Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “approximate computing”

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 883 records · Page 49
All sourcesNASA NTRSDOE OSTI
NASA NTRSDate not supplied

Uncertainty Quantification using Deep Ensembles for Decision Making in Cyber-Physical-Human Systems

In this paper and its companion, Differential Equation Approximation Using Gradient-Boosted Quantile Regression, Robison et al., we examine an approach to quantifying model uncertainty with the aim of increasing the trustworthiness of computational models in human-machine interactions. In Differential Equation Approximation Using Gradient-Boosted Quantile Regression, we focus on gradient-boosted decision trees, while in this one, we give more details about deep ensembles. Uncertainty quantification is crucial for building trustworthy autonomous decision-making agents in human-machine teams. There are two types of uncertainties: aleatoric and epistemic. The former is related to the inherent stochasticity (noise) of the process, whereas the latter is associated with the lack of knowledge or representation capability of models, such as neural networks. By lack of knowledge, we mean the model’s inability to accurately predict outputs for all possible inputs. The aleatory uncertainty can be estimated fairly easily with, for example, filters, whereas epistemic uncertainty is challenging to compute. This paper uses deep ensembles to quantify both aleatory and epistemic uncertainty. It can act as an uncertainty-aware surrogate transition model for decision-making frameworks. "Uncertainty-aware" means that the surrogate transition model should make predictions along with confidence in those predictions. In the context of decision-making, the transition models are ordinary differential equations (ODEs). Since ODEs can be simulated to make one-step or multi-step predictions, a good surrogate model for them should perform reasonably well in both modes. In a multi-step approach, the trajectory sampling method TS∞ was used to propagate uncertainty over multiple steps. The cartpole dynamical system was selected to demonstrate the ability of deep ensembles as good surrogate transition models for decision-making frameworks. The deep ensembles modeled the dynamics of cartpole ODEs and made uncertainty-aware predictions in single-step and multi-step transition modes.

CPH systems↗
NASA NTRS1997

An X-Ray Analysis Database of Photoionization Cross Sections Including Variable Ionization

Results of research efforts in the following areas are discussed: review of the major theoretical and experimental data of subshell photoionization cross sections and ionization edges of atomic ions to assess the accuracy of the data, and to compile the most reliable of these data in our own database; detailed atomic physics calculations to complement the database for all ions of 17 cosmically abundant elements; reconciling the data from various sources and our own calculations; and fitting cross sections with functional approximations and incorporating these functions into a compact computer code.Also, efforts included adapting an ionization equilibrium code, tabulating results, and incorporating them into the overall program and testing the code (both ionization equilibrium and opacity codes) with existing observational data. The background and scientific applications of this work are discussed. Atomic physics cross section models and calculations are described. Calculation results are compared with available experimental data and other theoretical data. The functional approximations used for fitting cross sections are outlined and applications of the database are discussed.

Wang, Ping↗
NASA NTRS1988

Some numerical and physical aspects of unsteady Navier-Stokes computations over airfoils using dynamic meshes

An upwind-biased implicit approximate factorization algorithm is applied to several unsteady flows on dynamic meshes. The thin-layer form of the compressible Navier-Stokes equations is used to solve both laminar and turbulent flows over airfoils pitching about the quarter chord. Numerical aspects of the solutions are investigated, including grid and time step effects. Two methods for determining fluxes - flux-vector splitting and flux-difference splitting - are compared. Flux-difference splitting predicts results more accurately than flux-vector splitting on a coarse mesh, but both methods agree on a fine mesh. Physical aspects of the computations are also examined. An equilibrium turbulent boundary layer model computes generally better unsteady results in comparison with experiment than a nonequilibrium model for the transonic case analyzed. Also, the size and location of the primary shed vortex for an airfoil pitching up at a constant rate is calculated in good agreement with experiment for two pitch rates.

Rumsey, Christopher L.↗
NASA NTRS1993

Advancing-Front Algorithm For Delaunay Triangulation

Efficient algorithm performs Delaunay triangulation to generate unstructured grids for use in computing two-dimensional flows. Once grid generated, one can optionally call upon additional subalgorithm that removes diagonal lines from quadrilateral cells nearly rectangular. Resulting approximately rectangular grid reduces cost per iteration of flow-computing algorithm.

Merriam, Marshal L.↗
NASA NTRS1992

Numerical analysis of turbine blade tip treatments

Three-dimensional solutions of the Navier-Stokes equations for a turbine blade with a turning angle of 180 degrees have been computed, including blade tip treatments involving cavities. The geometry approximates a preliminary design for the GGOT (Generic Gas Oxidizer Turbine). The data presented here will be compared with experimental data to be obtained from a linear cascade using original GGOT blades. Results have been computed for a blade with 1 percent clearance, based on chord, and three different cavity sizes. All tests were conducted at a Reynolds number of 4 x 10 exp 7. The grid contains 39,440 points with 10 spanwise planes in the tip clearance region of 5.008E-04 m. Streamline plots and velocity vectors together with velocity divergence plots reveal the general flow behavior in the clearance region. Blade tip temperature calculations suggest placement of a cavity close to the upstream side of the blade tip for reduction of overall blade tip temperature. The solutions do not account for the relative motion between the endwall and the turbine blade. The solutions obtained are generally consistent with previous work done in this area,

Gopalaswamy, Nath S.↗
NASA NTRS1998

Shallow Water Optical Water Quality Buoy

This NASA grant was funded as a result of an unsolicited proposal submission to Kennedy Space Center. The proposal proposed the development and testing of a shallow water optical water quality buoy. The buoy is meant to work in shallow aquatic systems (ponds, rivers, lagoons, and semi-enclosed water areas where strong wind wave action is not a major environmental During the project period of three years, a demonstration of the buoy was conducted. The last demonstration during the project period was held in November, 1996 when the buoy was demonstrated as being totally operational with no tethered communications line. During the last year of the project the buoy was made to be solar operated by large gel cell batteries. Fund limitations did not permit the batteries in metal enclosures as hoped for higher wind conditions, however the system used to date has worked continuously for in- situ operation of over 18 months continuous deployment. The system needs to have maintenance and somewhat continuous operational attention since various components have limited lifetime ages. For example, within the last six months the onboard computer has had to be repaired as it did approximately 6 months after deployment. The spectrograph had to be repaired and costs for repairs was covered by KB Science since no ftmds were available for this purpose after the grant expired. Most recently the computer web page server failed and it is currently being repaired by KB Science. In addition, the cell phone operation is currently being ftmded by Dr. Bostater in order to maintain the system's operation. The above points need to be made to allow NASA to understand that like any sophisticated measuring system in a lab or in the field, necessary funding and maintenance is needed to insure the system's operational state and to obtain quality factor. The proposal stated that the project was based upon the integration of a proprietary and confidential sensor and probe design that was developed by KB Science and Engineering and is currently patented by KB Science. The buoy's purpose was to collected hyperspectral optical signatures for analysis and resulting estimation of water quality parameters such as chlorophyll-a, seston and dissolved organic matter (DOC). The ultimate goal of the project was to develop a buoy that would integrate a probe to measure upwelling light from a source and thus relate this backscattered light to water quality parameters.

Bostater, Charles↗
NASA NTRS2001

B-spline Method in Fluid Dynamics

B-spline functions are bases for piecewise polynomials that possess attractive properties for complex flow simulations : they have compact support, provide a straightforward handling of boundary conditions and grid nonuniformities, and yield numerical schemes with high resolving power, where the order of accuracy is a mere input parameter. This paper reviews the progress made on the development and application of B-spline numerical methods to computational fluid dynamics problems. Basic B-spline approximation properties is investigated, and their relationship with conventional numerical methods is reviewed. Some fundamental developments towards efficient complex geometry spline methods are covered, such as local interpolation methods, fast solution algorithms on cartesian grid, non-conformal block-structured discretization, formulation of spline bases of higher continuity over triangulation, and treatment of pressure oscillations in Navier-Stokes equations. Application of some of these techniques to the computation of viscous incompressible flows is presented.

Botella, Olivier↗
NASA NTRS1983

Nonlinear transformat

A technique for designing automatic flight controllers for aircraft which utilizes the transformation theory of nonlinear systems to linear systems is presently being developed at NASA Ames Research Center. A method is considered in which a given nonlinear is transformed to a controllable linear system in Brunovsky canonical form. A linear approximation is introduced to the nonlinear system called the modified tangent model. This model is easily computed. Constructing the transformation for this model enables the designer to find an approximate transformation for the nonlinear system.

Ford, H.↗
NASA NTRS1988

A comparison of experimental and computational results for 5 and 10 degree cones at high Mach numbers

Experiments were performed on 5 and 10 deg slender cones at a velocity of approximately 5 km/sec in the NASA-Ames ballistic ranges. The flowfields for the cones were computed using ideal-gas and chemical nonequilibrium-air parabolized Navier-Stokes codes. Experimentally determined drag coefficients and shock shapes are compared with the results of the computer codes. Both the flight-data analysis methods and the computational codes are examined to achieve the most meaningful comparison. Under the conditions of the experiments, skin-friction drag makes up approximately 50 percent of the total drag for the 5 deg cone and 30 percent of the total drag for the 10 deg cone. Computed drag coefficients of the 10 deg cone agree well with the experimental values; predictions fall below the experimental values for the 5 deg cone.

Strawa, A. W.↗
NASA NTRS1989

Optical Communication with Semiconductor Laser Diode

Theoretical and experimental performance limits of a free-space direct detection optical communication system were studied using a semiconductor laser diode as the optical transmitter and a silicon avalanche photodiode (APD) as the receiver photodetector. Optical systems using these components are under consideration as replacements for microwave satellite communication links. Optical pulse position modulation (PPM) was chosen as the signal format. An experimental system was constructed that used an aluminum gallium arsenide semiconductor laser diode as the transmitter and a silicon avalanche photodiode photodetector. The system used Q=4 PPM signaling at a source data rate of 25 megabits per second. The PPM signal format requires regeneration of PPM slot clock and word clock waveforms in the receiver. A nearly exact computational procedure was developed to compute receiver bit error rate without using the Gaussion approximation. A transition detector slot clock recovery system using a phase lock loop was developed and implemented. A novel word clock recovery system was also developed. It was found that the results of the nearly exact computational procedure agreed well with actual measurements of receiver performance. The receiver sensitivity achieved was the closest to the quantum limit yet reported for an optical communication system of this type.

Davidson, Frederic↗
NASA NTRS2005

Scattering Properties and Brightness Temperatures Associated with Solid Precipitation

In the past few years, early solid precipitation detection and retrieval algorithms have been developed and shown to be applicable for snowing clouds and blizzards. NOAA has an operational snow versus rain classifier based on AMSU-B observations. Solid precipitation retrieval algorithms reported in the literature over the past two years include those that rely on neural nets, statistics, or physical relationships. All of the algorithms require the use of millimeter-wave radiometer observations. The millimeter-wave frequencies are especially sensitive to the scattering and emission properties of frozen particles due to the ice particle refractive index. Passive radiometric channels respond to both the integrated particle mass throughout the volume and field of view, and to the amount, location, and size distribution of the frozen (and liquid) particles with the sensitivity varying for different frequencies and hydrometeor types. This investigation probes the sensitivity of scattering and absorption coefficients, and hence computed brightness temperatures, resulting from variations in solid precipitation cloud profiles. The first study compares the single scattering, absorption, and asymmetry parameters associated with snow particles in clouds. Several methodologies are used to convert the physical characteristics (e.g., shape, size distributions, ice-air-water ratios) of ice particles to electromagnetic properties (e.g., absorption, scattering, and asymmetry factors). These methodologies include: conversion to solid ice particles, homogeneous dielectric mixing, or discrete dipole approximation. Changes in the conversion methodology can produce computed brightness temperature differences greater than 50 Kelvin.

Skofronick-Jackson, Gail M.↗
NASA NTRS2013

Spectrally Consistent Scattering, Absorption, and Polarization Properties of Atmospheric Ice Crystals at Wavelengths from 0.2 to 100 um

A data library is developed containing the scattering, absorption, and polarization properties of ice particles in the spectral range from 0.2 to 100 microns. The properties are computed based on a combination of the Amsterdam discrete dipole approximation (ADDA), the T-matrix method, and the improved geometric optics method (IGOM). The electromagnetic edge effect is incorporated into the extinction and absorption efficiencies computed from the IGOM. A full set of single-scattering properties is provided by considering three-dimensional random orientations for 11 ice crystal habits: droxtals, prolate spheroids, oblate spheroids, solid and hollow columns, compact aggregates composed of eight solid columns, hexagonal plates, small spatial aggregates composed of 5 plates, large spatial aggregates composed of 10 plates, and solid and hollow bullet rosettes. The maximum dimension of each habit ranges from 2 to 10,000 microns in 189 discrete sizes. For each ice crystal habit, three surface roughness conditions (i.e., smooth, moderately roughened, and severely roughened) are considered to account for the surface texture of large particles in the IGOM applicable domain. The data library contains the extinction efficiency, single-scattering albedo, asymmetry parameter, six independent nonzero elements of the phase matrix (P11, P12, P22, P33, P43, and P44), particle projected area, and particle volume to provide the basic single-scattering properties for remote sensing applications and radiative transfer simulations involving ice clouds. Furthermore, a comparison of satellite observations and theoretical simulations for the polarization characteristics of ice clouds demonstrates that ice cloud optical models assuming severely roughened ice crystals significantly outperform their counterparts assuming smooth ice crystals.

Optical properties↗
NASA NTRS2019

High-Performance Computing Optimization for Aladyn – Adaptive Neural Network Molecular Dynamics Mini-Application

This report provides a description and performance evaluation of the optimization techniques for high performance computing (HPC) implementation of the open source Computational Materials mini-application Aladyn (https://github.com/nasa/aladyn). Aladyn is a basic molecular dynamics code written in FORTRAN 2003, which is designed to demonstrate the use of adaptive neural networks (ANNs) in atomistic simulations. The role of ANNs is to efficiently reproduce the very complex energy landscape resulting from the atomic interactions in materials with the accuracy of the more expensive quantum mechanics-based calculations. The ANN is trained on a large set of atomic structures calculated using the density functional theory (DFT) method. While achieving orders of magnitude faster computational performance than DFT, the ANN-based approach was still very computationally demanding compared to the conventional approach of using empirically fitted energy functions. After its initial development, Aladyn was evaluated and optimized by experts at the NASA Advanced Supercomputing (NAS) division to exploit modern supercomputer architectures. The code has been optimized for execution on multicore central processing units (CPUs), including Intel® Skylake microarchitecture, and on graphic accelerators, such as Nvidia® V100 graphic processing units (GPUs), using Open Multi-Processing (OpenMP) and Open Accelerators (OpenACC) programming interfaces. The optimization achieved a speedup of 4.7 times the baseline version on CPU performance and an additional 2.4 times on CPU+GPU performance. Atomistic computer simulations are a fundamental tool in materials research to model material properties form physics-based first principles. Atomic interaction, governed by Quantum Mechanics (QM) require sophisticated and highly computationally demanding mathematical models to calculate [1]. Classical methods use approximate functional forms, empirically fitted through a set of variable parameters to emulate atomic energies as direct functions of atomic coordinates [2]. While empirical potentials are computationally much simpler, allowing simulations of large-scale systems of up to a trillion (1012) atoms [3], they are substantially less accurate compared to quantum calculations and applicable only to very specific atomic configurations or predefined crystallographic phases. A recently suggested approach is to use heuristic machine learning methods [4], such as those based on Adaptive Neural Networks (ANNs) to predict atomic energies, after being trained on a sufficiently large database of QM-calculated structures [5,6]. This approach reduces significantly the computational complexity, allowing for simulations of orders of magnitude larger systems compared to QM-based methods without compromising accuracy. Still, compared to classical methods using empirical energy functions, ANN methods remain two- to three orders of magnitude more computationally demanding. Hence, the computational cost of simulations, together with the need for extensive training of ANNs, still makes the practical implementation of ANN-based methods quite challenging. The purpose of the Aladyn mini-application software [7], available as open source at https://github.com/nasa/aladyn, is to be a testbed for exploring possible optimization strategies to develop highly scalable parallel algorithms for ANN-based atomistic simulations. Aladyn is aimed at utilizing the architecture of the high-end modern highperformance computing (HPC) hardware based on multicore central processing units (CPUs) equipped with graphic processing unit (GPU) accelerators. Specifically, the goal is to optimize the performance on a single HPC compute node, before implementing scaling to multi-node parallelization using message passing interface (MPI). At the same time, the open source code of Aladyn can serve as a training model for students and professors in academia.

Yamakov, Vesselin I.↗
NASA NTRS1993

Combining qualitative and quantitative spatial and temporal information in a hierarchical structure: Approximate reasoning for plan execution monitoring

The problem of plan generation (PG) and the problem of plan execution monitoring (PEM), including updating, queries, and resource-bounded replanning, have different reasoning and representation requirements. PEM requires the integration of qualitative and quantitative information. PEM is the receiving of data about the world in which a plan or agent is executing. The problem is to quickly determine the relevance of the data, the consistency of the data with respect to the expected effects, and if execution should continue. Only spatial and temporal aspects of the plan are addressed for relevance in this work. Current temporal reasoning systems are deficient in computational aspects or expressiveness. This work presents a hybrid qualitative and quantitative system that is fully expressive in its assertion language while offering certain computational efficiencies. In order to proceed, methods incorporating approximate reasoning using hierarchies, notions of locality, constraint expansion, and absolute parameters need be used and are shown to be useful for the anytime nature of PEM.

Hoebel, Louis J.↗
NASA NTRS1983

Helicopter rotor loads using discretized matched asymptotic expansions

The numerical practicality of a matched asymptotic expansion approach for the computation of unsteady three dimensional airloads on a helicopter rotor was improved. This effort utilizes a discretized repesentation of the doublet strength distribution and helical streamlines to decrease the computational requirements of the original analysis. The continuous variation of the doublet strength was approximated by piecewise constant or piecewise quadratic distributions, and the helical trajectory of a fluid particle was approximated by connected straight line segments. As a direct result of these simplified representations the computational time required for the execution of a typical flight condition was reduced by an order of magnitude with respect to the requirements of the original analysis. Airloads which were computed using the discretized method for a two bladed model rotor and a full scale four bladed rotor are in close agreement with measured results and airloads from the original asymptotic analysis. For conditions characterized by significant rotor/wake interaction the piecewise constant representation requires a reduced azimuth spacing to maintain acceptable accuracy.

Pierce, G. A.↗
NASA NTRS1986

Unified control/structure design and modeling research

To demonstrate the applicability of the control theory for distributed systems to large flexible space structures, research was focused on a model of a space antenna which consists of a rigid hub, flexible ribs, and a mesh reflecting surface. The space antenna model used is discussed along with the finite element approximation of the distributed model. The basic control problem is to design an optimal or near-optimal compensator to suppress the linear vibrations and rigid-body displacements of the structure. The application of an infinite dimensional Linear Quadratic Gaussian (LQG) control theory to flexible structure is discussed. Two basic approaches for robustness enhancement were investigated: loop transfer recovery and sensitivity optimization. A third approach synthesized from elements of these two basic approaches is currently under development. The control driven finite element approximation of flexible structures is discussed. Three sets of finite element basic vectors for computing functional control gains are compared. The possibility of constructing a finite element scheme to approximate the infinite dimensional Hamiltonian system directly, instead of indirectly is discussed.

Mingori, D. L.↗
NASA NTRS1988

Gaussian approximation versus nearly exact performance analysis of optical communication systems with PPM signaling and APD receivers

A 25-Mb/s direct-detection optical communication system that used Q = 4 PPM (pulse-position modulation) signaling was constructed and its performance measured under laboratory conditions. The system used a single-mode AlGaAs laser diode (lambda = 834 nm) and low-noise silicon avalanche photodiode (APD). A procedure for numerically computing system performance is given that uses the nearly exact Webb approximation of the true Conradi distribution for the APD output and does not require excessive amounts of computer time. Comparison revealed that modeling the APD ouptut as a Gaussian process under conditions of negligible background radiation and low (less than 10 to the -12 A) APD bulk leakage currents leads to substantial underestimates of optimal APD gain and overestimates of system bit error probability. Examples illustrate the breakdown of the Gaussian approximation in assessing system performance. The measured performance was in excellent agreement with that predicted by the computational procedure. The system achieved a bit error probability of 10 to the -6 at a received signal energy, corresponding to an average of 60 absorbed photons/bit and optimal APD gain of 700.

Davidson, Frederic M.↗