Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “high performance space computing”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 631 records · Page 35

Development of a Hydronic Rooftop Unit-HyPak-MA

The majority of U.S. commercial floor space is cooled by rooftop HVAC units (RTUs). RTU popularity derives chiefly from their low initial cost and relative ease of service access without disturbing building occupants. Unfortunately, current RTUs are inherently inefficient due to a combination of characteristics that unnecessarily increase cooling loads and energy use. 36% percent of annual U.S. energy, and two-thirds of electricity, is consumed in and by buildings. Commercial buildings consume approximately 4.2 quads of energy each year at a cost of $230 billion per year, with HVAC equipment consuming 1.2 quads of electricity. More than half of all U.S. commercial floor space is cooled by packaged HVAC units, most of which are rooftop units (RTUs). Inefficient RTUs create an estimated 3.5% of U.S. CO{sub 2} emissions, thus contributing significantly to global warming5. Also, RTUs often fail to maintain adequate ventilation air and air filtration, reducing indoor air quality. This is the second HyPak project to be supported by DOE through NETL. The prior project, referred to as HyPak-1 in this report, had two rounds of prototype fabrication and testing as well as computer modeling and market research. The HyPak-1 prototypes demonstrated the high performance capabilities of the HyPak concept, but made it clear that further development was required to reduce heat exchanger cost and improve system reliability before HyPak commercialization can commence. The HyPak-1 prototypes were limited to about 25% ventilation air fraction, limiting performance and marketability. The current project is intended to develop a 'mixed-air' product that is capable of full 0-100% modulation in ventilation air fraction, hence it was referred to as HyPak-MA in the proposal. (For simplicity, the -MA has been dropped when referencing the current project.) The objective of the HyPak Project is to design, develop and test a hydronic RTU that provides a quantum improvement over conventional RTU performance. Our proposal targeted 60% and 50% reduction in electrical energy use by the HyPak RTU for dry and humid climates, respectively, when compared with a conventional unit, and reduction in peak energy consumption of 50% and 33% respectively. In addition to performance targets, our goal is to develop a production-ready design with durability, reliability and maintainability similar to air-cooled packaged equipment, and that can be commercialized immediately following the conclusion of this project.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

Defects in SiC for Quantum Computing

Many novel materials are being actively considered for quantum information science and for realizing high-performance qubit operation at room temperature. It is known that deep defects in wide-band gap semiconductors can have spin states and long coherence times suitable for qubit operation. We theoretically investigate from ab-initio density functional theory (DFT) that the defect states in the hexagonal silicon carbide (4H-SiC) are potential qubit materials. The DFT supercell calculations were performed with the local-orbital and pseudopotential methods including hybrid exchange-correlation functionals. Di-vacancies in SiC supercells yielded defect levels in the gap consisting of closely spaced doublet just above the valence band edge, and higher levels in the band gap. The divacancy with a spin state of 1 is charge neutral. Furthermore, the divacancy is characterized by C-dangling bonds and a Si-dangling bonds. Jahn-teller distortions and formation energies as a function of the Fermi level and single photon interactions with these defect levels will be discussed. In contrast, the anti-site defects where C, Si are interchanged have high formation energies of 5.4 eV and have just a single shallow defect level close to the valence band edge, with no spin. We will compare results including the defect levels from both the electronic structure approaches.

36 MATERIALS SCIENCE↗

Are ultra-faint galaxies at z = 6–8 responsible for cosmic reionization? Combined constraints from the Hubble Frontier Fields clusters and parallels

We use deep Hubble Space Telescope imaging of the Frontier Fields to accurately measure the galaxy rest-frame ultraviolet luminosity function (UV LF) in the redshift range z ∼ 6–8. We combine observations in three lensing clusters, A2744, MACS 0416, and MACS 0717, and their associated parallel fields to select high-redshift dropout candidates. We use the latest lensing models to estimate the flux magnification and the effective survey volume in combination with completeness simulations performed in the source plane. We report the detection of 227 galaxy candidates at z = 6–7 and 25 candidates at z ∼ 8. While the total survey area is about 4 arcmin{sup 2} in each parallel field, it drops to about 0.6–1 arcmin{sup 2} in the cluster core fields because of the strong lensing. We compute the UV LF at z ∼ 7 using the combined galaxy sample and perform Monte Carlo simulations to determine the best-fit Schechter parameters. We are able to reliably constrain the LF down to an absolute magnitude of M{sub UV} = −15.25, which corresponds to 0.005 L{sup ⋆}. More importantly, we find that the faint-end slope remains steep down to this magnitude limit with α=−2.04{sub −0.17}{sup +0.13}. We find a characteristic magnitude of M{sup ⋆}=−20.89{sub −0.72}{sup +0.60} and log(ϕ{sup ⋆}) = −3.54{sub −0.45}{sup +0.48}. Our results confirm the most recent results in deep blank fields but extend the LF measurements more than two magnitudes deeper. The UV LF at z ∼ 8 is not very well constrained below M{sub UV} = −18 owing to the small number statistics and incompleteness uncertainties. To assess the contribution of galaxies to cosmic reionization, we derive the UV luminosity density at z ∼ 7 by integrating the UV LF down to an observational limit of M{sub UV} = −15. We show that our determination of log(ρ{sub UV}) = 26.2 ± 0.13 (erg s{sup −1} Hz{sup −1} Mpc{sup −3}) can be sufficient to maintain reionization with an escape fraction of ionizing radiation of f{sub esc} = 10%–15%. Future Hubble Frontier Fields observations will certainly improve the constraints on the UV LF at the epoch of reionization, paving the way to more ambitious programs using cosmic telescopes with the next generation of large aperture telescopes such as the James Webb Space Telescope and the European Extremely Large Telescope.

79 ASTRONOMY AND ASTROPHYSICS↗

Calculation of Friction Factors and Nusselt Numbers for Twisted Elliptical Tube Heat Exchangers Using Nek5000

Twisted elliptical tube geometries for heat exchanger design are expected to offer significant enhancement in heat transfer with a marginal increase in frictional losses. The presented work focuses on computational fluid dynamics (CFD) simulations of heated molten salt flows through twisted elliptical tube geometries at low modified Froude numbers. In this work, the objectives are to evaluate the available correlations at low Froude numbers and to determine the impact of using a small gap to resolve contact points. The spectral element CFD code Nek5000 was used for all simulations, which were performed in periodic domains of triangular and square unit cells surrounding a single tube through a complete twist using an explicit filtering large eddy simulation (LES) method. The simulation methodology was validated by comparing to the experimental data of Tan et al. Excellent agreement was observed for this comparison. The Dzyubenko correlation for transitional flow was evaluated for a range of Froude numbers. This correlation demonstrated good agreement with LES results at high modified Froude number. As Froude number was decreased toward the bounds of the correlation, the agreement worsened. Cases were then simulated at low Froude number, testing the effects of tube spacing. It was determined that the laminar case for the square unit cell is the most affected by increasing gap size, which should be minimized to mitigate the effect. In the triangular unit cell, the laminar flow regime is also more significantly impacted by increasing gap size compared to the turbulent flow regime which was only marginally impacted.

42 ENGINEERING↗

Second-Order Møller–Plesset Perturbation Theory in the Condensed Phase: An Efficient and Massively Parallel Gaussian and Plane Waves Approach

A novel algorithm, based on a hybrid Gaussian and plane waves (GPW) approach, is developed for the canonical second-order Møller–Plesset perturbation energy (MP 2 ) of finite and extended systems. The key aspect of the method is that the electron repulsion integrals (ia|λσ) are computed by direct integration between the products of Gaussian basis functions λσ and the electrostatic potential arising from a given occupied-virtual pair density ia. The electrostatic potential is obtained in a plane waves basis set after solving the Poisson equation in Fourier space. In particular, for condensed phase systems, this scheme is highly efficient. Furthermore, our implementation has low memory requirements and displays excellent parallel scalability up to 100 000 processes. In this way, canonical MP 2 calculations for condensed phase systems containing hundreds of atoms or more than 5000 basis functions can be performed within minutes, while systems up to 1000 atoms and 10 000 basis functions remain feasible. Solid LiH has been employed as a benchmark to study basis set and system size convergence. Lattice constants and cohesive energies of various molecular crystals have been studied with MP 2 and double-hybrid functionals.

Chemistry↗

Report on ISR-1 High-Altitude Balloon Flight

To test small technologies at lower cost for space science applications, LANL has developed a small high altitude balloon payload that could, in the future, be regularly and inexpensively launched from LANL. A neutron detector, NEMO, was integrated to evaluate its performance in a space-like mixed-radiation environment and collect neutron data in the atmosphere. In collaboration with EES-14, a high-altitude balloon payload was launched from LANL Technical Area 51 on February 27, 2023 and April 17, 2023. For real-time geolocation, a SAM-M8Q M8 GNSS module was used to get position and time, and an Iridium RockBLOCK 9603 was used to communicate with the ground using the Iridium satellite fleet. These modules were all controlled using an Iteaduino Mega microcontroller board. Finally, a High Altitude Science Eagle Flight Computer with a temperature pressure sensor ran independently, writing data to an SD card. All of these modules were powered by a 5 mAh lithium polymer battery. The battery was attached to the bottom of the payload while the remaining electronics were embedded in the underside of the top of the payload. These modules were wired as seen in Figure 1-2. The Iteaduino Mega microcontroller board was programmed to use the RockBLOCK to send a message once every 10 minutes containing neutron and GPS data read off the NEMO and SAM M8Q, respectively. Once the message send attempt finished, the RockBLOCK would be slept for the rest of the 10 minute interval. The Eagle Flight Computer ran continuously throughout the flight, taking data every 6 seconds. The RockBLOCK message data was set up to be delivered from the Iridium satellite fleet to a website, where it was stored and parsed to create live maps and plots for analysis and balloon retrieval. The RockBLOCK message data was additionally configured to be sent to an email as a fail-safe. The payload was ground-tested successfully for over 50 hours, with multiple revisions occurring to best prepare for conditions at altitude and improve the software and firmware to fix any issues that cropped up with the data pipeline. Additional to the balloon payload, the flight had an attached iMet-4 radiosonde and Garmin T5 GPS Dog Collar. The radiosonde provided GPS and meteorological data. The T5 dog collar is used along with a Garmin Astro 430 to track the balloon at a range of up to 9 miles for retrieval. The balloon itself was initially a 1600 g meteorological balloon with an attached High Altitude Science parachute, both of which can be seen in Figure 1-3. After the first flight, the EES team swapped to a Rocketman parachute.

42 ENGINEERING↗

Initial Development of Fusion Magnet Simulation Capabilities for Performance and Safety Evaluation Using the MOOSE Framework

Fusion energy holds the promise of being a transformative technology as a carbon-neutral, sustainable source of energy. Whole device modeling and the development of fusion digital twins will be increasingly important for emerging fusion device concepts at both national laboratories and within the commercial fusion industry. However, meeting the challenge of whole device modeling of fusion energy devices requires robust, multiphysics, multiscale modeling and simulation technologies capable of running on large-scale supercomputers. Detailed analysis of individual systems at-scale is also required to ensure safe and efficient operation as well as provide the safety basis for future device designs and licensing activities. In a tokamak, toroidal and poloidal magnets confine and shape the fusion plasma to promote the fusion reaction. High plasma temperatures and high magnetic field requirements in modern design concepts (leading to high amounts of energy stored within each magnet) impose electrical, thermal, and mechanical loads on the magnet components, which in turn impacts the safety considerations of the magnet and their supporting systems. Idaho National Laboratory (INL) has a history of working in this space, including development and benchmarking of the Magnetic System Circuitry Analysis Program (MSCAP) and Magnet Arcing (MAGARC) codes to study magnet quench events; notably, MAGARC was used to study quenching during the ITER Engineering Design Activity. However, these legacy codes and capabilities are not parallel and scalable, and new tools are required for future advances in this area, which leads to the INL-developed Multiphysics Object-Oriented Simulation Environment (MOOSE) framework. Developed originally for fission reactor systems under United States Department of Energy, Office of Nuclear Energy modeling and simulation programs, the MOOSE framework has been well-suited to multiscale, multiphysics modeling and simulation needs for nuclear systems. The framework is open-source, well-tested, under continuous development and deployment, and developed to a Nuclear Quality Assurance, Level 1 software quality standard. MOOSE has also been used in the fusion space previously in several projects: INL’s Tritium Migration Analysis Program, Version 8 (TMAP8) for tritium migration, UK Atomic Energy Authority’s A Unified Resource for OpenMC (fusion) Reactor Applications (AURORA) code for fusion thermo-mechanical and neutronics analysis, and Argonne National Laboratory’s Cardinal for high-fidelity computational fluid dynamics and neutronics. However, to model superconducting magnets, several MOOSE enhancements are required: additions to the current MOOSE electromagnetic capabilities, new material libraries for superconductors of interest (such as YBCO), as well as fusion-specific models for thermo-mechanics. This talk will discuss initial development activities to build these capabilities in MOOSE, focusing on initial validation and benchmarking activities. Proposed coupling workflows and future work to support the simulation of fusion magnets and magnet structural assemblies for performance and safety evaluation in MOOSE will also be discussed.

70 - PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Thermofield Theory for Finite-Temperature Electronic Structure

Wave function methods have offered a robust, systematically improvable means to study ground-state properties in quantum many-body systems. Theories like coupled cluster and their derivatives provide highly accurate approximations to the energy landscape at a reasonable computational cost. Analogues of such methods to study thermal properties, though highly desirable, have been lacking because evaluating thermal properties involve a trace over the entire Hilbert space, which is a formidable task. Besides, excited-state theories are generally not as well studied as ground-state ones. In this mini-review, we present an overview of a finite-temperature wave function formalism based on thermofield dynamics to overcome these difficulties. Thermofield dynamics allows us to map the equilibrium thermal density matrix to a pure state, i.e., a single wave function, albeit in an expanded Hilbert space. Ensemble averages become expectation values over this so-called thermal state. Around this thermal state, we have developed a procedure to generalize ground-state wave function theories to finite temperatures. As explicit examples, we highlight formulations of mean-field, configuration interaction, and coupled cluster theories for thermal properties of Fermions in the grand-canonical ensemble. To assess the quality of these approximations, we also show benchmark studies for the one-dimensional Hubbard model, while comparing against exact results. We will see that the thermal methods perform similarly to their ground-state counterparts, while merely adding a prefactor to the asymptotic computational cost. Furthermore, they also inherit all the properties, good or bad, from the ground-state methods, signifying the robustness of our formalism and the scope for future development.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

R and DE Robotic Sensor Intern Team: Advanced Sensor Suite Prototype

The Department of Energy (DOE) nuclear complex includes many hazardous and challenging environments that must be inspected remotely and routinely for structural integrity evaluation and long-term planning purposes. The H-Canyon Air Exhaust (HCAEX) Tunnel is an example of such an environment where a biannual inspection is currently performed. Presently inspections are performed by teleoperated robotic crawlers which provide low-res video feedback to the structural engineers. Advanced sensors: including Lidar and panospheric hi-res cameras have the potential to provide additional valuable data in future inspections. The primary goal of the sensor suite is to provide an internal 3D map of the HCAEX tunnel. Possible sensors were a stereo camera, a 3D Lidar, or a 2D Lidar that would need to be spun around. The issues with the stereo camera were that it does not have a large field-of-view and that its depth information was not nearly as accurate as even the baseline Lidars. The Hokuyo UTM-30LX-EW shown below provides a 270 deg. FOV with high accuracy and resolution while the largest FOV on a 3D Lidar is 40 deg. This was the reason a 2D Lidar was chosen over the 3D Lidar. Even though the 2D Lidar requires additional components and software to generate a 3D map, its FOV advantage is too great given that the HCAEX tunnel is a relatively confined space. Along with the Lidar, a camera sensor, IMU sensor, pan-tit unit, and on board computer needed to be identified. The two panospheric cameras allow for full 360 deg. FOV. The IMU will give data on the state of the robot inside the tunnel. The computer is used to gather and process data on board and to control the pan-tilt system. Mounts were designed and 3D printed to build the prototype of the sensor suite. It was designed to have the Lidar on top With its full FOV free while having the camera below and as close to together as possible while still leaving much of their FOV visible. The design also managed the cables such that they were unlikely to be damaged if in tension and so that they did not intrude much on the cameras vision. The sensor tree developed for the H-Canyon tunnel inspection project produces a large quantity of camera and Lidar data. a postprocessing pipeline was created to facilitate the analysis of these data. The pipeline follows the following routine: 1) Registration - Salient local features in subsequent scans are identified and used to align the scans With one another. This allows production of a single cohesive map of the tunnel. 2) Sensor Fusion - RGB camera data is 'painted' onto the depth data created by the Lidar. 3) Segmentation - Individual surfaces from within the scene are segmented out separately for individual analysis. 4) Degradation Analysis - Automatic mapping of local deviation from the plane in position or surface normal is performed, as is detection of exposed rebar using Lidar return intensity. 2) Obstacle Detection - Lidar data is also used in real time to detect obstacles in the environment. This could give feedback to the operator to successfully drive in the tunnel. 6) Graphic User Interface - a custom GUI was created to facilitate customer use of the system. The system provides real-time feedback from both the cameras and Lidar. The Lidar point-cloud data is also used for obstacle detection. As the Lidar spins, various algorithms are run to analyze the current 'slice' and obstacles in the environment are determined and registered on a cost-map. A Graphical User Interface (GUI) was developed to allow the user to easily command the crawler to set the rotational speed of the Lidar, perform a scan, and register and save the 3D generated cloud after a scan is done. For the good of the crawler, a critical aspect of the GUI design is to allow the user to cancel a scan any time after it is commanded and before it is saved. The sensor trees depth accuracy was evaluated in comparison to a high end commercial FARO Terrestrial Laser Scanner (TLS) with specifications for millimeter accuracy at distances above 10 m. For the purposes of these tests, the depth measurements produced by the TLS were taken as ground truth. Example test wall panels were scanned with the TLS and postprocessed with its proprietary software, and subsequently scanned with the sensor tree system. The two scans were manually aligned and for each point in the output cloud from the sensor tree, the distance to the nearest ground truth point in the FARO cloud was determined. 98.7% of all produced points were within the specified target range of 6 mm, with an average deviation of 1.75 mm. The test walls used have an overall in-wall depth variation of around 40 mm.

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗

Large Eddy Simulation of Convective Heat Transfer in a Random Pebble Bed Using the Spectral Element Method

The development of fluoride-cooled high-temperature reactors has drastically increased the demand for an in-depth understanding of the heat transfer (HT) in packed beds cooled by liquid salts. The complex flow fields and space-dependent porosity found in a pebble bed require a detailed understanding to ensure the proper cooling of the reactor core during normal and accident conditions. As detailed experimental data are complicated to obtain for these configurations, high-fidelity simulation such as large eddy simulation and direct numerical simulation (DNS) can be employed to create a high-resolution heat transfer numerical database that can assist in addressing industrial-driven issues associated with the heat transfer behavior of fluoride-cooled high-temperature reactors. In this paper, we performed a series of large eddy simulation using computational fluid dynamics (CFD) code NekRS to investigate the heat transfer for a bed of 1741 pebbles. Further, the characteristics of the flow, such as average, rms, and time series of velocity and temperature, have been analyzed. Porous media averages have also been performed. The simulation results show a good agreement between non-conjugate heat transfer and conjugate heat transfer. The generated data will be used to benchmark heat transfer modeling methods and local maxima/minima of heat transfer parameters. It will also be used for supporting convective heat transfer quantification for Kairos Power and benchmarking lower fidelity models.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

Lorentz group equivariant autoencoders

Abstract There has been significant work recently in developing machine learning (ML) models in high energy physics (HEP) for tasks such as classification, simulation, and anomaly detection. Often these models are adapted from those designed for datasets in computer vision or natural language processing, which lack inductive biases suited to HEP data, such as equivariance to its inherent symmetries. Such biases have been shown to make models more performant and interpretable, and reduce the amount of training data needed. To that end, we develop the Lorentz group autoencoder (LGAE), an autoencoder model equivariant with respect to the proper, orthochronous Lorentz group $$\textrm{SO}^+(3,1)$$ SO + ( 3 , 1 ) , with a latent space living in the representations of the group. We present our architecture and several experimental results on jets at the LHC and find it outperforms graph and convolutional neural network baseline models on several compression, reconstruction, and anomaly detection metrics. We also demonstrate the advantage of such an equivariant model in analyzing the latent space of the autoencoder, which can improve the explainability of potential anomalies discovered by such ML models.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

A Quantum Analog Coprocessor for Correlated Electron Systems Simulation

Analog quantum simulation is an approach for studying physical systems that might otherwise be computationally intractable to simulate on classical high-performance computing (HPC) systems. The key idea behind analog quantum simulation is the realization of a physical system with a low-energy effective Hamiltonian that is the same as the low-energy effective Hamiltonian of some target system to be studied. Purpose-built nanoelectronic devices are a natural candidate for implementing the analog quantum simulation of strongly correlated materials that are otherwise challenging to study using classical HPC systems. However, realizing devices that are sufficiently large to study the properties of a non-trivial material system (e.g., those described by a Fermi-Hubbard model) will eventually require the fabrication, control, and measurement of at least 0(10) quantum dots, or other engineered quantum impurities. As a step toward large-scale analog or digital quantum simulation platforms based on nanoelectronic devices, we propose a new approach to analog quantum simulation that makes use of the large Hilbert space dimension of the electronic baths that are used to adjust the occupancy of one or a few engineered quantum impurities. This approach to analog quantum simulation allows us to study a wide array of quantum impurity models. We can further augment the computational power of such an approach by combining it with a classical computer to facilitate dynamical mean-field theory (DMFT) calculations. DMFT replaces the solution of a lattice impurity problem with the solution of a family of localized impurity problems with bath couplings that are adjusted to satisfy a self-consistency condition between the two models. In DMFT, the computationally challenging task is the high-accuracy solution of an instance of a quantum impurity model that is determined self-consistently in coordination with a mean-field calculation. We propose using one or a few engineered quantum impurities with adjustable couplings to baths to realize an analog quantum coprocessor that effects the solution of such a model through measurements of a physical quantum impurity, operating in coordination with a classical computer to achieve a self-consistent solution to a DMFT calculation. We focus on implementation details relevant to a number of technologies for which Sandia has design, fabrication, and measurement expertise. The primary technical advances outlined in this report concern the development of a supporting modeling capability. As with all analog quantum simulation platforms, the successful design and operation of individual devices depends critically on one's ability to predict the effective low-energy Hamiltonian governing its dynamics Our project has made this possible and lays the foundation for future experimental implementations.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

GAINN: The Galaxy Assembly and Interaction Neural Networks for High-redshift JWST Observations

We present the Galaxy Assembly and Interaction Neural Networks (Gainn), a series of artificial neural networks for predicting the redshift, stellar mass, halo mass, and mass-weighted age of simulated galaxies based on James Webb Space Telescope (JWST) photometry. Our goal is to determine the best neural network for predicting these variables at 11 < z < 15. The parameters of the optimal neural network can then be used to estimate these variables for real, observed galaxies. The inputs of the neural networks are JWST filter magnitudes of a subset of five broadband filters (F150W, F200W, F277W, F356W, and F444W) and two medium-band filters (F162M and F182M). We compare the performance of the neural networks using different combinations of these filters, as well as different activation functions and numbers of layers. The best neural network predicted redshift with a normalized rms error of $0.010^{+0.003}_{-0.001}$, stellar mass with rms = $0.089^{+0.044}_{-0.022}$, halo mass with a mean-squared error of $0.022^{+0.014}_{-0.008}$, and mass-weighted age with rms = $12.466^{+5.065}_{-2.408}$. We also test the performance of Gainn on real data from MACS0647JD, an object observed by JWST. Predictions from Gainn for the first projection of the object (JD1) have normalized bias $\langle$Δz$\rangle$ < 0.00228, which is significantly smaller than found with template-fitting methods. We find that the optimal filter combination is F277W, F356W, F162M, and F200W when considering both theoretical accuracy and observational resources from JWST.

97 MATHEMATICS AND COMPUTING↗

Problematic projection to the in-sample subspace for a kernelized anomaly detector

We examine the properties and performance of kernelized anomaly detectors, with an emphasis on the Mahalanobis-distance-based kernel RX (KRX) algorithm. Although the detector generally performs well for high-bandwidth Gaussian kernels, it exhibits problematic (in some cases, catastrophic) performance for distances that are large compared to the bandwidth. By comparing KRX to two other anomaly detectors, we can trace the problem to a projection in feature space, which arises when a pseudoinverse is used on the covariance matrix in that feature space. Here, we show that a regularized variant of KRX overcomes this difficulty and achieves superior performance over a wide range of bandwidths.

58 GEOSCIENCES↗

Spectral Analysis of Regular Material Point Method and its Application to Study High Pressure Reverse Osmosis Membrane Compaction and Embossing

Material Point Method (MPM) is gaining widespread interest in applied continuum mechanics. The fact that all the continuum properties are stored on the particles (or material points) and the governing equations are solved on these material points makes MPM extremely suited to problems involving severe material deformations, such as crack propagation, soil movement, and fluid flows. Despite its popularity, only a few studies have focused on the numerical properties of MPM. This presentation introduces a global spectral analysis of the regular material point method. Contrary to previous studies, the analysis focuses on the numerical properties of the method in the spectral space. The amplification factor is derived as a function of the non- dimensional wave numbers. It provides insights into the stability and dissipative properties of the method for various CFL and Fourier numbers. The effect of the grid shape functions, number of particles per cell and their locations inside the grid cell are also analyzed. The EXAGOOP MPM solver (https://github.com/NREL/Exagoop.git) is developed at the National Renewable Energy Laboratory as a part of the NAWI UHPRO project and is based on the AMReX framework. A single-level, uniform cartesian grid is used as the background mesh, while the particle class in AMReX is used to manage the material point operations. Linear hat and B-splines are used as grid shape functions, while the time integration is performed using explicit Euler time integration. EXAGOOP is both CPU and GPU compatible and has been demonstrated to work well on multiple compute architectures. The performance of EXAGOOP on various computing architectures is presented along with its application to study compaction and embossing of high-pressure reverse osmosis membranes. The MPM solution accurately reproduces the membrane deformation. The deformed pore size and structure simulated using MPM also agree well with experimental SEM images.

material point method↗

Iterative sampling of expensive simulations for faster deep surrogate training

Deep neural network (DNN) surrogates of expensive physics simulations are enabling a rapid change in the way that common experimental design and analysis tasks are approached. Surrogate models allow simulations to be performed in parallel and separately from downstream tasks, thereby enabling analyses that would be impossible with the simulation in-the-loop; surrogates based on DNNs can effectively emulate diverse non-scalar data of the types collected in fusion and laboratory-astrophysics experiments. The challenge is in training the surrogate model, for which large ensembles of physics simulations must be run, preferably without wasting computational effort on uninteresting simulations. Here, in this paper, we present an iterative sampling scheme that can preferentially propose simulations in interesting regions of parameter space without neglecting unexplored regions, allowing high-quality and wide-ranging surrogate models to be trained using 2–3 times fewer simulations compare to space-filling designs. Our approach uses an explicit importance function defined on the simulation output space, balanced against a measure of simulation density which serves as a proxy for surrogate accuracy. It is easy to implement and can be tuned to find interesting simulations early in the study, allowing surrogates to be trained quickly and refined as new simulations become available; this represents an important step towards the routine generation of deep surrogate models quickly enough to be truly relevant to experimental work.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Multi-fidelity physics-informed machine learning for probabilistic damage diagnosis

Machine learning (ML) models are gaining popularity in structural health monitoring (SHM) because of their ability to learn the complex relationship between damage and sensor data. However, the lack of sufficient experimental data for structures with different degrees of damage is a key problem in training ML models for SHM. This problem can be alleviated by using physics-based models to generate the required training data to build physics-informed ML (PIML) models for SHM. However, it takes significant computational effort to perform enough high-fidelity simulations of the diagnostic test. It is thus desirable to know whether the available computational resource budget should be expended on numerous low-fidelity physics simulations, or a small number of high-fidelity simulations, or their combination. In this paper, we investigate this aspect of generating adequate training data for PIML, by constructing multi-fidelity PIML models. We evaluate the performance of several PIML models, trained with different amounts of low-fidelity and high-fidelity data, in locating hidden cracks in concrete structures using a nonlinear dynamics-based diagnosis technique. Here, we find that high-fidelity physics simulations that do not cover the (test and damage) parameter space do not improve the performance of diagnostic PIML models built using data from many low-fidelity physics simulations.

42 ENGINEERING↗

ENHANCED LENSING RATE BY CLUSTERING OF MASSIVE GALAXIES: NEWLY DISCOVERED SYSTEMS IN THE SLACS FIELDS

Galaxy-scale strong gravitational lens systems are useful for a variety of astrophysical applications. However, their use is limited by the relatively small samples of lenses known to date. It is thus important to develop efficient ways to discover new systems both in present and forthcoming data sets. For future large high-resolution imaging surveys we anticipate an ever-growing need for efficiency and for independence from spectroscopic data. In this paper, we exploit the clustering of massive galaxies to perform a high-efficiency imaging search for gravitational lenses. Our data set comprises 44 fields imaged by the Hubble Space Telescope (HST) Advanced Camera for Surveys (ACS), each of which is centered on a lens discovered by the Strong Lens ACS Survey (SLACS). We compare four different search methods: (1) automated detection with the HST Archive Galaxy-scale Gravitational Lens Survey (HAGGLeS) robot, (2) examining cutout images of bright galaxies (BGs) after subtraction of a smooth galaxy light distribution, (3) examining the unsubtracted BG cutouts, and (4) performing a full-frame visual inspection of the ACS images. We compute purity and completeness and consider investigator time for the four algorithms, using the main SLACS lenses as a testbed. The first and second algorithms perform the best. We present the four new lens systems discovered during this comprehensive search, as well as one other likely candidate. For each new lens we use the fundamental plane to estimate the lens velocity dispersion and predict, from the resulting lens geometry, the redshifts of the lensed sources. Two of these new systems are found in galaxy clusters, which include the SLACS lenses in the two respective fields. Overall we find that the enhanced lens abundance (30{sup +24} {sub -8} lenses deg{sup -2}) is higher than expected for random fields (12{sup +4} {sub -2} lenses deg{sup -2} for the COSMOS survey). Additionally, we find that the gravitational lenses we detect are qualitatively different from those in the parent SLACS sample: this imaging survey is largely probing higher redshift, and lower mass, early-type galaxies.

79 ASTRONOMY AND ASTROPHYSICS↗