Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “clustering 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 559 records · Page 31
All sourcesNASA NTRSDOE OSTI
DOE OSTI2020

Scalable molecular dynamics on CPU and GPU architectures with NAMD

NAMD is a molecular dynamics program designed for high-performance simulations of very large biological objects on CPU- and GPU-based architectures. NAMD offers scalable performance on petascale parallel supercomputers consisting of hundreds of thousands of cores, as well as on inexpensive commodity clusters commonly found in academic environments. It is written in C++ and leans on Charm++ parallel objects for optimal performance on low-latency architectures. NAMD is a versatile, multipurpose code that gathers state-of-the-art algorithms to carry out simulations in apt thermodynamic ensembles, using the widely popular CHARMM, AMBER, OPLS, and GROMOS biomolecular force fields. Here, we review the main features of NAMD that allow both equilibrium and enhanced-sampling molecular dynamics simulations with numerical efficiency. We describe the underlying concepts utilized by NAMD and their implementation, most notably for handling long-range electrostatics; controlling the temperature, pressure, and pH; applying external potentials on tailored grids; leveraging massively parallel resources in multiple-copy simulations; and hybrid quantum-mechanical/molecular-mechanical descriptions. We detail the variety of options offered by NAMD for enhanced-sampling simulations aimed at determining free-energy differences of either alchemical or geometrical transformations and outline their applicability to specific problems. Last, we discuss the roadmap for the development of NAMD and our current efforts toward achieving optimal performance on GPU-based architectures, for pushing back the limitations that have prevented biologically realistic billion-atom objects to be fruitfully simulated, and for making large-scale simulations less expensive and easier to set up, run, and analyze. NAMD is distributed free of charge with its source code at www.ks.uiuc.edu.

high-performance computing↗
DOE OSTI2022

CMB/kSZ and Compton-y Maps from 2500 deg 2 of SPT-SZ and Planck Survey Data

Abstract We present component-separated maps of the primary cosmic microwave background/kinematic Sunyaev–Zel’dovich (SZ) amplitude and the thermal SZ Compton- y parameter, created using data from the South Pole Telescope (SPT) and the Planck satellite. These maps, which cover the ∼2500 deg 2 of the southern sky imaged by the SPT-SZ survey, represent a significant improvement over previous such products available in this region by virtue of their higher angular resolution ( 1 .′ 25 for our highest-resolution Compton- y maps) and lower noise at small angular scales. In this work we detail the construction of these maps using linear combination techniques, including our method for limiting the correlation of our lowest-noise Compton- y map products with the cosmic infrared background. We perform a range of validation tests on these data products to test our sky modeling and combination algorithms, and we find good performance in all of these tests. Recognizing the potential utility of these data products for a wide range of astrophysical and cosmological analyses, including studies of the gas properties of galaxies, groups, and clusters, we make these products publicly available at http://pole.uchicago.edu/public/data/sptsz_ymap and on the NASA/LAMBDA website.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2022

Diffusion Monte Carlo approaches for studying nuclear quantum effects in fluxional molecules

Abstract Diffusion quantum Monte Carlo (DMC) provides a powerful approach for obtaining the ground state energy and wave function of molecules, ions, and molecular clusters. The approach is uniquely well suited for studies of fluxional molecules, which undergo large amplitude vibrational motions even in their ground state. In contrast to the electronic structure problem, where the wave function must be antisymmetric with respect to exchange of any pair of electrons, the wave function for the ground vibrational state is nodeless. This greatly simplifies the application of DMC for vibrational problems. Because there is not a single potential function that can be used to describe the intramolecular and intermolecular interactions in all molecular systems, most methods that are used to describe nuclear quantum effects rely on a carefully chosen zero‐order description of the molecular vibrations. In contrast, DMC calculations can be performed in Cartesian coordinates, making the DMC algorithm easily transferable between different chemical systems. In this contribution, the theory that underlies DMC will be discussed along with important considerations for performing DMC calculations. Extensions for evaluating vibrationally excited states and molecular properties are also discussed. Insights that can be obtained from DMC calculations are illustrated in the context of the protonated water clusters. This article is categorized under: Molecular and Statistical Mechanics > Molecular Dynamics and Monte‐Carlo Methods Theoretical and Physical Chemistry > Spectroscopy

Chemistry↗
DOE OSTI2025

Personalized Tucker Decomposition: Modeling Commonality and Peculiarity on Tensor Data

In this paper, we propose a personalized Tucker decomposition (perTucker) to address the limitations of traditional tensor decomposition methods in capturing heterogeneity across different datasets. perTucker decomposes tensor data into shared global components and personalized local components. We introduce an order orthogonality assumption and develop a proximal gradient regularized block coordinate descent algorithm guaranteed to converge to a stationary point. The unique and common representations learned by perTucker reveal intrinsic statistical patterns in data and provide valuable information for a wide range of downstream analytics, including anomaly detection, source classification, and clustering. We demonstrate perTucker’s effectiveness through a simulation study and two case studies on solar flare detection and tonnage signal classification.

14 SOLAR ENERGY↗
DOE OSTI2022

Probing the Galactic halo with RR Lyrae stars – II. The substructures of the Milky Way

ABSTRACT We identify substructures of the Galactic halo using 3003 type ab RR Lyrae stars with six-dimensional position–velocity information from the Sloan Digital Sky Survey, the Large Sky Area Multi-Object Fiber Spectroscopic Telescope and the Gaia Early Data Release 3. Based on this information, we define the separation of any two of the stars in the integrals of motion space and we identify substructures by utilizing the friends-of-friends algorithm. We identify members belonging to several known substructures: the Sagittarius stream, the Gaia-Enceladus-Sausage (GES), the Sequoia and the Helmi streams. In addition to these known substructures, there are three other substructures possibly associated with globular clusters NGC 5272, 6656 and 5024, respectively. Finally, we also find three remaining unknown substructures, one of which has large angular momentum and a mean metallicity −2.13 dex, which may be a new substructure. As for the GES, we find that it accounts for a large part of substructures in the inner halo and the apocentre distance is in the range 10–34 kpc, which suggests that the GES is mainly distributed in the inner halo. The near one-third proportion of the GES and the peak value of 20 kpc of the apocentre distance suggest that the GES could account for the break in the density profile of the Galactic halo at the Galactocentric distance ~20–25 kpc. The similarity when comparing the kinematic properties of the GES with the Hercules–Aquila Cloud and Virgo Overdensity suggests that the three substructures may have similar origins.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2020

Parallel algorithms for hyperdynamics and local hyperdynamics

Hyperdynamics (HD) is a method for accelerating the timescale of standard molecular dynamics (MD). It can be used for simulations of systems with an energy potential landscape that is a collection of basins, separated by barriers, where transitions between basins are infrequent. HD enables the system to escape from a basin more quickly while enabling a statistically accurate renormalization of the simulation time, thus effectively boosting the timescale of the simulation. In [Kim, Perez, Voter, J Chem Phys, 139:144110, 2013)1, a local version of HD was formulated, which exploits the intrinsic locality characteristic typical of most systems to mitigate the poor scaling properties of standard HD as the system size is increased. In this paper, we discuss how both HD and local HD can be formulated to run efficiently in parallel. We have implemented these ideas in the LAMMPS MD code, which means HD can be used with any interatomic potential LAMMPS supports. Together, these parallel methods allow simulations of any size to achieve the time acceleration offered by HD (which can be orders of magnitude), at a cost 3-5x that of standard MD. As examples, we performed two simulations of a million-atom system to model the diffusion and clustering of Pt adatoms on a large patch of Pt(100) surface for 80 and 160 μs.

74 ATOMIC AND MOLECULAR PHYSICS↗
DOE OSTI2021

Autonomous Experiments in Scanning Probe Microscopy and Spectroscopy: Choosing Where to Explore Polarization Dynamics in Ferroelectrics

Polarization dynamics in ferroelectric materials are explored via automated experiment in piezoresponse force microscopy/spectroscopy (PFM/S). A Bayesian optimization (BO) framework for imaging is developed, and its performance for a variety of acquisition and pathfinding functions is explored using previously acquired data. The optimized algorithm is then deployed on an operational scanning probe microscope (SPM) for finding areas of large electromechanical response in a thin film of PbTiO 3 , with results showing that, with just 20% of the area sampled, most high-response clusters were captured. Furthermore, this approach can allow performing more complex spectroscopies in SPM that were previously not possible due to time constraints and sample stability. Improvements to the framework to enable the incorporation of more prior information and improve efficiency further are modeled and discussed.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2021

Dynamics-Based Peptide–MHC Binding Optimization by a Convolutional Variational Autoencoder: A Use-Case Model for CASTELO

An unsolved challenge in the development of antigen-specific immunotherapies is determining the optimal antigens to target. Comprehension of antigen–major histocompatibility complex (MHC) binding is paramount toward achieving this goal. Here, we apply CASTELO, a combined machine learning-molecular dynamics (ML-MD) approach, to identify per-residue antigen binding contributions and then design novel antigens of increased MHC-II binding affinity for a type 1 diabetes-implicated system. We build upon a small-molecule lead optimization algorithm by training a convolutional variational autoencoder (CVAE) on MD trajectories of 48 different systems across four antigens and four HLA serotypes. We develop several new machine learning metrics including a structure-based anchor residue classification model as well as cluster comparison scores. ML-MD predictions agree well with experimental binding results and free energy perturbation-predicted binding affinities. Moreover, ML-MD metrics are independent of traditional MD stability metrics such as contact area and root-mean-square fluctuations (RMSF), which do not reflect binding affinity data. Finally, our work supports the role of structure-based deep learning techniques in antigen-specific immunotherapy design.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2023

Projecting Future Energy Production from Operating Wind Farms in North America. Part II: Statistical Downscaling

Abstract Capacity factors (CFs) derived from daily expected power at 22 operating wind farms in different regions of North America are used as predictands to train statistical downscaling algorithms using output from ERA5. The statistical downscaling models are then used to make CF projections for a suite of CMIP6 Earth System Models (ESMs). Downscaling is performed using a hybrid statistical approach that employs synoptic types derived using k -means clustering applied to sea level pressure fields with variance corrections applied as a function of the pressure gradient intensity. ESMs exhibit marked variability in terms of the skill with which the frequency of synoptic types and pressure gradients are reproduced relative to ERA5, and that differential skill is used to infer differential credibility in the associated CF projections. Projections of median annual mean CF [P50(CF)] in each 20-yr period from 1980 to 2099 show evidence of declines at most wind farms except in parts of the southern Great Plains, although the magnitude of the changes is strongly dependent on the ESM. For example, P50(CF) in 2080–99 deviate from those in 1980–99 by from −3.1 to +0.2 percentage points in the Northeast. The largest-magnitude declines in P50(CF) ranging from −3.9 to −2 percentage points are projected for the southern West Coast. CF trends exhibit marked seasonality and are strongly linked to changes in the relative intensity of future synoptic patterns, with much less impact from shifts in the occurrence of synoptic types over time. Internal climate modes continue to play a significant role in inducing interannual variability in wind power production, even under high radiative forcing scenarios. Significance Statement We describe how future climate changes may affect wind resources and wind power generation. Near-term changes in projected wind power electricity generation potential at operating wind farms over North America are small, but by the end of the current century electricity production is projected to decrease in many areas but may increase in parts of the southern Great Plains. The amount of change in projected wind power production is a strong function of the Earth system model that is downscaled and also depends on the continued presence of internally forced climate variability. An additional dependence on the amount of greenhouse gas–induced global warming indicates the transition of the energy sector to low-carbon sources may assist in maintaining the abundant U.S. wind resource.

Meteorology & Atmospheric Sciences↗
DOE OSTI2022

A Deep Reinforcement Learning-based Reserve Optimization in Active Distribution Systems for Tertiary Frequency Regulation

Federal Energy Regulatory Commission (FERC)Orders 841 and 2222 have recommended that distributed energy resources (DERs) should participate in energy and reserve markets; therefore, a mechanism needs to be developed to facilitate DERs’ participation at the distribution level. Although the available reserve from a single distribution system may not be sufficient for tertiary frequency regulation, stacked and coordinated contributions from several distribution systems can enable them participate in tertiary frequency regulation at scale. This paper proposes a deep reinforcement learning (DRL)-based approach for optimization of requested aggregated reserves by system operators among the clusters of DERs. The co-optimization of cost of reserve, distribution network loss, and voltage regulation of the feeders are considered while optimizing the reserves among participating DERs. The proposed framework adopts deep deterministic policy gradient (DDPG), which is an algorithm based on an actor-critic method. The effectiveness of the proposed method for allocating reserves among DERs is demonstrated through case studies on a modified IEEE 34-node distribution system.

deep reinforcement learning, distributed energy re↗
DOE OSTI2021

Design of Broadly Cross-Reactive M Protein–Based Group A Streptococcal Vaccines

Group A streptococcal infections are a significant cause of global morbidity and mortality. A leading vaccine candidate is the surface M protein, a major virulence determinant and protective Ag. An obstacle to the development of M protein–based vaccines is the >200 different M types defined by the N-terminal sequences that contain protective epitopes. Despite sequence variability, M proteins share coiled-coil structural motifs that bind host proteins required for virulence. In this study, we exploit this potential Achilles heel of conserved structure to predict cross-reactive M peptides that could serve as broadly protective vaccine Ags. Combining sequences with structural predictions, six heterologous M peptides in a sequence-related cluster were predicted to elicit cross-reactive Abs with the remaining five nonvaccine M types in the cluster. The six-valent vaccine elicited Abs in rabbits that reacted with all 11 M peptides in the cluster and functional opsonic Abs against vaccine and nonvaccine M types in the cluster. We next immunized mice with four sequence-unrelated M peptides predicted to contain different coiled-coil propensities and tested the antisera for cross-reactivity against 41 heterologous M peptides. Based on these results, we developed an improved algorithm to select cross-reactive peptide pairs using additional parameters of coiled-coil length and propensity. The revised algorithm accurately predicted cross-reactive Ab binding, improving the Matthews correlation coefficient from 0.42 to 0.74. These results form the basis for selecting the minimum number of N-terminal M peptides to include in potentially broadly efficacious multivalent vaccines that could impact the overall global burden of group A streptococcal diseases.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2022

Performance and power modeling and prediction using MuMMI and 10 machine learning methods

Energy-efficient scientific applications require insight into how high performance computing system features impact the applications' power and performance. This insight can result from the development of performance and power models. Here, in this article, we use the modeling and prediction tool MuMMI (Multiple Metrics Modeling Infrastructure) and 10 machine learning methods to model and predict performance and power consumption and compare their prediction error rates. We use an algorithm-based fault-tolerant linear algebra code and a multilevel checkpointing fault-tolerant heat distribution code to conduct our modeling and prediction study on the Cray XC40 Theta and IBM BG/Q Mira at Argonne National Laboratory and the Intel Haswell cluster Shepard at Sandia National Laboratories. Our experimental results show that the prediction error rates in performance and power using MuMMI are less than 10% for most cases. By utilizing the models for runtime, node power, CPU power, and memory power, we identify the most significant performance counters for potential application optimizations, and we predict theoretical outcomes of the optimizations. Based on two collected datasets, we analyze and compare the prediction accuracy in performance and power consumption using MuMMI and 10 machine learning methods.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

FedCSpc: A Cross-Silo Federated Learning System With Error-Bounded Lossy Parameter Compression

Cross-Silo federated learning is widely used for scaling deep neural network (DNN) training over data silos from different locations worldwide while guaranteeing data privacy. Communication has been identified as the main bottleneck when training large-scale models due to large-volume model parameters and gradient transmission across public networks with limited bandwidth. Most previous works focus on gradient compression, while limited work tries to compress parameters that can not be ignored and extremely affect communication performance during the training. Here, to bridge this gap, we propose FedCSpc: an efficient cross-silo federated learning system with an XAI-driven adaptive parameter compression strategy for large-scale model training. Our work substantially differs from existing gradient compression techniques due to the distinct data features of gradient and parameter. The key contributions of this paper are fourfold. (1) Our designed FedCSpc proposes to compress the parameter during the training using the state-of-the-art error-bounded lossy compressor – SZ3. (2) We develop an adaptive compression error bound adjustment algorithm to guarantee the model accuracy effectively. (3) We exploit an efficient approach to utilize the idle CPU resources of clients to compress the parameters. (4) We perform a comprehensive evaluation with a wide range of models and benchmarks on a GPU cluster with 65 GPUs. Results show that FedCSpc can achieve the same model accuracy as FedAvg while reducing the data volume of parameters and gradients in communication by up to 7.39× and 288×, respectively. With 32 clients on a 4 Gb size model, FedCSpc significantly outperforms FedAvg in wall-clock time in the emulated WAN environment (at the bandwidth of 1 Gbps or lower without loss of generality).

SZ3↗
DOE OSTI2022

Basement Fault Activation before Larger Earthquakes in Oklahoma and Kansas

Oklahoma and Kansas experienced unprecedented seismic activity over the past decade due to earthquakes associated with unconventional hydrocarbon development. The modest natural seismicity and incomplete knowledge of the fault network in the region made it difficult to anticipate the locations of earthquakes with larger magnitudes (M w ≥ 4). Here, we show that monitoring of microearthquakes at regional scale using a pretrained neural phase picker and an earthquake relocation algorithm can illuminate unknown fault structures, and deliver information that can be synthesized for earthquake forecasting. We found that 80% of the larger earthquakes that occurred in the past decade could have been anticipated based on the spatial extent of the seismicity clusters that were formed before these earthquakes occurred. We also found that once a seismicity cluster with a length scale enough to host a larger earthquake was formed, there was a ~5% chance that it would host one or more larger earthquakes within a year. This probability is nearly an order of magnitude higher than one based on Gutenberg–Richter statistics and preceding seismicity. Applying our approach in practice can provide critical information on seismic hazards for risk management and regulatory decision making.

58 GEOSCIENCES↗
DOE OSTI2021

First-Principles Grand-Canonical Simulations of Water Adsorption in Proton-Exchanged Zeolites

Water appears by design or as impurities in many important reactive systems. For those catalyzed by porous solid acids, such as widely used zeolites, experimentally quantifying the amount or elucidating the structure of adsorbed water clusters at reaction conditions is challenging, while computational studies (e.g., first-principles molecular dynamics simulations) often need to assume the loading to examine solvation effects. Furthermore, we perform first-principles grand-canonical simulations to predict water adsorption to H-ZSM-5 zeolites under specified experimental conditions. Presampling with inexpensive force fields and a pool-based parallelization algorithm are used to improve simulation efficiency, while molecular dynamics is used to sample configurations involving hydronium species. We observe that H + exchange dramatically increases the hydrophilicity of zeolite MFI and an appreciable amount of water is present at very low relative humidities. At all conditions examined, the zeolitic protons are found to dissociate readily from the surface basic sites and become mobile by participating in the hydrogen-bonded chains of adsorbed water molecules.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2025

Complexity Reduction Methods for Large-Scale Spatially Explicit Biofuels Network Design

The size and complexity of energy system optimization models have increased significantly in recent years, driven by the availability of high-resolution spatial data. We present complexity reduction and solution methods that enable us to efficiently represent high-resolution spatial data in the network design of large-scale energy systems. We aim to reduce the size and enhance the computational efficiency of network design models without sacrificing solution accuracy. Specifically, we first present how to aggregate highly granular data into larger resolutions without averaging out their specific properties through a composite-curve-based approach and then develop a method to linearly represent these curves. Second, we utilize a general clustering method to determine groups of geographically proximate biomass fields and establish a single transportation arc for all of them, reducing the number of transportation-related variables while maintaining an accurate representation of the system. Finally, we introduce a two-step algorithm that decomposes large-scale network design problems into two smaller, more manageable subproblems. We demonstrate the application of our methods using a case study of switchgrass-to-biofuels network design in the eight states of the U.S. Midwest, using realistic and highly explicit spatial data.

09 BIOMASS FUELS↗
DOE OSTI2021

Attention Enabled Multi-Agent DRL for Decentralized Volt-VAR Control of Active Distribution System Using PV Inverters and SVCs

This paper proposes attention enabled multi-agent deep reinforcement learning (MADRL) framework for active distribution network decentralized Volt-VAR control. Using the unsupervised clustering, the whole distribution system can be decomposed into several sub-networks according to the voltage and reactive power sensitivity relationships. Then, the distributed control problem of each sub-network is modeled as Markov games and solved by the improved MADRL algorithm, where each sub-network is modeled as an adaptive agent. An attention mechanism is developed to help each agent focus on specific information that is mostly related to the reward. All agents are centrally trained offline to learn the optimal coordinated Volt-VAR control strategy and executed in a decentralized manner to make online decisions with only local information. Compared with other distributed control approaches, the proposed method can effectively deal with uncertainties, achieve fast decision makings, and significantly reduce the communication requirements. Comparison results with model-based and other data-driven methods on IEEE 33-bus and 123-bus systems demonstrate the benefits of the proposed approach.

distribution network↗
DOE OSTI2024

Scalable semidefinite programming approach to variational embedding for quantum many-body problems

In quantum embedding theories, a quantum many-body system is divided into localized clusters of sites which are treated with an accurate ‘high-level’ theory and glued together self-consistently by a less accurate ‘low-level’ theory at the global scale. The recently introduced variational embedding approach for quantum many-body problems combines the insights of semidefinite relaxation and quantum embedding theory to provide a lower bound on the ground-state energy that improves as the cluster size is increased. The variational embedding method is formulated as a semidefinite program (SDP), which can suffer from poor computational scaling when treated with black-box solvers. Here, we exploit the interpretation of this SDP as an embedding method to develop an algorithm which alternates parallelizable local updates of the high-level quantities with updates that enforce the low-level global constraints. Moreover, we show how translation invariance in lattice systems can be exploited to reduce the complexity of projecting a key matrix to the positive semidefinite cone.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗