Deep learning analysis of deeply virtual exclusive photoproduction
Not Available
72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
SEARCH · Engineering Papers
Results for “physical 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.
Improved loss functions for machine-learned atomic potentials
Machine learning (ML) has become an invaluable tool across a wide array of domains in science as researchers find new ways to leverage its predictive power. This is especially true in chemistry, where ML is used to fit chemical properties or desirable attributes to the local structure of molecules and materials. In the pursuit of greater accuracy, it is relatively simple to increase the size or complexity of such models, although this often requires simultaneously seeking larger datasets in order to both fit and interpret the larger number of parameters. However, it is equally important to assess the quality and relative importance of the data and how these factors impact the training process. We, therefore, investigate the impact of using different loss functions for training neural network potentials (NNPs), as the loss function defines the error and parameter gradients used to train the NNP. In particular, we test the mean-squared error and Huber loss functions and, using insight from these functions, derive a new loss function based on the Asinh function, which yields significant improvement in the accuracy and generality of NNPs. We show that by discounting/minimizing errors and anomalies in the optimization process, both the Huber and Asinh loss functions improve the training of NNPs, leading to a final potential with a greater effective dimensionality.
37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
Quantum simulations in effective model spaces: Hamiltonian-learning variational quantum eigensolver using digital quantum computers and application to the Lipkin-Meshkov-Glick model
Quantum simulations offer the potential to predict the structure and dynamics of nuclear many-body systems that are beyond the capabilities of classical computing. Generally, preparing the ground state of strongly-interacting many-body systems relevant to nuclear physics is however inefficient, even using ideal quantum computers. In addition, currently available NISQ-era quantum devices possess modest numbers of qubits, limiting the size of quantum many-body systems that can be simulated. In this context, a reformulation of the quantum many-body problems using truncated model spaces and Hamiltonians is desirable to make them more amenable to near-term quantum computers. The importance of symmetries in low-energy theories, including effective field theories (EFTs), lattice quantum chromodynamics (QCD), and effective model spaces for nuclear systems, in particular their interplay with the reduction of active Hilbert spaces, is well known. Lesser known is the fact that the non-commutivity of some symmetries and truncations of the model space can be profitably combined with variational calculations to rearrange the entanglement into localized structures and enable more efficient simulations. Here, the goal of the present study is to explore and utilize the non-commutivity of symmetries and model-space truncations of quantum many-body systems important to nuclear physics, particularly in combination with variational algorithms for quantum simulations and effective Hamiltonian learning. We introduce an iterative hybrid classical-quantum algorithm, Hamiltonian learning variational quantum eigensolver (HL-VQE), that simultaneously optimizes an effective Hamiltonian, thereby rearranging entanglement into the effective model space, and the associated ground-state wavefunction. Quantum simulations, using classical computers and IBM's superconducting-qubit quantum computers, are performed to demonstrate the HL-VQE algorithm, in the context of the Lipkin-Meshkov-Glick (LMG) model of interacting fermions, where the Hamiltonian transformation corresponds to an orbital rotation. We use a mapping where the number of qubits scales with the $\log$ of the size of the effective model space, rather than the particle number. HL-VQE is found to provide an exponential improvement in LMG-model calculations of the ground-state energy and wavefunction, compared to naive truncations without Hamiltonian learning, throughout a significant fraction of the Hilbert space. In the context of EFT, this corresponds to counterterms scaling exponentially with the cut-off as opposed to power law. Implementations on IBM's QExperience quantum computers and simulators for 1- and 2-qubit effective model spaces are shown to provide accurate and precise results, reproducing classical predictions. For a range of parameters defining the LMG model, the HL-VQE algorithm is found to have better scaling of quantum resources requirements than previously explored algorithms. In particular, the HL-VQE scales efficiently over a large fraction of the model space, in contrast to VQE alone. This work constitutes a step in the development of entanglement-driven quantum algorithms for descriptions of nuclear many-body systems. This, in part, leverages the potential of noisy intermediate-scale quantum (NISQ) devices. The exponential scaling of counterterms observed in this study suggests the possibility of more general applicability to other non-perturbative EFTs.
73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗
Domain Aware Deep-learning Algorithms Integrated with Scientific-computing Technologies (DADAIST)
This technical report summarized the contribution of the DADAIST project funded by the Data Model Convergence Initiative via the Laboratory Directed Research and Development (LDRD) investments at Pacific Northwest National Laboratory (PNNL). Specifically, we report the development of the NeuroMANCER (Neural Modules with Adaptive Nonlinear Constraints and Efficient Regularizations), a new open-source Scientific Machine Learning library for formulating and solving parametric constrained optimization problems, physics-informed system identification, and parametric optimal control problems. NeuroMANCER is using differentiable programming to combine modern data-driven models and optimization modeling language into a coherent algorithmic and software framework. NeuroMANCER is a Pytorch-based framework and adopts much of its philosophy focused on research and development, rapid prototyping, and streamlined deployment. Strong emphasis is given to extensibility, interoperability with the PyTorch ecosystem, and quick adaptability to custom domain problems. Neuromancer repository contains a comprehensive library of differentiable modules, including custom activation functions, matrix factorizations, deep learning architectures, neural differential equations, differential equation solvers, implicit layers such as iterative solvers, high-level API for symbolic expressions, API for modeling and control of dynamical systems, and extensive set of tutorial code examples in the form of python scripts and jupyter notebooks.
97 MATHEMATICS AND COMPUTING↗
Robust impurity detection and tracking for tokamaks
A robust impurity detection and tracking code, able to generate large sets of dust tracks from tokamak camera footage, is presented. This machine learning–based code is tested with cameras from the Joint European Torus, Doublet-III-D, and Magnum-PSI and is able to generate dust tracks with a 65 – 100 % classification accuracy. Moreover, the number dust particles detected from a single camera shot can be up to the order of 1000. Several areas of improvement for the code are highlighted, such as generating more significant training data sets and accounting for selection biases. Finally, although the code is tested with dust in single two-dimensional camera views, it could easily be applied to multiple-camera stereoscopic reconstruction or nondust impurities.
70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
Upscaling Reactive Transport and Clogging in Shale Microcracks by Deep Learning
Fracture networks in shales exhibit multiscale features. A rock system may contain a few main fractures and thousands of microcracks, whose length and aperture are orders of magnitude smaller than the former. It is computationally prohibitive to resolve all the fractures explicitly for such multiscale fracture networks. One traditional approach is to model the small-scale features (e.g., microcracks in shales) as an effective medium. Although this fracture-matrix conceptualization significantly reduces the problem complexity, there are classes of physical processes that cannot be accurately upscaled by effective medium approximations, e.g., microcrack clogging during mineral reactions. Here, we employ deep learning in place of effective medium theory to upscale physical processes in small-scale features. Specifically, we consider reactive transport in a fracture-microcrack network where microcracks can be clogged by precipitation. A deep learning multiscale algorithm is developed, in which the microcracks are upscaled as a wall boundary condition of the main fractures. The wall boundary condition is constructed by recurrent neural networks, which take concentration histories as input and predict the solute transport from main fractures to microcracks. The deep learning multiscale algorithm is firstly employed in specific scenarios, then a general model is developed which can work under various conditions. The new approach is validated against fully resolved simulations and an analytical solution, providing a reliable and efficient solution for problems that cannot be upscaled by effective medium models.
58 GEOSCIENCES↗
Smart pixel sensors Towards on-sensor filtering of pixel clusters with deep learning
High granularity silicon pixel sensors are at the heart of energy frontier particle physics collider experiments. At an collision rate of 40\,MHz, these detectors create massive amounts of data. Signal processing that handles data incoming at those rate and intelligently reduces the data within the pixelated region of the detector \textit{at rate} will enhance physics performance and enable physics analyses that are not currently possible. Using the shape of charge clusters deposited in an array of small pixels, the physical properties of the traversing particle can be extracted with locally customized neural networks. In this first work, we present a neural network that can be embedded into the on-sensor readout and filter out hits from low momentum tracks, reducing the detector's data volume by 54.4-75.4\%. The network is designed and simulated as a custom readout integrated circuit with 28\,nm CMOS technology and is expected to operate at less than 300\,$\mu W$ with an area of less than 0.2\,mm$^2$.
43 PARTICLE ACCELERATORS↗
Exploring the Neutron Substructure with Advanced Polarized Helium-3 Targets (Or: How I Learned to Stop Worrying and Love Spectroscopy)
As we seek to understand the smallest, physical aspects of our universe, we cannot simply rely on our senses to probe the world around us as we did in the past. The smallest physical elements of our universe behave in strange, probabilistic ways and are completely invisible to the naked eye/ear/etc. So, we design clever experiments (such as scattering experiments) to probe these minute realms. Then, just as with the larger, observable world, we devise models and equations to describe what we think is happening. Due to the nature of the physical universe at the quantum scale and with the aid of symmetries such as Lorentz invariance, we can write down equations that describe the scattering, but the expressions contain functions, which we call ?form factors? and ?structure functions?, that we cannot compute from first principles. We can, however, formulate models that make predictions for these functions. By comparing our predictions with the observed data, we can gain insight into the validity of our models and thus a better physical understanding of what is happening at these minuscule scales. Studying the constituents inside of the nucleus of an atom adds another layer of difficulty if we can?t remove those components from the nucleus. This is the case with the neutron. When not bound in the nucleus with protons and other neutrons, the neutron will decay into a proton after about 15 minutes. So, we?re forced to study the neutron while it is still bound in the nucleus of an atom such as helium-3 (3He). For the last 1,000 years (rounding up), our group has developed high quality, polarized 3He targets made of an aluminosilicate glass. These targets are made in order to perform experiments at Jefferson Lab (JLab), experiments which let us determine the form factors and structure functions of the neutron by scattering polarized electrons from polarized neutrons (or rather polarized 3He). The specific experiments reported on in this thesis push the bounds of our understanding of the internal structure of the neutron. Good science is often about pushing experimental techniques to a new level. Toward that goal we study our polarized 3He targets both to advance the technology and to choose the best ones for our experiments. We do this using a process called nuclear magnetic resonance (NMR) to gauge the maximum polarization of a target and how fast the polarization decays with time. While these tests primarily provide us information that make analysis of our experimental scattering data possible, they also let us determine whether or not a target-cell is useful or even, dare I say, of spectacular quality. Our latest targets utilize a novel convection design allowing 3He to be polarized and quickly moved in front of the electron-beam, making it possible to use larger targets with higher electron-beam currents than ever before. This means more electrons scatter and we get more data. And by studying our targets in detail prior to using them in our experiments, we have found techniques to take effects which could have been detrimental to target quality and turn them to our advantage! It?s a real case of making lemonade out of lemons. We also use laser spectroscopy to study the absorption lines of alkali-metals in the target (potassium and rubidium, specifically). We add these alkali-metals to our target to facilitate polarizing the 3He. We can use the measurement of these pressure broadened absorption lines to determine the 3He density inside of the target with great precision. Historically, we understood the width of these lines would be dependent on the temperature of the target. Specifically, if I raise the temperature, the width should get bigger. I found that was not the case, which was very confusing at first, though very exciting now that I realize the data are self-consistent and suggestive of unexpected behavior. This thesis details the development of high quality, glass, polarized 3He targets for the 2020 An 1 /dn 2 and 2023 Gn E experiments, which utilized the first 3He convection targets and broke records in target quality. This thesis also covers the initial development of metal windows for the next-generation of 3He target-cells. Finally, this thesis documents the temperature dependence of the width of potassium (K) and rubidium (Rb) absorption lines as measured with laser spectroscopy.
Jantzi, Christopher↗
Deep-learning-enabled Bayesian inference of fuel magnetization in magnetized liner inertial fusion
Fuel magnetization in magneto-inertial fusion (MIF) experiments improves charged burn product confinement, reducing requirements on fuel areal density and pressure to achieve self-heating. By elongating the path length of 1.01 MeV tritons produced in a pure deuterium fusion plasma, magnetization enhances the probability for deuterium–tritium reactions producing 11.8-17.1 MeV neutrons. Nuclear diagnostics thus enable a sensitive probe of magnetization. Characterization of magnetization, including uncertainty quantification, is crucial for understanding the physics governing target performance in MIF platforms, such as magnetized liner inertial fusion (MagLIF) experiments conducted at Sandia National Laboratories, Z-facility. We demonstrate a deep-learned surrogate of a physics-based model of nuclear measurements. A single model evaluation is reduced from O ( 10 – 100 ) (10–100) CPU hours on a high-performance computing cluster down to O ( 10 – 100 ) (10) ms on a laptop. This enables a Bayesian inference of magnetization, rigorously accounting for uncertainties from surrogate modeling and noisy nuclear measurements. The approach is validated by testing on synthetic data and comparing with a previous study. We analyze a series of MagLIF experiments systematically varying preheat, resulting in the first ever systematic experimental study of magnetic confinement properties of the fuel plasma as a function of fundamental inputs on any neutron-producing MIF platform. We demonstrate that magnetization decreases from B R ∼ 0.2 ~0.5 to B R ∼ 0.2 ~0.2 MG cm as laser preheat energy deposited increases from E preheat ∼ 460 ~460 J to E preheat ∼ 460 ~1.4 kJ. This trend is consistent with 2D LASNEX simulations showing Nernst advection of the magnetic field out of the hot fuel and diffusion into the target liner.
70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
Machine Learning Modeling Pipeline for Extracting Nuclear Proliferation Events of Interest from Open Data Sources (U)
In FY2020, the Savannah River National Laboratory (SRNL) and the Sanghani Center for Artificial Intelligence and Data Analytics at Virginia Polytechnic Institute and State University entered a collaboration funded by Department of Energy’s (DOE) Office of Defense Nuclear Nonproliferation Research and Development. The project’s mission was to take the first steps toward developing a demonstration prototype system that uses multiple machine learning and data analytics methods on largescale open data sources to identify new, developing, and/or undeclared nuclear programs. Given the SRNL team’s on-site perspective of events culminating in the DOE’s decision to pursue the Savannah River Plutonium Processing Facility (SRPPF), the team targeted the identification of events and indicators in retrospective datasets that pointed to the activity of “fissile core fabrication at the Savannah River Site” prior to the official announcement in May of 2018. A preliminary modeling pipeline was developed in FY20 that showed the datasets contained adequate signal for continuation of efforts. In FY21, a modular demonstration prototype modeling pipeline has continued in development for two text-based data sources: a broad internet archive (Webhose Ltd.) and a decahose Twitter database (i.e., a global sampling of one in every ten Tweets). The techniques that have been developed rely on graph theory and anomaly detection to identify contextual shifts in key words and phrases at various points in time such that indicators of events of interest could be identified and subsequently, events could be extracted from the corpuses. The foundational concept behind the approaches is that contextual shifts in key words and phrases can act as indicators of events of interest. Both datasets have proven successful in extracting events of interest related to pit production at the Savannah River Site prior to the official announcement. In addition, the pipelines have generated a wide range of events broadly summarized as: the awarding of DOE contracts at major sites, DOE investments in various programs, accidents at DOE national laboratories, speculations about the fate of pit production in the DOE complex, domestic and international shipments and receipts of nuclear materials at DOE sites, termination of non-proliferation agreements with Russia, termination of MOX, new weapons development approvals/testing, nuclear posture reviews, major DOE cleanup/production milestones, political opinions, and nuclear watch groups’ opinions, among many others.
97 MATHEMATICS AND COMPUTING↗
Attention-based explainability for structure–property relationships
Machine learning methods are emerging as a universal paradigm for constructing correlative structure–property relationships in materials science based on multimodal characterization. However, this necessitates the development of methods for the physical interpretability of the resulting correlative models. Here, we demonstrate the potential of attention-based neural networks for revealing structure–property relationships and the underlying physical mechanisms, using the ferroelectric properties of PbTiO3 thin films as a case study. Through the analysis of attention scores, we disentangle the influence of distinct domain patterns on the polarization switching process. The attention-based Transformer model is explored both as a direct interpretability tool and as a surrogate for explaining representations learned via unsupervised machine learning, enabling the identification of physically grounded correlations. We compare attention-derived interpretability scores with classical SHapley Additive exPlanations analysis and show that, in contrast to applications in natural language processing, attention mechanisms in materials science exhibit high efficiency in highlighting meaningful structural features.
Slautin, Boris [Independent Researcher]↗
Application of machine learning approaches in the analysis of mass absorption cross-section of black carbon aerosols: Aerosol composition dependencies and sensitivity analyses
Physics-based models typically require an in-depth understanding of a phenomenon and assumptions of the underlying process(es), which are often hard to obtain in practice, whereas data-driven machine learning models learn the structure and patterns in the training data without any prior theoretical assumptions and then use inference to develop useful predictions. A novel machine learning-based algorithm has been previously developed for the prediction of black carbon mass absorption cross-section (MAC BC ) and applied to a variety of different atmospheric environments. In contrast to light scattering theories which require assumptions about the underlying physics, this algorithm uses time-series data of aerosol properties to estimate the temporally-varying MAC BC at 870 nm. Here, we analyze our algorithm and discuss the influence of aerosol optical properties (such as Ångström exponents and single scattering albedo) and chemical composition on the model outputs and the associated accuracy. Additionally, we conduct sensitivity analyses on our models to understand how the predictions change in response to different sets of input variables. Our support vector machine (SVM) for regression model is the least sensitive to variations in the input variables, although all models tend to exhibit a degradation to their accuracy when scattering Ångström exponents are less than one.
54 ENVIRONMENTAL SCIENCES↗
Anatomy of Continuous Mars SEIS and Pressure Data from Unsupervised Learning
The seismic noise recorded by the Interior Exploration using Seismic Investigations, Geodesy, and Heat Transport (InSight) seismometer (Seismic Experiment for Interior Structure [SEIS]) has a strong daily quasi-periodicity and numerous transient microevents, associated mostly with an active Martian environment with wind bursts, pressure drops, in addition to thermally induced lander and instrument cracks. That noise is far from the Earth’s microseismic noise. Quantifying the importance of nonstochasticity and identifying these microevents is mandatory for improving continuous data quality and noise analysis techniques, including autocorrelation. Cataloging these events has so far been made with specific algorithms and operator’s visual inspection. We investigate here the continuous data with an unsupervised deep-learning approach built on a deep scattering network. This leads to the successful detection and clustering of these microevents as well as better determination of daily cycles associated with changes in the intensity and color of the background noise. We first provide a description of our approach, and then present the learned clusters followed by a study of their origin and associated physical phenomena. We show that the clustering is robust over several Martian days, showing distinct types of glitches that repeat at a rate of several tens per sol with stable time differences. We show that the clustering and detection efficiency for pressure drops and glitches is comparable to or better than manual or targeted detection techniques proposed to date, noticeably with an unsupervised approach. Finally, here we discuss the origin of other clusters found, especially glitch sequences with stable time offsets that might generate artifacts in autocorrelation analyses. We conclude with presenting the potential of unsupervised learning for long-term space mission operations, in particular, for geophysical and environmental observatories.
58 GEOSCIENCES↗
Artificial Intelligence Applications in Renewable Energy
Addressing new methodologies in deep learning (DL), machine learning (ML) and artificial intelligence (AI), the webinar speakers will provide an overview of the literature spanning these three overlapping fields as applied to energy systems research. The audience will learn how developments in these areas have added new capabilities for pattern recognition and predictive modeling that are complementary to more traditional modeling approaches used in energy systems research. The speakers will illustrate several use cases of these new approaches in the energy space, such as physics-guided neural networks to improve ML regressions for solar data and how AI agents can be used to explore power plant operations.
40 EE - Geothermal Technologies Office (EE-4G)↗
Multi-Area Distribution System State Estimation Using Decentralized Physics-Aware Neural Networks
The development of active distribution grids requires more accurate and lower computational cost state estimation. In this paper, the authors investigate a decentralized learning-based distribution system state estimation (DSSE) approach for large distribution grids. The proposed approach decomposes the feeder-level DSSE into subarea-level estimation problems that can be solved independently. The proposed method is decentralized pruned physics-aware neural network (D-P2N2). The physical grid topology is used to parsimoniously design the connections between different hidden layers of the D-P2N2. Monte Carlo simulations based on one-year of load consumption data collected from smart meters for a three-phase distribution system power flow are developed to generate the measurement and voltage state data. The IEEE 123-node system is selected as the test network to benchmark the proposed algorithm against the classic weighted least squares and state-of-the-art learning-based DSSE approaches. Numerical results show that the D-P2N2 outperforms the state-of-the-art methods in terms of estimation accuracy and computational efficiency.
24 POWER TRANSMISSION AND DISTRIBUTION↗
Active and Transfer Learning of High-Dimensional Neural Network Potentials for Transition Metals
Classical molecular dynamics (MD) simulations represent a very popular and powerful tool for materials modeling and design. The predictive power of MD hinges on the ability of the interatomic potential to capture the underlying physics and chemistry. There have been decades of seminal work on developing interatomic potentials, albeit with a focus predominantly on capturing the properties of bulk materials. Such physics-based models, while extensively deployed for predicting the dynamics and properties of nanoscale systems over the past two decades, tend to perform poorly in predicting nanoscale potential energy surfaces (PESs) when compared to high-fidelity first-principles calculations. These limitations stem from the lack of flexibility in such models, which rely on a predefined functional form. Machine learning (ML) models and approaches have emerged as a viable alternative to capture the diverse size-dependent cluster geometries, nanoscale dynamics, and the complex nanoscale PESs, without sacrificing the bulk properties. Here, in this study, we introduce an ML workflow that combines transfer and active learning strategies to develop high-dimensional neural networks (NNs) for capturing the cluster and bulk properties for several different transition metals with applications in catalysis, microelectronics, and energy storage, to name a few. Our NN first learns the bulk PES from the high-quality physics-based models in literature and subsequently augments this learning via retraining with a higher-fidelity first-principles training data set to concurrently capture both the nanoscale and bulk PES. Our workflow departs from status-quo in its ability to learn from a sparsely sampled data set that nonetheless covers a diverse range of cluster configurations from near-equilibrium to highly nonequilibrium as well as learning strategies that iteratively improve the fingerprinting depending on model fidelity. All the developed models are rigorously tested against an extensive first-principles data set of energies and forces of cluster configurations as well as several properties of bulk configurations for 10 different transition metals. Our approach is material agnostic and provides a methodology to transfer and build upon the learnings from decades of seminal work in molecular simulations on to a new generation of ML-trained potentials to accelerate materials discovery and design.
36 MATERIALS SCIENCE↗
Machine learning for the advancement of membrane science and technology: A critical review
Not Available
37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
Communication Network Layer State Estimation Measurement Model for a Cyber-Secure Smart Grid
Network communication has been proven to be a very important tool and a key factor in the recent development and progress of the power grid operation. It is also considered as the foundation for the smart grid because information and communication are integrated into electricity distribution to achieve reliable and accurate knowledge of the power grid. In previous years, absorbing energy from substations and delivering it to customers was the only type of interaction we knew between utility companies and customers. Presently, the growing connections of small distributed generation units caused by the cost reduction of most of the technologies used in generation and storage of electrical energy, along with the potential benefits of renewable energy have pushed many researchers to look into the improvement of information and communication technologies (ICT) in order to ensure a bidirectional flow of power and data. Moreover, the evolution of information and communication technologies and its applications to smart grid have converted the smart grid into a cyber-physical system where vulnerabilities and additional security challenges such as cyber-threats and cyber-attacks have emerged. Previously, we have demonstrated that using machine learning-based processing on data gathered from communication networks and the power grid was a promising solution for detecting cyber threats by implementing a co-simulation of cyber-security for cross-layer strategy. Since the majority of the challenges observed can only be solved in the network communication layer, we present in this work a physics-based state estimation model of the communication network system towards enhanced cyber-physical security of the smart grid. Information integration with the previously developed machine learning model is developed, providing a enhanced cyber-physical security application for the smart grid. Easy-to-implement model, without hard-to-derive parameters, highlight potential aspects of the model for real-life applications.
Mathieu, Reynold↗