, and kinematic hardening. Modeled deformational behavior attending stress paths relevant for several reservoir production scenarios are described.

02 PETROLEUM↗

Growth of Jupiter: Formation in disks of gas and solids and evolution to the present epoch

The formation of Jupiter is modeled via core-nucleated accretion, and the planet's evolution is simulated up to the present epoch. Throughout the phases when the planet acquires most of its heavy-element content, the calculation of solids' accretion accounts for interactions with an evolving disk of planetesimals. The phase of growth from an embryo of a few hundred kilometers in radius until the time when the accretion of gas overtakes solids' accretion was presented by D'Angelo et al., and the same numerical methods are applied here. Those calculations followed the formation for about 4 × 10 5 years, until the epoch when the heavy-element and hydrogen/helium masses were $M_Z ≈ 7.3$ and $M_{XY} ≈ 0.15$ Earth's masses ($M_⊕$), respectively, and $\dot {M} _{XY} ≈ $$\dot {M}_Z$. In this work, the calculation is continued through the phase when $M_{XY}$ grows to equal $M_Z$ , at which age, about 2.4 × 10 6 years, the total mass of the planet is $M_p ≈ 20 M_⊕$. About 9 × 10 5 years later, $M_p$ is approximately $60 M_⊕$ and $M_Z ≈ 16 M_⊕$, three-quarters of which are delivered by planetesimals larger than 10 km in radius. Around this epoch, the contraction of the envelope dictates gas accretion rates a few times 10 –3 $M_⊕$ per year, initiating the regime of disk-limited accretion, whereby the planet can accrete all the gas provided by the disk, and its evolution is therefore tied to disk's evolution. Growth is continued by constructing simplified models of protosolar accretion disks that evolve through viscous diffusion, winds, and accretion on the planet. Jupiter's formation ends after ≈ 3.4–4.2 Myr, depending on the applied disk viscosity parameter, when nebula gas disperses. The young Jupiter is 4.5–5.5 times as voluminous as it is presently and thousands of times as luminous, ~10 –5 $L_⊙$. The heavy-element mass is ≈ 20 $M_⊕$. The evolution proceeds through the cooling and contraction phase, in isolation except for solar irradiation. After 4570 Myr, the age of the solar system, radius and luminosity of the planet are within 10$\%$ of current values, accounting also for uncertainties in the power absorbed from the Sun. During formation, and soon thereafter, the planet exhibits features, e.g., luminosity and effective temperature, that may probe aspects of the latter stages of formation, if observable. These possibly distinctive features, however, seem to disappear within a few tens of Myr.

79 ASTRONOMY AND ASTROPHYSICS↗

A SWIFT SURVEY OF ACCRETION ONTO STELLAR-MASS BLACK HOLES

We present a systemic analysis of all of the stellar-mass black hole binaries (confirmed and candidate) observed by the Swift observatory up to 2010 June. The broad Swift bandpass enables a trace of disk evolution over an unprecedented range in flux and temperature. The final data sample consists of 476 X-ray spectra containing greater than 100 counts, in the 0.6-10 keV band. This is the largest sample of high-quality CCD spectra of accreting black holes published to date. In addition, strictly simultaneous data at optical/UV wavelengths are available for 255 (54%) of these observations. The data are modeled with a combination of an accretion disk and a hard spectral component. For the hard component we consider both a simple power-law model and a thermal Comptonization model. An accretion disk is detected at greater than the 5{sigma} confidence level in 61% of the observations. Light curves and color-color diagrams are constructed for each system. Hardness-luminosity and disk fraction-luminosity diagrams are constructed and are observed to be consistent with those typically observed by RXTE, noting the sensitivity below 2 keV provided by Swift. The observed spectra have an average luminosity of {approx}1% Eddington, though we are sensitive to accretion disks down to a luminosity of 10{sup -3} L{sub Edd}. Thus, this is also the largest sample of such cool accretion disks studied to date. The accretion disk temperature distribution displays two peaks consistent with the classical hard and soft spectral states, with a smaller number of disks distributed between these. The distribution of inner disk radii is observed to be continuous regardless of which model is used to fit the hard continua. There is no evidence for large-scale truncation of the accretion disk in the hard state (at least for L{sub x} {approx}> 10{sup -3} L{sub Edd}), with all of the accretion disks having radii {approx}< 40 R{sub g} . Plots of the accretion disk inner radius versus hardness ratio reveal the disk radius to be decreasing as the spectrum hardens, i.e., enters the hard state. This is in contrast to expectations from the standard disk truncation paradigm and points toward a contribution from spectral hardening. The availability of simultaneous X-ray and optical/UV data for a subset of observations facilitates a critical examination of the role of disk irradiation via a modified disk model with a variable emissivity profile (i.e., T(r){proportional_to}r {sup -p}). The broadband spectra (X-ray-optical/UV) reveal irradiation of the accretion disk to be an important effect at all luminosities sampled herein, i.e., p {approx}< 0.75 for luminosities {approx}> 10{sup -3} L{sub Edd}. The accretion disk is found to dominate the UV emission irrespective of the assumed hard spectral component. Overall, we find the broadband soft-state spectra to be consistent with an irradiated accretion disk plus a corona, but we are unable to make conclusive statements regarding the nature of the hard-state accretion flow (e.g., ADAF/corona versus jet). Finally, the Swift data reveal a relation between the flux emitted by the accretion disk and that emitted by the corona, for this sample of stellar-mass black holes, that is found to be in broad agreement with the observed disk-corona relationship in Seyfert galaxies, suggesting a scale invariant coupling between the accretion disk and the corona.

79 ASTRONOMY AND ASTROPHYSICS↗

Evaluation of the VIIRS BRDF, Albedo and NBAR Products Suite and an Assessment of Continuity with the Long Term MODIS Record

Bidirectional Reflectance Distribution Function (BRDF) model parameters, Albedo quantities, and Nadir BRDF Adjusted Reflectance (NBAR) products derived from the Visible Infrared Imaging Radiometer Suite (VIIRS), on the Suomi-NPP (National Polar-orbiting Partnership) satellite are evaluated with spatially representative in situ tower albedometer measurements and through comparison with the MODerate Resolution Imaging Spectroradiometer (MODIS) long term record. White Sky Albedo (WSA), Black Sky Albedo (BSA), NBAR, and Quality Assurance (QA) results show that the VIIRS algorithm is performing well at the global scale and is able to provide global data products comparable with the heritage MODIS (with an absolute bias of 0.0069 for shortwave broadband WSA) despite the spectral, angular, and effective spatial resolution differences between the two sensors. Both VIIRS and MODIS albedo are shown to agree well with in situ albedo measurements at a number of spatially representative sites. This evaluation provides confidence that the high quality, daily VIIRS BRDF model parameters, Albedo, and NBAR products will be able to provide the long term continuity required by the modeling and monitoring communities.

VIIRS↗

Project: Modeling Relativistic Electrons from Nuclear Explosions in the Magnetosphere [Slides]

We present a summary of the FY12 activities for DTRA-funded project 'Modeling Relativistic Electrons from Nuclear Explosions in the Magnetosphere'. We briefly review the outstanding scientific questions and discuss the work done in the last year to try to answer these questions. We then discuss the agenda for this Technical Meeting with the DTRA sponsors. In the last year, we have continued our efforts to understand artificial radiation belts from several different perspectives: (1) Continued development of Electron Source Model (ESM) and comparison to HANE test data; (2) Continued studies of relativistic electron scattering by waves in the natural radiation belts; (3) Began study of self-generated waves from the HANE electrons; and (4) Began modeling for the UCLA laser experiment.

45 MILITARY TECHNOLOGY, WEAPONRY, AND NATIONAL DEF↗

Phase field-volumetric lattice Boltzmann model of ion uptake in porous nuclear waste form materials under continuous flow

The flow field within the mesopores of sorbent particles plays a crucial role in radionuclide diffusion and ion uptake kinetics, thus, impacting the overall performance of porous nuclear waste form materials. To fundamentally understand the influence of microstructures and material properties on the radionuclide absorption and retention processes requires a coupled multi-physics model that considers the advection and diffusion within the flow field, the reaction at liquid-solid interfaces, and finally, the solid-state diffusion within a complex nanoporous medium. Here, this study employs the volumetric lattice Boltzmann method (VLBM) to accurately and efficiently calculate the steady state velocity field inside the mesopores of sorbent particles. The obtained velocity field is then utilized to calculate the advection of ions in the steady flow. A phase field (PF) model of ion uptake is used to describe the reaction occurring at the solid-liquid interface and diffusion inside the porous medium. The integrated PF-VLBM model is verified in terms of the mass conservation and numerical efficiency and validated qualitatively with experimental observation data. Then, it is applied to study the influence of thermodynamic and kinetic properties, as well as flow field conditions on the ion uptake kinetics. The numerical results demonstrate that the ion uptake kinetics in porous particles has three distinct stages, which is in agreement with the observations in continuous flow experiments. In the first stage, the kinetics is predominantly controlled by the flow field and ion diffusivity in the liquid phase. The kinetics in the second stage is primarily governed by ion diffusivity in the solid phase. In the third stage the system reaches a dynamic equilibrium with a net zero uptake flux at the interface. It is also found that porous structures significantly affect the efficiency and capacity of ion uptake. The simulation results can help to understand the physics behind the observed ion uptake kinetics in experiments and to facilitate the development of constitutive equations that can account for heterogeneous microstructures in engineering performance codes.

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗

Verification of New Floating Capabilities in FAST v8: Preprint

In the latest release of NREL's wind turbine aero-hydro-servo-elastic simulation software, FAST v8, several new capabilities and major changes were introduced. FAST has been significantly altered to improve the simulator's modularity and to include new functionalities in the form of modules in the FAST v8 framework. This paper is focused on the improvements made for the modeling of floating offshore wind systems. The most significant change was to the hydrodynamic load calculation algorithms, which are embedded in the HydroDyn module. HydroDyn is now capable of applying strip-theory (via an extension of Morison's equation) at the member level for user-defined geometries. Users may now use a strip-theory-only approach for applying the hydrodynamic loads, as well as the previous potential-flow (radiation/diffraction) approach and a hybrid combination of both methods (radiation/diffraction and the drag component of Morison's equation). Second-order hydrodynamic implementations in both the wave kinematics used by the strip-theory solution and the wave-excitation loads in the potential-flow solution were also added to HydroDyn. The new floating capabilities were verified through a direct code-to-code comparison. We conducted a series of simulations of the International Energy Agency Wind Task 30 Offshore Code Comparison Collaboration Continuation (OC4) floating semisubmersible model and compared the wind turbine response predicted by FAST v8, the corresponding FAST v7 results, and results from other participants in the OC4 project. We found good agreement between FAST v7 and FAST v8 when using the linear radiation/diffraction modeling approach. The strip-theory-based approach inherently differs from the radiation/diffraction approach used in FAST v7 and we identified and characterized the differences. Enabling the second-order effects significantly improved the agreement between FAST v8 and the other OC4 participants.

17 WIND ENERGY↗

RAP: Resource-aware Automated GPU Sharing for Multi-GPU Recommendation Model Training and Input Preprocessing

Ensuring high-quality recommendations for newly onboarded users requires the continuous retraining of Deep Learning Recommendation Models (DLRMs) with freshly generated data. To serve the online DLRM retraining, existing solutions use hundreds of CPU computing nodes designated for input preprocessing, causing significant power consumption that surpasses even the power usage of GPU trainers. To this end, we propose RAP, an end-to-end DLRM training framework that supports Resource-aware Automated GPU sharing for DLRM input Preprocessing and Training. The core idea of RAP is to accurately capture the remaining GPU computing resources during DLRM training for input preprocessing, achieving superior training efficiency without requiring additional resources. Specifically, RAP utilizes a co-running cost model to efficiently assess the costs of various input preprocessing operations, and it implements a resource-aware horizontal fusion technique that adaptively merges smaller kernels according to GPU availability, circumventing any interference with DLRM training. In addition, RAP leverages a heuristic searching algorithm that jointly optimizes both the input preprocessing graph mapping and the co-running schedule to maximize the end-to-end DLRM training throughput. The comprehensive evaluation shows that RAP achieves 78.3× speedup on average over CPU-based DLRM input preprocessing frameworks. In addition, the end-to-end training throughput of RAP is only 2.04% lower than the ideal case, which has no input preprocessing overhead.

Wang, Zheng↗