Search NASA⌕ Search

SEARCH · Search NASA

Results for “differentiable simulation”

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 163 records · Page 9
All sourcesNASA NTRSDOE OSTI
DOE OSTI2025

QCD Theory Meets Information Theory

We present a novel technique to incorporate precision calculations from quantum chromodynamics into fully differential particle-level Monte Carlo simulations. By minimizing an information-theoretic quantity subject to constraints, our reweighted Monte Carlo incorporates systematic uncertainties absent in individual Monte Carlo predictions, achieving consistency with the theory input in precision and its estimated systematic uncertainties. Our method can be applied to arbitrary observables known from precision calculations, including multiple observables simultaneously. It generates strictly positive weights, thus offering a clear path to statistically powerful and theoretically precise computations for current and future collider experiments. As a proof of concept, we apply our technique to event-shape observables at electron-positron colliders, leveraging existing precision calculations of thrust. Our analysis highlights the importance of logarithmic moments of event shapes, which have not been previously studied in the collider physics literature.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2021

GeV-scale neutrinos: interactions with mesons and DUNE sensitivity

The simplest extension of the SM to account for the observed neutrino masses and mixings is the addition of at least two singlet fermions (or right-handed neutrinos). If their masses lie at or below the GeV scale, such new fermions would be produced in meson decays. Similarly, provided they are sufficiently heavy, their decay channels may involve mesons in the final state. Although the couplings between mesons and heavy neutrinos have been computed previously, significant discrepancies can be found in the literature. The aim of this paper is to clarify such discrepancies and provide consistent expressions for all relevant effective operators involving mesons with masses up to 2 GeV. Moreover, the effective Lagrangians obtained for both the Dirac and Majorana scenarios are made publicly available as FeynRules models so that fully differential event distributions can be easily simulated. As an application of our setup, we numerically compute the expected sensitivity of the DUNE near detector to these heavy neutral leptons.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗
DOE OSTI2022

Neuromorphic Computing is Turing-Complete

Neuromorphic computing is a non-von Neumann computing paradigm that performs computation by emulating the human brain. Neuromorphic systems are extremely energy-efficient and known to consume thousands of times less power than CPUs and GPUs. They have the potential to drive critical use cases such as autonomous vehicles, edge computing and internet of things in the future. For this reason, they are sought to be an indispensable part of the future computing landscape. Neuromorphic systems are mainly used for spike-based machine learning applications, although there are some non-machine learning applications in graph theory, differential equations, and spike-based simulations. These applications suggest that neuromorphic computing might be capable of general-purpose computing. However, general-purpose computability of neuromorphic computing has not been established yet. In this work, we prove that neuromorphic computing is Turing-complete and therefore capable of general-purpose computing. Specifically, we present a model of neuromorphic computing, with just two neuron parameters (threshold and leak), and two synaptic parameters (weight and delay). We devise neuromorphic circuits for computing all the μ-recursive functions (i.e., constant, successor and projection functions) and all the μ-recursive operators (i.e., composition, primitive recursion and minimization operators). Given that the μ-recursive functions and operators are precisely the ones that can be computed using a Turing machine, this work establishes the Turing-completeness of neuromorphic computing.

Date, Prasanna↗
DOE OSTI2020

MEUMAPPS (Microstructure Evolution Using Massively Parallel Phase-field Simulations)

The software “MEUMAPPS” is a h high-performance computing code used to simulate microstructure evolution associated with diffusional solid-state transformations in structural alloys. Understanding microstructure evolution during thermo-mechanical processing of structural alloys is the first step towards designing and processing alloys for specific applications by meeting property requirements demanded by the application. In this instance, the code is an important component in predicting processing-structure linkages during additive manufacturing of structural alloys as a function of processing parameters and alloy composition used in an additive manufacturing process. The code provides a detailed three-dimensional distribution of different types of phases / constituents and their morphologies, as well as the compositions of these phases that make up the microstructure. The microstructure is simulated by solving the governing partial differential equations using a Fourier Spectral Method.

Radhakrishnan, Balasubram↗
DOE OSTI2021

pnnl/pecblocks

ConverterBlocks uses the output of detailed electromagnetic transient (EMT) simulations to produce generalized block diagram models of power electronic systems. The process uses deep learning with customized block architectures. The outputs are reduced-order models that meet specified accuracy requirements, while providing important advantages over the original EMT models: ConverterBlock models have fewer nodes and can take longer time steps, resulting in shorter simulation times ConverterBlock models are continuously differentiable, making them compatible with control design methodsGeneralized block diagram modeling of power electronic converters for grid solar and storage applications.

McDermott, Tom↗
DOE OSTI2026

Fortran mimetic abstraction language (Formal) v0.1.

The Fortran mimetic abstraction language ("Formal") is a domain-specific language (DSL) embedded in Fortran 202Y [1]. Formal provides novel software abstractions for simulating phenomena governed by the partial differential equations (PDEs) of vector and tensor calculus. Such equations model an extremely broad set of physical phenomena, ranging from atmospheric winds to light propagation. Formal's data structures and algorithms mimic in form and behavior continuous functions and operators. Formal supports these mathematical constructs using mimetic discretizations that define a discrete calculus satisfying various tensor calculus theorems, thereby ensuring high-fidelity representations of the physics being modeled. [2] Formal 0.1.0 also lays a foundation for the future use of Fortran 202Y type-safe templates to facilitate the formal verification of tensor contractions in computational physics and artificial intelligence [3]. [1] "Fortran 202Y" is Fortran standard committee's informal designation for the next Fortran revision, which will likely be "Fortran 2028". [2] Corbino, J. and Castillo, J. (2020) Journal of Computational and Applied Mathematics, https://doi.org/10.1016/j.cam.2019.06.042. [3] Haveraaen, M., Järvi, J., & Rouson, D. (2019). Reflecting on Generics for Fortran. https://j3-fortran.org/doc/year/19/19-188.pdf.

Rouson, Damian [Lawrence Berkeley National Laborat↗
DOE OSTI2021

Position Papers for the ASCR Workshop on Cybersecurity and Privacy for Scientific Computing Ecosystems

At the request of the Department of Energy's (DOE) Office of Advanced Scientific Computing Research (ASCR), this program committee has been tasked with organizing a workshop to identify basic research needs in cybersecurity and privacy to better support DOE's science and energy mission. As part of the process, the program committee is soliciting community input in the form of position papers to help identify significant use cases, facility issues, and other barriers to enabling verifiably trustworthy computational science while preserving data confidentiality as appropriate for scientific workflows of interest to DOE. The program committee will review these position papers and based on the fit of their area of expertise and interest, selected contributors will have the opportunity to participate in the workshop currently planned as a virtual event November 3-5th, 2021. The thrust areas that will be explored by this workshop are the following: (1) Algorithms for secure, scalable, privacy-enhancing technologies and frameworks, including: Federated AI/ML, Differential privacy, Randomized algorithms, Adversarial modeling & simulation, Graph algorithms, and Formal methods; (2) Platforms to support the entire scientific-computing ecosystem, including edge computing for large-scale experiments, focusing on heterogeneous systems and distributed systems, including: Heterogeneous computing systems, Distributed computing systems, and Secure data architectures; and (3) Data workflows to allow agile use of data while preserving integrity and privacy, making the important properties verifiable either at runtime or post-computation, including: Integrity and provenance and Data management infrastructure. Topics that are out-of-scope for the workshop include discussing specific proposed solutions or areas that are clearly out of DOE's fundamental and applied-sciences mission scope, e.g., cryptography, enterprise security, and general-operations technology.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2021

Application of Scalar Filtered Density Function to Turbulent Flows Under Supercritical Condition

Abstract The scalar filtered density function (FDF) methodology is extended and employed for large eddy simulation (LES) of turbulent flows under supercritical condition. To describe real fluid behavior, the extended methodology incorporates the generalized heat and mass diffusion models along with real fluid thermodynamic relations which are derived using the cubic Peng–Robinson equation of state. These models are implemented within the stochastic differential equations comprising the scalar FDF transport. Simulations are conducted of a temporally developing mixing layer under supercritical condition and the results are assessed by comparing with data generated by direct numerical simulation (DNS) of the same layer. The consistency of the proposed FDF methodology is assessed. The LES-FDF predictions are shown to agree favorably with the DNS data and exhibit several key features pertaining to supercritical turbulent flows.

Energy & Fuels↗
DOE OSTI2025

Simulating High-Frequency Seismograms in Realistic Earth Models to Better Understand Source Discrimination Based on Differential Magnitudes ( M L− M c)

Discriminating low-yield underground nuclear explosions from small earthquakes is a key task in monitoring nuclear test ban treaties. P/S amplitude ratios have been an effective discriminant for moderate-sized events recorded at regional distances, but it is unclear if they are as effective in discriminating small seismic events recorded at local distances (<150 km). The difference between local magnitude (M L ) and coda duration magnitude (M c ) has been proposed as a new discriminant that may complement P/S amplitude ratios at local distances. Here, in this work, we calculate high-frequency (up to ∼4 Hz) synthetic seismograms at epicentral distances of 0–30 km in realistic models of the Salt Lake basin (Utah, United States) to better understand how variations in source type and depth affect M L −M c values. The Earth models incorporate simplified 1D and deterministic 3D structures, small-wavelength stochastic velocity perturbations, and surface topography. Coda waves are enhanced for the more complicated models compared to the base 1D model, but still underpredict observed durations by about a factor of two, which results in overprediction of amplitude to duration ratios (i.e., M L −M c values) for a near-surface explosion and a 7 km deep earthquake. For both source types, the predicted M L and M c values decrease as source depth increases, and M L −M c shows only minor variation with depth; however, M L −M c is on average ∼0.5 units smaller for explosions than earthquakes. This finding may imply that M L −M c has sensitivity to source type, in addition to being a depth discriminant, but more modeling is needed given the limitations of the current study. Future modeling should incorporate higher-frequency (≳5 Hz) simulations over a larger distance range (0–150 km), where M L and M c are commonly measured, while honoring low shear velocities (<300 m/s) near the surface and sampling a wider range of earthquake and explosion source mechanisms.

Hutchings, Sean J. [Univ. of Utah, Salt Lake City,↗
DOE OSTI2021

Deep Learning Approaches to Surrogates for Solving the Diffusion Equation for Mechanistic Real-World Simulations

In many mechanistic medical, biological, physical, and engineered spatiotemporal dynamic models the numerical solution of partial differential equations (PDEs), especially for diffusion, fluid flow and mechanical relaxation, can make simulations impractically slow. Biological models of tissues and organs often require the simultaneous calculation of the spatial variation of concentration of dozens of diffusing chemical species. One clinical example where rapid calculation of a diffusing field is of use is the estimation of oxygen gradients in the retina, based on imaging of the retinal vasculature, to guide surgical interventions in diabetic retinopathy. Furthermore, the ability to predict blood perfusion and oxygenation may one day guide clinical interventions in diverse settings, i.e., from stent placement in treating heart disease to BOLD fMRI interpretation in evaluating cognitive function (Xie et al., 2019; Lee et al., 2020). Since the quasi-steady-state solutions required for fast-diffusing chemical species like oxygen are particularly computationally costly, we consider the use of a neural network to provide an approximate solution to the steady-state diffusion equation. Machine learning surrogates, neural networks trained to provide approximate solutions to such complicated numerical problems, can often provide speed-ups of several orders of magnitude compared to direct calculation. Surrogates of PDEs could enable use of larger and more detailed models than are possible with direct calculation and can make including such simulations in real-time or near-real time workflows practical. Creating a surrogate requires running the direct calculation tens of thousands of times to generate training data and then training the neural network, both of which are computationally expensive. Often the practical applications of such models require thousands to millions of replica simulations, for example for parameter identification and uncertainty quantification, each of which gains speed from surrogate use and rapidly recovers the up-front costs of surrogate generation. We use a Convolutional Neural Network to approximate the stationary solution to the diffusion equation in the case of two equal-diameter, circular, constant-value sources located at random positions in a two-dimensional square domain with absorbing boundary conditions. Such a configuration caricatures the chemical concentration field of a fast-diffusing species like oxygen in a tissue with two parallel blood vessels in a cross section perpendicular to the two blood vessels. To improve convergence during training, we apply a training approach that uses roll-back to reject stochastic changes to the network that increase the loss function. The trained neural network approximation is about 1000 times faster than the direct calculation for individual replicas. Because different applications will have different criteria for acceptable approximation accuracy, we discuss a variety of loss functions and accuracy estimators that can help select the best network for a particular application. We briefly discuss some of the issues we encountered with overfitting, mismapping of the field values and the geometrical conditions that lead to large absolute and relative errors in the approximate solution.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2026

Differentiable multiphase flow model for physics-informed machine learning in reservoir pressure management

Accurate subsurface reservoir pressure control is extremely challenging due to geological heterogeneity and multiphase fluid-flow dynamics. Predicting behavior in this setting relies on high-fidelity physics-based simulations that are computationally expensive. Yet, the uncertain, heterogeneous properties that control these flows make it necessary to perform many of these expensive simulations, which is often prohibitive. To address these challenges, we introduce a physics-informed machine learning workflow that couples a fully differentiable multiphase flow simulator, which is implemented in the DPFEHM framework with a convolutional neural network (CNN). The CNN learns to predict fluid extraction rates from heterogeneous permeability fields to enforce pressure limits at critical reservoir locations. By incorporating transient multiphase flow physics into the training process, our method enables more practical and accurate predictions for realistic injection-extraction scenarios compared to previous works. To speed up training, we pretrain the model on single-phase, steady-state simulations and then finetune it on full multiphase scenarios, which dramatically reduces the computational cost. We demonstrate that high-accuracy training can be achieved with fewer than three thousand full-physics multiphase flow simulations – compared to previous estimates requiring up to ten million. This drastic reduction in the number of simulations is achieved by leveraging transfer learning from much less expensive single phase simulations.

25 ENERGY STORAGE↗
DOE OSTI2020

Analysis into Asymptotic Convergence to Full Nonlinear Solutions and Exploration of the Implication of Numerical Operator Mutation of Differential Systems

A robust, sufficiently accurate and practical hydrodynamic simulation toolset is required as a key component of the modeling and simulation of air-gap electrostatic discharge events. This work was performed to complement these ongoing efforts. In particular, hydrodynamic simulations must be vetted to ensure they are robust and sufficiently accurate over relevant characteristic scales. Verification models were generated in order to cultivate the technical knowledge and expertise needed to properly create, implement and execute numerical simulations. Furthermore, this effort was utilized extensively to educate students on the mathematical and numerical principles underlying hydrodynamic simulations. This education opportunity, provided in a holistic and rigorous manner, has greatly benefited developing scientists and engineers with the necessary understandings and toolsets required to excel at accomplishing the task at hand, and, more generally, it has enabled them to generate key programmatic deliverables. This report articulates several subtilties; specifically, how perturbations, nonlinear behavior, and dissipative mechanisms influence numerical stability, how to properly structure mathematical and numerical solutions, and how to properly generate error estimation/assignment. A more rigorous discussion of the consequences of such topics can be found in the body of this report in Chapters 2 and 3 with qualitative findings discussed in Chapter 4.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2024

Organic Evaporation, Oxidation, and Hydrolysis Testing in Support of Hanford Sample-and-Send

The Hanford site has approximately 54 to 56 million gallons of radioactive mixed waste stored in 156 unretrieved underground storage tanks. The Hanford Waste Treatment and Immobilization Plant (WTP) is being built to treat and immobilize the tank waste. The baseline method for immobilization of Low Activity Waste (LAW) through the WTP is vitrification, but additional immobilization capacity is needed to supplement the initial LAW melters. An alternative cementitious waste form is being investigated for that future immobilization method. However, one impediment to a cementitious waste form is the presence of Land Disposal Restricted (LDR) organic chemicals in tank waste, which are regulated on a concentration based standard in the final waste form. Hence, if the quantity of organics in LAW is high enough, they must be destroyed or removed to make a waste form compatible with disposal in a mixed low level waste landfill. This work evaluates potential avenues for treatment of LDR organics to eliminate the impediment and permit possible use of a cementitious waste form. Vacuum evaporation testing to remove LDR organics consisted of preparing a non-radioactive LAW simulant, spiking that simulant with organic chemicals, and evaporating the mixture via differential distillation. The apparatus was a laboratory-scale vacuum evaporator operated at 60 ±5 torr absolute (vacuum evaporation). The LAW simulant represented the liquid expected to be retrieved from the Hanford tank farms at approximately 4.0 M [Na+] total sodium ion concentration. The concentration of the organic chemicals added was significantly higher than typically found in the tank waste samples since the higher levels were necessary to assist in analytical measurement and tracking of the spiked species.

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗
DOE OSTI2023

Learning effective SDEs from Brownian dynamic simulations of colloidal particles

We construct a reduced, data-driven, parameter dependent effective stochastic differential equation (eSDE) for electric-field mediated colloidal crystallization using data obtained from Brownian dynamics simulations. We use diffusion maps (a manifold learning algorithm) to identify a set of useful latent observables. In this latent space we identify an eSDE using a deep learning architecture inspired by numerical stochastic integrators and compare it with the traditional Kramers–Moyal expansion estimation. We show that the obtained variables and the learned dynamics accurately encode the physics of the Brownian dynamic simulations. We further illustrate that our reduced model captures the dynamics of corresponding experimental data. Further, our dimension reduction/reduced model identification approach can be easily ported to a broad class of particle systems dynamics experiments/models.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2022

A new moment-based general-relativistic neutrino-radiation transport code: Methods and first applications to neutron star mergers

ABSTRACT We present a new moment-based energy-integrated neutrino transport code for neutron star merger simulations in general relativity. In the merger context, ours is the first code to include Doppler effects at all orders in υ/c, retaining all non-linear neutrino–matter coupling terms. The code is validated with a stringent series of tests. We show that the inclusion of full neutrino–matter coupling terms is necessary to correctly capture the trapping of neutrinos in relativistically moving media, such as in differentially rotating merger remnants. We perform preliminary simulations proving the robustness of the scheme in simulating ab-initio mergers to black hole collapse and long-term neutron star remnants up to ${\sim }70\,$ ms. The latter is the longest dynamical space-time, 3D, general relativistic simulations with full neutrino transport to date. We compare results obtained at different resolutions and using two different closures for the moment scheme. We do not find evidences of significant out-of-thermodynamic equilibrium effects, such as bulk viscosity, on the post-merger dynamics or gravitational wave emission. Neutrino luminosities and average energies are in good agreement with theory expectations and previous simulations by other groups using similar schemes. We compare dynamical and early wind ejecta properties obtained with M1 and with our older neutrino treatment. We find that the M1 results have systematically larger proton fractions. However, the differences in the nucleosynthesis yields are modest. This work sets the basis for future detailed studies spanning a wider set of neutrino reactions, binaries, and equations of state.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2024

GrainGNN: A dynamic graph neural network for predicting 3D grain microstructure

We propose GrainGNN, a surrogate model for the evolution of polycrystalline grain structure under rapid solidification conditions in metal additive manufacturing. High fidelity simulations of solidification microstructures are typically performed using multicomponent partial differential equations (PDEs) with moving interfaces. The inherent randomness of the PDE initial conditions (grain seeds) necessitates ensemble simulations to predict microstructure statistics, e.g., grain size, aspect ratio, and crystallographic orientation. Here, currently such ensemble simulations are prohibitively expensive and surrogates are necessary.In GrainGNN, we use a dynamic graph to represent interface motion and topological changes due to grain coarsening. We use a reduced representation of the microstructure using hand-crafted features; we combine pattern finding and altering graph algorithms with two neural networks, a classifier (for topological changes) and a regressor (for interface motion). Both networks have an encoder-decoder architecture; the encoder has a multi-layer transformer long-short-term-memory architecture; the decoder is a single layer perceptron.We evaluate GrainGNN by comparing it to high-fidelity phase field simulations for in-distribution and out-of-distribution grain configurations for solidification under laser power bed fusion conditions. GrainGNN results in 80%–90% pointwise accuracy; and nearly identical distributions of scalar quantities of interest (QoI) between phase field and GrainGNN simulations compared using Kolmogorov-Smirnov test. GrainGNN's inference speedup (PyTorch on single x86 CPU) over a high-fidelity phase field simulation (CUDA on a single NVIDIA A100 GPU) is 150×–2000× for 100-initial grain problem. Further, using GrainGNN, we model the formation of 11,600 grains in 220 seconds on a single CPU core.

36 MATERIALS SCIENCE↗
DOE OSTI2022

Computational Methods to Characterize Panel Loading Conditions for Accelerated Testing

Panel cracking and degradation due to wind loading are known to have a detrimental effect on power output. Recreating these damaging conditions in a controlled experimental setting requires an understanding of the loads being generated at the panel surface as a function of both wind speed, wind direction, and panel orientation. To better understand these relationships, a computational fluid dynamics (CFD) simulation package was constructed using an open-source Python library for solving partial differential equations with the finite element method. This CFD package allows the simulated wind speed and panel orientation of a multi-panel array to be easily changed and provides the capability to measure the traction forces at discrete points along the sun-facing and ground-facing surface of an interior panel residing in the wake of one or more upstream panels. A parameter exploration was performed in which the panel angle was varied in even increments from -70 deg to 70 deg and the wind speed was varied from 2 to 30 m/s. During post processing, the measured traction along the panel surface was averaged spatially and interpreted as a time-varying signal, where further processing of this signal yielded the root mean square amplitude and a characteristic frequency associated with the loading. This study found that higher wind speeds are generally associated with increased amplitude of loading and that the panel orientation angle can significantly exacerbate or mitigate this loading. These outcomes are presented along with current work on higher-fidelity verification simulations and recommendations for performing accelerated experimental testing.

loading↗
DOE OSTI2022

Modeling Dust Production, Growth, and Destruction in Reionization-era Galaxies with the CROC Simulations: Methods and Parameter Exploration

We introduce a model for the explicit evolution of interstellar dust in a cosmological galaxy formation simulation. We post-process a simulation from the Cosmic Reionization on Computers project, integrating an ordinary differential equation for the evolution of the dust-to-gas ratio along pathlines in the simulation sampled with a tracer particle technique. This model incorporates the effects of dust grain production in asymptotic giant branch star winds and supernovae, grain growth due to the accretion of heavy elements from the gas phase of the interstellar medium, and grain destruction due to thermal sputtering in the high-temperature gas of supernova remnants due to thermal sputtering. A main conclusion of our analysis is the importance of a carefully chosen dust destruction model, for which different reasonable parameterizations can predict very different values at the ∼100 pc resolution of the interstellar medium (ISM) in our simulations. We run this dust model on the single most massive galaxy in a 10 h$^{−1}$ comoving megaparsec box, which attains a stellar mass of ∼2 × 10$^{9}$ M $_{⊙}$ by z = 5. We find that the model is capable of reproducing dust masses and dust-sensitive observable quantities broadly consistent with existing data from high-redshift galaxies. The total dust mass in the simulated galaxy is somewhat sensitive to parameter choices for the dust model, especially the timescale for grain growth due to accretion in the ISM. Consequently, observable quantity observations that can constrain galaxy dust masses at these epochs are potentially useful for placing constraints on dust physics.

79 ASTRONOMY AND ASTROPHYSICS↗