Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Distributed computing, machine learning”

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 181 records · Page 10

Accelerating Ab Initio Simulation via Nested Monte Carlo and Machine Learned Reference Potentials

As a corollary of the rapid advances in computing, ab initio simulation is playing an increasingly important role in modeling materials at the atomic scale. Two strategies are possible, ab initio Monte Carlo (AIMC) and molecular dynamics (AIMD) simulation. The former benefits from exact sampling from the correct thermodynamic distribution, while the latter is typically more efficient with its collective all-atom coordinate updates. In this study, using a relatively simple test model comprised of helium and argon, we show that AIMC can be brought up to, and even above, the performance levels of AIMD via a hybrid nested sampling/machine learning (ML) strategy. Here, ML provides an accurate classical reference potential (up to three-body explicit interactions) that can pilot long collective Monte Carlo moves that are accepted or rejected in toto à la nested Monte Carlo (NMC); this is in contrast to the single move nature of a naive implementation. Our proposed method only requires a small up front expense from evaluating the ab initio energies and forces of (100) random configurations for training. Importantly, our method does not totally rely on the trained, assuredly imperfect, interaction. We show that high performance and exact sampling at the desired level of theory can be realized even when the trained interaction has appreciable differences from the ab initio potential. Remarkably, at the highest levels of performance realized via our approach, a pair of statistically uncorrelated atomic configurations can be generated with (1) ab initio calculations.

74 ATOMIC AND MOLECULAR PHYSICS↗

Scalable training of trustworthy and energy-efficient predictive graph foundation models for atomistic materials modeling: a case study with HydraGNN

We present our work on developing and training scalable, trustworthy, and energy-efficient predictive graph foundation models (GFMs) using HydraGNN, a multi-headed graph convolutional neural network architecture. HydraGNN expands the boundaries of graph neural network (GNN) computations in both training scale and data diversity. It abstracts over message passing algorithms, allowing both reproduction of and comparison across algorithmic innovations that define nearest-neighbor convolution in GNNs. This work discusses a series of optimizations that have allowed scaling up the GFMs training to tens of thousands of GPUs on datasets consisting of hundreds of millions of graphs. Our GFMs use multitask learning (MTL) to simultaneously learn graph-level and node-level properties of atomistic structures, such as energy and atomic forces. Using over 154 million atomistic structures for training, we illustrate the performance of our approach along with the lessons learned on two state-of-the-art US Department of Energy (US-DOE) supercomputers, namely the Perlmutter petascale system at the National Energy Research Scientific Computing Center and the Frontier exascale system at Oak Ridge Leadership Computing Facility. The HydraGNN architecture enables the GFM to achieve near-linear strong scaling performance using more than 2000 GPUs on Perlmutter and 16,000 GPUs on Frontier.

97 MATHEMATICS AND COMPUTING↗

DEEP Solar: Data DrivEn Modeling and Analytics for Enhanced System Layer ImPlementation

Realizing the SETO 2030 mission of reducing solar energy costs to 3-5 c/kWh will require innovative enabling research on effective, cost-efficient integration of local PV within distribution systems. However, the intermittent and variable nature of PVs compels operators to impose conservative hosting capacity constraints. Given the extremely high variability of (intermittent and unpredictable) solar energy generation, relaxing the capacity constraints (which are currently around 15%) and achieving 100% or greater integration of renewables will require a fundamental transformation of the power grid via the utilization of exponentially larger amounts of AMI enabled fine-grained data. To address the challenges in increasing the penetration of renewable energy based DERs, this project envisions an Enhanced System Layer (ESL) at the distribution network level that is reliable, cost-effective and scalable to millions of Distributed Energy Resources (DERs)/devices. This includes developing: 1) Transformative and highly scalable machine learning based predictive analytics tools that plug into distribution system planning and provide real-time situational awareness at the distribution level for short and long-term operational planning. The tools will be built using novel data-driven energy models of millions of active nodes with AMI, 2) Adaptive stochastic analysis and optimization algorithms for real-time grid operations, 3) Dynamic Scenario Analysis using parallel Cloudenabled implementations with < 1 minute computational cycle times.

14 SOLAR ENERGY↗

Informed unsupervised machine learning analysis of dislocation microstructure from high-resolution differential aperture X-ray structural microscopy data

This study leverages high-resolution differential-aperture X-ray structural microscopy (DAXM) to probe the local dislocation structure in deformed 304L-stainless steel at small strain, by measuring the lattice rotation and deviatoric elastic strain with a sub-micron resolution. For a single grain in a polycrystalline specimen, the measured lattice rotation field over the measured volume exhibited a multimodal distribution while the deviatoric elastic strain showed a single-mode distribution. An unsupervised Cauchy mixture machine learning model was developed to resolve the multimodal distribution of the lattice rotation. By mapping the lattice rotation data associated with each Cauchy peak in the model back onto the measured volume, we identify contiguous regions of the crystal rotated near the average values corresponding to the peaks of the overall rotation distribution. These regions represent the grain subdivision in the microstructure. Finally, the dislocation density tensor was also computed and its norm was laid over the rotation field to detect the subgrain boundaries. This step provided a validation of the Cauchy mixture model for the analysis of the lattice rotation distribution. The current study highlights the integration of advanced X-ray microscopy techniques with data-driven analysis methods to uncover detailed microstructure scales in deformed crystals.

Machine learning; Lattice rotation; High-energy X-↗

Data-driven, Interpretable Photometric Redshifts Trained on Heterogeneous and Unrepresentative Data

We present a new method for inferring photometric redshifts in deep galaxy and quasar surveys, based on a data-driven model of latent spectral energy distributions (SEDs) and a physical model of photometric fluxes as a function of redshift. This conceptually novel approach combines the advantages of both machine learning methods and template fitting methods by building template SEDs directly from the spectroscopic training data. This is made computationally tractable with Gaussian processes operating in flux–redshift space, encoding the physics of redshifts and the projection of galaxy SEDs onto photometric bandpasses. This method alleviates the need to acquire representative training data or to construct detailed galaxy SED models; it requires only that the photometric bandpasses and calibrations be known or have parameterized unknowns. The training data can consist of a combination of spectroscopic and deep many-band photometric data with reliable redshifts, which do not need to entirely spatially overlap with the target survey of interest or even involve the same photometric bands. We showcase the method on the i -magnitude-selected, spectroscopically confirmed galaxies in the COSMOS field. The model is trained on the deepest bands (from SUBARU and HST ) and photometric redshifts are derived using the shallower SDSS optical bands only. We demonstrate that we obtain accurate redshift point estimates and probability distributions despite the training and target sets having very different redshift distributions, noise properties, and even photometric bands. Our model can also be used to predict missing photometric fluxes or to simulate populations of galaxies with realistic fluxes and redshifts, for example.

79 ASTRONOMY AND ASTROPHYSICS↗

Automatic log analysis with NLP for the CMS workflow handling

The central Monte-Carlo production of the CMS experiment utilizes the WLCG infrastructure and manages daily thousands of tasks, each up to thousands of jobs. The distributed computing system is bound to sustain a certain rate of failures of various types, which are currently handled by computing operators a posteriori. Within the context of computing operations, and operation intelligence, we propose a Machine Learning technique to learn from the operators with a view to reduce the operational workload and delays. This work is in continuation of CMS work on operation intelligence to try and reach accurate predictions with Machine Learning. We present an approach to consider the log files of the workflows as regular text to leverage modern techniques from Natural Language Processing (NLP). In general, log files contain a substantial amount of text that is not human language. Therefore, different log parsing approaches are studied in order to map the log files’ words to high dimensional vectors. These vectors are then exploited as feature space to train a model that predicts the action that the operator has to take. This approach has the advantage that the information of the log files is extracted automatically and the format of the logs can be arbitrary. In this work the performance of the log file analysis with NLP is presented and compared to previous approaches.

Layer, Lukas↗

Machine-learning prediction for quasiparton distribution function matrix elements

There have been rapid developments in the direct calculation in lattice QCD (LQCD) of the Bjorken-x dependence of hadron structure through large-momentum effective theory (LaMET). LaMET overcomes the previous limitation of LQCD to moments (that is, integrals over Bjorken x) of hadron structure, allowing LQCD to directly provide the kinematic regions where the experimental values are least known. LaMET requires large-momentum hadron states to minimize its systematics and allow us to reach small- x reliably. This means that very fine lattice spacing to minimize lattice artifacts at order $(P_za)^n$ will become crucial for next-generation LaMET-like structure calculations. Furthermore, such calculations require operators with long Wilson-link displacements, especially in finer lattice units, increasing the communication costs relative to that of the propagator inversion. In this work, we explore whether machine-learning algorithms can make predictions of correlators to reduce the computational cost of these LQCD calculations. We consider two algorithms, gradient-boosting decision tree and linear models, applied to LaMET data, the matrix elements needed to determine the kaon and $η_s$ unpolarized parton distribution functions (PDFs), meson distribution amplitude (DA), and the nucleon gluon PDF. We find that both algorithms can reliably predict the target observables with different prediction accuracy and systematic errors. The predictions from smaller displacement z to larger ones work better than those for momentum p due to the higher correlation among the data.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Smart Methane Emission Detection System Development (Final Report)

Working with the Department of Energy's National Energy Technology Laboratory, Southwest Research Institute® (SwRI®) developed a system to identify methane leaks reliably, accurately, and autonomously at critical midstream sections of the natural gas distribution network in real-time for the purpose of mitigating methane emissions using Optical Gas Imaging (OGI) cameras. SwRI's Smart Leak Detection – Methane (SLED/M) adds a high degree of automation to the process of methane leak detection to minimize sources of human error, minimize response time to a leak event, and maximize midstream visibility. Furthermore, SwRI has been working towards integrating Quantitative OGI (QOGI) capabilities into this existing technology. By leveraging Deep Learning, SwRI now has the capability to estimate fugitive emission leak rates quickly and reliably, which allows operators to detect emissions, quantify leak rate, prioritize repairs, and validate the repairs in a single instrument. The next generation QOGI technology leverages the same cameras used in Leak Detection and Repair (LDAR) programs, with improvements in safety and speed for traditional quantification-based repairs, ultimately leading to less overhead cost for the operators. The goals for this research were to develop two types of models with the following goals: Run in real-time on the edge (≥ 12 Hz), Classification: Achieve less than 5% false positive detection, Classification: Achieve ≥ 95% methane plume detection rate, Regression: achieve ≤ 10 standard cubic feet per hour (scfh) prediction > 70% of the time. In order to achieve these results, multiple infrared (IR) and other sensors were investigated in tandem with the midwave IR (MWIR) OGI to provide additional information to train the underlying models. Information on atmospheric conditions including humidity, temperature, pressure, and solar radiation was provided by a weather station. Several machine learning and deep learning architectures and methods, including looking at quantized classification networks and regressions networks, were explored. As further data was collected, curated, and labeled, it allowed for more refined regressive networks to be adequately trained, leading to better insight into the true flow rates being observed. An important valuable deliverable of this research effort was the development of an advanced network which underwent multiple iterations capable of giving a continuous output. The current network has a predicted mean average percentage error (MAPE) of 12.3% just outside our target goal of 10.00%, but an accuracy of 97.78% at ±50 scfh, well within the overall goal for the Department of Energy (DOE) program. Upon closer inspection, it was observed that more than 10% of datapoints contributing to the MAPE predictions were the result of low flow rate predictions and are beyond the sensitivity of instrument measurement as a result of normal operational variation and noise.

03 NATURAL GAS↗

Quantifying dispersity in size and shape of nanoparticles from small-angle scattering data using machine learning based CREASE

Here, we use machine learning (ML) enhanced computational reverse engineering analysis of scattering experiments (CREASE) to interpret small-angle X-ray scattering (SAXS) data obtained from a system of nanoparticles without a priori knowledge of their exact shapes (e.g. spheres or ellipsoids), sizes (0.5–50 nm) and distributions. The SAXS measurements yielded three categories of scattering profiles exhibiting 'strong', 'weak' and 'no' features. Diminishing features (e.g. broadening or disappearing peaks) in scattering profiles have always been attributed to the presence of significant dispersity in the system. Such featureless SAXS data are not suitable for traditional analysis using analytical models. If one were to fit a relevant analytical model (e.g. the lmfit analytical model for polydisperse spheres) to these 'weak' and 'no' SAXS profiles from our nanoparticle systems, one would obtain non-unique interpretations of the data. Relying on electron microscopy to identify the distributions of nanoparticle shapes and sizes is also unfeasible, especially in high-throughput synthesis and characterization loops. In such situations, to identify the distributions of particle sizes and shapes that could be present in the sample, one must rely on methods like ML-CREASE to interpret the data quickly and output all relevant interpretations about the structure present in the system. The ML-CREASE optimization loop takes the experimental scattering profile as input and outputs multiple candidate solutions whose computed scattering profiles match the SAXS profile input. The ML-CREASE method outputs distributions of relevant structural features, such as the volume fraction of the nanoparticles in the system and the mean and standard deviation of the particle size and aspect ratio, assuming a type of distribution (e.g. normal, log-normal) for size and aspect ratio. We find that, for the SAXS profiles analyzed here, accounting for the shape dispersity along with size dispersity of the nanoparticles using ML-CREASE improved the match between the computed scattering profiles and input experimental profiles.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

End-to-end microgrid protection using distributed data-driven methods

This paper introduces an end-to-end microgrid protection framework that offers real-time system monitoring, fault-related decision making, and circuit breaker control. This is achieved through the design of distributed data-driven techniques based on the support vector machine method, where each relay is responsible for distributed data collection, fault detection, fault localization, and fault isolation. Local communication is established among neighboring relays, fostering cooperative fault localization and isolation. This decentralized design not only reduces the computational and communication requirements but also enables the adaptability of each relay under varying operational dynamics. The proposed end-to-end protection framework was validated using MATLAB/Simulink simulations on a 100% renewable microgrid, achieving an accuracy of 93.1% with response time of 0.0523 s, in protecting against a range of fault scenarios that are characterized by various types, locations, impedances, load conditions, photovoltaic power levels, and microgrid operating modes.

24 POWER TRANSMISSION AND DISTRIBUTION↗

Solving differential‐algebraic equations in power system dynamic analysis with quantum computing

Abstract Power system dynamics are generally modeled by high dimensional non‐linear differential‐algebraic equations (DAEs) given a large number of components forming the network. These DAEs' complexity can grow exponentially due to the increasing penetration of distributed energy resources, whereas their computation time becomes sensitive due to the increasing interconnection of the power grid with other energy systems. This paper demonstrates the use of quantum computing algorithms to solve DAEs for power system dynamic analysis. We leverage a symbolic programming framework to equivalently convert the power system's DAEs into ordinary differential equations (ODEs) using index reduction methods and then encode their data into qubits using amplitude encoding. The system non‐linearity is captured by Hamiltonian simulation with truncated Taylor expansion so that state variables can be updated by a quantum linear equation solver. Our results show that quantum computing can solve the power system's DAEs accurately with a computational complexity polynomial in the logarithm of the system dimension. We also illustrate the use of recent advanced tools in scientific machine learning for implementing complex computing concepts, that is, Taylor expansion, DAEs/ODEs transformation, and quantum computing solver with abstract representation for power engineering applications.

computational complexity↗

Investigating the influence of particle distribution on force and torque statistics using hierarchical machine learning

An accurate representation of hydrodynamic force and torque experienced by every particle in a distribution can be obtained from particle resolved (PR) simulations. These unique quantities are influenced by the deterministic position of surrounding particles. However, systems simulated with this methodology are typically limited to particles due to the involved computational cost. This resource requirement is a major bottleneck in analyzing the effect of variations in particle distribution. Here, this article attempts to address this bottleneck by availing relatively inexpensive deep learning models. The surrogate models that we employ in this article use a physics‐based hierarchical framework and symmetry‐preserving neural networks to achieve robustness with limited training data. This article first performs additional generalizability tests on PR data of distinct distributions that are not involved in the training process. The models are then deployed on several different particle distributions. Impact of clustering and structure on the observed statistics are investigated.

42 ENGINEERING↗

Learning Optimal Solutions for Extremely Fast AC Optimal Power Flow

We develop, in this paper, a machine learning approach to optimize the real-time operation of electric power grids. In particular, we learn feasible solutions to the AC optimal power flow (OPF) problem with negligible optimality gaps. The AC OPF problem aims at identifying optimal operational conditions of the power grids that minimize power losses and/or generation costs. Due to the computational challenges with solving this nonconvex problem, many efforts have focused on linearizing or approximating the problem in order to solve the AC OPF on faster timescales. However, many of these approximations can be fairly poor representations of the actual system state and still require solving an optimization problem, which can be time consuming for large networks. In this work, we learn a mapping between the system loading and optimal generation values, enabling us to find near-optimal and feasible AC OPF solutions. This allows us to bypass solving the traditionally nonconvex AC OPF problem, resulting in a significant decrease in computational burden for grid operators.

machine learning↗

A Comprehensive Analysis of PINNs for Power System Transient Stability

The integration of machine learning in power systems, particularly in stability and dynamics, addresses the challenges brought by the integration of renewable energies and distributed energy resources (DERs). Traditional methods for power system transient stability, involving solving differential equations with computational techniques, face limitations due to their time-consuming and computationally demanding nature. This paper introduces physics-informed Neural Networks (PINNs) as a promising solution for these challenges, especially in scenarios with limited data availability and the need for high computational speed. PINNs offer a novel approach for complex power systems by incorporating additional equations and adapting to various system scales, from a single bus to multi-bus networks. Our study presents the first comprehensive evaluation of physics-informed Neural Networks (PINNs) in the context of power system transient stability, addressing various grid complexities. Additionally, we introduce a novel approach for adjusting loss weights to improve the adaptability of PINNs to diverse systems. Our experimental findings reveal that PINNs can be efficiently scaled while maintaining high accuracy. Furthermore, these results suggest that PINNs significantly outperform the traditional ode45 method in terms of efficiency, especially as the system size increases, showcasing a progressive speed advantage over ode45.

97 MATHEMATICS AND COMPUTING↗

A Comprehensive Analysis of PINNs for Power System Transient Stability

The integration of machine learning in power systems, particularly in stability and dynamics, addresses the challenges brought by the integration of renewable energies and distributed energy resources (DERs). Traditional methods for power system transient stability, involving solving differential equations with computational techniques, face limitations due to their time-consuming and computationally demanding nature. This paper introduces physics-informed Neural Networks (PINNs) as a promising solution for these challenges, especially in scenarios with limited data availability and the need for high computational speed. PINNs offer a novel approach for complex power systems by incorporating additional equations and adapting to various system scales, from a single bus to multi-bus networks. Our study presents the first comprehensive evaluation of physics-informed Neural Networks (PINNs) in the context of power system transient stability, addressing various grid complexities. Additionally, we introduce a novel approach for adjusting loss weights to improve the adaptability of PINNs to diverse systems. Our experimental findings reveal that PINNs can be efficiently scaled while maintaining high accuracy. Furthermore, these results suggest that PINNs significantly outperform the traditional ode45 method in terms of efficiency, especially as the system size increases, showcasing a progressive speed advantage over ode45.

24 POWER TRANSMISSION AND DISTRIBUTION↗

Learning Optimal Solutions for Extremely Fast AC Optimal Power Flow: Preprint

We develop, in this paper, a machine learning approach to optimize the real-time operation of electric power grids. In particular, we learn feasible solutions to the AC optimal power flow (OPF) problem with negligible optimality gaps. The AC OPF problem aims at identifying optimal operational conditions of the power grids that minimize power losses and/or generation costs. Due to the computational challenges with solving this nonconvex problem, many efforts have focused on linearizing or approximating the problem in order to solve the AC OPF on faster timescales. However, many of these approximations can be fairly poor representations of the actual system state and still require solving an optimization problem, which can be time consuming for large networks. In this work, we learn a mapping between the system loading and optimal generation values, enabling us to find near-optimal and feasible AC OPF solutions. This allows us to bypass solving the traditionally nonconvex AC OPF problem, resulting in a significant decrease in computational burden for grid operators.

machine learning↗

Toward a smart data transfer node

Scientific computing systems are becoming significantly more complex, with distributed teams and complex workflows spanning resources from telescopes and light sources to fast networks and Internet of Things sensor systems. In such settings, no single, centralized administrative team and software stack can coordinate and manage all resources used by a single application. Indeed, we have reached a critical limit in manageability using current human-in-the-loop techniques. We therefore argue that resources must begin to respond automatically, adapting and tuning their behavior in response to observed properties of scientific workflows. Over time, machine learning methods can be used to identify effective strategies for autonomic, goal-driven management behaviors that can be applied end-to-end across the scientific computing landscape. Using the data transfer nodes that are widely deployed in modern research networks as an example, we explore the architecture, methods, and algorithms needed for a smart data transfer node to support future scientific computing systems that self-tune and self-manage.

97 MATHEMATICS AND COMPUTING↗

Machine Learning-enabled Scalable Performance Prediction of Scientific Codes

Hardware architectures become increasingly complex as the compute capabilities grow to exascale. Here, we present the Analytical Memory Model with Pipelines (AMMP) of the Performance Prediction Toolkit (PPT). PPT-AMMP takes high-level source code and hardware architecture parameters as input and predicts runtime of that code on the target hardware platform, which is defined in the input parameters. PPT-AMMP transforms the code to an (architecture-independent) intermediate representation, then (i) analyzes the basic block structure of the code, (ii) processes architecture-independent virtual memory access patterns that it uses to build memory reuse distance distribution models for each basic block, and (iii) runs detailed basic-block level simulations to determine hardware pipeline usage. PPT-AMMP uses machine learning and regression techniques to build the prediction models based on small instances of the input code, then integrates into a higher-order discrete-event simulation model of PPT running on Simian PDES engine. We validate PPT-AMMP on four standard computational physics benchmarks and present a use case of hardware parameter sensitivity analysis to identify bottleneck hardware resources on different code inputs. We further extend PPT-AMMP to predict the performance of a scientific application code, namely, the radiation transport mini-app SNAP. To this end, we analyze multi-variate regression models that accurately predict the reuse profiles and the basic block counts. We validate predicted SNAP runtimes against actual measured times.

97 MATHEMATICS AND COMPUTING↗