Search NASA⌕ Search

SEARCH · Search NASA

Results for “partitioned algorithm”

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 451 records · Page 25

An n -material thresholding method for improving integerness of solutions in topology optimization

It is common in solving topology optimization problems to replace an integer-valued characteristic function design field with the material volume fraction field, a real-valued approximation of the design field that permits "fictitious" mixtures of materials during intermediate iterations in the optimization process. This is reasonable so long as one can interpolate properties for such materials and so long as the final design is integer valued. For this purpose, we present a method for smoothly thresholding the volume fractions of an arbitrary number of material phases which specify the design. This method is trivial for two-material design problems, for example, the canonical topology design problem of specifying the presence or absence of a single material within a domain, but it becomes more complex when three or more materials are used, as often occurs in material design problems. We take advantage of the similarity in properties between the volume fractions and the barycentric coordinates on a simplex to derive a thresholding, method which is applicable to an arbitrary number of materials. As we show in a sensitivity analysis, this method has smooth derivatives, allowing it to be used in gradient-based optimization algorithms. Finally, we present results, which show synergistic effects when used with Solid Isotropic Material with Penalty and Rational Approximation of Material Properties material interpolation functions, popular methods of ensuring integerness of solutions.

42 ENGINEERING↗

Adaptive time stepping for fluid-structure interaction solvers

In this work, a novel adaptive time stepping scheme for fluid-structure interaction (FSI) problems is proposed that allows for controlling the accuracy of the time-discrete solution. Furthermore, it eases practical computations by providing an efficient and very robust time step size selection. This has proven to be very useful, especially when addressing new physical problems, where no educated guess for an appropriate time step size is available. The fluid and the structure field, but also the fluid-structure interface are taken into account for the purpose of a posteriori error estimation, rendering it easy to implement and only adding negligible additional cost. The adaptive time stepping scheme is incorporated into a monolithic solution framework, but can straightforwardly be applied to partitioned solvers as well. The basic idea can be extended to the coupling of an arbitrary number of physical models. Accuracy and efficiency of the proposed method are studied in a variety of numerical examples ranging from academic benchmark tests to complex biomedical applications like the pulsatile blood flow through an abdominal aortic aneurysm. Finally, the demonstrated accuracy of the time-discrete solution in combination with reduced computational cost make this algorithm very appealing in all kinds of FSI applications.

97 MATHEMATICS AND COMPUTING↗

A Massively Parallel Adaptive Fast-Multipole Method on Heterogeneous Architectures

We present new scalable algorithms and an implementation of the kernel-independent fast multiple method (KIFMM), employing hybrid distributed memory message passing (via MPI) and shared memory/streaming using graphics processing unit (GPU) acceleration to rapidly evaluate two-body non-oscillatory potentials. On traditional CPU-only systems, our implementation scales well up to 30 billion unknowns on 65k cores (AMD/CRAY-based Kraken system at NSF/NICS) on tree data structures with 25 levels between leaves. On GPU-enabled systems, we achieve 30 x speedup for problems of up to 256 million points on 256 GPUs (Lincoln at NSF/NCSA) over a comparable CPU-only implementation. Both of these demonstrations represent the largest and fastest of their kind of which we are aware. We achieve scalability at extreme core counts by extending the initial work of Ying et al. (ACM/IEEE SC 03) with a new approach to scalable MPI-based tree construction and partitioning. For the sub-components of KIFMM, which direct- and approximate-interactions, target evaluation, and source-to-multipole translations, we use CUDA-based GPU-acceleration to achieve excellent performance. To do so requires carefully constructed data structure transformations, which we describe, and whose cost we show is minor. Taken together, these components show promise for ultrascalable FMM in the petascale era and beyond.

99 GENERAL AND MISCELLANEOUS↗

A learning-based measurement framework for traffic matrix inference in software defined networks

Here, we propose an intelligent framework for Traffic Matrix (TM) inference in Software Defined Networks (SDN) where the Ternary Content Addressable Memory (TCAM) entries of switches are partitioned into two parts to: (1) effectively aggregate part of incoming flows for aggregate measurements, and (2) de-aggregate and directly measure the most informative flows for per-flow measurements. These measurements are then processed to effectively estimate the size of network flows. Under hard resource constraints of limited TCAM sizes, we show how to design the optimal and efficient-compressed flow aggregation matrices. We propose an optimal Multi-Armed Bandit (MAB) based algorithm to adaptively measure the most rewarding flows. We evaluate the performance of our framework using real traffic traces from different network environments and by considering two main applications: TM estimation and Heavy Hitter (HH) detection. Moreover, we have implemented a prototype of our framework in Mininet to demonstrate its effectiveness.

42 ENGINEERING↗

Real-time streamflow forecasting: AI vs. Hydrologic insights

In this paper, we propose a set of simple benchmarks for the evaluation of data-based models for real-time streamflow forecasting, such as those developed with sophisticated Artificial Intelligence (AI) algorithms. The benchmarks are also data-based and provide context to judge incremental improvements in the performance metrics from the more complicated approaches. The benchmarks include temporal and spatial persistence, persistence corrected for baseflow and streamflow, as well as river distance weighted runoff obtained from space-time distributed rainfall. In the development of the benchmarks, we use basic hydrologic insights such as flow aggregation by the river network, scale-dependence in basin response, streamflow partitioning into quick flow and baseflow, water travel time, and rainfall averaging by the basin width function. The study uses 140 streamflow gauges in Iowa that cover a range of basin scales between 7 and 37,000 km 2 . The data cover 17 years. This work demonstrates that the proposed benchmarks can provide good performance according to several commonly used metrics. For example, streamflow forecasting at half of the test locations across years achieves a Kling-Gupta Efficiency (KGE) score of 0.6 or higher at one-day ahead lead time, and 20% of cases reach the KGE of 0.8 or higher. The proposed benchmarks are easy to implement and should prove useful for developers of data-based as well as physics-based hydrologic models and real-time data assimilation techniques.

54 ENVIRONMENTAL SCIENCES↗

Investigation of particle and vapor wall-loss effects on controlled wood-smoke smog-chamber experiments

Here, smog chambers are extensively used to study processes that drive gas and particle evolution in the atmosphere. A limitation of these experiments is that particles and gas-phase species may be lost to chamber walls on shorter timescales than the timescales of the atmospheric processes being studied in the chamber experiments. These particle and vapor wall losses have been investigated in recent studies of secondary organic aerosol (SOA) formation, but they have not been systematically investigated in experiments of primary emissions from combustion. The semi-volatile nature of combustion emissions (e.g. from wood smoke) may complicate the behavior of particle and vapor wall deposition in the chamber over the course of the experiments due to the competition between gas/particle and gas/wall partitioning. Losses of vapors to the walls may impact particle evaporation in these experiments, and potential precursors for SOA formation from combustion may be lost to the walls, causing underestimations of aerosol yields. Here, we conduct simulations to determine how particle and gas-phase wall losses contributed to the observed evolution of the aerosol during experiments in the third Fire Lab At Missoula Experiment (FLAME III). We use the TwO-Moment Aerosol Sectional (TOMAS) microphysics algorithm coupled with the organic volatility basis set (VBS) and wall-loss formulations to examine the predicted extent of particle and vapor wall losses. We limit the scope of our study to the dark periods in the chamber before photo-oxidation to simplify the aerosol system for this initial study. Our model simulations suggest that over one-third of the initial particle-phase organic mass (41 %) was lost during the experiments, and over half of this particle-organic mass loss was from direct particle wall loss (65 % of the loss) with the remainder from evaporation of the particles driven by vapor losses to the walls (35 % of the loss). We perform a series of sensitivity tests to understand uncertainties in our simulations. Uncertainty in the initial wood-smoke volatility distribution contributes 18 % uncertainty to the final particle-organic mass remaining in the chamber (relative to base-assumption simulation). We show that the total mass loss may depend on the effective saturation concentration of vapor with respect to the walls as these values currently vary widely in the literature. The details of smoke dilution during the filling of smog chambers may influence the mass loss to the walls, and a dilution of ~ 25:1 during the experiments increased particle-organic mass loss by 33 % compared to a simulation where we assume the particles and vapors are initially in equilibrium in the chamber. Finally, we discuss how our findings may influence interpretations of emission factors and SOA production in wood-smoke smog-chamber experiments.

54 ENVIRONMENTAL SCIENCES↗

The all configuration mean energy multiconfiguration self-consistent-field method. I. Equal configuration weights

The All Configuration Mean Energy (ACME) conditions are a special case of state averaging for Multiconfigurational Self-Consistent-Field (MCSCF) orbital optimisation. The method is formulated using the Graphical Unitary Group Approach (GUGA) in which the Configuration State Function (CSF) basis is represented as walks within a Shavitt graph. This graphical formulation leads to efficient recursive algorithms for the energy and reduced density matrices (RDM) that are independent of the CSF dimension and that scale only as O(n2) where n is the number of occupied orbitals. The Hamiltonian matrix diagonalization step is obviated and the CSF expansion coefficients are neither referenced nor required. This allows MCSCF orbital optimisation to be performed for essentially unlimited numbers of active orbitals and arbitrarily large CSF expansions. The discussion includes various types of CSF expansion spaces, the partitioning of the essential and redundant orbital optimisation parameters, the computation of the spin-density, and the formulation of state-specific analytic gradients and nonadiabatic coupling for high-level electronic structure methods that use the ACME MCSCF orbitals.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Enhancing scalability of a matrix-free eigensolver for studying many-body localization

We propose several techniques to enhance the parallel scalability of a matrix-free eigensolver designed for studying many-body localization (MBL) of quantum spin chain models with nearest-neighbor interactions and on-site disorder. This type of problem is computationally challenging because the dimension of the associated Hamiltonian matrix grows exponentially with respect to the number of spins L, and we need to average over different realizations of the random disorder to obtain relevant statistical behavior. For each disorder realization, we need to compute eigenvalues from different regions of the spectrum and their corresponding eigenvectors. In previous work, the interior eigenstates for a single eigenvalue problem are computed via the shift-and-invert Lanczos algorithm. Due to the extremely high memory footprint of the LU factorizations, this technique is not well suited for large L’s. For example, we need thousands of compute nodes on modern high performance computing infrastructures to go beyond L = 24. The matrix-free approach does not suffer from this memory bottleneck, however, its scalability is limited by a computation and communication load imbalance. To reduce this imbalance and to significantly enhance the scalability of the matrix-free eigensolver, we reorder the matrix and leverage the consistent space runtime, CSPACER. We also show its efficiency in managing irregular communication patterns at scale compared to optimized MPI non-blocking two-sided and one-sided RMA implementation variants. This effort enables us to study MBL for spin chains with a larger number of spins. The efficiency and effectiveness of the proposed algorithm is demonstrated by computing eigenstates on a massively parallel many-core high performance computer.

METIS↗

Intelligent Partitioning based Fully Parallel AC Security-Constrained Optimal Power Flow

Today’s power grid is becoming more diverse and integrated with high-level distributed energy resources and smart control technologies that is creating a new set of grid management challenges in terms of large-scale, nonlinear, and non-convex problem modeling, complex and time-consuming computation, as well as difficult uncertainty handling. This project focused on solving a challenging multi-period security-constrained generation scheduling problem, which is of great importance for maximizing the social welfare of real-time dispatch, day-ahead market, as well as weekly planning of power systems. Our developed software explored parallel optimization algorithms for complex and realistic power system models, and develop fast, efficient, and robust grid optimization solutions on the high-performance computing platform that will enable increased grid economics, flexibility, resilience, as well as energy security in the United States.

24 POWER TRANSMISSION AND DISTRIBUTION↗

Ab Initio Direct Dynamics

The reactivity and dynamics of molecular systems can be explored computationally by classical trajectory calculations. The traditional approach involves fitting a functional form of a potential energy surface (PES) to the energies from a large number of electronic structure calculations and then integrating numerous trajectories on this fitted PES to model the molecular dynamics. The ever-decreasing cost of computing and continuing advances in computational chemistry software have made it possible to use electronic structure calculations directly in molecular dynamics simulations without first having to construct a fitted PES. In this “on-the-fly” approach, every time the energy and its derivatives are needed for the integration of the equations of motion, they are obtained directly from quantum chemical calculations. This approach started to become practical in the mid-1990s as a result of increased availability of inexpensive computer resources and improved computational chemistry software. The application of direct dynamics calculations has grown rapidly over the last 25 years and would require a lengthy review article. The present Account is limited to some of our contributions to methods development and various applications. To improve the efficiency of direct dynamics calculations, we developed a Hessian-based predictor-corrector algorithm for integrating classical trajectories. Hessian updating made this even more efficient. Furthermore, this approach was also used to improve algorithms for following the steepest descent reaction paths. For larger molecular systems, we developed an extended Lagrangian approach in which the electronic structure is propagated along with the molecular structure. Strong field chemistry is a rapidly growing area, and to improve the accuracy of molecular dynamics in intense laser fields, we included the time-varying electric field in a novel predictor-corrector trajectory integration algorithm. Since intense laser fields can excite and ionize molecules, we extended our studies to include electron dynamics. Specifically, we developed code for time-dependent configuration interaction electron dynamics to simulate strong field ionization by intense laser pulses. Our initial application of ab initio direct dynamics in 1994 was to CH 2 O → H 2 + CO; the calculated vibrational distributions in the products were in very good agreement with experiment. In the intervening years, we have used direct dynamics to explore energy partitioning in various dissociation reactions, unimolecular dissociations yielding three fragments, reactions with branching after the transition state, nonstatistical dynamics of chemically activated molecules, dynamics of molecular fragmentation by intense infrared laser pulses, selective activation of specific dissociation channels by aligned intense infrared laser fields, angular dependence of strong field ionization, and simulation of sequential double ionization.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

An asynchronous traversal engine for graph-based rich metadata management

Rich metadata in high-performance computing (HPC) systems contains extended information about users, jobs, data files, and their relationships. Property graphs are a promising data model to represent heterogeneous rich metadata flexibly. Specifically, a property graph can use vertices to represent different entities and edges to record the relationships between vertices with unique annotations. The high-volume HPC use case, with millions of entities and relationships, naturally requires an out-of-core distributed property graph database, which must support live updates (to ingest production information in real time), low-latency point queries (for frequent metadata operations such as permission checking), and large-scale traversals (for provenance data mining). Among these needs, large-scale property graph traversals are particularly challenging for distributed graph storage systems. Most existing graph systems implement a "level synchronous" breadth-first search algorithm that relies on global synchronization in each traversal step. This performs well in many problem domains; but a rich metadata management system is characterized by imbalanced graphs, long traversal lengths, and concurrent workloads, each of which has the potential to introduce or exacerbate stragglers (i.e., abnormally slow steps or servers in a graph traversal) that lead to low overall throughput for synchronous traversal algorithms. Previous research indicated that the straggler problem can be mitigated by using asynchronous traversal algorithms, and many graph-processing frameworks have successfully demonstrated this approach. Such systems require the graph to be loaded into a separate batch-processing framework instead of being iteratively accessed, however. In this work, we investigate a general asynchronous graph traversal engine that can operate atop a rich metadata graph in its native format. We outline a traversal-aware query language and key optimizations (traversal-affiliate caching and execution merging) necessary for efficient performance. We further explore the effect of different graph partitioning strategies on the traversal performance for both synchronous and asynchronous traversal engines. Our experiments show that the asynchronous graph traversal engine is more efficient than its synchronous counterpart in the case of HPC rich metadata processing, where more servers are involved and larger traversals are needed. Furthermore, the asynchronous traversal engine is more adaptive to different graph partitioning strategies.

96 KNOWLEDGE MANAGEMENT AND PRESERVATION↗

Risk of Distant Metastasis in Parathyroid Carcinoma and Its Effect on Survival: A Retrospective Review from a High-Volume Center

Background: Development of distant metastases (DM) is associated with markedly decreased survival in parathyroid carcinoma (PC). We sought to identify factors associated with development of DM and to quantify the effect that development of DM had on overall survival (OS). Methods: Patients with surgically resected local/regional PC treated or surveilled at a tertiary-referral cancer hospital from 1980 to 2017 were included. We assessed the association between biochemical and clinicopathologic factors (preoperative parathyroid hormone (PTH) levels, sex, race, age, preoperative serum calcium levels, serum calcium levels at 6 months postop, tumor size, and extent of resection) with the development of DM. We also assessed the effect of development of DM on OS. Results: Seventy-five patients with PC were assessed; 17 (22.7%) developed DM at a median follow-up of 77 months. The cumulative incidence of DM in the cohort was 20, 30, and 38% at 5, 10, and 20 years respectively. Tumor size > 3.2 cm based on recursive partitioning analysis was the only significant predictor for development of DM (hazard ratio (HR) = 3.51; 95% confidence interval [CI] 1.04–11.91; p = 0.04). Median OS for the entire cohort was 17 years compared with 40 months for the cohort who developed DM. The HR for death after distant metastasis was 9.6 (95% CI 4.2–22.3; p < 0.0001). Conclusions: Development of distant metastasis during surveillance is associated with decreased OS, including late recurrences. Primary tumor size should be considered in future interval surveillance and development of treatment algorithms.

62 RADIOLOGY AND NUCLEAR MEDICINE↗

Information content of spatially distributed ground-based measurements for hydrologic-parameter calibration in mixed rain-snow mountain headwaters

Parameters in hydrologic models used in mixed rain-snow regions are often uncertain to calibrate and overfitted on streamflow. To contribute addressing these challenges, we used an algorithm that assesses modeling performances through time (Dynamic Identifiability Analysis) to quantify the information content of spatially distributed ground-based measurements for identifying optimal parameter values in the Precipitation Runoff Modeling System (PRMS) model. Including spatially distributed ground-based measurements in Identifiability Analysis allowed us to unambiguously estimate more parameter values than only using streamflow (seven parameters instead of two out of a pool of thirty-three). Peaks in information gain were obtained when using dew-point temperature to identify precipitation phase-partitioning parameters. Multi-attribute identifiability analysis also yielded optimal parameter values that were temporally less variable than those estimated using streamflow alone. Overall, identifying parameter values using ground-based measurements improved the simulation of key drivers of the surface-water budget, such as air temperature and precipitation-phase partitioning. However, parameters simulating surface-to-subsurface mass fluxes like snow accumulation and melt or evapotranspiration were poorly identified by any attribute and so emerged as key sources of predictive uncertainty for this distributed-parameter hydrologic model. In conclusion, this work demonstrates the value of expanded ground-based measurements for identifying parameters in distributed-parameter hydrologic models and so diagnosing their conceptual uncertainty across the water budget.

54 ENVIRONMENTAL SCIENCES↗

OPER: Optimality-Guided Embedding Table Parallelization for Large-scale Recommendation Model

With the sharp increasing volume of user data, Deep Learning Recommendation Model (DLRM) becomes an indispensable infrastructure in large technology companies. However, large-scale DLRM on the multi-GPU platform is still inefficient due to unbalanced workload partitioning and intensive inter-GPU communication. To this end, we propose OPER, an OPtimality guided Embedding table placement for large-scale Recommendation model training and inference. OPER explores the potential of mitigating remote memory access latency in DLRM through fine-grained embedding table placement. Specifically, OPER proposes a theoretical modeling that builds up the relationship between EMT placement and the embedding communication latency in both training and inference. OPER proves the NP hardness of finding the optimal embedding table placement and proposes a heuristic algorithm that yields near optimal placement. OPER implements a SHMEM-based embedding table training system and a unified embedding index mapping to support fine-grained embedding table sharding and placement. Comprehensive experiments reveal that OPER achieves on average 3.4× and 5.1× speedup on training and inference respectively over state-of-the-art DLRM frameworks.

Wang, Zheng↗

Dynamic load balance scheme for the DSMC algorithm

The direct simulation Monte Carlo (DSMC) algorithm, devised by Bird, has been used over a wide range of various rarified flow problems in the past 40 years. While the DSMC is suitable for the parallel implementation on powerful multi-processor architecture, it also introduces a large load imbalance across the processor array, even for small examples. The load imposed on a processor by a DSMC calculation is determined to a large extent by the total of simulator particles upon it. Since most flows are impulsively started with initial distribution of particles which is surely quite different from the steady state, the total of simulator particles will change dramatically. The load balance based upon an initial distribution of particles will break down as the steady state of flow is reached. The load imbalance and huge computational cost of DSMC has limited its application to rarefied or simple transitional flows. In this paper, by taking advantage of METIS, a software for partitioning unstructured graphs, and taking the total of simulator particles in each cell as a weight information, the repartitioning based upon the principle that each processor handles approximately the equal total of simulator particles has been achieved. The computation must pause several times to renew the total of simulator particles in each processor and repartition the whole domain again. Thus the load balance across the processors array holds in the duration of computation. The parallel efficiency can be improved effectively. The benchmark solution of a cylinder submerged in hypersonic flow has been simulated numerically. Besides, hypersonic flow past around a complex wing-body configuration has also been simulated. The results have displayed that, for both of cases, the computational time can be reduced by about 50%.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Automatic Regionalization Algorithm for Distributed State Estimation in Power Systems

The deregulation of the power system and the incorporation of generation from renewable energy sources recessitates faster state estimation in the smart grid. Distributed state estimation (DSE) has become a promising and scalable solution to this urgent demand. In this paper, we investigate the regionalization algorithms for the power system, a necessary step before distributed state estimation can be performed. To the best of the authors' knowledge, this is the first investigation on automatic regionalization (AR). We propose three spectral clustering based AR algorithms. Simulations show that our proposed algorithms outperform the two investigated manual regionalization cases. With the help of AR algorithms, we also show how the number of regions impacts the accuracy and convergence speed of the DSE and conclude that the number of regions needs to be chosen carefully to improve the convergence speed of DSEs.

24 POWER TRANSMISSION AND DISTRIBUTION↗

Flexible classification with spatial quantile comparison and novel statistical features applied to spent nuclear fuel analysis

Multivariate classification algorithms are a common tool in contemporary data science that have begun to be applied to nuclear forensic analyses. Using a multivariate generalization of quantile–quantile plots for comparing unknown statistical distributions, a new model-free multivariate classifier called the Quantile–Quantile Comparator has been developed and tested on the analysis of simulated irradiated nuclear fuel. A method to estimate the sample-specific probability of having made an incorrect classification decision by analyzing the distribution of quantile comparisons has been developed, and methods have been developed to analyze the partition of the known data library used to analyze unknown data. By analyzing this partition, the analyst may identify classes that are likely to be confused, identify test data that are inconsistent with the data library and return a “None-of-the-Above” classification decision, and optimize the class boundaries. In this paper, the development of these capabilities is described, and a case study using real experimental assay data collected from nuclear fuel extracted from the newly-released In conclusion, Spent Fuel Isotopic Composition Database to demonstrate the efficacy of these capabilities is presented.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

Development of formalisms based on locally coupled open subsystems for calculations in molecular electronic structure and dynamics

Quantum embedding theories model a collection of interacting molecules as a set of subsystems, where each can be treated with a particular electronic structure method (wave function or density functional theory, for example); these theories can lead to computationally efficient and accurate algorithms. Motivated by challenges in the field, we previously described a formalism (that models two subsystems), which we call “locally coupled open subsystems” (LCOS), for the computation of ground-state energies, fractional electron-occupation numbers of the subsystems, and the size-consistent limit of subsystem dissociation. In this work we present the full (nonrelativistic) LCOS theory and the following extensions of our previous work: (i) the framework to study systems composed of multiple subsystems and a procedure to spin-adapt the auxiliary wave function that describes the partitioned system, so that its spin state matches that of the real system of interest; (ii) potential functionals and ideas to employ machine learning for the computation of ground-state densities and energies; (iii) formulation of two LCOS ground-state approaches where the fragments are assigned Kohn-Sham wave functions; and (iv) a time-dependent (TD) extension of these two ground-state formalisms in which the state of the subsystems evolves according to a unitary propagation; from this evolution we can extract TD electron populations of the fragments, for instance. We also discuss potential applications of the TD LCOS theory to linear photoabsorption and Raman spectroscopy. As a result, the developments presented in this work can lead to ground-state and TD electronic structure calculations where the computational scaling can be controlled, depending on the level of theory and the accuracy desired to model each one of the subsystems and their coupling.

74 ATOMIC AND MOLECULAR PHYSICS↗