Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Neural Network”

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 847 records · Page 47

Neural Networks Used to Compare Designed and Measured Time-Average Patterns

Electronic time-average holograms are convenient for comparing the measured vibration modes of fan blades with those calculated by finite-element models. At the NASA Lewis Research Center, neural networks recently were trained to perform what had been a simple visual comparison of the predictions of the design models with the measurements. Finite-element models were used to train neural networks to recognize damage and strain information encoded in subtle changes in the time-average patterns of cantilevers. But the design-grade finite element models were unable to train the neural networks to detect damage in complex blade shapes. The design-model-generated patterns simply did not agree well enough with the measured patterns. Instead, hybrid-training records, with measured time-average patterns as the input and model-generated strain information as the output, were used to effect successful training.

Decker, Arthur J.↗

Neural Network Control of a Magnetically Suspended Rotor System

Abstract Magnetic bearings offer significant advantages because of their noncontact operation, which can reduce maintenance. Higher speeds, no friction, no lubrication, weight reduction, precise position control, and active damping make them far superior to conventional contact bearings. However, there are technical barriers that limit the application of this technology in industry. One of them is the need for a nonlinear controller that can overcome the system nonlinearity and uncertainty inherent in magnetic bearings. This paper discusses the use of a neural network as a nonlinear controller that circumvents system nonlinearity. A neural network controller was well trained and successfully demonstrated on a small magnetic bearing rig. This work demonstrated the feasibility of using a neural network to control nonlinear magnetic bearings and systems with unknown dynamics.

Choi, Benjamin↗

Deployment of Dynamic Neural Network Optimization to Minimize Heat Rate During Ramping for Coal Power Plants (Final Technical Report)

Much success was achieved throughout the course of this project. A successful implementation of Dynamic Neural Network Optimization (D-NNO) was coupled with Adaptive Predictive Controls (APC) and a novel hardware installation comprised of an advanced sensor network (ASN) measuring mass-weighted averages of flue gas constituents above the horizontal superheater of a coal-fired utility boiler. From 2019 through 2023 (including an extension due to COVID delays), the team was able to prototype, evaluate, deploy, iterate, and ultimately finalize an advanced closed-loop control D-NNO system which demonstrated the ability to: •improve unit efficiency ~2.0% relative to unoptimized operation (represented as total fuel fired per MWh generated) •improve unit NOx emission rates 10%+ beyond static optimization baselines •improve unit temperature stability as much as 58% and on average 12% •improve operating load stability as much as 35% The culmination of this project has generated an advanced methodology of deploying specially designed recurrent neural networks (long short-term memory, gated recurrent unit, encoder-decoder networks, transformers, etc.), customized trajectory planning and closed-loop optimization modules capable of adapting to live electric grid responses and demands, self-tuning and adaptive expert controls constantly adjusting prediction parameters to real-time unit behavior, and a hardware/software package able to reliably calculate net unit heat rate (NUHR) in real-time using flue gas constituents, machine learning, and known combustion relationships. Through this real-time NUHR value, immediate feedback on system adjustments relative to operating efficiency was available, allowing for rapid improvements to system performance. In addition to development and deployment of the advanced D-NNO system, the approach methodology has been readily commercialized through the project platform Griffin Open Systems, LLC, the D-NNO software platform host. Similar methodologies to those developed by this project have already been deployed at 5 other units across the United States, with another 6 implementations scheduled, and more expected. Over the course of the project, multiple academic papers were submitted and accepted for publication within esteemed academic journals, and PhD students were trained and graduated, as well as undergraduate students becoming involved and participating to project objectives.

01 COAL, LIGNITE, AND PEAT↗

An optimization of the FPGA trigger based on the artificial neural network for a detection of neutrino-origin showers

Observations of ultra-high energy neutrinos became a priority in experimental astro-particle physics. Up to now, the Pierre Auger Observatory did not find any candidate on a neutrino event. This imposes competitive limits to the diffuse flux of ultra-high energy neutrinos in the EeV range and above. A very low rate of events potentially generated by neutrinos is a significant challenge for a detection technique and requires both sophisticated algorithms and high-resolution hardware. A trigger based on a artificial neural network was implemented into the Cyclone{sup R} V E FPGA 5CEFA9F31I7. The prototype Front-End boards for Auger-Beyond-2015 with Cyclone{sup R} V E can test the neural network algorithm in real pampas conditions in 2015. Showers for muon and tau neutrino initiating particles on various altitudes, angles and energies were simulated in CORSICA and Offline platforms giving pattern of ADC traces in Auger water Cherenkov detectors. The 3-layer 12-10-1 neural network was taught in MATLAB by simulated ADC traces according the Levenberg-Marquardt algorithm. Results show that a probability of a ADC traces generation is very low due to a small neutrino cross-section. Nevertheless, ADC traces, if occur, for 1-10 EeV showers are relatively short and can be analyzed by 16-point input algorithm. For 100 EeV range traces are much longer, but with significantly higher amplitudes, which can be detected by standard threshold algorithms. We optimized the coefficients from MATLAB to get a maximal range of potentially registered events and for fixed-point FPGA processing to minimize calculation errors. Currently used Front-End boards based on no-more produced ACEXR PLDs and obsolete Cyclone{sup R} FPGAs allow an implementation of relatively simple threshold algorithms for triggers. New sophisticated trigger implemented in Cyclone{sup R} V E FPGAs with large amount of DSP blocks, embedded memory running with 120 - 160 MHz sampling may support to discover neutrino events in the Pierre Auger Observatory. (authors)

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Limitations of opto-electronic neural networks

Consideration is given to the limitations of implementing neurons, weights, and connections in neural networks for electronics and optics. It is shown that the advantages of each technology are utilized when electronically fabricated neurons are included and a combination of optics and electronics are employed for the weights and connections. The relationship between the types of neural networks being constructed and the choice of technologies to implement the weights and connections is examined.

Yu, Jeffrey↗

Reducing model bias in a deep learning classifier using domain adversarial neural networks in the MINERvA experiment

We present a simulation-based study using deep convolutional neural networks (DCNNs) to identify neutrino interaction vertices in the MINERvA passive targets region, and illustrate the application of domain adversarial neural networks (DANNs) in this context. DANNs are designed to be trained in one domain (simulated data) but tested in a second domain (physics data) and utilize unlabeled data from the second domain so that during training only features which are unable to discriminate between the domains are promoted. MINERvA is a neutrino-nucleus scattering experiment using the NuMI beamline at Fermilab. A-dependent cross sections are an important part of the physics program, and these measurements require vertex finding in complicated events. To illustrate the impact of the DANN we used a modified set of simulation in place of physics data during the training of the DANN and then used the label of the modified simulation during the evaluation of the DANN. In conclusion, we find that deep learning based methods offer significant advantages over our prior track-based reconstruction for the task of vertex finding, and that DANNs are able to improve the performance of deep networks by leveraging available unlabeled data and by mitigating network performance degradation rooted in biases in the physics models used for training.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Neural networks application to divergence-based passive ranging

The purpose of this report is to summarize the state of knowledge and outline the planned work in divergence-based/neural networks approach to the problem of passive ranging derived from optical flow. Work in this and closely related areas is reviewed in order to provide the necessary background for further developments. New ideas about devising a monocular passive-ranging system are then introduced. It is shown that image-plan divergence is independent of image-plan location with respect to the focus of expansion and of camera maneuvers because it directly measures the object's expansion which, in turn, is related to the time-to-collision. Thus, a divergence-based method has the potential of providing a reliable range complementing other monocular passive-ranging methods which encounter difficulties in image areas close to the focus of expansion. Image-plan divergence can be thought of as some spatial/temporal pattern. A neural network realization was chosen for this task because neural networks have generally performed well in various other pattern recognition applications. The main goal of this work is to teach a neural network to derive the divergence from the imagery.

Barniv, Yair↗

Surface daytime net radiation estimation using artificial neural networks

Net all-wave surface radiation (R n ) is one of the most important fundamental parameters in various applications. However, conventional R n measurements are difficult to collect because of the high cost and ongoing maintenance of recording instruments. Therefore, various empirical R n estimation models have been developed. This study presents the results of two artificial neural network (ANN) models (general regression neural networks (GRNN) and Neuroet) to estimate R n globally from multi-source data, including remotely sensed products, surface measurements, and meteorological reanalysis products. R n estimates provided by the two ANNs were tested against in-situ radiation measurements obtained from 251 global sites between 1991–2010 both in global mode (all data were used to fit the models) and in conditional mode (the data were divided into four subsets and the models were fitted separately). Based on the results obtained from extensive experiments, it has been proved that the two ANNs were superior to linear-based empirical models in both global and conditional modes and that the GRNN performed better and was more stable than Neuroet. The GRNN estimates had a determination coefficient (R 2 ) of 0.92, a root mean square error (RMSE) of 34.27 W·m –2 , and a bias of –0.61 W·m –2 in global mode based on the validation dataset. In conclusion, ANN methods are a potentially powerful tool for global R n estimation.

14 SOLAR ENERGY↗

Approximation of nearly-periodic symplectic maps via structure-preserving neural networks

A continuous-time dynamical system with parameter ε is nearly-periodic if all its trajectories are periodic with nowhere-vanishing angular frequency as ε approaches 0. Nearly-periodic maps are discrete-time analogues of nearly-periodic systems, defined as parameter-dependent diffeomorphisms that limit to rotations along a circle action, and they admit formal U(1) symmetries to all orders when the limiting rotation is non-resonant. For Hamiltonian nearly-periodic maps on exact presymplectic manifolds, the formal U(1) symmetry gives rise to a discrete-time adiabatic invariant. In this paper, we construct a novel structure-preserving neural network to approximate nearly-periodic symplectic maps. This neural network architecture, which we call symplectic gyroceptron, ensures that the resulting surrogate map is nearly-periodic and symplectic, and that it gives rise to a discrete-time adiabatic invariant and a long-time stability. This new structure-preserving neural network provides a promising architecture for surrogate modeling of non-dissipative dynamical systems that automatically steps over short timescales without introducing spurious instabilities.

97 MATHEMATICS AND COMPUTING↗

Backflow Transformations via Neural Networks for Quantum Many-Body Wave Functions

Obtaining an accurate ground state wave function is one of the great challenges in the quantum many-body problem. In this Letter, we propose a new class of wave functions, neural network backflow (NNB). Here, the backflow approach, pioneered originally by Feynman and Cohen [Phys. Rev. 102, 1189 (1956)], adds correlation to a mean-field ground state by transforming the single-particle orbitals in a configuration-dependent way. NNB uses a feed-forward neural network to learn the optimal transformation via variational Monte Carlo calculations. NNB directly dresses a mean-field state, can be systematically improved, and directly alters the sign structure of the wave function. It generalizes the standard backflow [L. F. Tocchio et al., Phys. Rev. B 78, 041101(R) (2008)], which we show how to explicitly represent as a NNB. We benchmark the NNB on Hubbard models at intermediate doping, finding that it significantly decreases the relative error, restores the symmetry of both observables and single-particle orbitals, and decreases the double-occupancy density. Finally, we illustrate interesting patterns in the weights and bias of the optimized neural network.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗

Poster Abstract: Investigating Occupancy Profiles Using Convolutional Neural Networks: Preprint

In this paper, we implement a convolutional neural networks based autoencoder to investigate occupancy profiles. We use the American time use survey data, which contains 191,558 schedules with binary occupancy information. Our results suggest that the trained filters provide an important insight of occupancy profiles (i.e., dominant and distinct patterns), and the latent space compresses the profiles with representative information.

14 SOLAR ENERGY↗

Data efficiency and extrapolation trends in neural network interatomic potentials

Abstract Recently, key architectural advances have been proposed for neural network interatomic potentials (NNIPs), such as incorporating message-passing networks, equivariance, or many-body expansion terms. Although modern NNIP models exhibit small differences in test accuracy, this metric is still considered the main target when developing new NNIP architectures. In this work, we show how architectural and optimization choices influence the generalization of NNIPs, revealing trends in molecular dynamics (MD) stability, data efficiency, and loss landscapes. Using the 3BPA dataset, we uncover trends in NNIP errors and robustness to noise, showing these metrics are insufficient to predict MD stability in the high-accuracy regime. With a large-scale study on NequIP, MACE, and their optimizers, we show that our metric of loss entropy predicts out-of-distribution error and data efficiency despite being computed only on the training set. This work provides a deep learning justification for probing extrapolation and can inform the development of next-generation NNIPs.

36 MATERIALS SCIENCE↗

Neural Network Prediction of Strong Lensing Systems with Domain Adaptation and Uncertainty Quantification

Modeling strong gravitational lenses is computationally expensive for the complex data from modern and next-generation cosmic surveys. Deep learning has emerged as a promising approach for finding lenses and predicting lensing parameters, such as the Einstein radius. Mean-variance Estimators (MVEs) are a common approach for obtaining aleatoric (data) uncertainties from a neural network prediction. However, neural networks have not been demonstrated to perform well on out-of-domain target data successfully - e.g., when trained on simulated data and applied to real, observational data. In this work, we perform the first study of the efficacy of MVEs in combination with unsupervised domain adaptation (UDA) on strong lensing data. The source domain data is noiseless, and the target domain data has noise mimicking modern cosmology surveys. We find that adding UDA to MVE increases the accuracy on the target data by a factor of about two over an MVE model without UDA. Including UDA also permits much more well-calibrated aleatoric uncertainty predictions. Advancements in this approach may enable future applications of MVE models to real observational data.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Accelerating defect predictions in semiconductors using graph neural networks

First-principles computations reliably predict the energetics of point defects in semiconductors but are constrained by the expense of using large supercells and advanced levels of theory. Machine learning models trained on computational data, especially ones that sufficiently encode defect coordination environments, can be used to accelerate defect predictions. Here, we develop a framework for the prediction and screening of native defects and functional impurities in a chemical space of group IV, III–V, and II–VI zinc blende semiconductors, powered by crystal Graph-based Neural Networks (GNNs) trained on high-throughput density functional theory (DFT) data. Using an innovative approach of sampling partially optimized defect configurations from DFT calculations, we generate one of the largest computational defect datasets to date, containing many types of vacancies, self-interstitials, anti-site substitutions, impurity interstitials and substitutions, as well as some defect complexes. We applied three types of established GNN techniques, namely crystal graph convolutional neural network, materials graph network, and Atomistic Line Graph Neural Network (ALIGNN), to rigorously train models for predicting defect formation energy (DFE) in multiple charge states and chemical potential conditions. We find that ALIGNN yields the best DFE predictions with root mean square errors around 0.3 eV, which represents a prediction accuracy of 98% given the range of values within the dataset, improving significantly on the state-of-the-art. We further show that GNN-based defective structure optimization can take us close to DFT-optimized geometries at a fraction of the cost of full DFT. The current models are based on the semi-local generalized gradient approximation-Perdew–Burke–Ernzerhof (PBE) functional but are highly promising because of the correlation of computed energetics and defect levels with higher levels of theory and experimental data, the accuracy and necessity of discovering novel metastable and low energy defect structures at the PBE level of theory before advanced methods could be applied, and the ability to train multi-fidelity models in the future with new data from non-local functionals. The DFT-GNN models enable prediction and screening across thousands of hypothetical defects based on both unoptimized and partially optimized defective structures, helping identify electronically active defects in technologically important semiconductors.

Rahman, Md Habibur (ORCID:000000027705984X)↗

Compressing deep neural networks on FPGAs to binary and ternary precision with hls4ml

We present the implementation of binary and ternary neural networks in the hls4ml library, designed to automatically convert deep neural network models to digital circuits with FPGA firmware. Starting from benchmark models trained with floating point precision, we investigate different strategies to reduce the network's resource consumption by reducing the numerical precision of the network parameters to binary or ternary. We discuss the trade-off between model accuracy and resource consumption. In addition, we show how to balance between latency and accuracy by retaining full precision on a selected subset of network components. As an example, we consider two multiclass classification tasks: handwritten digit recognition with the MNIST data set and jet identification with simulated proton-proton collisions at the CERN Large Hadron Collider. The binary and ternary implementation has similar performance to the higher precision implementation while using drastically fewer FPGA resources.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Fully convolutional neural network for removing background in noisy images of uranium bearing particles

A fully convolutional neural network (FCN) was developed to supersede automatic or manual thresholding algorithms used for tabulating SIMS particle search data. The FCN was designed to perform a binary classification of pixels in each image belonging to a particle or not, thereby effectively removing background signal without manually or automatically determining an intensity threshold. Using 8,000 images from 28 different particle screening analyses, the FCN was trained to accurately predict pixels belonging to a particle with near 99% accuracy. Background eliminated images were then segmented using a watershed technique in order to determine isotopic ratios of particles. A comparison of the isotopic distributions of an independent data set segmented using the neural network, compared to a commercially available automated particle measurement (APM) program developed by CAMECA, highlighted the necessity for effective background removal to ensure that resulting particle identification is not only accurate, but preserves valuable signal that could be lost due to improper segmentation. The FCN approach improves the robustness of current state-of-the-art particle searching algorithms by reducing user input biases, resulting in an improved absolute signal per particle and decreased uncertainty of the determined isotope ratios.

Tarolli, Jay G. (ORCID:0000000307369057)↗

Bayesian Optimized Deep Ensemble for Uncertainty Quantification of Deep Neural Networks: a System Safety Case Study on Sodium Fast Reactor Thermal Stratification Modeling

Deep neural networks (DNNs) are increasingly important to scientific computing and engineering system simulations. Accurate uncertainty quantification (UQ) for DNNs is critical in safety-sensitive engineering domains. Traditional Deep Ensemble (DE) methods, while easy to implement, frequently suffer from poorly calibrated uncertainty estimates and limited predictive accuracy due to reliance on fixed architectures with varied weight initializations. To address these issues, we introduce a workflow that combines Bayesian Optimization (BO) and DE. The workflow is modular, scalable, and integrates parallel BO initialized with Sobol sequences to individually optimize the hyperparameters of each ensemble member. This method enhances ensemble diversity, improves predictive accuracy, and provides reliable uncertainty estimates. We evaluate the proposed BODE approach in a sodium fast reactor thermal stratification modeling case study, where we used a densely connected convolutional neural network to predict turbulent viscosity during the reactor transient with consideration of data noise. We benchmark its performance against several optimization approaches, including baseline deep ensemble, evolutionary algorithm-optimized ensemble, ensemble formed via random search combined with greedy selection, and a BO ensemble using random initialization. Here, our results demonstrate superior performance of the developed BODE approach. In noise-free scenarios, BODE notably reduces incorrect aleatoric uncertainty and significantly enhances predictive accuracy. Under conditions of 5% and 10% Gaussian noise, BODE adaptively quantifies uncertainty proportional to data noise, achieving up to an 80% reduction in root mean square error compared to baseline methods and producing well-calibrated prediction intervals.

Bayesian optimization↗

Benchmarking the Performance of Neuromorphic and Spiking Neural Network Simulators

Software simulators play a critical role in the development of new algorithms and system architectures in any field of engineering. Neuromorphic computing, which has shown potential in building brain-inspired energy-efficient hardware, suffers a slow-down in the development cycle due to a lack of flexible and easy-to-use simulators of either neuromorphic hardware itself or of spiking neural networks (SNNs), the type of neural network computation executed on most neuromorphic systems. While there are several openly available neuromorphic or SNN simulation packages developed by a variety of research groups, they have mostly targeted computational neuroscience simulations, and only a few have targeted small-scale machine learning tasks with SNNs. Evaluations or comparisons of these simulators have often targeted computational neuroscience-style workloads. In this work, we seek to evaluate the performance of several publicly available SNN simulators with respect to non-computational neuroscience workloads, in terms of speed, flexibility, and scalability. We evaluate the performance of the NEST, Brian2, Brian2GeNN, BindsNET and Nengo packages under a common front-end neuromorphic framework. Our evaluation tasks include a variety of different network architectures and workload types to mimic the computation common in different algorithms, including feed-forward network inference, genetic algorithms, and reservoir computing. We also study the scalability of each of these simulators when running on different computing hardware, from single core CPU workstations to multi-node supercomputers. Our results show that the BindsNET simulator has the best speed and scalability for most of the SNN workloads (sparse, dense, and layered SNN architectures) on a single core CPU. However, when comparing the simulators leveraging the GPU capabilities, Brian2GeNN outperforms the others for these workloads in terms of scalability. NEST performs the best for small sparse networks and is also the most flexible simulator in terms of reconfiguration capability NEST shows a speedup of at least 2x compared to the other packages when running evolutionary algorithms for SNNs. The multi-node and multi-thread capabilities of NEST show at least 2x speedup compared to the rest of the simulators (single core CPU or GPU based simulators) for large and sparse networks. We conclude our work by providing a set of recommendations on the suitability of employing these simulators for different tasks and scales of operations. We also present the characteristics for a future generic ideal SNN simulator for different neuromorphic computing workloads.

97 MATHEMATICS AND COMPUTING↗