Search NASA⌕ Search

SEARCH · Search NASA

Results for “Deep generative model”

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 649 records · Page 36

Extracting low energy signals from raw LArTPC waveforms using deep learning techniques -- A proof of concept

We investigate the feasibility of using deep learning techniques, in the form of a one-dimensional convolutional neural network (1D-CNN), for the extraction of signals from the raw waveforms produced by the individual channels of liquid argon time projection chamber (LArTPC) detectors. A minimal generic LArTPC detector model is developed to generate realistic noise and signal waveforms used to train and test the 1D-CNN, and evaluate its performance on low-level signals. We demonstrate that our approach overcomes the inherent shortcomings of traditional cut-based methods by extending sensitivity to signals with ADC values below their imposed thresholds. This approach exhibits great promise in enhancing the capabilities of future generation neutrino experiments like DUNE to carry out their low-energy neutrino physics programs.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Data-based filtered dissipation rate modelling for multi-modal turbulent combustion: evaluating a priori model generalizability

Manifold-based models offer a computationally efficient alternative to directly transporting the thermochemical state in computational simulations of turbulent reacting flows, projecting the high-dimensional thermochemical state-space onto a low-dimensional manifold. Recent efforts have yielded a manifold-based model applicable to multi-modal combustion, enabling reconstruction of the thermochemical state from solutions to two-dimensional manifold equations in mixture fraction and generalized progress variable that are parameterised by three scalar dissipation rates. In coarse-grained simulations such as Large Eddy Simulation (LES), closure of the multi-modal manifold equations and subfilter variances/covariance requires closure of three filtered scalar dissipation rates. Here, the present work adopts a data-based approach, providing closure for the three filtered scalar dissipation rates via deep neural networks (DNNs). High-fidelity datasets corresponding to an autoigniting n-dodecane jet flame and a bluff body swirl-stabilized confined lifted spray flame of two aviation fuels (Jet-A and C1) with different ignition propensities are leveraged to generate training data that spans a diverse range of thermodynamic conditions and combustion modes, including low- and high-temperature ignition regimes in addition to premixed and nonpremixed behaviour. A final DNN model is trained to enforce inherent physical constraints by learning nonlinear functional transformations of the three filtered scalar dissipation rates. The generalizability of this constrained DNN model is demonstrated a priori via conditional statistics evaluated on the lifted spray flame with C1–a configuration that had not been included in the training data. Excellent DNN agreement with conditional DNS statistics is observed, and integrated gradients are computed to identify the most sensitive input variables. The similarity of the marginal PDFs of the most informative input variables and outputs across configurations are quantified via the Wasserstein metric, demonstrating that data-based models may successfully generalize to unseen parametric conditions so long as the most informative input variables share similar distributions across training and testing datasets.

Data-based modelling↗

Quasi-Classical Trajectory Calculation of Rate Constants Using an Ab Initio Trained Machine Learning Model (aML-MD) with Multifidelity Data

Machine learning (ML) provides a great opportunity for the construction of models with improved accuracy in classical molecular dynamics (MD). However, the accuracy of a ML trained model is limited by the quality and quantity of the training data. Generating large sets of accurate ab initio training data can require significant computational resources. Furthermore, inconsistent or incompatible data with different accuracies obtained using different methods may lead to biased or unreliable ML models that do not accurately represent the underlying physics. Recently, transfer learning showed its potential for avoiding these problems as well as for improving the accuracy, efficiency, and generalization of ML models using multifidelity data. In this work, ab initio trained ML-based MD (aML-MD) models are developed through transfer learning using DFT and multireference data from multiple sources with varying accuracy within the Deep Potential MD framework. Further, the accuracy of the force field is demonstrated by calculating rate constants for the H + HO 2 → H 2 + 3 O 2 reaction using quasi-classical trajectories. We show that the aML-MD model with transfer learning can accurately predict the rate constants while reducing the computational cost by more than five times compared to the use of more expensive quantum chemistry training data sets. Hence, the aML-MD model with transfer learning shows great potential in using multifidelity data to reduce the computational cost involved in generating the training set for these potentials.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Demand-side solutions in the US building sector could achieve deep emissions reductions and avoid over $100 billion in power sector costs

Buildings are energy-intensive and a primary source of US end-use sector carbon emissions. Although building emissions today are 25% below their 2005 peak, far deeper reductions are needed to reach the US 2050 net-zero emissions goal. However, plausible decarbonization pathways that consider both buildings and their interactions with the power grid remain poorly understood. Here, we couple detailed modeling of building energy use and the grid to quantify building decarbonization potential and associated grid impacts. We find up to a 91% reduction in building CO 2 emissions from 2005 levels by 2050 using a portfolio of building efficiency, demand flexibility, and electrification measures alongside rapid grid decarbonization. Building efficiency and flexibility could generate up to $107 billion in annual power system cost savings by 2050, offsetting over a third of the incremental cost of full grid decarbonization. Our results underscore multiple benefits of demand-side solutions for deep decarbonization of US buildings.

29 ENERGY PLANNING, POLICY, AND ECONOMY↗

Heterogeneous energetic material damage simulator (HEDS): A deep learning approach to simulate damage–sensitivity linkages

Damage in the microstructures of energetic materials (EMs), such as propellants and plastic bonded explosives (PBXs), can significantly alter their response to external loads. Both sensitization and desensitization can occur, causing concerns with safety and performance in the field; predictive models that connect damage and the sensitivity of EMs can enable design and provide confidence in their robustness and reliability. However, modeling of damage evolution is challenging for real microstructures of EMs; samples of damaged EMs are difficult to obtain, thereby hindering experiments and direct numerical simulations to determine the sensitivity of EMs at various stages of damage. Here, we develop an approach to generate synthetic, i.e., in silico produced, damaged microstructures for use in simulations to connect damage levels to sensitivity. The development of the present workflow to generate and impose varying levels of damage in microstructures, known as HEDS (Heterogeneous Energetic Material Damage Simulator), begins with a small set of images of damaged PBXs and combines a collection of deep neural network techniques to generate microstructures with varying levels of damage. By making the synthetic microstructures conform closely to those observed in available real, imaged microstructures, we develop an ensemble of damaged microstructures that can be used for in silico shock experiments. HEDS develops these microstructure ensembles as level set fields, which are directly employed in a sharp interface Eulerian hydrocode where shock simulations are performed to quantify the energy release rate from hotspot fields generated in the microstructure. These capabilities can be useful for the analysis and assessment of changes in the sensitivity of EMs and to design formulations that are less susceptible to damage-induced changes in sensitivity and performance.

Fang, Irene (ORCID:0009000844557122)↗

Software design space exploration for exascale combustion co-design

The design of hardware for next-generation exascale computing systems will require a deep understanding of how software optimizations impact hardware design trade-offs. In order to characterize how co-tuning hardware and software parameters affects the performance of combustion simulation codes, we created ExaSAT, a compiler-driven static analysis and performance modeling framework. Our framework can evaluate hundreds of hardware/software configurations in seconds, providing an essential speed advantage over simulators and dynamic analysis techniques during the co-design process. Our analytic performance model shows that advanced code transformations, such as cache blocking and loop fusion, can have a significant impact on choices for cache and memory architecture. Our modeling helped us identify tuned configurations that achieve a 90% reduction in memory traffic, which could significantly improve performance and reduce energy consumption. These techniques will also be useful for the development of advanced programming models and runtimes, which must reason about these optimizations to deliver better performance and energy efficiency.

97 MATHEMATICS AND COMPUTING↗

Automatic Detection of Defects in High-Reliability Components

Disastrous consequences can result from defects in manufactured parts—particularly the high consequence parts developed at Sandia. Identifying flaws in as-built parts can be done with nondestructive means, such as X-ray Computed Tomography (CT). However, due to artifacts and complex imagery, the task of analyzing the CT images falls to humans. Human analysis is inherently unreproducible, unscalable, and can easily miss subtle flaws. We hypothesized that deep learning methods could improve defect identification, increase the number of parts that can effectively be analyzed, and do it in a reproducible manner. We pursued two methods: 1) generating a defect-free version of a scan and looking for differences (PandaNet), and 2) using pre-trained models to develop a statistical model of normality (Feature-based Anomaly Detection System: FADS). Both PandaNet and FADS provide good results, are scalable, and can identify anomalies in imagery. In particular, FADS enables zero-shot (training-free) identification of defects for minimal computational cost and expert time. It significantly outperforms prior approaches in computational cost while achieving comparable results. FADS’ core concept has also shown utility beyond anomaly detection by providing feature extraction for downstream tasks.

47 OTHER INSTRUMENTATION↗

Search for an Exotic S=-2, Q=-2 Baryon Resonance in Proton-Proton Interactions at $\sqrt{s_{NN}}$ = 17.3 GeV

Searches and analyses of strong gravitational lenses are challenging due to the rarity and image complexity of these astronomical objects. Next-generation surveys (both ground- and space-based) will provide more opportunities to derive science from these objects, but only if they can be analyzed on realistic time-scales. Currently, these analyses are expensive. In this work, we present a regression analysis with uncertainty estimates using deep learning models to measure four parameters of strong gravitational lenses in simulated Dark Energy Survey data. Using only $gri$-band images, we predict Einstein Radius, lens velocity dispersion, lens redshift to within $10-15\%$ of truth values and source redshift to $30\%$ of truth values, along with predictive uncertainties. This work helps to take a step along the path of faster analyses of strong lenses with deep learning frameworks.

79 ASTRONOMY AND ASTROPHYSICS↗

From Femtoseconds to Gigaseconds: The SolDeg Project to Analyze and Mitigate the Performance Degradation of a-Si:H/c-Si Heterojunction Solar Cells

We pursued two major projects. In the first, we studied the degradation of a-Si/c-Si Heterojunction cells. We substantially expanded our previously developed SolDeg platform, and incorporated hydrogen into it. We also expanded the capabilities of SolDeg by developing a three channel model for a very efficient way to determine long term junction dynamics. The first main result was that we identified hydrogen drifting away from the interface as a key degradation channel. The second was that we developed and proposed the concept of a “reverse silicon density gradient” in the a-Si layer that had the promise of reducing degradation by up to 80%. In the second project, we studied the degradation of TOPCon solar cells. We again identified a key degradation channel, which in these cells was the formation of pinholes across the oxide layer. We then again were able to develop and to propose a way to dramatically reduce degradation by discovering that tuning the hydrogen concentration forms a sharp and deep minimum of the recombination current.

14 SOLAR ENERGY↗

Evaluation of low-exergy heating and cooling systems and topology optimization for deep energy savings at the urban district level

District energy systems have the potential to achieve deep energy savings by leveraging the density and diversity of loads in urban districts. However, planning and adoption of district thermal energy systems is hindered by the analytical burden and high infrastructure costs. It is hypothesized that network topology optimization would enable wider adoption of advanced (ambient temperature) district thermal energy systems, resulting in energy savings. In this study, energy modeling is used to compare the energy performance of “conventional” and “advanced” district thermal energy systems at the urban district level, and a partial exhaustive search is used to evaluate a heuristic for the topology optimization problem. For the prototypical district considered, advanced district thermal energy systems mated with low-exergy building heating and cooling systems achieved a source energy use intensity that was 49% lower than that of conventional systems. The minimal spanning tree heuristic was demonstrated to be effective for the network topology optimization problem in the context of a prototypical district, and contributes to mitigating the problem’s computational complexity. The work presented in this paper demonstrates the potential of advanced district thermal energy systems to achieve deep energy savings, and advances to addressing barriers to their adoption through topology optimization.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

Hardening of DOE UQ Codes for Ultra-Deep Proactive Geosteering in Oilfield Service

The primary objective of this project was to harden the DOE uncertainty quantification (UQ) code to solve the inverse problem for the geosteering service in the oilfield drilling and exploration. The primary focus is to verify the feasibility of using QUESO to solve geosteering inverse problems through statistical inversion. QUESO is a software library developed by QUEST, a DOE's Scientific Discovery through Advanced Computing (SciDAC) institute. In the process of drilling a borehole, geosteering is the act of adjusting the well trajectory on the fly to reach the geological targets or to keep the wellbore within desired formation. Geosteering inversion is to infer underground formation structure using measurements such as azimuthal resistivity logging data. The geosteering process relies on the real-time reconstruction of the subsurface earth model, and a robust and efficient inversion algorithm is always necessary to satisfy the job requirement. The new generation of azimuthal resistivity geosteering tool is capable of detecting multiple formation boundaries 20-100 ft away, and the corresponding geosteering inversion problem can be very challenging as many parameters need to be inverted within such a large depth of investigation. The success of the project will support US energy security by helping to advance subsurface modeling and characterization technologies in oilfield applications. An effective subsurface inversion algorithm and the implementation as a software package will promote the usage of ultra-deep azimuthal resistivity measurements, which can help well placement in complicated underground environments, reduce costs and risks of drilling, and increase oil and gas production. It will help medium and small size independent service companies and tool manufactures to enter the more advanced and lucrative directional drilling area. In addition, by hardening the QUESO code and building a transparent interface for easy manipulation of the QUESO functions, we provide a user-friendly tool for a larger scientific and industry communities that rely heavily on modeling and simulation tools for forecasting, parameter studies, design, and decision making. These are all activities that can strongly benefit from meaningful representations of modeling uncertainty supported by our product.

58 GEOSCIENCES↗

A Comparative Study on Formability of the Third-Generation Automotive Medium-Mn Steel and 22MnB5 Steel

Third-generation advanced automotive medium-Mn steel, which can replace 22MnB5 steel, was newly developed to improve the lightweight and crashworthiness of automobile. Studies on the formability and simulation method of medium-Mn steel have just been initiated. In this study, finite element simulation models of square-cup deep drawing were established based on various material property experiments and validated by experiments. The effects of blank holder force (BHF), fillet radii of tools (die and punch) on the maximum drawing depth (MDD), thickness distribution of the formed products, and the microstructure before and after forming were investigated and compared with those on 22MnB5 steel. Results show that the MDD of the two steels decreased with increased BHF but increased with the fillet radius of punch; however, the fillet radius of die showed no significant effect on the MDD for both steels. Compared with hot-formed 22MnB5 steel, the martensitic transformation of the hot-formed medium-Mn steel is rarely influenced by the process parameters; thus, it holds the complete, fine-grained, and uniform martensitic microstructure. Moreover, the medium-Mn has better formability, lower initial blank temperature, and smaller impact of BHF and fillet radius of tools on the hot-formed product. Thus, a theoretical basis for the replacement of 22MnB5 steel by medium-Mn steel in hot forming process is provided.

36 MATERIALS SCIENCE↗

Accelerating Traction Motor Optimization Design with AI Surrogate Models

The advancement of artificial intelligence systems enables the use of data-driven physics-based surrogate models to explore design spaces rapidly and deeply for engineering projects. This work presents a surrogate model workflow that accelerates electric traction motor design optimization by replacing finite element analysis (FEA) with an artificial neural network (ANN) and using this model in a genetic algorithm for design optimization. A baseline interior permanent-magnet motor is parameterized and sampled to generate FEA-labeled training data, after which a feed-forward ANN predicts key outputs (e.g., loss components and weight). The validated surrogate enables genetic-algorithm optimization and deep search over the design space without new FEA runs, producing Pareto-optimal trade-offs between weight and losses and set of optimized designs for rapid downselection of manufacturable motor designs.

Ribeiro, Pedro [ORNL] (ORCID:0009000921026641)↗

Constraining the Stellar-to-Halo Mass Relation with Galaxy Clustering and Weak Lensing from DES Year 3 Data

We develop a framework to study the relation between the stellar mass of a galaxy and the total mass of its host dark matter halo using galaxy clustering and galaxy-galaxy lensing measurements. We model a wide range of scales, roughly from $\sim 100 \; {\rm kpc}$ to $\sim 100 \; {\rm Mpc}$, using a theoretical framework based on the Halo Occupation Distribution and data from Year 3 of the Dark Energy Survey (DES) dataset. The new advances of this work include: 1) the generation and validation of a new stellar mass-selected galaxy sample in the range of $\log M_\star/M_\odot \sim 9.6$ to $\sim 11.5$; 2) the joint-modeling framework of galaxy clustering and galaxy-galaxy lensing that is able to describe our stellar mass-selected sample deep into the 1-halo regime; and 3) stellar-to-halo mass relation (SHMR) constraints from this dataset. In general, our SHMR constraints agree well with existing literature with various weak lensing measurements. We constrain the free parameters in the SHMR functional form $\log M_\star (M_h) = \log(εM_1) + f\left[ \log\left( M_h / M_1 \right) \right] - f(0)$, with $f(x) \equiv -\log(10^{αx}+1) + δ[\log(1+\exp(x))]^γ/ [1+\exp(10^{-x})]$, to be $\log M_1 = 11.506^{+0.325}_{-0.404}$, $\log ε= -1.632^{+0.306}_{-0.181}$, $α= -1.638^{+0.108}_{-0.099}$, $γ= 0.596^{+0.251}_{-0.210}$ and $δ= 3.810^{+2.045}_{-1.811}$. The inferred average satellite fraction is within $\sim 5-35\%$ for our fiducial results and we do not see any clear trends with redshift or stellar mass. Furthermore, we find that the inferred average galaxy bias values follow the generally expected trends with stellar mass and redshift. Our study is the first SHMR in DES in this mass range, and we expect the stellar mass sample to be of general interest for other science cases.

Zacharegkas, G. [Argonne; Chicago U., KICP] (ORCID↗

DL-TODA: A Deep Learning Tool for Omics Data Analysis

Metagenomics is a technique for genome-wide profiling of microbiomes; this technique generates billions of DNA sequences called reads. Given the multiplication of metagenomic projects, computational tools are necessary to enable the efficient and accurate classification of metagenomic reads without needing to construct a reference database. The program DL-TODA presented here aims to classify metagenomic reads using a deep learning model trained on over 3000 bacterial species. A convolutional neural network architecture originally designed for computer vision was applied for the modeling of species-specific features. Using synthetic testing data simulated with 2454 genomes from 639 species, DL-TODA was shown to classify nearly 75% of the reads with high confidence. The classification accuracy of DL-TODA was over 0.98 at taxonomic ranks above the genus level, making it comparable with Kraken2 and Centrifuge, two state-of-the-art taxonomic classification tools. DL-TODA also achieved an accuracy of 0.97 at the species level, which is higher than 0.93 by Kraken2 and 0.85 by Centrifuge on the same test set. Application of DL-TODA to the human oral and cropland soil metagenomes further demonstrated its use in analyzing microbiomes from diverse environments. Compared to Centrifuge and Kraken2, DL-TODA predicted distinct relative abundance rankings and is less biased toward a single taxon.

59 BASIC BIOLOGICAL SCIENCES↗

Mitigating Algorithmic Bias in Cancer Site Classification Models

Purpose Integrating artificial intelligence in cancer diagnostics has improved tumor classification beyond rule-based systems. Despite these advancements, these models may still encode demographic biases. We conducted a large-scale, applied bias-probing study of a deep learning–based cancer site classifier to quantify race information encoded in document embeddings. We then evaluated how performance changes when race-correlated embedding dimensions are removed in a post-training sensitivity analysis. Methods The cancer site classifier was trained using 3.5 million electronic cancer pathology reports from six of the National Cancer Institute's SEER registries. We trained a hierarchical self-attention network to generate 400-dimensional document embeddings. These embeddings were used to train two downstream, gradient-boosted decision tree classifiers: one to classify the cancer sites and another to predict racial categories. We identified overlapping features by intersecting the top 50 feature-importance rankings from the site and race models and computed their cumulative feature importance in each model. As a post hoc sensitivity analysis, we progressively pruned these overlapping dimensions, retrained the site model, and compared overall macro-F1 and accuracy, race-stratified macro-F1, and group fairness metrics on the basis of demographic parity and equalized odds before and after pruning. Results The analysis revealed minimal feature overlap between the cancer site and race prediction models, and the cumulative importance scores indicated a negligible influence of racial information on clinical predictions. Post-training pruning of overlapping features did not compromise the models' diagnostic accuracy, with a 0.07% loss in accuracy. Conclusion Our findings demonstrate that HiSAN-generated embeddings from SEER data can be used effectively in cancer site classification without significant demographic bias influencing the outcomes. Post-training pruning therefore functions as a practical audit and sensitivity check.

Shivanna, Abhishek [ORNL] (ORCID:0009000665228593)↗

Modern deep neural networks for Direct Normal Irradiance forecasting: A classification approach

The escalating energy demand and the adverse environmental impacts of fossil-fuel use necessitate a shift towards cleaner and renewable alternatives. Concentrated Solar Power (CSP) technology emerges as a promising solution, offering a carbon-free alternative for power generation. The efficiency and profitability of CSP depend on the Direct Normal Irradiance (DNI) component of solar radiation; hence, accurate DNI forecasting can help optimize CSP plants’ operations and performance. The unpredictable nature of weather phenomena, particularly cloud cover, introduces uncertainty into DNI projections. Existing DNI forecasting models use meteorological factors, which are both challenging to estimate numerically over short prediction windows and expensive to model through data at a sufficiently high spatial and temporal resolution. This research addresses the challenge by presenting a novel approach that formulates DNI prediction as a multi-class classification problem, departing from conventional regression-based methods. The primary objective of this classification framework is to identify optimal periods aligning with specific operational thresholds for CSP plants, contributing to enhanced dispatch optimization strategies. We model the DNI classification problem using four advanced deep neural networks – rectified linear unit (ReLU) networks, 1D residual networks (ResNets), bidirectional long short-term memory (BiLSTM) networks, and transformers – achieving accuracies up to 93.5% without requiring meteorological parameters.

14 SOLAR ENERGY↗